Add copies of the JPEG, PNG, and Z libraries...

git-svn-id: file:///fltk/svn/fltk/branches/branch-1.1@3635 ea41ed52-d2ee-0310-a9c1-e6b18d33e121

135 files changed, 71836 insertions, 0 deletions

diff --git a/CHANGES b/CHANGES
index 147cd9de9..6a8ea585e 100644
--- a/CHANGES
+++ b/CHANGES
@@ -1,6 +1,9 @@
 CHANGES IN FLTK 1.1.5rc2
 
 	- Documentation updates (STR #365, STR #399, STR #412)
+	- FLTK now includes copies of the PNG, JPEG, and ZLIB
+	  libraries for platforms that do not have them (STR
+	  #441)
 	- The fltk-config script did not include the
 	  "-mno-cygwin" option under CygWin (STR #434)
 	- Fl_Help_View::find() did not check for a NULL value
diff --git a/jpeg/Makefile b/jpeg/Makefile
new file mode 100644
index 000000000..b02b1d66a
--- /dev/null
+++ b/jpeg/Makefile
@@ -0,0 +1,85 @@
+#
+# "$Id: Makefile,v 1.1.2.1 2004/07/05 14:12:36 easysw Exp $"
+#
+# JPEG library makefile for the Fast Light Toolkit (FLTK).
+#
+# Copyright 1997-2004 by Easy Software Products.
+#
+# This library is free software; you can redistribute it and/or
+# modify it under the terms of the GNU Library General Public
+# License as published by the Free Software Foundation; either
+# version 2 of the License, or (at your option) any later version.
+#
+# This library is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# Library General Public License for more details.
+#
+# You should have received a copy of the GNU Library General Public
+# License along with this library; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+# USA.
+#
+# Please report all bugs and problems to "fltk-bugs@fltk.org".
+#
+
+include ../makeinclude
+
+
+#
+# Object files...
+#
+
+OBJS	=	jmemnobs.o \
+		jcapimin.o jcapistd.o jccoefct.o jccolor.o jcdctmgr.o \
+		jchuff.o jcinit.o jcmainct.o jcmarker.o jcmaster.o jcomapi.o \
+		jcparam.o jcphuff.o jcprepct.o jcsample.o jctrans.o \
+		jdapimin.o jdapistd.o jdatadst.o jdatasrc.o jdcoefct.o \
+		jdcolor.o jddctmgr.o jdhuff.o jdinput.o jdmainct.o jdmarker.o \
+		jdmaster.o jdmerge.o jdphuff.o jdpostct.o jdsample.o \
+		jdtrans.o jerror.o jfdctflt.o jfdctfst.o jfdctint.o \
+		jidctflt.o jidctfst.o jidctint.o jidctred.o jquant1.o \
+		jquant2.o jutils.o jmemmgr.o
+
+
+#
+# Make all targets...
+#
+
+all:	libjpeg.a
+
+
+#
+# Clean all targets and object files...
+#
+
+clean:
+	$(RM) $(OBJS)
+	$(RM) libjpeg.a
+
+
+#
+# libjpeg.a
+#
+
+libjpeg.a:	$(OBJS)
+	echo Archiving $@...
+	$(RM) $@
+	$(AR) $(ARFLAGS) $@ $(OBJS)
+	$(RANLIB) $@
+
+
+#
+# Make dependencies...
+#
+
+depend:	$(OBJS:.o=.c)
+	makedepend -Y -I.. -f makedepend $(OBJS:.o=.c)
+
+include makedepend
+
+$(OBJS):	../makeinclude
+
+#
+# End of "$Id: Makefile,v 1.1.2.1 2004/07/05 14:12:36 easysw Exp $".
+#
diff --git a/jpeg/README b/jpeg/README
new file mode 100644
index 000000000..86cc20669
--- /dev/null
+++ b/jpeg/README
@@ -0,0 +1,385 @@
+The Independent JPEG Group's JPEG software
+==========================================
+
+README for release 6b of 27-Mar-1998
+====================================
+
+This distribution contains the sixth public release of the Independent JPEG
+Group's free JPEG software.  You are welcome to redistribute this software and
+to use it for any purpose, subject to the conditions under LEGAL ISSUES, below.
+
+Serious users of this software (particularly those incorporating it into
+larger programs) should contact IJG at jpeg-info@uunet.uu.net to be added to
+our electronic mailing list.  Mailing list members are notified of updates
+and have a chance to participate in technical discussions, etc.
+
+This software is the work of Tom Lane, Philip Gladstone, Jim Boucher,
+Lee Crocker, Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi,
+Guido Vollbeding, Ge' Weijers, and other members of the Independent JPEG
+Group.
+
+IJG is not affiliated with the official ISO JPEG standards committee.
+
+
+DOCUMENTATION ROADMAP
+=====================
+
+This file contains the following sections:
+
+OVERVIEW            General description of JPEG and the IJG software.
+LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
+REFERENCES          Where to learn more about JPEG.
+ARCHIVE LOCATIONS   Where to find newer versions of this software.
+RELATED SOFTWARE    Other stuff you should get.
+FILE FORMAT WARS    Software *not* to get.
+TO DO               Plans for future IJG releases.
+
+Other documentation files in the distribution are:
+
+User documentation:
+  install.doc       How to configure and install the IJG software.
+  usage.doc         Usage instructions for cjpeg, djpeg, jpegtran,
+                    rdjpgcom, and wrjpgcom.
+  *.1               Unix-style man pages for programs (same info as usage.doc).
+  wizard.doc        Advanced usage instructions for JPEG wizards only.
+  change.log        Version-to-version change highlights.
+Programmer and internal documentation:
+  libjpeg.doc       How to use the JPEG library in your own programs.
+  example.c         Sample code for calling the JPEG library.
+  structure.doc     Overview of the JPEG library's internal structure.
+  filelist.doc      Road map of IJG files.
+  coderules.doc     Coding style rules --- please read if you contribute code.
+
+Please read at least the files install.doc and usage.doc.  Useful information
+can also be found in the JPEG FAQ (Frequently Asked Questions) article.  See
+ARCHIVE LOCATIONS below to find out where to obtain the FAQ article.
+
+If you want to understand how the JPEG code works, we suggest reading one or
+more of the REFERENCES, then looking at the documentation files (in roughly
+the order listed) before diving into the code.
+
+
+OVERVIEW
+========
+
+This package contains C software to implement JPEG image compression and
+decompression.  JPEG (pronounced "jay-peg") is a standardized compression
+method for full-color and gray-scale images.  JPEG is intended for compressing
+"real-world" scenes; line drawings, cartoons and other non-realistic images
+are not its strong suit.  JPEG is lossy, meaning that the output image is not
+exactly identical to the input image.  Hence you must not use JPEG if you
+have to have identical output bits.  However, on typical photographic images,
+very good compression levels can be obtained with no visible change, and
+remarkably high compression levels are possible if you can tolerate a
+low-quality image.  For more details, see the references, or just experiment
+with various compression settings.
+
+This software implements JPEG baseline, extended-sequential, and progressive
+compression processes.  Provision is made for supporting all variants of these
+processes, although some uncommon parameter settings aren't implemented yet.
+For legal reasons, we are not distributing code for the arithmetic-coding
+variants of JPEG; see LEGAL ISSUES.  We have made no provision for supporting
+the hierarchical or lossless processes defined in the standard.
+
+We provide a set of library routines for reading and writing JPEG image files,
+plus two sample applications "cjpeg" and "djpeg", which use the library to
+perform conversion between JPEG and some other popular image file formats.
+The library is intended to be reused in other applications.
+
+In order to support file conversion and viewing software, we have included
+considerable functionality beyond the bare JPEG coding/decoding capability;
+for example, the color quantization modules are not strictly part of JPEG
+decoding, but they are essential for output to colormapped file formats or
+colormapped displays.  These extra functions can be compiled out of the
+library if not required for a particular application.  We have also included
+"jpegtran", a utility for lossless transcoding between different JPEG
+processes, and "rdjpgcom" and "wrjpgcom", two simple applications for
+inserting and extracting textual comments in JFIF files.
+
+The emphasis in designing this software has been on achieving portability and
+flexibility, while also making it fast enough to be useful.  In particular,
+the software is not intended to be read as a tutorial on JPEG.  (See the
+REFERENCES section for introductory material.)  Rather, it is intended to
+be reliable, portable, industrial-strength code.  We do not claim to have
+achieved that goal in every aspect of the software, but we strive for it.
+
+We welcome the use of this software as a component of commercial products.
+No royalty is required, but we do ask for an acknowledgement in product
+documentation, as described under LEGAL ISSUES.
+
+
+LEGAL ISSUES
+============
+
+In plain English:
+
+1. We don't promise that this software works.  (But if you find any bugs,
+   please let us know!)
+2. You can use this software for whatever you want.  You don't have to pay us.
+3. You may not pretend that you wrote this software.  If you use it in a
+   program, you must acknowledge somewhere in your documentation that
+   you've used the IJG code.
+
+In legalese:
+
+The authors make NO WARRANTY or representation, either express or implied,
+with respect to this software, its quality, accuracy, merchantability, or
+fitness for a particular purpose.  This software is provided "AS IS", and you,
+its user, assume the entire risk as to its quality and accuracy.
+
+This software is copyright (C) 1991-1998, Thomas G. Lane.
+All Rights Reserved except as specified below.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+software (or portions thereof) for any purpose, without fee, subject to these
+conditions:
+(1) If any part of the source code for this software is distributed, then this
+README file must be included, with this copyright and no-warranty notice
+unaltered; and any additions, deletions, or changes to the original files
+must be clearly indicated in accompanying documentation.
+(2) If only executable code is distributed, then the accompanying
+documentation must state that "this software is based in part on the work of
+the Independent JPEG Group".
+(3) Permission for use of this software is granted only if the user accepts
+full responsibility for any undesirable consequences; the authors accept
+NO LIABILITY for damages of any kind.
+
+These conditions apply to any software derived from or based on the IJG code,
+not just to the unmodified library.  If you use our work, you ought to
+acknowledge us.
+
+Permission is NOT granted for the use of any IJG author's name or company name
+in advertising or publicity relating to this software or products derived from
+it.  This software may be referred to only as "the Independent JPEG Group's
+software".
+
+We specifically permit and encourage the use of this software as the basis of
+commercial products, provided that all warranty or liability claims are
+assumed by the product vendor.
+
+
+ansi2knr.c is included in this distribution by permission of L. Peter Deutsch,
+sole proprietor of its copyright holder, Aladdin Enterprises of Menlo Park, CA.
+ansi2knr.c is NOT covered by the above copyright and conditions, but instead
+by the usual distribution terms of the Free Software Foundation; principally,
+that you must include source code if you redistribute it.  (See the file
+ansi2knr.c for full details.)  However, since ansi2knr.c is not needed as part
+of any program generated from the IJG code, this does not limit you more than
+the foregoing paragraphs do.
+
+The Unix configuration script "configure" was produced with GNU Autoconf.
+It is copyright by the Free Software Foundation but is freely distributable.
+The same holds for its supporting scripts (config.guess, config.sub,
+ltconfig, ltmain.sh).  Another support script, install-sh, is copyright
+by M.I.T. but is also freely distributable.
+
+It appears that the arithmetic coding option of the JPEG spec is covered by
+patents owned by IBM, AT&T, and Mitsubishi.  Hence arithmetic coding cannot
+legally be used without obtaining one or more licenses.  For this reason,
+support for arithmetic coding has been removed from the free JPEG software.
+(Since arithmetic coding provides only a marginal gain over the unpatented
+Huffman mode, it is unlikely that very many implementations will support it.)
+So far as we are aware, there are no patent restrictions on the remaining
+code.
+
+The IJG distribution formerly included code to read and write GIF files.
+To avoid entanglement with the Unisys LZW patent, GIF reading support has
+been removed altogether, and the GIF writer has been simplified to produce
+"uncompressed GIFs".  This technique does not use the LZW algorithm; the
+resulting GIF files are larger than usual, but are readable by all standard
+GIF decoders.
+
+We are required to state that
+    "The Graphics Interchange Format(c) is the Copyright property of
+    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
+    CompuServe Incorporated."
+
+
+REFERENCES
+==========
+
+We highly recommend reading one or more of these references before trying to
+understand the innards of the JPEG software.
+
+The best short technical introduction to the JPEG compression algorithm is
+	Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+	Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
+(Adjacent articles in that issue discuss MPEG motion picture compression,
+applications of JPEG, and related topics.)  If you don't have the CACM issue
+handy, a PostScript file containing a revised version of Wallace's article is
+available at ftp://ftp.uu.net/graphics/jpeg/wallace.ps.gz.  The file (actually
+a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
+omits the sample images that appeared in CACM, but it includes corrections
+and some added material.  Note: the Wallace article is copyright ACM and IEEE,
+and it may not be used for commercial purposes.
+
+A somewhat less technical, more leisurely introduction to JPEG can be found in
+"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
+M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
+good explanations and example C code for a multitude of compression methods
+including JPEG.  It is an excellent source if you are comfortable reading C
+code but don't know much about data compression in general.  The book's JPEG
+sample code is far from industrial-strength, but when you are ready to look
+at a full implementation, you've got one here...
+
+The best full description of JPEG is the textbook "JPEG Still Image Data
+Compression Standard" by William B. Pennebaker and Joan L. Mitchell, published
+by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.  Price US$59.95, 638 pp.
+The book includes the complete text of the ISO JPEG standards (DIS 10918-1
+and draft DIS 10918-2).  This is by far the most complete exposition of JPEG
+in existence, and we highly recommend it.
+
+The JPEG standard itself is not available electronically; you must order a
+paper copy through ISO or ITU.  (Unless you feel a need to own a certified
+official copy, we recommend buying the Pennebaker and Mitchell book instead;
+it's much cheaper and includes a great deal of useful explanatory material.)
+In the USA, copies of the standard may be ordered from ANSI Sales at (212)
+642-4900, or from Global Engineering Documents at (800) 854-7179.  (ANSI
+doesn't take credit card orders, but Global does.)  It's not cheap: as of
+1992, ANSI was charging $95 for Part 1 and $47 for Part 2, plus 7%
+shipping/handling.  The standard is divided into two parts, Part 1 being the
+actual specification, while Part 2 covers compliance testing methods.  Part 1
+is titled "Digital Compression and Coding of Continuous-tone Still Images,
+Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
+10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
+Continuous-tone Still Images, Part 2: Compliance testing" and has document
+numbers ISO/IEC IS 10918-2, ITU-T T.83.
+
+Some extensions to the original JPEG standard are defined in JPEG Part 3,
+a newer ISO standard numbered ISO/IEC IS 10918-3 and ITU-T T.84.  IJG
+currently does not support any Part 3 extensions.
+
+The JPEG standard does not specify all details of an interchangeable file
+format.  For the omitted details we follow the "JFIF" conventions, revision
+1.02.  A copy of the JFIF spec is available from:
+	Literature Department
+	C-Cube Microsystems, Inc.
+	1778 McCarthy Blvd.
+	Milpitas, CA 95035
+	phone (408) 944-6300,  fax (408) 944-6314
+A PostScript version of this document is available by FTP at
+ftp://ftp.uu.net/graphics/jpeg/jfif.ps.gz.  There is also a plain text
+version at ftp://ftp.uu.net/graphics/jpeg/jfif.txt.gz, but it is missing
+the figures.
+
+The TIFF 6.0 file format specification can be obtained by FTP from
+ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
+found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
+IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
+Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
+(Compression tag 7).  Copies of this Note can be obtained from ftp.sgi.com or
+from ftp://ftp.uu.net/graphics/jpeg/.  It is expected that the next revision
+of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
+Although IJG's own code does not support TIFF/JPEG, the free libtiff library
+uses our library to implement TIFF/JPEG per the Note.  libtiff is available
+from ftp://ftp.sgi.com/graphics/tiff/.
+
+
+ARCHIVE LOCATIONS
+=================
+
+The "official" archive site for this software is ftp.uu.net (Internet
+address 192.48.96.9).  The most recent released version can always be found
+there in directory graphics/jpeg.  This particular version will be archived
+as ftp://ftp.uu.net/graphics/jpeg/jpegsrc.v6b.tar.gz.  If you don't have
+direct Internet access, UUNET's archives are also available via UUCP; contact
+help@uunet.uu.net for information on retrieving files that way.
+
+Numerous Internet sites maintain copies of the UUNET files.  However, only
+ftp.uu.net is guaranteed to have the latest official version.
+
+You can also obtain this software in DOS-compatible "zip" archive format from
+the SimTel archives (ftp://ftp.simtel.net/pub/simtelnet/msdos/graphics/), or
+on CompuServe in the Graphics Support forum (GO CIS:GRAPHSUP), library 12
+"JPEG Tools".  Again, these versions may sometimes lag behind the ftp.uu.net
+release.
+
+The JPEG FAQ (Frequently Asked Questions) article is a useful source of
+general information about JPEG.  It is updated constantly and therefore is
+not included in this distribution.  The FAQ is posted every two weeks to
+Usenet newsgroups comp.graphics.misc, news.answers, and other groups.
+It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
+and other news.answers archive sites, including the official news.answers
+archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
+If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
+with body
+	send usenet/news.answers/jpeg-faq/part1
+	send usenet/news.answers/jpeg-faq/part2
+
+
+RELATED SOFTWARE
+================
+
+Numerous viewing and image manipulation programs now support JPEG.  (Quite a
+few of them use this library to do so.)  The JPEG FAQ described above lists
+some of the more popular free and shareware viewers, and tells where to
+obtain them on Internet.
+
+If you are on a Unix machine, we highly recommend Jef Poskanzer's free
+PBMPLUS software, which provides many useful operations on PPM-format image
+files.  In particular, it can convert PPM images to and from a wide range of
+other formats, thus making cjpeg/djpeg considerably more useful.  The latest
+version is distributed by the NetPBM group, and is available from numerous
+sites, notably ftp://wuarchive.wustl.edu/graphics/graphics/packages/NetPBM/.
+Unfortunately PBMPLUS/NETPBM is not nearly as portable as the IJG software is;
+you are likely to have difficulty making it work on any non-Unix machine.
+
+A different free JPEG implementation, written by the PVRG group at Stanford,
+is available from ftp://havefun.stanford.edu/pub/jpeg/.  This program
+is designed for research and experimentation rather than production use;
+it is slower, harder to use, and less portable than the IJG code, but it
+is easier to read and modify.  Also, the PVRG code supports lossless JPEG,
+which we do not.  (On the other hand, it doesn't do progressive JPEG.)
+
+
+FILE FORMAT WARS
+================
+
+Some JPEG programs produce files that are not compatible with our library.
+The root of the problem is that the ISO JPEG committee failed to specify a
+concrete file format.  Some vendors "filled in the blanks" on their own,
+creating proprietary formats that no one else could read.  (For example, none
+of the early commercial JPEG implementations for the Macintosh were able to
+exchange compressed files.)
+
+The file format we have adopted is called JFIF (see REFERENCES).  This format
+has been agreed to by a number of major commercial JPEG vendors, and it has
+become the de facto standard.  JFIF is a minimal or "low end" representation.
+We recommend the use of TIFF/JPEG (TIFF revision 6.0 as modified by TIFF
+Technical Note #2) for "high end" applications that need to record a lot of
+additional data about an image.  TIFF/JPEG is fairly new and not yet widely
+supported, unfortunately.
+
+The upcoming JPEG Part 3 standard defines a file format called SPIFF.
+SPIFF is interoperable with JFIF, in the sense that most JFIF decoders should
+be able to read the most common variant of SPIFF.  SPIFF has some technical
+advantages over JFIF, but its major claim to fame is simply that it is an
+official standard rather than an informal one.  At this point it is unclear
+whether SPIFF will supersede JFIF or whether JFIF will remain the de-facto
+standard.  IJG intends to support SPIFF once the standard is frozen, but we
+have not decided whether it should become our default output format or not.
+(In any case, our decoder will remain capable of reading JFIF indefinitely.)
+
+Various proprietary file formats incorporating JPEG compression also exist.
+We have little or no sympathy for the existence of these formats.  Indeed,
+one of the original reasons for developing this free software was to help
+force convergence on common, open format standards for JPEG files.  Don't
+use a proprietary file format!
+
+
+TO DO
+=====
+
+The major thrust for v7 will probably be improvement of visual quality.
+The current method for scaling the quantization tables is known not to be
+very good at low Q values.  We also intend to investigate block boundary
+smoothing, "poor man's variable quantization", and other means of improving
+quality-vs-file-size performance without sacrificing compatibility.
+
+In future versions, we are considering supporting some of the upcoming JPEG
+Part 3 extensions --- principally, variable quantization and the SPIFF file
+format.
+
+As always, speeding things up is of great interest.
+
+Please send bug reports, offers of help, etc. to jpeg-info@uunet.uu.net.
diff --git a/jpeg/change.log b/jpeg/change.log
new file mode 100644
index 000000000..74102c0db
--- /dev/null
+++ b/jpeg/change.log
@@ -0,0 +1,217 @@
+CHANGE LOG for Independent JPEG Group's JPEG software
+
+
+Version 6b  27-Mar-1998
+-----------------------
+
+jpegtran has new features for lossless image transformations (rotation
+and flipping) as well as "lossless" reduction to grayscale.
+
+jpegtran now copies comments by default; it has a -copy switch to enable
+copying all APPn blocks as well, or to suppress comments.  (Formerly it
+always suppressed comments and APPn blocks.)  jpegtran now also preserves
+JFIF version and resolution information.
+
+New decompressor library feature: COM and APPn markers found in the input
+file can be saved in memory for later use by the application.  (Before,
+you had to code this up yourself with a custom marker processor.)
+
+There is an unused field "void * client_data" now in compress and decompress
+parameter structs; this may be useful in some applications.
+
+JFIF version number information is now saved by the decoder and accepted by
+the encoder.  jpegtran uses this to copy the source file's version number,
+to ensure "jpegtran -copy all" won't create bogus files that contain JFXX
+extensions but claim to be version 1.01.  Applications that generate their
+own JFXX extension markers also (finally) have a supported way to cause the
+encoder to emit JFIF version number 1.02.
+
+djpeg's trace mode reports JFIF 1.02 thumbnail images as such, rather
+than as unknown APP0 markers.
+
+In -verbose mode, djpeg and rdjpgcom will try to print the contents of
+APP12 markers as text.  Some digital cameras store useful text information
+in APP12 markers.
+
+Handling of truncated data streams is more robust: blocks beyond the one in
+which the error occurs will be output as uniform gray, or left unchanged
+if decoding a progressive JPEG.  The appearance no longer depends on the
+Huffman tables being used.
+
+Huffman tables are checked for validity much more carefully than before.
+
+To avoid the Unisys LZW patent, djpeg's GIF output capability has been
+changed to produce "uncompressed GIFs", and cjpeg's GIF input capability
+has been removed altogether.  We're not happy about it either, but there
+seems to be no good alternative.
+
+The configure script now supports building libjpeg as a shared library
+on many flavors of Unix (all the ones that GNU libtool knows how to
+build shared libraries for).  Use "./configure --enable-shared" to
+try this out.
+
+New jconfig file and makefiles for Microsoft Visual C++ and Developer Studio.
+Also, a jconfig file and a build script for Metrowerks CodeWarrior
+on Apple Macintosh.  makefile.dj has been updated for DJGPP v2, and there
+are miscellaneous other minor improvements in the makefiles.
+
+jmemmac.c now knows how to create temporary files following Mac System 7
+conventions.
+
+djpeg's -map switch is now able to read raw-format PPM files reliably.
+
+cjpeg -progressive -restart no longer generates any unnecessary DRI markers.
+
+Multiple calls to jpeg_simple_progression for a single JPEG object
+no longer leak memory.
+
+
+Version 6a  7-Feb-96
+--------------------
+
+Library initialization sequence modified to detect version mismatches
+and struct field packing mismatches between library and calling application.
+This change requires applications to be recompiled, but does not require
+any application source code change.
+
+All routine declarations changed to the style "GLOBAL(type) name ...",
+that is, GLOBAL, LOCAL, METHODDEF, EXTERN are now macros taking the
+routine's return type as an argument.  This makes it possible to add
+Microsoft-style linkage keywords to all the routines by changing just
+these macros.  Note that any application code that was using these macros
+will have to be changed.
+
+DCT coefficient quantization tables are now stored in normal array order
+rather than zigzag order.  Application code that calls jpeg_add_quant_table,
+or otherwise manipulates quantization tables directly, will need to be
+changed.  If you need to make such code work with either older or newer
+versions of the library, a test like "#if JPEG_LIB_VERSION >= 61" is
+recommended.
+
+djpeg's trace capability now dumps DQT tables in natural order, not zigzag
+order.  This allows the trace output to be made into a "-qtables" file
+more easily.
+
+New system-dependent memory manager module for use on Apple Macintosh.
+
+Fix bug in cjpeg's -smooth option: last one or two scanlines would be
+duplicates of the prior line unless the image height mod 16 was 1 or 2.
+
+Repair minor problems in VMS, BCC, MC6 makefiles.
+
+New configure script based on latest GNU Autoconf.
+
+Correct the list of include files needed by MetroWerks C for ccommand().
+
+Numerous small documentation updates.
+
+
+Version 6  2-Aug-95
+-------------------
+
+Progressive JPEG support: library can read and write full progressive JPEG
+files.  A "buffered image" mode supports incremental decoding for on-the-fly
+display of progressive images.  Simply recompiling an existing IJG-v5-based
+decoder with v6 should allow it to read progressive files, though of course
+without any special progressive display.
+
+New "jpegtran" application performs lossless transcoding between different
+JPEG formats; primarily, it can be used to convert baseline to progressive
+JPEG and vice versa.  In support of jpegtran, the library now allows lossless
+reading and writing of JPEG files as DCT coefficient arrays.  This ability
+may be of use in other applications.
+
+Notes for programmers:
+* We changed jpeg_start_decompress() to be able to suspend; this makes all
+decoding modes available to suspending-input applications.  However,
+existing applications that use suspending input will need to be changed
+to check the return value from jpeg_start_decompress().  You don't need to
+do anything if you don't use a suspending data source.
+* We changed the interface to the virtual array routines: access_virt_array
+routines now take a count of the number of rows to access this time.  The
+last parameter to request_virt_array routines is now interpreted as the
+maximum number of rows that may be accessed at once, but not necessarily
+the height of every access.
+
+
+Version 5b  15-Mar-95
+---------------------
+
+Correct bugs with grayscale images having v_samp_factor > 1.
+
+jpeg_write_raw_data() now supports output suspension.
+
+Correct bugs in "configure" script for case of compiling in
+a directory other than the one containing the source files.
+
+Repair bug in jquant1.c: sometimes didn't use as many colors as it could.
+
+Borland C makefile and jconfig file work under either MS-DOS or OS/2.
+
+Miscellaneous improvements to documentation.
+
+
+Version 5a  7-Dec-94
+--------------------
+
+Changed color conversion roundoff behavior so that grayscale values are
+represented exactly.  (This causes test image files to change.)
+
+Make ordered dither use 16x16 instead of 4x4 pattern for a small quality
+improvement.
+
+New configure script based on latest GNU Autoconf.
+Fix configure script to handle CFLAGS correctly.
+Rename *.auto files to *.cfg, so that configure script still works if
+file names have been truncated for DOS.
+
+Fix bug in rdbmp.c: didn't allow for extra data between header and image.
+
+Modify rdppm.c/wrppm.c to handle 2-byte raw PPM/PGM formats for 12-bit data.
+
+Fix several bugs in rdrle.c.
+
+NEED_SHORT_EXTERNAL_NAMES option was broken.
+
+Revise jerror.h/jerror.c for more flexibility in message table.
+
+Repair oversight in jmemname.c NO_MKTEMP case: file could be there
+but unreadable.
+
+
+Version 5  24-Sep-94
+--------------------
+
+Version 5 represents a nearly complete redesign and rewrite of the IJG
+software.  Major user-visible changes include:
+  * Automatic configuration simplifies installation for most Unix systems.
+  * A range of speed vs. image quality tradeoffs are supported.
+    This includes resizing of an image during decompression: scaling down
+    by a factor of 1/2, 1/4, or 1/8 is handled very efficiently.
+  * New programs rdjpgcom and wrjpgcom allow insertion and extraction
+    of text comments in a JPEG file.
+
+The application programmer's interface to the library has changed completely.
+Notable improvements include:
+  * We have eliminated the use of callback routines for handling the
+    uncompressed image data.  The application now sees the library as a
+    set of routines that it calls to read or write image data on a
+    scanline-by-scanline basis.
+  * The application image data is represented in a conventional interleaved-
+    pixel format, rather than as a separate array for each color channel.
+    This can save a copying step in many programs.
+  * The handling of compressed data has been cleaned up: the application can
+    supply routines to source or sink the compressed data.  It is possible to
+    suspend processing on source/sink buffer overrun, although this is not
+    supported in all operating modes.
+  * All static state has been eliminated from the library, so that multiple
+    instances of compression or decompression can be active concurrently.
+  * JPEG abbreviated datastream formats are supported, ie, quantization and
+    Huffman tables can be stored separately from the image data.
+  * And not only that, but the documentation of the library has improved
+    considerably!
+
+
+The last widely used release before the version 5 rewrite was version 4A of
+18-Feb-93.  Change logs before that point have been discarded, since they
+are not of much interest after the rewrite.
diff --git a/jpeg/coderules.doc b/jpeg/coderules.doc
new file mode 100644
index 000000000..0ab5d9bd3
--- /dev/null
+++ b/jpeg/coderules.doc
@@ -0,0 +1,118 @@
+IJG JPEG LIBRARY:  CODING RULES
+
+Copyright (C) 1991-1996, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+Since numerous people will be contributing code and bug fixes, it's important
+to establish a common coding style.  The goal of using similar coding styles
+is much more important than the details of just what that style is.
+
+In general we follow the recommendations of "Recommended C Style and Coding
+Standards" revision 6.1 (Cannon et al. as modified by Spencer, Keppel and
+Brader).  This document is available in the IJG FTP archive (see
+jpeg/doc/cstyle.ms.tbl.Z, or cstyle.txt.Z for those without nroff/tbl).
+
+Block comments should be laid out thusly:
+
+/*
+ *  Block comments in this style.
+ */
+
+We indent statements in K&R style, e.g.,
+	if (test) {
+	  then-part;
+	} else {
+	  else-part;
+	}
+with two spaces per indentation level.  (This indentation convention is
+handled automatically by GNU Emacs and many other text editors.)
+
+Multi-word names should be written in lower case with underscores, e.g.,
+multi_word_name (not multiWordName).  Preprocessor symbols and enum constants
+are similar but upper case (MULTI_WORD_NAME).  Names should be unique within
+the first fifteen characters.  (On some older systems, global names must be
+unique within six characters.  We accommodate this without cluttering the
+source code by using macros to substitute shorter names.)
+
+We use function prototypes everywhere; we rely on automatic source code
+transformation to feed prototype-less C compilers.  Transformation is done
+by the simple and portable tool 'ansi2knr.c' (courtesy of Ghostscript).
+ansi2knr is not very bright, so it imposes a format requirement on function
+declarations: the function name MUST BEGIN IN COLUMN 1.  Thus all functions
+should be written in the following style:
+
+LOCAL(int *)
+function_name (int a, char *b)
+{
+    code...
+}
+
+Note that each function definition must begin with GLOBAL(type), LOCAL(type),
+or METHODDEF(type).  These macros expand to "static type" or just "type" as
+appropriate.  They provide a readable indication of the routine's usage and
+can readily be changed for special needs.  (For instance, special linkage
+keywords can be inserted for use in Windows DLLs.)
+
+ansi2knr does not transform method declarations (function pointers in
+structs).  We handle these with a macro JMETHOD, defined as
+	#ifdef HAVE_PROTOTYPES
+	#define JMETHOD(type,methodname,arglist)  type (*methodname) arglist
+	#else
+	#define JMETHOD(type,methodname,arglist)  type (*methodname) ()
+	#endif
+which is used like this:
+	struct function_pointers {
+	  JMETHOD(void, init_entropy_encoder, (int somearg, jparms *jp));
+	  JMETHOD(void, term_entropy_encoder, (void));
+	};
+Note the set of parentheses surrounding the parameter list.
+
+A similar solution is used for forward and external function declarations
+(see the EXTERN and JPP macros).
+
+If the code is to work on non-ANSI compilers, we cannot rely on a prototype
+declaration to coerce actual parameters into the right types.  Therefore, use
+explicit casts on actual parameters whenever the actual parameter type is not
+identical to the formal parameter.  Beware of implicit conversions to "int".
+
+It seems there are some non-ANSI compilers in which the sizeof() operator
+is defined to return int, yet size_t is defined as long.  Needless to say,
+this is brain-damaged.  Always use the SIZEOF() macro in place of sizeof(),
+so that the result is guaranteed to be of type size_t.
+
+
+The JPEG library is intended to be used within larger programs.  Furthermore,
+we want it to be reentrant so that it can be used by applications that process
+multiple images concurrently.  The following rules support these requirements:
+
+1. Avoid direct use of file I/O, "malloc", error report printouts, etc;
+pass these through the common routines provided.
+
+2. Minimize global namespace pollution.  Functions should be declared static
+wherever possible.  (Note that our method-based calling conventions help this
+a lot: in many modules only the initialization function will ever need to be
+called directly, so only that function need be externally visible.)  All
+global function names should begin with "jpeg_", and should have an
+abbreviated name (unique in the first six characters) substituted by macro
+when NEED_SHORT_EXTERNAL_NAMES is set.
+
+3. Don't use global variables; anything that must be used in another module
+should be in the common data structures.
+
+4. Don't use static variables except for read-only constant tables.  Variables
+that should be private to a module can be placed into private structures (see
+the system architecture document, structure.doc).
+
+5. Source file names should begin with "j" for files that are part of the
+library proper; source files that are not part of the library, such as cjpeg.c
+and djpeg.c, do not begin with "j".  Keep source file names to eight
+characters (plus ".c" or ".h", etc) to make life easy for MS-DOSers.  Keep
+compression and decompression code in separate source files --- some
+applications may want only one half of the library.
+
+Note: these rules (particularly #4) are not followed religiously in the
+modules that are used in cjpeg/djpeg but are not part of the JPEG library
+proper.  Those modules are not really intended to be used in other
+applications.
diff --git a/jpeg/filelist.doc b/jpeg/filelist.doc
new file mode 100644
index 000000000..e14982ca5
--- /dev/null
+++ b/jpeg/filelist.doc
@@ -0,0 +1,210 @@
+IJG JPEG LIBRARY:  FILE LIST
+
+Copyright (C) 1994-1998, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+Here is a road map to the files in the IJG JPEG distribution.  The
+distribution includes the JPEG library proper, plus two application
+programs ("cjpeg" and "djpeg") which use the library to convert JPEG
+files to and from some other popular image formats.  A third application
+"jpegtran" uses the library to do lossless conversion between different
+variants of JPEG.  There are also two stand-alone applications,
+"rdjpgcom" and "wrjpgcom".
+
+
+THE JPEG LIBRARY
+================
+
+Include files:
+
+jpeglib.h	JPEG library's exported data and function declarations.
+jconfig.h	Configuration declarations.  Note: this file is not present
+		in the distribution; it is generated during installation.
+jmorecfg.h	Additional configuration declarations; need not be changed
+		for a standard installation.
+jerror.h	Declares JPEG library's error and trace message codes.
+jinclude.h	Central include file used by all IJG .c files to reference
+		system include files.
+jpegint.h	JPEG library's internal data structures.
+jchuff.h	Private declarations for Huffman encoder modules.
+jdhuff.h	Private declarations for Huffman decoder modules.
+jdct.h		Private declarations for forward & reverse DCT subsystems.
+jmemsys.h	Private declarations for memory management subsystem.
+jversion.h	Version information.
+
+Applications using the library should include jpeglib.h (which in turn
+includes jconfig.h and jmorecfg.h).  Optionally, jerror.h may be included
+if the application needs to reference individual JPEG error codes.  The
+other include files are intended for internal use and would not normally
+be included by an application program.  (cjpeg/djpeg/etc do use jinclude.h,
+since its function is to improve portability of the whole IJG distribution.
+Most other applications will directly include the system include files they
+want, and hence won't need jinclude.h.)
+
+
+C source code files:
+
+These files contain most of the functions intended to be called directly by
+an application program:
+
+jcapimin.c	Application program interface: core routines for compression.
+jcapistd.c	Application program interface: standard compression.
+jdapimin.c	Application program interface: core routines for decompression.
+jdapistd.c	Application program interface: standard decompression.
+jcomapi.c	Application program interface routines common to compression
+		and decompression.
+jcparam.c	Compression parameter setting helper routines.
+jctrans.c	API and library routines for transcoding compression.
+jdtrans.c	API and library routines for transcoding decompression.
+
+Compression side of the library:
+
+jcinit.c	Initialization: determines which other modules to use.
+jcmaster.c	Master control: setup and inter-pass sequencing logic.
+jcmainct.c	Main buffer controller (preprocessor => JPEG compressor).
+jcprepct.c	Preprocessor buffer controller.
+jccoefct.c	Buffer controller for DCT coefficient buffer.
+jccolor.c	Color space conversion.
+jcsample.c	Downsampling.
+jcdctmgr.c	DCT manager (DCT implementation selection & control).
+jfdctint.c	Forward DCT using slow-but-accurate integer method.
+jfdctfst.c	Forward DCT using faster, less accurate integer method.
+jfdctflt.c	Forward DCT using floating-point arithmetic.
+jchuff.c	Huffman entropy coding for sequential JPEG.
+jcphuff.c	Huffman entropy coding for progressive JPEG.
+jcmarker.c	JPEG marker writing.
+jdatadst.c	Data destination manager for stdio output.
+
+Decompression side of the library:
+
+jdmaster.c	Master control: determines which other modules to use.
+jdinput.c	Input controller: controls input processing modules.
+jdmainct.c	Main buffer controller (JPEG decompressor => postprocessor).
+jdcoefct.c	Buffer controller for DCT coefficient buffer.
+jdpostct.c	Postprocessor buffer controller.
+jdmarker.c	JPEG marker reading.
+jdhuff.c	Huffman entropy decoding for sequential JPEG.
+jdphuff.c	Huffman entropy decoding for progressive JPEG.
+jddctmgr.c	IDCT manager (IDCT implementation selection & control).
+jidctint.c	Inverse DCT using slow-but-accurate integer method.
+jidctfst.c	Inverse DCT using faster, less accurate integer method.
+jidctflt.c	Inverse DCT using floating-point arithmetic.
+jidctred.c	Inverse DCTs with reduced-size outputs.
+jdsample.c	Upsampling.
+jdcolor.c	Color space conversion.
+jdmerge.c	Merged upsampling/color conversion (faster, lower quality).
+jquant1.c	One-pass color quantization using a fixed-spacing colormap.
+jquant2.c	Two-pass color quantization using a custom-generated colormap.
+		Also handles one-pass quantization to an externally given map.
+jdatasrc.c	Data source manager for stdio input.
+
+Support files for both compression and decompression:
+
+jerror.c	Standard error handling routines (application replaceable).
+jmemmgr.c	System-independent (more or less) memory management code.
+jutils.c	Miscellaneous utility routines.
+
+jmemmgr.c relies on a system-dependent memory management module.  The IJG
+distribution includes the following implementations of the system-dependent
+module:
+
+jmemnobs.c	"No backing store": assumes adequate virtual memory exists.
+jmemansi.c	Makes temporary files with ANSI-standard routine tmpfile().
+jmemname.c	Makes temporary files with program-generated file names.
+jmemdos.c	Custom implementation for MS-DOS (16-bit environment only):
+		can use extended and expanded memory as well as temp files.
+jmemmac.c	Custom implementation for Apple Macintosh.
+
+Exactly one of the system-dependent modules should be configured into an
+installed JPEG library (see install.doc for hints about which one to use).
+On unusual systems you may find it worthwhile to make a special
+system-dependent memory manager.
+
+
+Non-C source code files:
+
+jmemdosa.asm	80x86 assembly code support for jmemdos.c; used only in
+		MS-DOS-specific configurations of the JPEG library.
+
+
+CJPEG/DJPEG/JPEGTRAN
+====================
+
+Include files:
+
+cdjpeg.h	Declarations shared by cjpeg/djpeg/jpegtran modules.
+cderror.h	Additional error and trace message codes for cjpeg et al.
+transupp.h	Declarations for jpegtran support routines in transupp.c.
+
+C source code files:
+
+cjpeg.c		Main program for cjpeg.
+djpeg.c		Main program for djpeg.
+jpegtran.c	Main program for jpegtran.
+cdjpeg.c	Utility routines used by all three programs.
+rdcolmap.c	Code to read a colormap file for djpeg's "-map" switch.
+rdswitch.c	Code to process some of cjpeg's more complex switches.
+		Also used by jpegtran.
+transupp.c	Support code for jpegtran: lossless image manipulations.
+
+Image file reader modules for cjpeg:
+
+rdbmp.c		BMP file input.
+rdgif.c		GIF file input (now just a stub).
+rdppm.c		PPM/PGM file input.
+rdrle.c		Utah RLE file input.
+rdtarga.c	Targa file input.
+
+Image file writer modules for djpeg:
+
+wrbmp.c		BMP file output.
+wrgif.c		GIF file output (a mere shadow of its former self).
+wrppm.c		PPM/PGM file output.
+wrrle.c		Utah RLE file output.
+wrtarga.c	Targa file output.
+
+
+RDJPGCOM/WRJPGCOM
+=================
+
+C source code files:
+
+rdjpgcom.c	Stand-alone rdjpgcom application.
+wrjpgcom.c	Stand-alone wrjpgcom application.
+
+These programs do not depend on the IJG library.  They do use
+jconfig.h and jinclude.h, only to improve portability.
+
+
+ADDITIONAL FILES
+================
+
+Documentation (see README for a guide to the documentation files):
+
+README		Master documentation file.
+*.doc		Other documentation files.
+*.1		Documentation in Unix man page format.
+change.log	Version-to-version change highlights.
+example.c	Sample code for calling JPEG library.
+
+Configuration/installation files and programs (see install.doc for more info):
+
+configure	Unix shell script to perform automatic configuration.
+ltconfig	Support scripts for configure (from GNU libtool).
+ltmain.sh
+config.guess
+config.sub
+install-sh	Install shell script for those Unix systems lacking one.
+ckconfig.c	Program to generate jconfig.h on non-Unix systems.
+jconfig.doc	Template for making jconfig.h by hand.
+makefile.*	Sample makefiles for particular systems.
+jconfig.*	Sample jconfig.h for particular systems.
+ansi2knr.c	De-ANSIfier for pre-ANSI C compilers (courtesy of
+		L. Peter Deutsch and Aladdin Enterprises).
+
+Test files (see install.doc for test procedure):
+
+test*.*		Source and comparison files for confidence test.
+		These are binary image files, NOT text files.
diff --git a/jpeg/jcapimin.c b/jpeg/jcapimin.c
new file mode 100644
index 000000000..54fb8c58c
--- /dev/null
+++ b/jpeg/jcapimin.c
@@ -0,0 +1,280 @@
+/*
+ * jcapimin.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the compression half
+ * of the JPEG library.  These are the "minimum" API routines that may be
+ * needed in either the normal full-compression case or the transcoding-only
+ * case.
+ *
+ * Most of the routines intended to be called directly by an application
+ * are in this file or in jcapistd.c.  But also see jcparam.c for
+ * parameter-setup helper routines, jcomapi.c for routines shared by
+ * compression and decompression, and jctrans.c for the transcoding case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Initialization of a JPEG compression object.
+ * The error manager must already be set up (in case memory manager fails).
+ */
+
+GLOBAL(void)
+jpeg_CreateCompress (j_compress_ptr cinfo, int version, size_t structsize)
+{
+  int i;
+
+  /* Guard against version mismatches between library and caller. */
+  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  if (version != JPEG_LIB_VERSION)
+    ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
+  if (structsize != SIZEOF(struct jpeg_compress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
+	     (int) SIZEOF(struct jpeg_compress_struct), (int) structsize);
+
+  /* For debugging purposes, we zero the whole master structure.
+   * But the application has already set the err pointer, and may have set
+   * client_data, so we have to save and restore those fields.
+   * Note: if application hasn't set client_data, tools like Purify may
+   * complain here.
+   */
+  {
+    struct jpeg_error_mgr * err = cinfo->err;
+    void * client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, SIZEOF(struct jpeg_compress_struct));
+    cinfo->err = err;
+    cinfo->client_data = client_data;
+  }
+  cinfo->is_decompressor = FALSE;
+
+  /* Initialize a memory manager instance for this object */
+  jinit_memory_mgr((j_common_ptr) cinfo);
+
+  /* Zero out pointers to permanent structures. */
+  cinfo->progress = NULL;
+  cinfo->dest = NULL;
+
+  cinfo->comp_info = NULL;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++)
+    cinfo->quant_tbl_ptrs[i] = NULL;
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    cinfo->dc_huff_tbl_ptrs[i] = NULL;
+    cinfo->ac_huff_tbl_ptrs[i] = NULL;
+  }
+
+  cinfo->script_space = NULL;
+
+  cinfo->input_gamma = 1.0;	/* in case application forgets */
+
+  /* OK, I'm ready */
+  cinfo->global_state = CSTATE_START;
+}
+
+
+/*
+ * Destruction of a JPEG compression object
+ */
+
+GLOBAL(void)
+jpeg_destroy_compress (j_compress_ptr cinfo)
+{
+  jpeg_destroy((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Abort processing of a JPEG compression operation,
+ * but don't destroy the object itself.
+ */
+
+GLOBAL(void)
+jpeg_abort_compress (j_compress_ptr cinfo)
+{
+  jpeg_abort((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Forcibly suppress or un-suppress all quantization and Huffman tables.
+ * Marks all currently defined tables as already written (if suppress)
+ * or not written (if !suppress).  This will control whether they get emitted
+ * by a subsequent jpeg_start_compress call.
+ *
+ * This routine is exported for use by applications that want to produce
+ * abbreviated JPEG datastreams.  It logically belongs in jcparam.c, but
+ * since it is called by jpeg_start_compress, we put it here --- otherwise
+ * jcparam.o would be linked whether the application used it or not.
+ */
+
+GLOBAL(void)
+jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress)
+{
+  int i;
+  JQUANT_TBL * qtbl;
+  JHUFF_TBL * htbl;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    if ((qtbl = cinfo->quant_tbl_ptrs[i]) != NULL)
+      qtbl->sent_table = suppress;
+  }
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    if ((htbl = cinfo->dc_huff_tbl_ptrs[i]) != NULL)
+      htbl->sent_table = suppress;
+    if ((htbl = cinfo->ac_huff_tbl_ptrs[i]) != NULL)
+      htbl->sent_table = suppress;
+  }
+}
+
+
+/*
+ * Finish JPEG compression.
+ *
+ * If a multipass operating mode was selected, this may do a great deal of
+ * work including most of the actual output.
+ */
+
+GLOBAL(void)
+jpeg_finish_compress (j_compress_ptr cinfo)
+{
+  JDIMENSION iMCU_row;
+
+  if (cinfo->global_state == CSTATE_SCANNING ||
+      cinfo->global_state == CSTATE_RAW_OK) {
+    /* Terminate first pass */
+    if (cinfo->next_scanline < cinfo->image_height)
+      ERREXIT(cinfo, JERR_TOO_LITTLE_DATA);
+    (*cinfo->master->finish_pass) (cinfo);
+  } else if (cinfo->global_state != CSTATE_WRCOEFS)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any remaining passes */
+  while (! cinfo->master->is_last_pass) {
+    (*cinfo->master->prepare_for_pass) (cinfo);
+    for (iMCU_row = 0; iMCU_row < cinfo->total_iMCU_rows; iMCU_row++) {
+      if (cinfo->progress != NULL) {
+	cinfo->progress->pass_counter = (long) iMCU_row;
+	cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows;
+	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      }
+      /* We bypass the main controller and invoke coef controller directly;
+       * all work is being done from the coefficient buffer.
+       */
+      if (! (*cinfo->coef->compress_data) (cinfo, (JSAMPIMAGE) NULL))
+	ERREXIT(cinfo, JERR_CANT_SUSPEND);
+    }
+    (*cinfo->master->finish_pass) (cinfo);
+  }
+  /* Write EOI, do final cleanup */
+  (*cinfo->marker->write_file_trailer) (cinfo);
+  (*cinfo->dest->term_destination) (cinfo);
+  /* We can use jpeg_abort to release memory and reset global_state */
+  jpeg_abort((j_common_ptr) cinfo);
+}
+
+
+/*
+ * Write a special marker.
+ * This is only recommended for writing COM or APPn markers.
+ * Must be called after jpeg_start_compress() and before
+ * first call to jpeg_write_scanlines() or jpeg_write_raw_data().
+ */
+
+GLOBAL(void)
+jpeg_write_marker (j_compress_ptr cinfo, int marker,
+		   const JOCTET *dataptr, unsigned int datalen)
+{
+  JMETHOD(void, write_marker_byte, (j_compress_ptr info, int val));
+
+  if (cinfo->next_scanline != 0 ||
+      (cinfo->global_state != CSTATE_SCANNING &&
+       cinfo->global_state != CSTATE_RAW_OK &&
+       cinfo->global_state != CSTATE_WRCOEFS))
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
+  write_marker_byte = cinfo->marker->write_marker_byte;	/* copy for speed */
+  while (datalen--) {
+    (*write_marker_byte) (cinfo, *dataptr);
+    dataptr++;
+  }
+}
+
+/* Same, but piecemeal. */
+
+GLOBAL(void)
+jpeg_write_m_header (j_compress_ptr cinfo, int marker, unsigned int datalen)
+{
+  if (cinfo->next_scanline != 0 ||
+      (cinfo->global_state != CSTATE_SCANNING &&
+       cinfo->global_state != CSTATE_RAW_OK &&
+       cinfo->global_state != CSTATE_WRCOEFS))
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
+}
+
+GLOBAL(void)
+jpeg_write_m_byte (j_compress_ptr cinfo, int val)
+{
+  (*cinfo->marker->write_marker_byte) (cinfo, val);
+}
+
+
+/*
+ * Alternate compression function: just write an abbreviated table file.
+ * Before calling this, all parameters and a data destination must be set up.
+ *
+ * To produce a pair of files containing abbreviated tables and abbreviated
+ * image data, one would proceed as follows:
+ *
+ *		initialize JPEG object
+ *		set JPEG parameters
+ *		set destination to table file
+ *		jpeg_write_tables(cinfo);
+ *		set destination to image file
+ *		jpeg_start_compress(cinfo, FALSE);
+ *		write data...
+ *		jpeg_finish_compress(cinfo);
+ *
+ * jpeg_write_tables has the side effect of marking all tables written
+ * (same as jpeg_suppress_tables(..., TRUE)).  Thus a subsequent start_compress
+ * will not re-emit the tables unless it is passed write_all_tables=TRUE.
+ */
+
+GLOBAL(void)
+jpeg_write_tables (j_compress_ptr cinfo)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Initialize the marker writer ... bit of a crock to do it here. */
+  jinit_marker_writer(cinfo);
+  /* Write them tables! */
+  (*cinfo->marker->write_tables_only) (cinfo);
+  /* And clean up. */
+  (*cinfo->dest->term_destination) (cinfo);
+  /*
+   * In library releases up through v6a, we called jpeg_abort() here to free
+   * any working memory allocated by the destination manager and marker
+   * writer.  Some applications had a problem with that: they allocated space
+   * of their own from the library memory manager, and didn't want it to go
+   * away during write_tables.  So now we do nothing.  This will cause a
+   * memory leak if an app calls write_tables repeatedly without doing a full
+   * compression cycle or otherwise resetting the JPEG object.  However, that
+   * seems less bad than unexpectedly freeing memory in the normal case.
+   * An app that prefers the old behavior can call jpeg_abort for itself after
+   * each call to jpeg_write_tables().
+   */
+}
diff --git a/jpeg/jcapistd.c b/jpeg/jcapistd.c
new file mode 100644
index 000000000..c0320b1b1
--- /dev/null
+++ b/jpeg/jcapistd.c
@@ -0,0 +1,161 @@
+/*
+ * jcapistd.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the compression half
+ * of the JPEG library.  These are the "standard" API routines that are
+ * used in the normal full-compression case.  They are not used by a
+ * transcoding-only application.  Note that if an application links in
+ * jpeg_start_compress, it will end up linking in the entire compressor.
+ * We thus must separate this file from jcapimin.c to avoid linking the
+ * whole compression library into a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Compression initialization.
+ * Before calling this, all parameters and a data destination must be set up.
+ *
+ * We require a write_all_tables parameter as a failsafe check when writing
+ * multiple datastreams from the same compression object.  Since prior runs
+ * will have left all the tables marked sent_table=TRUE, a subsequent run
+ * would emit an abbreviated stream (no tables) by default.  This may be what
+ * is wanted, but for safety's sake it should not be the default behavior:
+ * programmers should have to make a deliberate choice to emit abbreviated
+ * images.  Therefore the documentation and examples should encourage people
+ * to pass write_all_tables=TRUE; then it will take active thought to do the
+ * wrong thing.
+ */
+
+GLOBAL(void)
+jpeg_start_compress (j_compress_ptr cinfo, boolean write_all_tables)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (write_all_tables)
+    jpeg_suppress_tables(cinfo, FALSE);	/* mark all tables to be written */
+
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Perform master selection of active modules */
+  jinit_compress_master(cinfo);
+  /* Set up for the first pass */
+  (*cinfo->master->prepare_for_pass) (cinfo);
+  /* Ready for application to drive first pass through jpeg_write_scanlines
+   * or jpeg_write_raw_data.
+   */
+  cinfo->next_scanline = 0;
+  cinfo->global_state = (cinfo->raw_data_in ? CSTATE_RAW_OK : CSTATE_SCANNING);
+}
+
+
+/*
+ * Write some scanlines of data to the JPEG compressor.
+ *
+ * The return value will be the number of lines actually written.
+ * This should be less than the supplied num_lines only in case that
+ * the data destination module has requested suspension of the compressor,
+ * or if more than image_height scanlines are passed in.
+ *
+ * Note: we warn about excess calls to jpeg_write_scanlines() since
+ * this likely signals an application programmer error.  However,
+ * excess scanlines passed in the last valid call are *silently* ignored,
+ * so that the application need not adjust num_lines for end-of-image
+ * when using a multiple-scanline buffer.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_write_scanlines (j_compress_ptr cinfo, JSAMPARRAY scanlines,
+		      JDIMENSION num_lines)
+{
+  JDIMENSION row_ctr, rows_left;
+
+  if (cinfo->global_state != CSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->next_scanline >= cinfo->image_height)
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->next_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->image_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Give master control module another chance if this is first call to
+   * jpeg_write_scanlines.  This lets output of the frame/scan headers be
+   * delayed so that application can write COM, etc, markers between
+   * jpeg_start_compress and jpeg_write_scanlines.
+   */
+  if (cinfo->master->call_pass_startup)
+    (*cinfo->master->pass_startup) (cinfo);
+
+  /* Ignore any extra scanlines at bottom of image. */
+  rows_left = cinfo->image_height - cinfo->next_scanline;
+  if (num_lines > rows_left)
+    num_lines = rows_left;
+
+  row_ctr = 0;
+  (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, num_lines);
+  cinfo->next_scanline += row_ctr;
+  return row_ctr;
+}
+
+
+/*
+ * Alternate entry point to write raw data.
+ * Processes exactly one iMCU row per call, unless suspended.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data,
+		     JDIMENSION num_lines)
+{
+  JDIMENSION lines_per_iMCU_row;
+
+  if (cinfo->global_state != CSTATE_RAW_OK)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->next_scanline >= cinfo->image_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->next_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->image_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Give master control module another chance if this is first call to
+   * jpeg_write_raw_data.  This lets output of the frame/scan headers be
+   * delayed so that application can write COM, etc, markers between
+   * jpeg_start_compress and jpeg_write_raw_data.
+   */
+  if (cinfo->master->call_pass_startup)
+    (*cinfo->master->pass_startup) (cinfo);
+
+  /* Verify that at least one iMCU row has been passed. */
+  lines_per_iMCU_row = cinfo->max_v_samp_factor * DCTSIZE;
+  if (num_lines < lines_per_iMCU_row)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Directly compress the row. */
+  if (! (*cinfo->coef->compress_data) (cinfo, data)) {
+    /* If compressor did not consume the whole row, suspend processing. */
+    return 0;
+  }
+
+  /* OK, we processed one iMCU row. */
+  cinfo->next_scanline += lines_per_iMCU_row;
+  return lines_per_iMCU_row;
+}
diff --git a/jpeg/jccoefct.c b/jpeg/jccoefct.c
new file mode 100644
index 000000000..1963ddb61
--- /dev/null
+++ b/jpeg/jccoefct.c
@@ -0,0 +1,449 @@
+/*
+ * jccoefct.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the coefficient buffer controller for compression.
+ * This controller is the top level of the JPEG compressor proper.
+ * The coefficient buffer lies between forward-DCT and entropy encoding steps.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* We use a full-image coefficient buffer when doing Huffman optimization,
+ * and also for writing multiple-scan JPEG files.  In all cases, the DCT
+ * step is run during the first pass, and subsequent passes need only read
+ * the buffered coefficients.
+ */
+#ifdef ENTROPY_OPT_SUPPORTED
+#define FULL_COEF_BUFFER_SUPPORTED
+#else
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+#define FULL_COEF_BUFFER_SUPPORTED
+#endif
+#endif
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_coef_controller pub; /* public fields */
+
+  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
+  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
+  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+
+  /* For single-pass compression, it's sufficient to buffer just one MCU
+   * (although this may prove a bit slow in practice).  We allocate a
+   * workspace of C_MAX_BLOCKS_IN_MCU coefficient blocks, and reuse it for each
+   * MCU constructed and sent.  (On 80x86, the workspace is FAR even though
+   * it's not really very big; this is to keep the module interfaces unchanged
+   * when a large coefficient buffer is necessary.)
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays.
+   */
+  JBLOCKROW MCU_buffer[C_MAX_BLOCKS_IN_MCU];
+
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
+} my_coef_controller;
+
+typedef my_coef_controller * my_coef_ptr;
+
+
+/* Forward declarations */
+METHODDEF(boolean) compress_data
+    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+METHODDEF(boolean) compress_first_pass
+    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+METHODDEF(boolean) compress_output
+    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+#endif
+
+
+LOCAL(void)
+start_iMCU_row (j_compress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (coef->iMCU_row_num < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->mcu_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_coef (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  coef->iMCU_row_num = 0;
+  start_iMCU_row(cinfo);
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (coef->whole_image[0] != NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_data;
+    break;
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+  case JBUF_SAVE_AND_PASS:
+    if (coef->whole_image[0] == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_first_pass;
+    break;
+  case JBUF_CRANK_DEST:
+    if (coef->whole_image[0] == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    coef->pub.compress_data = compress_output;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data in the single-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the image.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf contains a plane for each component in image,
+ * which we index according to the component's SOF position.
+ */
+
+METHODDEF(boolean)
+compress_data (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, bi, ci, yindex, yoffset, blockcnt;
+  JDIMENSION ypos, xpos;
+  jpeg_component_info *compptr;
+
+  /* Loop to write as much as one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num <= last_MCU_col;
+	 MCU_col_num++) {
+      /* Determine where data comes from in input_buf and do the DCT thing.
+       * Each call on forward_DCT processes a horizontal row of DCT blocks
+       * as wide as an MCU; we rely on having allocated the MCU_buffer[] blocks
+       * sequentially.  Dummy blocks at the right or bottom edge are filled in
+       * specially.  The data in them does not matter for image reconstruction,
+       * so we fill them with values that will encode to the smallest amount of
+       * data, viz: all zeroes in the AC entries, DC entries equal to previous
+       * block's DC value.  (Thanks to Thomas Kinsman for this idea.)
+       */
+      blkn = 0;
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+						: compptr->last_col_width;
+	xpos = MCU_col_num * compptr->MCU_sample_width;
+	ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  if (coef->iMCU_row_num < last_iMCU_row ||
+	      yoffset+yindex < compptr->last_row_height) {
+	    (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+					 input_buf[compptr->component_index],
+					 coef->MCU_buffer[blkn],
+					 ypos, xpos, (JDIMENSION) blockcnt);
+	    if (blockcnt < compptr->MCU_width) {
+	      /* Create some dummy blocks at the right edge of the image. */
+	      jzero_far((void FAR *) coef->MCU_buffer[blkn + blockcnt],
+			(compptr->MCU_width - blockcnt) * SIZEOF(JBLOCK));
+	      for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
+		coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn+bi-1][0][0];
+	      }
+	    }
+	  } else {
+	    /* Create a row of dummy blocks at the bottom of the image. */
+	    jzero_far((void FAR *) coef->MCU_buffer[blkn],
+		      compptr->MCU_width * SIZEOF(JBLOCK));
+	    for (bi = 0; bi < compptr->MCU_width; bi++) {
+	      coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn-1][0][0];
+	    }
+	  }
+	  blkn += compptr->MCU_width;
+	  ypos += DCTSIZE;
+	}
+      }
+      /* Try to write the MCU.  In event of a suspension failure, we will
+       * re-DCT the MCU on restart (a bit inefficient, could be fixed...)
+       */
+      if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->mcu_ctr = MCU_col_num;
+	return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+
+/*
+ * Process some data in the first pass of a multi-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the image.
+ * This amount of data is read from the source buffer, DCT'd and quantized,
+ * and saved into the virtual arrays.  We also generate suitable dummy blocks
+ * as needed at the right and lower edges.  (The dummy blocks are constructed
+ * in the virtual arrays, which have been padded appropriately.)  This makes
+ * it possible for subsequent passes not to worry about real vs. dummy blocks.
+ *
+ * We must also emit the data to the entropy encoder.  This is conveniently
+ * done by calling compress_output() after we've loaded the current strip
+ * of the virtual arrays.
+ *
+ * NB: input_buf contains a plane for each component in image.  All
+ * components are DCT'd and loaded into the virtual arrays in this pass.
+ * However, it may be that only a subset of the components are emitted to
+ * the entropy encoder during this first pass; be careful about looking
+ * at the scan-dependent variables (MCU dimensions, etc).
+ */
+
+METHODDEF(boolean)
+compress_first_pass (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION blocks_across, MCUs_across, MCUindex;
+  int bi, ci, h_samp_factor, block_row, block_rows, ndummy;
+  JCOEF lastDC;
+  jpeg_component_info *compptr;
+  JBLOCKARRAY buffer;
+  JBLOCKROW thisblockrow, lastblockrow;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Align the virtual buffer for this component. */
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[ci],
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, TRUE);
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (coef->iMCU_row_num < last_iMCU_row)
+      block_rows = compptr->v_samp_factor;
+    else {
+      /* NB: can't use last_row_height here, since may not be set! */
+      block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+    }
+    blocks_across = compptr->width_in_blocks;
+    h_samp_factor = compptr->h_samp_factor;
+    /* Count number of dummy blocks to be added at the right margin. */
+    ndummy = (int) (blocks_across % h_samp_factor);
+    if (ndummy > 0)
+      ndummy = h_samp_factor - ndummy;
+    /* Perform DCT for all non-dummy blocks in this iMCU row.  Each call
+     * on forward_DCT processes a complete horizontal row of DCT blocks.
+     */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      thisblockrow = buffer[block_row];
+      (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+				   input_buf[ci], thisblockrow,
+				   (JDIMENSION) (block_row * DCTSIZE),
+				   (JDIMENSION) 0, blocks_across);
+      if (ndummy > 0) {
+	/* Create dummy blocks at the right edge of the image. */
+	thisblockrow += blocks_across; /* => first dummy block */
+	jzero_far((void FAR *) thisblockrow, ndummy * SIZEOF(JBLOCK));
+	lastDC = thisblockrow[-1][0];
+	for (bi = 0; bi < ndummy; bi++) {
+	  thisblockrow[bi][0] = lastDC;
+	}
+      }
+    }
+    /* If at end of image, create dummy block rows as needed.
+     * The tricky part here is that within each MCU, we want the DC values
+     * of the dummy blocks to match the last real block's DC value.
+     * This squeezes a few more bytes out of the resulting file...
+     */
+    if (coef->iMCU_row_num == last_iMCU_row) {
+      blocks_across += ndummy;	/* include lower right corner */
+      MCUs_across = blocks_across / h_samp_factor;
+      for (block_row = block_rows; block_row < compptr->v_samp_factor;
+	   block_row++) {
+	thisblockrow = buffer[block_row];
+	lastblockrow = buffer[block_row-1];
+	jzero_far((void FAR *) thisblockrow,
+		  (size_t) (blocks_across * SIZEOF(JBLOCK)));
+	for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
+	  lastDC = lastblockrow[h_samp_factor-1][0];
+	  for (bi = 0; bi < h_samp_factor; bi++) {
+	    thisblockrow[bi][0] = lastDC;
+	  }
+	  thisblockrow += h_samp_factor; /* advance to next MCU in row */
+	  lastblockrow += h_samp_factor;
+	}
+      }
+    }
+  }
+  /* NB: compress_output will increment iMCU_row_num if successful.
+   * A suspension return will result in redoing all the work above next time.
+   */
+
+  /* Emit data to the entropy encoder, sharing code with subsequent passes */
+  return compress_output(cinfo, input_buf);
+}
+
+
+/*
+ * Process some data in subsequent passes of a multi-pass case.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the scan.
+ * The data is obtained from the virtual arrays and fed to the entropy coder.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf is ignored; it is likely to be a NULL pointer.
+ */
+
+METHODDEF(boolean)
+compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan.
+   * NB: during first pass, this is safe only because the buffers will
+   * already be aligned properly, so jmemmgr.c won't need to do any I/O.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, FALSE);
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	start_col = MCU_col_num * compptr->MCU_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+	    coef->MCU_buffer[blkn++] = buffer_ptr++;
+	  }
+	}
+      }
+      /* Try to write the MCU. */
+      if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->mcu_ctr = MCU_col_num;
+	return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+#endif /* FULL_COEF_BUFFER_SUPPORTED */
+
+
+/*
+ * Initialize coefficient buffer controller.
+ */
+
+GLOBAL(void)
+jinit_c_coef_controller (j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_coef_ptr coef;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_coef_controller));
+  cinfo->coef = (struct jpeg_c_coef_controller *) coef;
+  coef->pub.start_pass = start_pass_coef;
+
+  /* Create the coefficient buffer. */
+  if (need_full_buffer) {
+#ifdef FULL_COEF_BUFFER_SUPPORTED
+    /* Allocate a full-image virtual array for each component, */
+    /* padded to a multiple of samp_factor DCT blocks in each direction. */
+    int ci;
+    jpeg_component_info *compptr;
+
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor),
+	 (JDIMENSION) compptr->v_samp_factor);
+    }
+#else
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+  } else {
+    /* We only need a single-MCU buffer. */
+    JBLOCKROW buffer;
+    int i;
+
+    buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+    for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
+      coef->MCU_buffer[i] = buffer + i;
+    }
+    coef->whole_image[0] = NULL; /* flag for no virtual arrays */
+  }
+}
diff --git a/jpeg/jccolor.c b/jpeg/jccolor.c
new file mode 100644
index 000000000..0a8a4b5d1
--- /dev/null
+++ b/jpeg/jccolor.c
@@ -0,0 +1,459 @@
+/*
+ * jccolor.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_converter pub; /* public fields */
+
+  /* Private state for RGB->YCC conversion */
+  INT32 * rgb_ycc_tab;		/* => table for RGB to YCbCr conversion */
+} my_color_converter;
+
+typedef my_color_converter * my_cconvert_ptr;
+
+
+/**************** RGB -> YCbCr conversion: most common case **************/
+
+/*
+ * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
+ * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
+ * The conversion equations to be implemented are therefore
+ *	Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *	Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
+ *	Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
+ * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
+ * Note: older versions of the IJG code used a zero offset of MAXJSAMPLE/2,
+ * rather than CENTERJSAMPLE, for Cb and Cr.  This gave equal positive and
+ * negative swings for Cb/Cr, but meant that grayscale values (Cb=Cr=0)
+ * were not represented exactly.  Now we sacrifice exact representation of
+ * maximum red and maximum blue in order to get exact grayscales.
+ *
+ * To avoid floating-point arithmetic, we represent the fractional constants
+ * as integers scaled up by 2^16 (about 4 digits precision); we have to divide
+ * the products by 2^16, with appropriate rounding, to get the correct answer.
+ *
+ * For even more speed, we avoid doing any multiplications in the inner loop
+ * by precalculating the constants times R,G,B for all possible values.
+ * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table);
+ * for 12-bit samples it is still acceptable.  It's not very reasonable for
+ * 16-bit samples, but if you want lossless storage you shouldn't be changing
+ * colorspace anyway.
+ * The CENTERJSAMPLE offsets and the rounding fudge-factor of 0.5 are included
+ * in the tables to save adding them separately in the inner loop.
+ */
+
+#define SCALEBITS	16	/* speediest right-shift on some machines */
+#define CBCR_OFFSET	((INT32) CENTERJSAMPLE << SCALEBITS)
+#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
+#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+
+/* We allocate one big table and divide it up into eight parts, instead of
+ * doing eight alloc_small requests.  This lets us use a single table base
+ * address, which can be held in a register in the inner loops on many
+ * machines (more than can hold all eight addresses, anyway).
+ */
+
+#define R_Y_OFF		0			/* offset to R => Y section */
+#define G_Y_OFF		(1*(MAXJSAMPLE+1))	/* offset to G => Y section */
+#define B_Y_OFF		(2*(MAXJSAMPLE+1))	/* etc. */
+#define R_CB_OFF	(3*(MAXJSAMPLE+1))
+#define G_CB_OFF	(4*(MAXJSAMPLE+1))
+#define B_CB_OFF	(5*(MAXJSAMPLE+1))
+#define R_CR_OFF	B_CB_OFF		/* B=>Cb, R=>Cr are the same */
+#define G_CR_OFF	(6*(MAXJSAMPLE+1))
+#define B_CR_OFF	(7*(MAXJSAMPLE+1))
+#define TABLE_SIZE	(8*(MAXJSAMPLE+1))
+
+
+/*
+ * Initialize for RGB->YCC colorspace conversion.
+ */
+
+METHODDEF(void)
+rgb_ycc_start (j_compress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  INT32 * rgb_ycc_tab;
+  INT32 i;
+
+  /* Allocate and fill in the conversion tables. */
+  cconvert->rgb_ycc_tab = rgb_ycc_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(TABLE_SIZE * SIZEOF(INT32)));
+
+  for (i = 0; i <= MAXJSAMPLE; i++) {
+    rgb_ycc_tab[i+R_Y_OFF] = FIX(0.29900) * i;
+    rgb_ycc_tab[i+G_Y_OFF] = FIX(0.58700) * i;
+    rgb_ycc_tab[i+B_Y_OFF] = FIX(0.11400) * i     + ONE_HALF;
+    rgb_ycc_tab[i+R_CB_OFF] = (-FIX(0.16874)) * i;
+    rgb_ycc_tab[i+G_CB_OFF] = (-FIX(0.33126)) * i;
+    /* We use a rounding fudge-factor of 0.5-epsilon for Cb and Cr.
+     * This ensures that the maximum output will round to MAXJSAMPLE
+     * not MAXJSAMPLE+1, and thus that we don't have to range-limit.
+     */
+    rgb_ycc_tab[i+B_CB_OFF] = FIX(0.50000) * i    + CBCR_OFFSET + ONE_HALF-1;
+/*  B=>Cb and R=>Cr tables are the same
+    rgb_ycc_tab[i+R_CR_OFF] = FIX(0.50000) * i    + CBCR_OFFSET + ONE_HALF-1;
+*/
+    rgb_ycc_tab[i+G_CR_OFF] = (-FIX(0.41869)) * i;
+    rgb_ycc_tab[i+B_CR_OFF] = (-FIX(0.08131)) * i;
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ *
+ * Note that we change from the application's interleaved-pixel format
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
+ *
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+METHODDEF(void)
+rgb_ycc_convert (j_compress_ptr cinfo,
+		 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		 JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+		 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles RGB->grayscale conversion, which is the same
+ * as the RGB->Y portion of RGB->YCbCr.
+ * We assume rgb_ycc_start has been called (we only use the Y tables).
+ */
+
+METHODDEF(void)
+rgb_gray_convert (j_compress_ptr cinfo,
+		  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		  JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* Y */
+      outptr[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles Adobe-style CMYK->YCCK conversion,
+ * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the same
+ * conversion as above, while passing K (black) unchanged.
+ * We assume rgb_ycc_start has been called.
+ */
+
+METHODDEF(void)
+cmyk_ycck_convert (j_compress_ptr cinfo,
+		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		   JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2, outptr3;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    outptr3 = output_buf[3][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = MAXJSAMPLE - GETJSAMPLE(inptr[0]);
+      g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
+      b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
+      /* K passes through as-is */
+      outptr3[col] = inptr[3];	/* don't need GETJSAMPLE here */
+      inptr += 4;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+		 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles grayscale output with no conversion.
+ * The source can be either plain grayscale or YCbCr (since Y == gray).
+ */
+
+METHODDEF(void)
+grayscale_convert (j_compress_ptr cinfo,
+		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+		   JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+  int instride = cinfo->input_components;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      outptr[col] = inptr[0];	/* don't need GETJSAMPLE() here */
+      inptr += instride;
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles multi-component colorspaces without conversion.
+ * We assume input_components == num_components.
+ */
+
+METHODDEF(void)
+null_convert (j_compress_ptr cinfo,
+	      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+	      JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  register int ci;
+  int nc = cinfo->num_components;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    /* It seems fastest to make a separate pass for each component. */
+    for (ci = 0; ci < nc; ci++) {
+      inptr = *input_buf;
+      outptr = output_buf[ci][output_row];
+      for (col = 0; col < num_cols; col++) {
+	outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
+	inptr += nc;
+      }
+    }
+    input_buf++;
+    output_row++;
+  }
+}
+
+
+/*
+ * Empty method for start_pass.
+ */
+
+METHODDEF(void)
+null_method (j_compress_ptr cinfo)
+{
+  /* no work needed */
+}
+
+
+/*
+ * Module initialization routine for input colorspace conversion.
+ */
+
+GLOBAL(void)
+jinit_color_converter (j_compress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert;
+
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_color_converter));
+  cinfo->cconvert = (struct jpeg_color_converter *) cconvert;
+  /* set start_pass to null method until we find out differently */
+  cconvert->pub.start_pass = null_method;
+
+  /* Make sure input_components agrees with in_color_space */
+  switch (cinfo->in_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->input_components != 1)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_RGB:
+#if RGB_PIXELSIZE != 3
+    if (cinfo->input_components != RGB_PIXELSIZE)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+#endif /* else share code with YCbCr */
+
+  case JCS_YCbCr:
+    if (cinfo->input_components != 3)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  case JCS_CMYK:
+  case JCS_YCCK:
+    if (cinfo->input_components != 4)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+
+  default:			/* JCS_UNKNOWN can be anything */
+    if (cinfo->input_components < 1)
+      ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+    break;
+  }
+
+  /* Check num_components, set conversion method based on requested space */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->num_components != 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_GRAYSCALE)
+      cconvert->pub.color_convert = grayscale_convert;
+    else if (cinfo->in_color_space == JCS_RGB) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = rgb_gray_convert;
+    } else if (cinfo->in_color_space == JCS_YCbCr)
+      cconvert->pub.color_convert = grayscale_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_RGB && RGB_PIXELSIZE == 3)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_YCbCr:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_RGB) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = rgb_ycc_convert;
+    } else if (cinfo->in_color_space == JCS_YCbCr)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_CMYK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_CMYK)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_YCCK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    if (cinfo->in_color_space == JCS_CMYK) {
+      cconvert->pub.start_pass = rgb_ycc_start;
+      cconvert->pub.color_convert = cmyk_ycck_convert;
+    } else if (cinfo->in_color_space == JCS_YCCK)
+      cconvert->pub.color_convert = null_convert;
+    else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  default:			/* allow null conversion of JCS_UNKNOWN */
+    if (cinfo->jpeg_color_space != cinfo->in_color_space ||
+	cinfo->num_components != cinfo->input_components)
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    cconvert->pub.color_convert = null_convert;
+    break;
+  }
+}
diff --git a/jpeg/jcdctmgr.c b/jpeg/jcdctmgr.c
new file mode 100644
index 000000000..61fa79b9e
--- /dev/null
+++ b/jpeg/jcdctmgr.c
@@ -0,0 +1,387 @@
+/*
+ * jcdctmgr.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the forward-DCT management logic.
+ * This code selects a particular DCT implementation to be used,
+ * and it performs related housekeeping chores including coefficient
+ * quantization.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+
+/* Private subobject for this module */
+
+typedef struct {
+  struct jpeg_forward_dct pub;	/* public fields */
+
+  /* Pointer to the DCT routine actually in use */
+  forward_DCT_method_ptr do_dct;
+
+  /* The actual post-DCT divisors --- not identical to the quant table
+   * entries, because of scaling (especially for an unnormalized DCT).
+   * Each table is given in normal array order.
+   */
+  DCTELEM * divisors[NUM_QUANT_TBLS];
+
+#ifdef DCT_FLOAT_SUPPORTED
+  /* Same as above for the floating-point case. */
+  float_DCT_method_ptr do_float_dct;
+  FAST_FLOAT * float_divisors[NUM_QUANT_TBLS];
+#endif
+} my_fdct_controller;
+
+typedef my_fdct_controller * my_fdct_ptr;
+
+
+/*
+ * Initialize for a processing pass.
+ * Verify that all referenced Q-tables are present, and set up
+ * the divisor table for each one.
+ * In the current implementation, DCT of all components is done during
+ * the first pass, even if only some components will be output in the
+ * first scan.  Hence all components should be examined here.
+ */
+
+METHODDEF(void)
+start_pass_fdctmgr (j_compress_ptr cinfo)
+{
+  my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
+  int ci, qtblno, i;
+  jpeg_component_info *compptr;
+  JQUANT_TBL * qtbl;
+  DCTELEM * dtbl;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    qtblno = compptr->quant_tbl_no;
+    /* Make sure specified quantization table is present */
+    if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
+	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+      ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
+    qtbl = cinfo->quant_tbl_ptrs[qtblno];
+    /* Compute divisors for this quant table */
+    /* We may do this more than once for same table, but it's not a big deal */
+    switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+    case JDCT_ISLOW:
+      /* For LL&M IDCT method, divisors are equal to raw quantization
+       * coefficients multiplied by 8 (to counteract scaling).
+       */
+      if (fdct->divisors[qtblno] == NULL) {
+	fdct->divisors[qtblno] = (DCTELEM *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      DCTSIZE2 * SIZEOF(DCTELEM));
+      }
+      dtbl = fdct->divisors[qtblno];
+      for (i = 0; i < DCTSIZE2; i++) {
+	dtbl[i] = ((DCTELEM) qtbl->quantval[i]) << 3;
+      }
+      break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+    case JDCT_IFAST:
+      {
+	/* For AA&N IDCT method, divisors are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 * We apply a further scale factor of 8.
+	 */
+#define CONST_BITS 14
+	static const INT16 aanscales[DCTSIZE2] = {
+	  /* precomputed values scaled up by 14 bits */
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+	};
+	SHIFT_TEMPS
+
+	if (fdct->divisors[qtblno] == NULL) {
+	  fdct->divisors[qtblno] = (DCTELEM *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					DCTSIZE2 * SIZEOF(DCTELEM));
+	}
+	dtbl = fdct->divisors[qtblno];
+	for (i = 0; i < DCTSIZE2; i++) {
+	  dtbl[i] = (DCTELEM)
+	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
+				  (INT32) aanscales[i]),
+		    CONST_BITS-3);
+	}
+      }
+      break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+    case JDCT_FLOAT:
+      {
+	/* For float AA&N IDCT method, divisors are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 * We apply a further scale factor of 8.
+	 * What's actually stored is 1/divisor so that the inner loop can
+	 * use a multiplication rather than a division.
+	 */
+	FAST_FLOAT * fdtbl;
+	int row, col;
+	static const double aanscalefactor[DCTSIZE] = {
+	  1.0, 1.387039845, 1.306562965, 1.175875602,
+	  1.0, 0.785694958, 0.541196100, 0.275899379
+	};
+
+	if (fdct->float_divisors[qtblno] == NULL) {
+	  fdct->float_divisors[qtblno] = (FAST_FLOAT *)
+	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+					DCTSIZE2 * SIZEOF(FAST_FLOAT));
+	}
+	fdtbl = fdct->float_divisors[qtblno];
+	i = 0;
+	for (row = 0; row < DCTSIZE; row++) {
+	  for (col = 0; col < DCTSIZE; col++) {
+	    fdtbl[i] = (FAST_FLOAT)
+	      (1.0 / (((double) qtbl->quantval[i] *
+		       aanscalefactor[row] * aanscalefactor[col] * 8.0)));
+	    i++;
+	  }
+	}
+      }
+      break;
+#endif
+    default:
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+      break;
+    }
+  }
+}
+
+
+/*
+ * Perform forward DCT on one or more blocks of a component.
+ *
+ * The input samples are taken from the sample_data[] array starting at
+ * position start_row/start_col, and moving to the right for any additional
+ * blocks. The quantized coefficients are returned in coef_blocks[].
+ */
+
+METHODDEF(void)
+forward_DCT (j_compress_ptr cinfo, jpeg_component_info * compptr,
+	     JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+	     JDIMENSION start_row, JDIMENSION start_col,
+	     JDIMENSION num_blocks)
+/* This version is used for integer DCT implementations. */
+{
+  /* This routine is heavily used, so it's worth coding it tightly. */
+  my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
+  forward_DCT_method_ptr do_dct = fdct->do_dct;
+  DCTELEM * divisors = fdct->divisors[compptr->quant_tbl_no];
+  DCTELEM workspace[DCTSIZE2];	/* work area for FDCT subroutine */
+  JDIMENSION bi;
+
+  sample_data += start_row;	/* fold in the vertical offset once */
+
+  for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
+    /* Load data into workspace, applying unsigned->signed conversion */
+    { register DCTELEM *workspaceptr;
+      register JSAMPROW elemptr;
+      register int elemr;
+
+      workspaceptr = workspace;
+      for (elemr = 0; elemr < DCTSIZE; elemr++) {
+	elemptr = sample_data[elemr] + start_col;
+#if DCTSIZE == 8		/* unroll the inner loop */
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	*workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+#else
+	{ register int elemc;
+	  for (elemc = DCTSIZE; elemc > 0; elemc--) {
+	    *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
+	  }
+	}
+#endif
+      }
+    }
+
+    /* Perform the DCT */
+    (*do_dct) (workspace);
+
+    /* Quantize/descale the coefficients, and store into coef_blocks[] */
+    { register DCTELEM temp, qval;
+      register int i;
+      register JCOEFPTR output_ptr = coef_blocks[bi];
+
+      for (i = 0; i < DCTSIZE2; i++) {
+	qval = divisors[i];
+	temp = workspace[i];
+	/* Divide the coefficient value by qval, ensuring proper rounding.
+	 * Since C does not specify the direction of rounding for negative
+	 * quotients, we have to force the dividend positive for portability.
+	 *
+	 * In most files, at least half of the output values will be zero
+	 * (at default quantization settings, more like three-quarters...)
+	 * so we should ensure that this case is fast.  On many machines,
+	 * a comparison is enough cheaper than a divide to make a special test
+	 * a win.  Since both inputs will be nonnegative, we need only test
+	 * for a < b to discover whether a/b is 0.
+	 * If your machine's division is fast enough, define FAST_DIVIDE.
+	 */
+#ifdef FAST_DIVIDE
+#define DIVIDE_BY(a,b)	a /= b
+#else
+#define DIVIDE_BY(a,b)	if (a >= b) a /= b; else a = 0
+#endif
+	if (temp < 0) {
+	  temp = -temp;
+	  temp += qval>>1;	/* for rounding */
+	  DIVIDE_BY(temp, qval);
+	  temp = -temp;
+	} else {
+	  temp += qval>>1;	/* for rounding */
+	  DIVIDE_BY(temp, qval);
+	}
+	output_ptr[i] = (JCOEF) temp;
+      }
+    }
+  }
+}
+
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+METHODDEF(void)
+forward_DCT_float (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		   JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+		   JDIMENSION start_row, JDIMENSION start_col,
+		   JDIMENSION num_blocks)
+/* This version is used for floating-point DCT implementations. */
+{
+  /* This routine is heavily used, so it's worth coding it tightly. */
+  my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
+  float_DCT_method_ptr do_dct = fdct->do_float_dct;
+  FAST_FLOAT * divisors = fdct->float_divisors[compptr->quant_tbl_no];
+  FAST_FLOAT workspace[DCTSIZE2]; /* work area for FDCT subroutine */
+  JDIMENSION bi;
+
+  sample_data += start_row;	/* fold in the vertical offset once */
+
+  for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
+    /* Load data into workspace, applying unsigned->signed conversion */
+    { register FAST_FLOAT *workspaceptr;
+      register JSAMPROW elemptr;
+      register int elemr;
+
+      workspaceptr = workspace;
+      for (elemr = 0; elemr < DCTSIZE; elemr++) {
+	elemptr = sample_data[elemr] + start_col;
+#if DCTSIZE == 8		/* unroll the inner loop */
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	*workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+#else
+	{ register int elemc;
+	  for (elemc = DCTSIZE; elemc > 0; elemc--) {
+	    *workspaceptr++ = (FAST_FLOAT)
+	      (GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
+	  }
+	}
+#endif
+      }
+    }
+
+    /* Perform the DCT */
+    (*do_dct) (workspace);
+
+    /* Quantize/descale the coefficients, and store into coef_blocks[] */
+    { register FAST_FLOAT temp;
+      register int i;
+      register JCOEFPTR output_ptr = coef_blocks[bi];
+
+      for (i = 0; i < DCTSIZE2; i++) {
+	/* Apply the quantization and scaling factor */
+	temp = workspace[i] * divisors[i];
+	/* Round to nearest integer.
+	 * Since C does not specify the direction of rounding for negative
+	 * quotients, we have to force the dividend positive for portability.
+	 * The maximum coefficient size is +-16K (for 12-bit data), so this
+	 * code should work for either 16-bit or 32-bit ints.
+	 */
+	output_ptr[i] = (JCOEF) ((int) (temp + (FAST_FLOAT) 16384.5) - 16384);
+      }
+    }
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
+
+
+/*
+ * Initialize FDCT manager.
+ */
+
+GLOBAL(void)
+jinit_forward_dct (j_compress_ptr cinfo)
+{
+  my_fdct_ptr fdct;
+  int i;
+
+  fdct = (my_fdct_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_fdct_controller));
+  cinfo->fdct = (struct jpeg_forward_dct *) fdct;
+  fdct->pub.start_pass = start_pass_fdctmgr;
+
+  switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+  case JDCT_ISLOW:
+    fdct->pub.forward_DCT = forward_DCT;
+    fdct->do_dct = jpeg_fdct_islow;
+    break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  case JDCT_IFAST:
+    fdct->pub.forward_DCT = forward_DCT;
+    fdct->do_dct = jpeg_fdct_ifast;
+    break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  case JDCT_FLOAT:
+    fdct->pub.forward_DCT = forward_DCT_float;
+    fdct->do_float_dct = jpeg_fdct_float;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+
+  /* Mark divisor tables unallocated */
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    fdct->divisors[i] = NULL;
+#ifdef DCT_FLOAT_SUPPORTED
+    fdct->float_divisors[i] = NULL;
+#endif
+  }
+}
diff --git a/jpeg/jchuff.c b/jpeg/jchuff.c
new file mode 100644
index 000000000..f23525054
--- /dev/null
+++ b/jpeg/jchuff.c
@@ -0,0 +1,909 @@
+/*
+ * jchuff.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy encoding routines.
+ *
+ * Much of the complexity here has to do with supporting output suspension.
+ * If the data destination module demands suspension, we want to be able to
+ * back up to the start of the current MCU.  To do this, we copy state
+ * variables into local working storage, and update them back to the
+ * permanent JPEG objects only upon successful completion of an MCU.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jchuff.h"		/* Declarations shared with jcphuff.c */
+
+
+/* Expanded entropy encoder object for Huffman encoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  INT32 put_buffer;		/* current bit-accumulation buffer */
+  int put_bits;			/* # of bits now in it */
+  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+} savable_state;
+
+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest,src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest,src)  \
+	((dest).put_buffer = (src).put_buffer, \
+	 (dest).put_bits = (src).put_bits, \
+	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
+	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
+	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
+	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  savable_state saved;		/* Bit buffer & DC state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  int next_restart_num;		/* next restart number to write (0-7) */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  c_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+
+#ifdef ENTROPY_OPT_SUPPORTED	/* Statistics tables for optimization */
+  long * dc_count_ptrs[NUM_HUFF_TBLS];
+  long * ac_count_ptrs[NUM_HUFF_TBLS];
+#endif
+} huff_entropy_encoder;
+
+typedef huff_entropy_encoder * huff_entropy_ptr;
+
+/* Working state while writing an MCU.
+ * This struct contains all the fields that are needed by subroutines.
+ */
+
+typedef struct {
+  JOCTET * next_output_byte;	/* => next byte to write in buffer */
+  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+  savable_state cur;		/* Current bit buffer & DC state */
+  j_compress_ptr cinfo;		/* dump_buffer needs access to this */
+} working_state;
+
+
+/* Forward declarations */
+METHODDEF(boolean) encode_mcu_huff JPP((j_compress_ptr cinfo,
+					JBLOCKROW *MCU_data));
+METHODDEF(void) finish_pass_huff JPP((j_compress_ptr cinfo));
+#ifdef ENTROPY_OPT_SUPPORTED
+METHODDEF(boolean) encode_mcu_gather JPP((j_compress_ptr cinfo,
+					  JBLOCKROW *MCU_data));
+METHODDEF(void) finish_pass_gather JPP((j_compress_ptr cinfo));
+#endif
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ * If gather_statistics is TRUE, we do not output anything during the scan,
+ * just count the Huffman symbols used and generate Huffman code tables.
+ */
+
+METHODDEF(void)
+start_pass_huff (j_compress_ptr cinfo, boolean gather_statistics)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci, dctbl, actbl;
+  jpeg_component_info * compptr;
+
+  if (gather_statistics) {
+#ifdef ENTROPY_OPT_SUPPORTED
+    entropy->pub.encode_mcu = encode_mcu_gather;
+    entropy->pub.finish_pass = finish_pass_gather;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    entropy->pub.encode_mcu = encode_mcu_huff;
+    entropy->pub.finish_pass = finish_pass_huff;
+  }
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    if (gather_statistics) {
+#ifdef ENTROPY_OPT_SUPPORTED
+      /* Check for invalid table indexes */
+      /* (make_c_derived_tbl does this in the other path) */
+      if (dctbl < 0 || dctbl >= NUM_HUFF_TBLS)
+	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
+      if (actbl < 0 || actbl >= NUM_HUFF_TBLS)
+	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
+      /* Allocate and zero the statistics tables */
+      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
+      if (entropy->dc_count_ptrs[dctbl] == NULL)
+	entropy->dc_count_ptrs[dctbl] = (long *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      257 * SIZEOF(long));
+      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * SIZEOF(long));
+      if (entropy->ac_count_ptrs[actbl] == NULL)
+	entropy->ac_count_ptrs[actbl] = (long *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      257 * SIZEOF(long));
+      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * SIZEOF(long));
+#endif
+    } else {
+      /* Compute derived values for Huffman tables */
+      /* We may do this more than once for a table, but it's not expensive */
+      jpeg_make_c_derived_tbl(cinfo, TRUE, dctbl,
+			      & entropy->dc_derived_tbls[dctbl]);
+      jpeg_make_c_derived_tbl(cinfo, FALSE, actbl,
+			      & entropy->ac_derived_tbls[actbl]);
+    }
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_val[ci] = 0;
+  }
+
+  /* Initialize bit buffer to empty */
+  entropy->saved.put_buffer = 0;
+  entropy->saved.put_bits = 0;
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/*
+ * Compute the derived values for a Huffman table.
+ * This routine also performs some validation checks on the table.
+ *
+ * Note this is also used by jcphuff.c.
+ */
+
+GLOBAL(void)
+jpeg_make_c_derived_tbl (j_compress_ptr cinfo, boolean isDC, int tblno,
+			 c_derived_tbl ** pdtbl)
+{
+  JHUFF_TBL *htbl;
+  c_derived_tbl *dtbl;
+  int p, i, l, lastp, si, maxsymbol;
+  char huffsize[257];
+  unsigned int huffcode[257];
+  unsigned int code;
+
+  /* Note that huffsize[] and huffcode[] are filled in code-length order,
+   * paralleling the order of the symbols themselves in htbl->huffval[].
+   */
+
+  /* Find the input Huffman table */
+  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+  htbl =
+    isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+
+  /* Allocate a workspace if we haven't already done so. */
+  if (*pdtbl == NULL)
+    *pdtbl = (c_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(c_derived_tbl));
+  dtbl = *pdtbl;
+  
+  /* Figure C.1: make table of Huffman code length for each symbol */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    i = (int) htbl->bits[l];
+    if (i < 0 || p + i > 256)	/* protect against table overrun */
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    while (i--)
+      huffsize[p++] = (char) l;
+  }
+  huffsize[p] = 0;
+  lastp = p;
+  
+  /* Figure C.2: generate the codes themselves */
+  /* We also validate that the counts represent a legal Huffman code tree. */
+
+  code = 0;
+  si = huffsize[0];
+  p = 0;
+  while (huffsize[p]) {
+    while (((int) huffsize[p]) == si) {
+      huffcode[p++] = code;
+      code++;
+    }
+    /* code is now 1 more than the last code used for codelength si; but
+     * it must still fit in si bits, since no code is allowed to be all ones.
+     */
+    if (((INT32) code) >= (((INT32) 1) << si))
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    code <<= 1;
+    si++;
+  }
+  
+  /* Figure C.3: generate encoding tables */
+  /* These are code and size indexed by symbol value */
+
+  /* Set all codeless symbols to have code length 0;
+   * this lets us detect duplicate VAL entries here, and later
+   * allows emit_bits to detect any attempt to emit such symbols.
+   */
+  MEMZERO(dtbl->ehufsi, SIZEOF(dtbl->ehufsi));
+
+  /* This is also a convenient place to check for out-of-range
+   * and duplicated VAL entries.  We allow 0..255 for AC symbols
+   * but only 0..15 for DC.  (We could constrain them further
+   * based on data depth and mode, but this seems enough.)
+   */
+  maxsymbol = isDC ? 15 : 255;
+
+  for (p = 0; p < lastp; p++) {
+    i = htbl->huffval[p];
+    if (i < 0 || i > maxsymbol || dtbl->ehufsi[i])
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    dtbl->ehufco[i] = huffcode[p];
+    dtbl->ehufsi[i] = huffsize[p];
+  }
+}
+
+
+/* Outputting bytes to the file */
+
+/* Emit a byte, taking 'action' if must suspend. */
+#define emit_byte(state,val,action)  \
+	{ *(state)->next_output_byte++ = (JOCTET) (val);  \
+	  if (--(state)->free_in_buffer == 0)  \
+	    if (! dump_buffer(state))  \
+	      { action; } }
+
+
+LOCAL(boolean)
+dump_buffer (working_state * state)
+/* Empty the output buffer; return TRUE if successful, FALSE if must suspend */
+{
+  struct jpeg_destination_mgr * dest = state->cinfo->dest;
+
+  if (! (*dest->empty_output_buffer) (state->cinfo))
+    return FALSE;
+  /* After a successful buffer dump, must reset buffer pointers */
+  state->next_output_byte = dest->next_output_byte;
+  state->free_in_buffer = dest->free_in_buffer;
+  return TRUE;
+}
+
+
+/* Outputting bits to the file */
+
+/* Only the right 24 bits of put_buffer are used; the valid bits are
+ * left-justified in this part.  At most 16 bits can be passed to emit_bits
+ * in one call, and we never retain more than 7 bits in put_buffer
+ * between calls, so 24 bits are sufficient.
+ */
+
+INLINE
+LOCAL(boolean)
+emit_bits (working_state * state, unsigned int code, int size)
+/* Emit some bits; return TRUE if successful, FALSE if must suspend */
+{
+  /* This routine is heavily used, so it's worth coding tightly. */
+  register INT32 put_buffer = (INT32) code;
+  register int put_bits = state->cur.put_bits;
+
+  /* if size is 0, caller used an invalid Huffman table entry */
+  if (size == 0)
+    ERREXIT(state->cinfo, JERR_HUFF_MISSING_CODE);
+
+  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
+  
+  put_bits += size;		/* new number of bits in buffer */
+  
+  put_buffer <<= 24 - put_bits; /* align incoming bits */
+
+  put_buffer |= state->cur.put_buffer; /* and merge with old buffer contents */
+  
+  while (put_bits >= 8) {
+    int c = (int) ((put_buffer >> 16) & 0xFF);
+    
+    emit_byte(state, c, return FALSE);
+    if (c == 0xFF) {		/* need to stuff a zero byte? */
+      emit_byte(state, 0, return FALSE);
+    }
+    put_buffer <<= 8;
+    put_bits -= 8;
+  }
+
+  state->cur.put_buffer = put_buffer; /* update state variables */
+  state->cur.put_bits = put_bits;
+
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+flush_bits (working_state * state)
+{
+  if (! emit_bits(state, 0x7F, 7)) /* fill any partial byte with ones */
+    return FALSE;
+  state->cur.put_buffer = 0;	/* and reset bit-buffer to empty */
+  state->cur.put_bits = 0;
+  return TRUE;
+}
+
+
+/* Encode a single block's worth of coefficients */
+
+LOCAL(boolean)
+encode_one_block (working_state * state, JCOEFPTR block, int last_dc_val,
+		  c_derived_tbl *dctbl, c_derived_tbl *actbl)
+{
+  register int temp, temp2;
+  register int nbits;
+  register int k, r, i;
+  
+  /* Encode the DC coefficient difference per section F.1.2.1 */
+  
+  temp = temp2 = block[0] - last_dc_val;
+
+  if (temp < 0) {
+    temp = -temp;		/* temp is abs value of input */
+    /* For a negative input, want temp2 = bitwise complement of abs(input) */
+    /* This code assumes we are on a two's complement machine */
+    temp2--;
+  }
+  
+  /* Find the number of bits needed for the magnitude of the coefficient */
+  nbits = 0;
+  while (temp) {
+    nbits++;
+    temp >>= 1;
+  }
+  /* Check for out-of-range coefficient values.
+   * Since we're encoding a difference, the range limit is twice as much.
+   */
+  if (nbits > MAX_COEF_BITS+1)
+    ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
+  
+  /* Emit the Huffman-coded symbol for the number of bits */
+  if (! emit_bits(state, dctbl->ehufco[nbits], dctbl->ehufsi[nbits]))
+    return FALSE;
+
+  /* Emit that number of bits of the value, if positive, */
+  /* or the complement of its magnitude, if negative. */
+  if (nbits)			/* emit_bits rejects calls with size 0 */
+    if (! emit_bits(state, (unsigned int) temp2, nbits))
+      return FALSE;
+
+  /* Encode the AC coefficients per section F.1.2.2 */
+  
+  r = 0;			/* r = run length of zeros */
+  
+  for (k = 1; k < DCTSIZE2; k++) {
+    if ((temp = block[jpeg_natural_order[k]]) == 0) {
+      r++;
+    } else {
+      /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+      while (r > 15) {
+	if (! emit_bits(state, actbl->ehufco[0xF0], actbl->ehufsi[0xF0]))
+	  return FALSE;
+	r -= 16;
+      }
+
+      temp2 = temp;
+      if (temp < 0) {
+	temp = -temp;		/* temp is abs value of input */
+	/* This code assumes we are on a two's complement machine */
+	temp2--;
+      }
+      
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      nbits = 1;		/* there must be at least one 1 bit */
+      while ((temp >>= 1))
+	nbits++;
+      /* Check for out-of-range coefficient values */
+      if (nbits > MAX_COEF_BITS)
+	ERREXIT(state->cinfo, JERR_BAD_DCT_COEF);
+      
+      /* Emit Huffman symbol for run length / number of bits */
+      i = (r << 4) + nbits;
+      if (! emit_bits(state, actbl->ehufco[i], actbl->ehufsi[i]))
+	return FALSE;
+
+      /* Emit that number of bits of the value, if positive, */
+      /* or the complement of its magnitude, if negative. */
+      if (! emit_bits(state, (unsigned int) temp2, nbits))
+	return FALSE;
+      
+      r = 0;
+    }
+  }
+
+  /* If the last coef(s) were zero, emit an end-of-block code */
+  if (r > 0)
+    if (! emit_bits(state, actbl->ehufco[0], actbl->ehufsi[0]))
+      return FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(boolean)
+emit_restart (working_state * state, int restart_num)
+{
+  int ci;
+
+  if (! flush_bits(state))
+    return FALSE;
+
+  emit_byte(state, 0xFF, return FALSE);
+  emit_byte(state, JPEG_RST0 + restart_num, return FALSE);
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < state->cinfo->comps_in_scan; ci++)
+    state->cur.last_dc_val[ci] = 0;
+
+  /* The restart counter is not updated until we successfully write the MCU. */
+
+  return TRUE;
+}
+
+
+/*
+ * Encode and output one MCU's worth of Huffman-compressed coefficients.
+ */
+
+METHODDEF(boolean)
+encode_mcu_huff (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  working_state state;
+  int blkn, ci;
+  jpeg_component_info * compptr;
+
+  /* Load up working state */
+  state.next_output_byte = cinfo->dest->next_output_byte;
+  state.free_in_buffer = cinfo->dest->free_in_buffer;
+  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cinfo = cinfo;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! emit_restart(&state, entropy->next_restart_num))
+	return FALSE;
+  }
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+    if (! encode_one_block(&state,
+			   MCU_data[blkn][0], state.cur.last_dc_val[ci],
+			   entropy->dc_derived_tbls[compptr->dc_tbl_no],
+			   entropy->ac_derived_tbls[compptr->ac_tbl_no]))
+      return FALSE;
+    /* Update last_dc_val */
+    state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+  }
+
+  /* Completed MCU, so update state */
+  cinfo->dest->next_output_byte = state.next_output_byte;
+  cinfo->dest->free_in_buffer = state.free_in_buffer;
+  ASSIGN_STATE(entropy->saved, state.cur);
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Finish up at the end of a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+finish_pass_huff (j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  working_state state;
+
+  /* Load up working state ... flush_bits needs it */
+  state.next_output_byte = cinfo->dest->next_output_byte;
+  state.free_in_buffer = cinfo->dest->free_in_buffer;
+  ASSIGN_STATE(state.cur, entropy->saved);
+  state.cinfo = cinfo;
+
+  /* Flush out the last data */
+  if (! flush_bits(&state))
+    ERREXIT(cinfo, JERR_CANT_SUSPEND);
+
+  /* Update state */
+  cinfo->dest->next_output_byte = state.next_output_byte;
+  cinfo->dest->free_in_buffer = state.free_in_buffer;
+  ASSIGN_STATE(entropy->saved, state.cur);
+}
+
+
+/*
+ * Huffman coding optimization.
+ *
+ * We first scan the supplied data and count the number of uses of each symbol
+ * that is to be Huffman-coded. (This process MUST agree with the code above.)
+ * Then we build a Huffman coding tree for the observed counts.
+ * Symbols which are not needed at all for the particular image are not
+ * assigned any code, which saves space in the DHT marker as well as in
+ * the compressed data.
+ */
+
+#ifdef ENTROPY_OPT_SUPPORTED
+
+
+/* Process a single block's worth of coefficients */
+
+LOCAL(void)
+htest_one_block (j_compress_ptr cinfo, JCOEFPTR block, int last_dc_val,
+		 long dc_counts[], long ac_counts[])
+{
+  register int temp;
+  register int nbits;
+  register int k, r;
+  
+  /* Encode the DC coefficient difference per section F.1.2.1 */
+  
+  temp = block[0] - last_dc_val;
+  if (temp < 0)
+    temp = -temp;
+  
+  /* Find the number of bits needed for the magnitude of the coefficient */
+  nbits = 0;
+  while (temp) {
+    nbits++;
+    temp >>= 1;
+  }
+  /* Check for out-of-range coefficient values.
+   * Since we're encoding a difference, the range limit is twice as much.
+   */
+  if (nbits > MAX_COEF_BITS+1)
+    ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+  /* Count the Huffman symbol for the number of bits */
+  dc_counts[nbits]++;
+  
+  /* Encode the AC coefficients per section F.1.2.2 */
+  
+  r = 0;			/* r = run length of zeros */
+  
+  for (k = 1; k < DCTSIZE2; k++) {
+    if ((temp = block[jpeg_natural_order[k]]) == 0) {
+      r++;
+    } else {
+      /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+      while (r > 15) {
+	ac_counts[0xF0]++;
+	r -= 16;
+      }
+      
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      if (temp < 0)
+	temp = -temp;
+      
+      /* Find the number of bits needed for the magnitude of the coefficient */
+      nbits = 1;		/* there must be at least one 1 bit */
+      while ((temp >>= 1))
+	nbits++;
+      /* Check for out-of-range coefficient values */
+      if (nbits > MAX_COEF_BITS)
+	ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+      
+      /* Count Huffman symbol for run length / number of bits */
+      ac_counts[(r << 4) + nbits]++;
+      
+      r = 0;
+    }
+  }
+
+  /* If the last coef(s) were zero, emit an end-of-block code */
+  if (r > 0)
+    ac_counts[0]++;
+}
+
+
+/*
+ * Trial-encode one MCU's worth of Huffman-compressed coefficients.
+ * No data is actually output, so no suspension return is possible.
+ */
+
+METHODDEF(boolean)
+encode_mcu_gather (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int blkn, ci;
+  jpeg_component_info * compptr;
+
+  /* Take care of restart intervals if needed */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      /* Re-initialize DC predictions to 0 */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+	entropy->saved.last_dc_val[ci] = 0;
+      /* Update restart state */
+      entropy->restarts_to_go = cinfo->restart_interval;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+    htest_one_block(cinfo, MCU_data[blkn][0], entropy->saved.last_dc_val[ci],
+		    entropy->dc_count_ptrs[compptr->dc_tbl_no],
+		    entropy->ac_count_ptrs[compptr->ac_tbl_no]);
+    entropy->saved.last_dc_val[ci] = MCU_data[blkn][0][0];
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Generate the best Huffman code table for the given counts, fill htbl.
+ * Note this is also used by jcphuff.c.
+ *
+ * The JPEG standard requires that no symbol be assigned a codeword of all
+ * one bits (so that padding bits added at the end of a compressed segment
+ * can't look like a valid code).  Because of the canonical ordering of
+ * codewords, this just means that there must be an unused slot in the
+ * longest codeword length category.  Section K.2 of the JPEG spec suggests
+ * reserving such a slot by pretending that symbol 256 is a valid symbol
+ * with count 1.  In theory that's not optimal; giving it count zero but
+ * including it in the symbol set anyway should give a better Huffman code.
+ * But the theoretically better code actually seems to come out worse in
+ * practice, because it produces more all-ones bytes (which incur stuffed
+ * zero bytes in the final file).  In any case the difference is tiny.
+ *
+ * The JPEG standard requires Huffman codes to be no more than 16 bits long.
+ * If some symbols have a very small but nonzero probability, the Huffman tree
+ * must be adjusted to meet the code length restriction.  We currently use
+ * the adjustment method suggested in JPEG section K.2.  This method is *not*
+ * optimal; it may not choose the best possible limited-length code.  But
+ * typically only very-low-frequency symbols will be given less-than-optimal
+ * lengths, so the code is almost optimal.  Experimental comparisons against
+ * an optimal limited-length-code algorithm indicate that the difference is
+ * microscopic --- usually less than a hundredth of a percent of total size.
+ * So the extra complexity of an optimal algorithm doesn't seem worthwhile.
+ */
+
+GLOBAL(void)
+jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
+{
+#define MAX_CLEN 32		/* assumed maximum initial code length */
+  UINT8 bits[MAX_CLEN+1];	/* bits[k] = # of symbols with code length k */
+  int codesize[257];		/* codesize[k] = code length of symbol k */
+  int others[257];		/* next symbol in current branch of tree */
+  int c1, c2;
+  int p, i, j;
+  long v;
+
+  /* This algorithm is explained in section K.2 of the JPEG standard */
+
+  MEMZERO(bits, SIZEOF(bits));
+  MEMZERO(codesize, SIZEOF(codesize));
+  for (i = 0; i < 257; i++)
+    others[i] = -1;		/* init links to empty */
+  
+  freq[256] = 1;		/* make sure 256 has a nonzero count */
+  /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
+   * that no real symbol is given code-value of all ones, because 256
+   * will be placed last in the largest codeword category.
+   */
+
+  /* Huffman's basic algorithm to assign optimal code lengths to symbols */
+
+  for (;;) {
+    /* Find the smallest nonzero frequency, set c1 = its symbol */
+    /* In case of ties, take the larger symbol number */
+    c1 = -1;
+    v = 1000000000L;
+    for (i = 0; i <= 256; i++) {
+      if (freq[i] && freq[i] <= v) {
+	v = freq[i];
+	c1 = i;
+      }
+    }
+
+    /* Find the next smallest nonzero frequency, set c2 = its symbol */
+    /* In case of ties, take the larger symbol number */
+    c2 = -1;
+    v = 1000000000L;
+    for (i = 0; i <= 256; i++) {
+      if (freq[i] && freq[i] <= v && i != c1) {
+	v = freq[i];
+	c2 = i;
+      }
+    }
+
+    /* Done if we've merged everything into one frequency */
+    if (c2 < 0)
+      break;
+    
+    /* Else merge the two counts/trees */
+    freq[c1] += freq[c2];
+    freq[c2] = 0;
+
+    /* Increment the codesize of everything in c1's tree branch */
+    codesize[c1]++;
+    while (others[c1] >= 0) {
+      c1 = others[c1];
+      codesize[c1]++;
+    }
+    
+    others[c1] = c2;		/* chain c2 onto c1's tree branch */
+    
+    /* Increment the codesize of everything in c2's tree branch */
+    codesize[c2]++;
+    while (others[c2] >= 0) {
+      c2 = others[c2];
+      codesize[c2]++;
+    }
+  }
+
+  /* Now count the number of symbols of each code length */
+  for (i = 0; i <= 256; i++) {
+    if (codesize[i]) {
+      /* The JPEG standard seems to think that this can't happen, */
+      /* but I'm paranoid... */
+      if (codesize[i] > MAX_CLEN)
+	ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
+
+      bits[codesize[i]]++;
+    }
+  }
+
+  /* JPEG doesn't allow symbols with code lengths over 16 bits, so if the pure
+   * Huffman procedure assigned any such lengths, we must adjust the coding.
+   * Here is what the JPEG spec says about how this next bit works:
+   * Since symbols are paired for the longest Huffman code, the symbols are
+   * removed from this length category two at a time.  The prefix for the pair
+   * (which is one bit shorter) is allocated to one of the pair; then,
+   * skipping the BITS entry for that prefix length, a code word from the next
+   * shortest nonzero BITS entry is converted into a prefix for two code words
+   * one bit longer.
+   */
+  
+  for (i = MAX_CLEN; i > 16; i--) {
+    while (bits[i] > 0) {
+      j = i - 2;		/* find length of new prefix to be used */
+      while (bits[j] == 0)
+	j--;
+      
+      bits[i] -= 2;		/* remove two symbols */
+      bits[i-1]++;		/* one goes in this length */
+      bits[j+1] += 2;		/* two new symbols in this length */
+      bits[j]--;		/* symbol of this length is now a prefix */
+    }
+  }
+
+  /* Remove the count for the pseudo-symbol 256 from the largest codelength */
+  while (bits[i] == 0)		/* find largest codelength still in use */
+    i--;
+  bits[i]--;
+  
+  /* Return final symbol counts (only for lengths 0..16) */
+  MEMCOPY(htbl->bits, bits, SIZEOF(htbl->bits));
+  
+  /* Return a list of the symbols sorted by code length */
+  /* It's not real clear to me why we don't need to consider the codelength
+   * changes made above, but the JPEG spec seems to think this works.
+   */
+  p = 0;
+  for (i = 1; i <= MAX_CLEN; i++) {
+    for (j = 0; j <= 255; j++) {
+      if (codesize[j] == i) {
+	htbl->huffval[p] = (UINT8) j;
+	p++;
+      }
+    }
+  }
+
+  /* Set sent_table FALSE so updated table will be written to JPEG file. */
+  htbl->sent_table = FALSE;
+}
+
+
+/*
+ * Finish up a statistics-gathering pass and create the new Huffman tables.
+ */
+
+METHODDEF(void)
+finish_pass_gather (j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci, dctbl, actbl;
+  jpeg_component_info * compptr;
+  JHUFF_TBL **htblptr;
+  boolean did_dc[NUM_HUFF_TBLS];
+  boolean did_ac[NUM_HUFF_TBLS];
+
+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
+  MEMZERO(did_dc, SIZEOF(did_dc));
+  MEMZERO(did_ac, SIZEOF(did_ac));
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    if (! did_dc[dctbl]) {
+      htblptr = & cinfo->dc_huff_tbl_ptrs[dctbl];
+      if (*htblptr == NULL)
+	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->dc_count_ptrs[dctbl]);
+      did_dc[dctbl] = TRUE;
+    }
+    if (! did_ac[actbl]) {
+      htblptr = & cinfo->ac_huff_tbl_ptrs[actbl];
+      if (*htblptr == NULL)
+	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->ac_count_ptrs[actbl]);
+      did_ac[actbl] = TRUE;
+    }
+  }
+}
+
+
+#endif /* ENTROPY_OPT_SUPPORTED */
+
+
+/*
+ * Module initialization routine for Huffman entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_huff_encoder (j_compress_ptr cinfo)
+{
+  huff_entropy_ptr entropy;
+  int i;
+
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(huff_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
+  entropy->pub.start_pass = start_pass_huff;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL;
+#ifdef ENTROPY_OPT_SUPPORTED
+    entropy->dc_count_ptrs[i] = entropy->ac_count_ptrs[i] = NULL;
+#endif
+  }
+}
diff --git a/jpeg/jchuff.h b/jpeg/jchuff.h
new file mode 100644
index 000000000..a9599fc1e
--- /dev/null
+++ b/jpeg/jchuff.h
@@ -0,0 +1,47 @@
+/*
+ * jchuff.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains declarations for Huffman entropy encoding routines
+ * that are shared between the sequential encoder (jchuff.c) and the
+ * progressive encoder (jcphuff.c).  No other modules need to see these.
+ */
+
+/* The legal range of a DCT coefficient is
+ *  -1024 .. +1023  for 8-bit data;
+ * -16384 .. +16383 for 12-bit data.
+ * Hence the magnitude should always fit in 10 or 14 bits respectively.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MAX_COEF_BITS 10
+#else
+#define MAX_COEF_BITS 14
+#endif
+
+/* Derived data constructed for each Huffman table */
+
+typedef struct {
+  unsigned int ehufco[256];	/* code for each symbol */
+  char ehufsi[256];		/* length of code for each symbol */
+  /* If no code has been allocated for a symbol S, ehufsi[S] contains 0 */
+} c_derived_tbl;
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_make_c_derived_tbl	jMkCDerived
+#define jpeg_gen_optimal_table	jGenOptTbl
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+/* Expand a Huffman table definition into the derived format */
+EXTERN(void) jpeg_make_c_derived_tbl
+	JPP((j_compress_ptr cinfo, boolean isDC, int tblno,
+	     c_derived_tbl ** pdtbl));
+
+/* Generate an optimal table definition given the specified counts */
+EXTERN(void) jpeg_gen_optimal_table
+	JPP((j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[]));
diff --git a/jpeg/jcinit.c b/jpeg/jcinit.c
new file mode 100644
index 000000000..5efffe331
--- /dev/null
+++ b/jpeg/jcinit.c
@@ -0,0 +1,72 @@
+/*
+ * jcinit.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains initialization logic for the JPEG compressor.
+ * This routine is in charge of selecting the modules to be executed and
+ * making an initialization call to each one.
+ *
+ * Logically, this code belongs in jcmaster.c.  It's split out because
+ * linking this routine implies linking the entire compression library.
+ * For a transcoding-only application, we want to be able to use jcmaster.c
+ * without linking in the whole library.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Master selection of compression modules.
+ * This is done once at the start of processing an image.  We determine
+ * which modules will be used and give them appropriate initialization calls.
+ */
+
+GLOBAL(void)
+jinit_compress_master (j_compress_ptr cinfo)
+{
+  /* Initialize master control (includes parameter checking/processing) */
+  jinit_c_master_control(cinfo, FALSE /* full compression */);
+
+  /* Preprocessing */
+  if (! cinfo->raw_data_in) {
+    jinit_color_converter(cinfo);
+    jinit_downsampler(cinfo);
+    jinit_c_prep_controller(cinfo, FALSE /* never need full buffer here */);
+  }
+  /* Forward DCT */
+  jinit_forward_dct(cinfo);
+  /* Entropy encoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      jinit_phuff_encoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_encoder(cinfo);
+  }
+
+  /* Need a full-image coefficient buffer in any multi-pass mode. */
+  jinit_c_coef_controller(cinfo,
+		(boolean) (cinfo->num_scans > 1 || cinfo->optimize_coding));
+  jinit_c_main_controller(cinfo, FALSE /* never need full buffer here */);
+
+  jinit_marker_writer(cinfo);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Write the datastream header (SOI) immediately.
+   * Frame and scan headers are postponed till later.
+   * This lets application insert special markers after the SOI.
+   */
+  (*cinfo->marker->write_file_header) (cinfo);
+}
diff --git a/jpeg/jcmainct.c b/jpeg/jcmainct.c
new file mode 100644
index 000000000..a0d82683f
--- /dev/null
+++ b/jpeg/jcmainct.c
@@ -0,0 +1,293 @@
+/*
+ * jcmainct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the main buffer controller for compression.
+ * The main buffer lies between the pre-processor and the JPEG
+ * compressor proper; it holds downsampled data in the JPEG colorspace.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Note: currently, there is no operating mode in which a full-image buffer
+ * is needed at this step.  If there were, that mode could not be used with
+ * "raw data" input, since this module is bypassed in that case.  However,
+ * we've left the code here for possible use in special applications.
+ */
+#undef FULL_MAIN_BUFFER_SUPPORTED
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_main_controller pub; /* public fields */
+
+  JDIMENSION cur_iMCU_row;	/* number of current iMCU row */
+  JDIMENSION rowgroup_ctr;	/* counts row groups received in iMCU row */
+  boolean suspended;		/* remember if we suspended output */
+  J_BUF_MODE pass_mode;		/* current operating mode */
+
+  /* If using just a strip buffer, this points to the entire set of buffers
+   * (we allocate one for each component).  In the full-image case, this
+   * points to the currently accessible strips of the virtual arrays.
+   */
+  JSAMPARRAY buffer[MAX_COMPONENTS];
+
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+  /* If using full-image storage, this array holds pointers to virtual-array
+   * control blocks for each component.  Unused if not full-image storage.
+   */
+  jvirt_sarray_ptr whole_image[MAX_COMPONENTS];
+#endif
+} my_main_controller;
+
+typedef my_main_controller * my_main_ptr;
+
+
+/* Forward declarations */
+METHODDEF(void) process_data_simple_main
+	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
+	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+METHODDEF(void) process_data_buffer_main
+	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
+	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
+#endif
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_main (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+
+  /* Do nothing in raw-data mode. */
+  if (cinfo->raw_data_in)
+    return;
+
+  jmain->cur_iMCU_row = 0;	/* initialize counters */
+  jmain->rowgroup_ctr = 0;
+  jmain->suspended = FALSE;
+  jmain->pass_mode = pass_mode;	/* save mode for use by process_data */
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+    if (jmain->whole_image[0] != NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+    jmain->pub.process_data = process_data_simple_main;
+    break;
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+  case JBUF_SAVE_SOURCE:
+  case JBUF_CRANK_DEST:
+  case JBUF_SAVE_AND_PASS:
+    if (jmain->whole_image[0] == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    jmain->pub.process_data = process_data_buffer_main;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This routine handles the simple pass-through mode,
+ * where we have only a strip buffer.
+ */
+
+METHODDEF(void)
+process_data_simple_main (j_compress_ptr cinfo,
+			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+			  JDIMENSION in_rows_avail)
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+
+  while (jmain->cur_iMCU_row < cinfo->total_iMCU_rows) {
+    /* Read input data if we haven't filled the main buffer yet */
+    if (jmain->rowgroup_ctr < DCTSIZE)
+      (*cinfo->prep->pre_process_data) (cinfo,
+					input_buf, in_row_ctr, in_rows_avail,
+					jmain->buffer, &jmain->rowgroup_ctr,
+					(JDIMENSION) DCTSIZE);
+
+    /* If we don't have a full iMCU row buffered, return to application for
+     * more data.  Note that preprocessor will always pad to fill the iMCU row
+     * at the bottom of the image.
+     */
+    if (jmain->rowgroup_ctr != DCTSIZE)
+      return;
+
+    /* Send the completed row to the compressor */
+    if (! (*cinfo->coef->compress_data) (cinfo, jmain->buffer)) {
+      /* If compressor did not consume the whole row, then we must need to
+       * suspend processing and return to the application.  In this situation
+       * we pretend we didn't yet consume the last input row; otherwise, if
+       * it happened to be the last row of the image, the application would
+       * think we were done.
+       */
+      if (! jmain->suspended) {
+	(*in_row_ctr)--;
+	jmain->suspended = TRUE;
+      }
+      return;
+    }
+    /* We did finish the row.  Undo our little suspension hack if a previous
+     * call suspended; then mark the main buffer empty.
+     */
+    if (jmain->suspended) {
+      (*in_row_ctr)++;
+      jmain->suspended = FALSE;
+    }
+    jmain->rowgroup_ctr = 0;
+    jmain->cur_iMCU_row++;
+  }
+}
+
+
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+
+/*
+ * Process some data.
+ * This routine handles all of the modes that use a full-size buffer.
+ */
+
+METHODDEF(void)
+process_data_buffer_main (j_compress_ptr cinfo,
+			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+			  JDIMENSION in_rows_avail)
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+  int ci;
+  jpeg_component_info *compptr;
+  boolean writing = (jmain->pass_mode != JBUF_CRANK_DEST);
+
+  while (jmain->cur_iMCU_row < cinfo->total_iMCU_rows) {
+    /* Realign the virtual buffers if at the start of an iMCU row. */
+    if (jmain->rowgroup_ctr == 0) {
+      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	   ci++, compptr++) {
+	jmain->buffer[ci] = (*cinfo->mem->access_virt_sarray)
+	  ((j_common_ptr) cinfo, jmain->whole_image[ci],
+	   jmain->cur_iMCU_row * (compptr->v_samp_factor * DCTSIZE),
+	   (JDIMENSION) (compptr->v_samp_factor * DCTSIZE), writing);
+      }
+      /* In a read pass, pretend we just read some source data. */
+      if (! writing) {
+	*in_row_ctr += cinfo->max_v_samp_factor * DCTSIZE;
+	jmain->rowgroup_ctr = DCTSIZE;
+      }
+    }
+
+    /* If a write pass, read input data until the current iMCU row is full. */
+    /* Note: preprocessor will pad if necessary to fill the last iMCU row. */
+    if (writing) {
+      (*cinfo->prep->pre_process_data) (cinfo,
+					input_buf, in_row_ctr, in_rows_avail,
+					jmain->buffer, &jmain->rowgroup_ctr,
+					(JDIMENSION) DCTSIZE);
+      /* Return to application if we need more data to fill the iMCU row. */
+      if (jmain->rowgroup_ctr < DCTSIZE)
+	return;
+    }
+
+    /* Emit data, unless this is a sink-only pass. */
+    if (jmain->pass_mode != JBUF_SAVE_SOURCE) {
+      if (! (*cinfo->coef->compress_data) (cinfo, jmain->buffer)) {
+	/* If compressor did not consume the whole row, then we must need to
+	 * suspend processing and return to the application.  In this situation
+	 * we pretend we didn't yet consume the last input row; otherwise, if
+	 * it happened to be the last row of the image, the application would
+	 * think we were done.
+	 */
+	if (! jmain->suspended) {
+	  (*in_row_ctr)--;
+	  jmain->suspended = TRUE;
+	}
+	return;
+      }
+      /* We did finish the row.  Undo our little suspension hack if a previous
+       * call suspended; then mark the main buffer empty.
+       */
+      if (jmain->suspended) {
+	(*in_row_ctr)++;
+	jmain->suspended = FALSE;
+      }
+    }
+
+    /* If get here, we are done with this iMCU row.  Mark buffer empty. */
+    jmain->rowgroup_ctr = 0;
+    jmain->cur_iMCU_row++;
+  }
+}
+
+#endif /* FULL_MAIN_BUFFER_SUPPORTED */
+
+
+/*
+ * Initialize main buffer controller.
+ */
+
+GLOBAL(void)
+jinit_c_main_controller (j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_main_ptr jmain;
+  int ci;
+  jpeg_component_info *compptr;
+
+  jmain = (my_main_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_main_controller));
+  cinfo->main = (struct jpeg_c_main_controller *) jmain;
+  jmain->pub.start_pass = start_pass_main;
+
+  /* We don't need to create a buffer in raw-data mode. */
+  if (cinfo->raw_data_in)
+    return;
+
+  /* Create the buffer.  It holds downsampled data, so each component
+   * may be of a different size.
+   */
+  if (need_full_buffer) {
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+    /* Allocate a full-image virtual array for each component */
+    /* Note we pad the bottom to a multiple of the iMCU height */
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      jmain->whole_image[ci] = (*cinfo->mem->request_virt_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+	 compptr->width_in_blocks * DCTSIZE,
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor) * DCTSIZE,
+	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
+    }
+#else
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif
+  } else {
+#ifdef FULL_MAIN_BUFFER_SUPPORTED
+    jmain->whole_image[0] = NULL; /* flag for no virtual arrays */
+#endif
+    /* Allocate a strip buffer for each component */
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      jmain->buffer[ci] = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 compptr->width_in_blocks * DCTSIZE,
+	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
+    }
+  }
+}
diff --git a/jpeg/jcmarker.c b/jpeg/jcmarker.c
new file mode 100644
index 000000000..3d1e6c6d5
--- /dev/null
+++ b/jpeg/jcmarker.c
@@ -0,0 +1,664 @@
+/*
+ * jcmarker.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to write JPEG datastream markers.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+typedef enum {			/* JPEG marker codes */
+  M_SOF0  = 0xc0,
+  M_SOF1  = 0xc1,
+  M_SOF2  = 0xc2,
+  M_SOF3  = 0xc3,
+  
+  M_SOF5  = 0xc5,
+  M_SOF6  = 0xc6,
+  M_SOF7  = 0xc7,
+  
+  M_JPG   = 0xc8,
+  M_SOF9  = 0xc9,
+  M_SOF10 = 0xca,
+  M_SOF11 = 0xcb,
+  
+  M_SOF13 = 0xcd,
+  M_SOF14 = 0xce,
+  M_SOF15 = 0xcf,
+  
+  M_DHT   = 0xc4,
+  
+  M_DAC   = 0xcc,
+  
+  M_RST0  = 0xd0,
+  M_RST1  = 0xd1,
+  M_RST2  = 0xd2,
+  M_RST3  = 0xd3,
+  M_RST4  = 0xd4,
+  M_RST5  = 0xd5,
+  M_RST6  = 0xd6,
+  M_RST7  = 0xd7,
+  
+  M_SOI   = 0xd8,
+  M_EOI   = 0xd9,
+  M_SOS   = 0xda,
+  M_DQT   = 0xdb,
+  M_DNL   = 0xdc,
+  M_DRI   = 0xdd,
+  M_DHP   = 0xde,
+  M_EXP   = 0xdf,
+  
+  M_APP0  = 0xe0,
+  M_APP1  = 0xe1,
+  M_APP2  = 0xe2,
+  M_APP3  = 0xe3,
+  M_APP4  = 0xe4,
+  M_APP5  = 0xe5,
+  M_APP6  = 0xe6,
+  M_APP7  = 0xe7,
+  M_APP8  = 0xe8,
+  M_APP9  = 0xe9,
+  M_APP10 = 0xea,
+  M_APP11 = 0xeb,
+  M_APP12 = 0xec,
+  M_APP13 = 0xed,
+  M_APP14 = 0xee,
+  M_APP15 = 0xef,
+  
+  M_JPG0  = 0xf0,
+  M_JPG13 = 0xfd,
+  M_COM   = 0xfe,
+  
+  M_TEM   = 0x01,
+  
+  M_ERROR = 0x100
+} JPEG_MARKER;
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_marker_writer pub; /* public fields */
+
+  unsigned int last_restart_interval; /* last DRI value emitted; 0 after SOI */
+} my_marker_writer;
+
+typedef my_marker_writer * my_marker_ptr;
+
+
+/*
+ * Basic output routines.
+ *
+ * Note that we do not support suspension while writing a marker.
+ * Therefore, an application using suspension must ensure that there is
+ * enough buffer space for the initial markers (typ. 600-700 bytes) before
+ * calling jpeg_start_compress, and enough space to write the trailing EOI
+ * (a few bytes) before calling jpeg_finish_compress.  Multipass compression
+ * modes are not supported at all with suspension, so those two are the only
+ * points where markers will be written.
+ */
+
+LOCAL(void)
+emit_byte (j_compress_ptr cinfo, int val)
+/* Emit a byte */
+{
+  struct jpeg_destination_mgr * dest = cinfo->dest;
+
+  *(dest->next_output_byte)++ = (JOCTET) val;
+  if (--dest->free_in_buffer == 0) {
+    if (! (*dest->empty_output_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+}
+
+
+LOCAL(void)
+emit_marker (j_compress_ptr cinfo, JPEG_MARKER mark)
+/* Emit a marker code */
+{
+  emit_byte(cinfo, 0xFF);
+  emit_byte(cinfo, (int) mark);
+}
+
+
+LOCAL(void)
+emit_2bytes (j_compress_ptr cinfo, int value)
+/* Emit a 2-byte integer; these are always MSB first in JPEG files */
+{
+  emit_byte(cinfo, (value >> 8) & 0xFF);
+  emit_byte(cinfo, value & 0xFF);
+}
+
+
+/*
+ * Routines to write specific marker types.
+ */
+
+LOCAL(int)
+emit_dqt (j_compress_ptr cinfo, int index)
+/* Emit a DQT marker */
+/* Returns the precision used (0 = 8bits, 1 = 16bits) for baseline checking */
+{
+  JQUANT_TBL * qtbl = cinfo->quant_tbl_ptrs[index];
+  int prec;
+  int i;
+
+  if (qtbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, index);
+
+  prec = 0;
+  for (i = 0; i < DCTSIZE2; i++) {
+    if (qtbl->quantval[i] > 255)
+      prec = 1;
+  }
+
+  if (! qtbl->sent_table) {
+    emit_marker(cinfo, M_DQT);
+
+    emit_2bytes(cinfo, prec ? DCTSIZE2*2 + 1 + 2 : DCTSIZE2 + 1 + 2);
+
+    emit_byte(cinfo, index + (prec<<4));
+
+    for (i = 0; i < DCTSIZE2; i++) {
+      /* The table entries must be emitted in zigzag order. */
+      unsigned int qval = qtbl->quantval[jpeg_natural_order[i]];
+      if (prec)
+	emit_byte(cinfo, (int) (qval >> 8));
+      emit_byte(cinfo, (int) (qval & 0xFF));
+    }
+
+    qtbl->sent_table = TRUE;
+  }
+
+  return prec;
+}
+
+
+LOCAL(void)
+emit_dht (j_compress_ptr cinfo, int index, boolean is_ac)
+/* Emit a DHT marker */
+{
+  JHUFF_TBL * htbl;
+  int length, i;
+  
+  if (is_ac) {
+    htbl = cinfo->ac_huff_tbl_ptrs[index];
+    index += 0x10;		/* output index has AC bit set */
+  } else {
+    htbl = cinfo->dc_huff_tbl_ptrs[index];
+  }
+
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, index);
+  
+  if (! htbl->sent_table) {
+    emit_marker(cinfo, M_DHT);
+    
+    length = 0;
+    for (i = 1; i <= 16; i++)
+      length += htbl->bits[i];
+    
+    emit_2bytes(cinfo, length + 2 + 1 + 16);
+    emit_byte(cinfo, index);
+    
+    for (i = 1; i <= 16; i++)
+      emit_byte(cinfo, htbl->bits[i]);
+    
+    for (i = 0; i < length; i++)
+      emit_byte(cinfo, htbl->huffval[i]);
+    
+    htbl->sent_table = TRUE;
+  }
+}
+
+
+LOCAL(void)
+emit_dac (j_compress_ptr cinfo)
+/* Emit a DAC marker */
+/* Since the useful info is so small, we want to emit all the tables in */
+/* one DAC marker.  Therefore this routine does its own scan of the table. */
+{
+#ifdef C_ARITH_CODING_SUPPORTED
+  char dc_in_use[NUM_ARITH_TBLS];
+  char ac_in_use[NUM_ARITH_TBLS];
+  int length, i;
+  jpeg_component_info *compptr;
+  
+  for (i = 0; i < NUM_ARITH_TBLS; i++)
+    dc_in_use[i] = ac_in_use[i] = 0;
+  
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    compptr = cinfo->cur_comp_info[i];
+    dc_in_use[compptr->dc_tbl_no] = 1;
+    ac_in_use[compptr->ac_tbl_no] = 1;
+  }
+  
+  length = 0;
+  for (i = 0; i < NUM_ARITH_TBLS; i++)
+    length += dc_in_use[i] + ac_in_use[i];
+  
+  emit_marker(cinfo, M_DAC);
+  
+  emit_2bytes(cinfo, length*2 + 2);
+  
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    if (dc_in_use[i]) {
+      emit_byte(cinfo, i);
+      emit_byte(cinfo, cinfo->arith_dc_L[i] + (cinfo->arith_dc_U[i]<<4));
+    }
+    if (ac_in_use[i]) {
+      emit_byte(cinfo, i + 0x10);
+      emit_byte(cinfo, cinfo->arith_ac_K[i]);
+    }
+  }
+#endif /* C_ARITH_CODING_SUPPORTED */
+}
+
+
+LOCAL(void)
+emit_dri (j_compress_ptr cinfo)
+/* Emit a DRI marker */
+{
+  emit_marker(cinfo, M_DRI);
+  
+  emit_2bytes(cinfo, 4);	/* fixed length */
+
+  emit_2bytes(cinfo, (int) cinfo->restart_interval);
+}
+
+
+LOCAL(void)
+emit_sof (j_compress_ptr cinfo, JPEG_MARKER code)
+/* Emit a SOF marker */
+{
+  int ci;
+  jpeg_component_info *compptr;
+  
+  emit_marker(cinfo, code);
+  
+  emit_2bytes(cinfo, 3 * cinfo->num_components + 2 + 5 + 1); /* length */
+
+  /* Make sure image isn't bigger than SOF field can handle */
+  if ((long) cinfo->image_height > 65535L ||
+      (long) cinfo->image_width > 65535L)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) 65535);
+
+  emit_byte(cinfo, cinfo->data_precision);
+  emit_2bytes(cinfo, (int) cinfo->image_height);
+  emit_2bytes(cinfo, (int) cinfo->image_width);
+
+  emit_byte(cinfo, cinfo->num_components);
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    emit_byte(cinfo, compptr->component_id);
+    emit_byte(cinfo, (compptr->h_samp_factor << 4) + compptr->v_samp_factor);
+    emit_byte(cinfo, compptr->quant_tbl_no);
+  }
+}
+
+
+LOCAL(void)
+emit_sos (j_compress_ptr cinfo)
+/* Emit a SOS marker */
+{
+  int i, td, ta;
+  jpeg_component_info *compptr;
+  
+  emit_marker(cinfo, M_SOS);
+  
+  emit_2bytes(cinfo, 2 * cinfo->comps_in_scan + 2 + 1 + 3); /* length */
+  
+  emit_byte(cinfo, cinfo->comps_in_scan);
+  
+  for (i = 0; i < cinfo->comps_in_scan; i++) {
+    compptr = cinfo->cur_comp_info[i];
+    emit_byte(cinfo, compptr->component_id);
+    td = compptr->dc_tbl_no;
+    ta = compptr->ac_tbl_no;
+    if (cinfo->progressive_mode) {
+      /* Progressive mode: only DC or only AC tables are used in one scan;
+       * furthermore, Huffman coding of DC refinement uses no table at all.
+       * We emit 0 for unused field(s); this is recommended by the P&M text
+       * but does not seem to be specified in the standard.
+       */
+      if (cinfo->Ss == 0) {
+	ta = 0;			/* DC scan */
+	if (cinfo->Ah != 0 && !cinfo->arith_code)
+	  td = 0;		/* no DC table either */
+      } else {
+	td = 0;			/* AC scan */
+      }
+    }
+    emit_byte(cinfo, (td << 4) + ta);
+  }
+
+  emit_byte(cinfo, cinfo->Ss);
+  emit_byte(cinfo, cinfo->Se);
+  emit_byte(cinfo, (cinfo->Ah << 4) + cinfo->Al);
+}
+
+
+LOCAL(void)
+emit_jfif_app0 (j_compress_ptr cinfo)
+/* Emit a JFIF-compliant APP0 marker */
+{
+  /*
+   * Length of APP0 block	(2 bytes)
+   * Block ID			(4 bytes - ASCII "JFIF")
+   * Zero byte			(1 byte to terminate the ID string)
+   * Version Major, Minor	(2 bytes - major first)
+   * Units			(1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
+   * Xdpu			(2 bytes - dots per unit horizontal)
+   * Ydpu			(2 bytes - dots per unit vertical)
+   * Thumbnail X size		(1 byte)
+   * Thumbnail Y size		(1 byte)
+   */
+  
+  emit_marker(cinfo, M_APP0);
+  
+  emit_2bytes(cinfo, 2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1); /* length */
+
+  emit_byte(cinfo, 0x4A);	/* Identifier: ASCII "JFIF" */
+  emit_byte(cinfo, 0x46);
+  emit_byte(cinfo, 0x49);
+  emit_byte(cinfo, 0x46);
+  emit_byte(cinfo, 0);
+  emit_byte(cinfo, cinfo->JFIF_major_version); /* Version fields */
+  emit_byte(cinfo, cinfo->JFIF_minor_version);
+  emit_byte(cinfo, cinfo->density_unit); /* Pixel size information */
+  emit_2bytes(cinfo, (int) cinfo->X_density);
+  emit_2bytes(cinfo, (int) cinfo->Y_density);
+  emit_byte(cinfo, 0);		/* No thumbnail image */
+  emit_byte(cinfo, 0);
+}
+
+
+LOCAL(void)
+emit_adobe_app14 (j_compress_ptr cinfo)
+/* Emit an Adobe APP14 marker */
+{
+  /*
+   * Length of APP14 block	(2 bytes)
+   * Block ID			(5 bytes - ASCII "Adobe")
+   * Version Number		(2 bytes - currently 100)
+   * Flags0			(2 bytes - currently 0)
+   * Flags1			(2 bytes - currently 0)
+   * Color transform		(1 byte)
+   *
+   * Although Adobe TN 5116 mentions Version = 101, all the Adobe files
+   * now in circulation seem to use Version = 100, so that's what we write.
+   *
+   * We write the color transform byte as 1 if the JPEG color space is
+   * YCbCr, 2 if it's YCCK, 0 otherwise.  Adobe's definition has to do with
+   * whether the encoder performed a transformation, which is pretty useless.
+   */
+  
+  emit_marker(cinfo, M_APP14);
+  
+  emit_2bytes(cinfo, 2 + 5 + 2 + 2 + 2 + 1); /* length */
+
+  emit_byte(cinfo, 0x41);	/* Identifier: ASCII "Adobe" */
+  emit_byte(cinfo, 0x64);
+  emit_byte(cinfo, 0x6F);
+  emit_byte(cinfo, 0x62);
+  emit_byte(cinfo, 0x65);
+  emit_2bytes(cinfo, 100);	/* Version */
+  emit_2bytes(cinfo, 0);	/* Flags0 */
+  emit_2bytes(cinfo, 0);	/* Flags1 */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_YCbCr:
+    emit_byte(cinfo, 1);	/* Color transform = 1 */
+    break;
+  case JCS_YCCK:
+    emit_byte(cinfo, 2);	/* Color transform = 2 */
+    break;
+  default:
+    emit_byte(cinfo, 0);	/* Color transform = 0 */
+    break;
+  }
+}
+
+
+/*
+ * These routines allow writing an arbitrary marker with parameters.
+ * The only intended use is to emit COM or APPn markers after calling
+ * write_file_header and before calling write_frame_header.
+ * Other uses are not guaranteed to produce desirable results.
+ * Counting the parameter bytes properly is the caller's responsibility.
+ */
+
+METHODDEF(void)
+write_marker_header (j_compress_ptr cinfo, int marker, unsigned int datalen)
+/* Emit an arbitrary marker header */
+{
+  if (datalen > (unsigned int) 65533)		/* safety check */
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  emit_marker(cinfo, (JPEG_MARKER) marker);
+
+  emit_2bytes(cinfo, (int) (datalen + 2));	/* total length */
+}
+
+METHODDEF(void)
+write_marker_byte (j_compress_ptr cinfo, int val)
+/* Emit one byte of marker parameters following write_marker_header */
+{
+  emit_byte(cinfo, val);
+}
+
+
+/*
+ * Write datastream header.
+ * This consists of an SOI and optional APPn markers.
+ * We recommend use of the JFIF marker, but not the Adobe marker,
+ * when using YCbCr or grayscale data.  The JFIF marker should NOT
+ * be used for any other JPEG colorspace.  The Adobe marker is helpful
+ * to distinguish RGB, CMYK, and YCCK colorspaces.
+ * Note that an application can write additional header markers after
+ * jpeg_start_compress returns.
+ */
+
+METHODDEF(void)
+write_file_header (j_compress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+
+  emit_marker(cinfo, M_SOI);	/* first the SOI */
+
+  /* SOI is defined to reset restart interval to 0 */
+  marker->last_restart_interval = 0;
+
+  if (cinfo->write_JFIF_header)	/* next an optional JFIF APP0 */
+    emit_jfif_app0(cinfo);
+  if (cinfo->write_Adobe_marker) /* next an optional Adobe APP14 */
+    emit_adobe_app14(cinfo);
+}
+
+
+/*
+ * Write frame header.
+ * This consists of DQT and SOFn markers.
+ * Note that we do not emit the SOF until we have emitted the DQT(s).
+ * This avoids compatibility problems with incorrect implementations that
+ * try to error-check the quant table numbers as soon as they see the SOF.
+ */
+
+METHODDEF(void)
+write_frame_header (j_compress_ptr cinfo)
+{
+  int ci, prec;
+  boolean is_baseline;
+  jpeg_component_info *compptr;
+  
+  /* Emit DQT for each quantization table.
+   * Note that emit_dqt() suppresses any duplicate tables.
+   */
+  prec = 0;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    prec += emit_dqt(cinfo, compptr->quant_tbl_no);
+  }
+  /* now prec is nonzero iff there are any 16-bit quant tables. */
+
+  /* Check for a non-baseline specification.
+   * Note we assume that Huffman table numbers won't be changed later.
+   */
+  if (cinfo->arith_code || cinfo->progressive_mode ||
+      cinfo->data_precision != 8) {
+    is_baseline = FALSE;
+  } else {
+    is_baseline = TRUE;
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      if (compptr->dc_tbl_no > 1 || compptr->ac_tbl_no > 1)
+	is_baseline = FALSE;
+    }
+    if (prec && is_baseline) {
+      is_baseline = FALSE;
+      /* If it's baseline except for quantizer size, warn the user */
+      TRACEMS(cinfo, 0, JTRC_16BIT_TABLES);
+    }
+  }
+
+  /* Emit the proper SOF marker */
+  if (cinfo->arith_code) {
+    emit_sof(cinfo, M_SOF9);	/* SOF code for arithmetic coding */
+  } else {
+    if (cinfo->progressive_mode)
+      emit_sof(cinfo, M_SOF2);	/* SOF code for progressive Huffman */
+    else if (is_baseline)
+      emit_sof(cinfo, M_SOF0);	/* SOF code for baseline implementation */
+    else
+      emit_sof(cinfo, M_SOF1);	/* SOF code for non-baseline Huffman file */
+  }
+}
+
+
+/*
+ * Write scan header.
+ * This consists of DHT or DAC markers, optional DRI, and SOS.
+ * Compressed data will be written following the SOS.
+ */
+
+METHODDEF(void)
+write_scan_header (j_compress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+  int i;
+  jpeg_component_info *compptr;
+
+  if (cinfo->arith_code) {
+    /* Emit arith conditioning info.  We may have some duplication
+     * if the file has multiple scans, but it's so small it's hardly
+     * worth worrying about.
+     */
+    emit_dac(cinfo);
+  } else {
+    /* Emit Huffman tables.
+     * Note that emit_dht() suppresses any duplicate tables.
+     */
+    for (i = 0; i < cinfo->comps_in_scan; i++) {
+      compptr = cinfo->cur_comp_info[i];
+      if (cinfo->progressive_mode) {
+	/* Progressive mode: only DC or only AC tables are used in one scan */
+	if (cinfo->Ss == 0) {
+	  if (cinfo->Ah == 0)	/* DC needs no table for refinement scan */
+	    emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+	} else {
+	  emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
+	}
+      } else {
+	/* Sequential mode: need both DC and AC tables */
+	emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+	emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
+      }
+    }
+  }
+
+  /* Emit DRI if required --- note that DRI value could change for each scan.
+   * We avoid wasting space with unnecessary DRIs, however.
+   */
+  if (cinfo->restart_interval != marker->last_restart_interval) {
+    emit_dri(cinfo);
+    marker->last_restart_interval = cinfo->restart_interval;
+  }
+
+  emit_sos(cinfo);
+}
+
+
+/*
+ * Write datastream trailer.
+ */
+
+METHODDEF(void)
+write_file_trailer (j_compress_ptr cinfo)
+{
+  emit_marker(cinfo, M_EOI);
+}
+
+
+/*
+ * Write an abbreviated table-specification datastream.
+ * This consists of SOI, DQT and DHT tables, and EOI.
+ * Any table that is defined and not marked sent_table = TRUE will be
+ * emitted.  Note that all tables will be marked sent_table = TRUE at exit.
+ */
+
+METHODDEF(void)
+write_tables_only (j_compress_ptr cinfo)
+{
+  int i;
+
+  emit_marker(cinfo, M_SOI);
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++) {
+    if (cinfo->quant_tbl_ptrs[i] != NULL)
+      (void) emit_dqt(cinfo, i);
+  }
+
+  if (! cinfo->arith_code) {
+    for (i = 0; i < NUM_HUFF_TBLS; i++) {
+      if (cinfo->dc_huff_tbl_ptrs[i] != NULL)
+	emit_dht(cinfo, i, FALSE);
+      if (cinfo->ac_huff_tbl_ptrs[i] != NULL)
+	emit_dht(cinfo, i, TRUE);
+    }
+  }
+
+  emit_marker(cinfo, M_EOI);
+}
+
+
+/*
+ * Initialize the marker writer module.
+ */
+
+GLOBAL(void)
+jinit_marker_writer (j_compress_ptr cinfo)
+{
+  my_marker_ptr marker;
+
+  /* Create the subobject */
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_marker_writer));
+  cinfo->marker = (struct jpeg_marker_writer *) marker;
+  /* Initialize method pointers */
+  marker->pub.write_file_header = write_file_header;
+  marker->pub.write_frame_header = write_frame_header;
+  marker->pub.write_scan_header = write_scan_header;
+  marker->pub.write_file_trailer = write_file_trailer;
+  marker->pub.write_tables_only = write_tables_only;
+  marker->pub.write_marker_header = write_marker_header;
+  marker->pub.write_marker_byte = write_marker_byte;
+  /* Initialize private state */
+  marker->last_restart_interval = 0;
+}
diff --git a/jpeg/jcmaster.c b/jpeg/jcmaster.c
new file mode 100644
index 000000000..aab4020b8
--- /dev/null
+++ b/jpeg/jcmaster.c
@@ -0,0 +1,590 @@
+/*
+ * jcmaster.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains master control logic for the JPEG compressor.
+ * These routines are concerned with parameter validation, initial setup,
+ * and inter-pass control (determining the number of passes and the work 
+ * to be done in each pass).
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private state */
+
+typedef enum {
+	main_pass,		/* input data, also do first output step */
+	huff_opt_pass,		/* Huffman code optimization pass */
+	output_pass		/* data output pass */
+} c_pass_type;
+
+typedef struct {
+  struct jpeg_comp_master pub;	/* public fields */
+
+  c_pass_type pass_type;	/* the type of the current pass */
+
+  int pass_number;		/* # of passes completed */
+  int total_passes;		/* total # of passes needed */
+
+  int scan_number;		/* current index in scan_info[] */
+} my_comp_master;
+
+typedef my_comp_master * my_master_ptr;
+
+
+/*
+ * Support routines that do various essential calculations.
+ */
+
+LOCAL(void)
+initial_setup (j_compress_ptr cinfo)
+/* Do computations that are needed before master selection phase */
+{
+  int ci;
+  jpeg_component_info *compptr;
+  long samplesperrow;
+  JDIMENSION jd_samplesperrow;
+
+  /* Sanity check on image dimensions */
+  if (cinfo->image_height <= 0 || cinfo->image_width <= 0
+      || cinfo->num_components <= 0 || cinfo->input_components <= 0)
+    ERREXIT(cinfo, JERR_EMPTY_IMAGE);
+
+  /* Make sure image isn't bigger than I can handle */
+  if ((long) cinfo->image_height > (long) JPEG_MAX_DIMENSION ||
+      (long) cinfo->image_width > (long) JPEG_MAX_DIMENSION)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) JPEG_MAX_DIMENSION);
+
+  /* Width of an input scanline must be representable as JDIMENSION. */
+  samplesperrow = (long) cinfo->image_width * (long) cinfo->input_components;
+  jd_samplesperrow = (JDIMENSION) samplesperrow;
+  if ((long) jd_samplesperrow != samplesperrow)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* For now, precision must match compiled-in value... */
+  if (cinfo->data_precision != BITS_IN_JSAMPLE)
+    ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision);
+
+  /* Check that number of components won't exceed internal array sizes */
+  if (cinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	     MAX_COMPONENTS);
+
+  /* Compute maximum sampling factors; check factor validity */
+  cinfo->max_h_samp_factor = 1;
+  cinfo->max_v_samp_factor = 1;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
+	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+      ERREXIT(cinfo, JERR_BAD_SAMPLING);
+    cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
+				   compptr->h_samp_factor);
+    cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
+				   compptr->v_samp_factor);
+  }
+
+  /* Compute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Fill in the correct component_index value; don't rely on application */
+    compptr->component_index = ci;
+    /* For compression, we never do DCT scaling. */
+    compptr->DCT_scaled_size = DCTSIZE;
+    /* Size in DCT blocks */
+    compptr->width_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->height_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+    /* Size in samples */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) cinfo->max_h_samp_factor);
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) cinfo->max_v_samp_factor);
+    /* Mark component needed (this flag isn't actually used for compression) */
+    compptr->component_needed = TRUE;
+  }
+
+  /* Compute number of fully interleaved MCU rows (number of times that
+   * main controller will call coefficient controller).
+   */
+  cinfo->total_iMCU_rows = (JDIMENSION)
+    jdiv_round_up((long) cinfo->image_height,
+		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(void)
+validate_script (j_compress_ptr cinfo)
+/* Verify that the scan script in cinfo->scan_info[] is valid; also
+ * determine whether it uses progressive JPEG, and set cinfo->progressive_mode.
+ */
+{
+  const jpeg_scan_info * scanptr;
+  int scanno, ncomps, ci, coefi, thisi;
+  int Ss, Se, Ah, Al;
+  boolean component_sent[MAX_COMPONENTS];
+#ifdef C_PROGRESSIVE_SUPPORTED
+  int * last_bitpos_ptr;
+  int last_bitpos[MAX_COMPONENTS][DCTSIZE2];
+  /* -1 until that coefficient has been seen; then last Al for it */
+#endif
+
+  if (cinfo->num_scans <= 0)
+    ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, 0);
+
+  /* For sequential JPEG, all scans must have Ss=0, Se=DCTSIZE2-1;
+   * for progressive JPEG, no scan can have this.
+   */
+  scanptr = cinfo->scan_info;
+  if (scanptr->Ss != 0 || scanptr->Se != DCTSIZE2-1) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+    cinfo->progressive_mode = TRUE;
+    last_bitpos_ptr = & last_bitpos[0][0];
+    for (ci = 0; ci < cinfo->num_components; ci++) 
+      for (coefi = 0; coefi < DCTSIZE2; coefi++)
+	*last_bitpos_ptr++ = -1;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    cinfo->progressive_mode = FALSE;
+    for (ci = 0; ci < cinfo->num_components; ci++) 
+      component_sent[ci] = FALSE;
+  }
+
+  for (scanno = 1; scanno <= cinfo->num_scans; scanptr++, scanno++) {
+    /* Validate component indexes */
+    ncomps = scanptr->comps_in_scan;
+    if (ncomps <= 0 || ncomps > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, ncomps, MAX_COMPS_IN_SCAN);
+    for (ci = 0; ci < ncomps; ci++) {
+      thisi = scanptr->component_index[ci];
+      if (thisi < 0 || thisi >= cinfo->num_components)
+	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+      /* Components must appear in SOF order within each scan */
+      if (ci > 0 && thisi <= scanptr->component_index[ci-1])
+	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+    }
+    /* Validate progression parameters */
+    Ss = scanptr->Ss;
+    Se = scanptr->Se;
+    Ah = scanptr->Ah;
+    Al = scanptr->Al;
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      /* The JPEG spec simply gives the ranges 0..13 for Ah and Al, but that
+       * seems wrong: the upper bound ought to depend on data precision.
+       * Perhaps they really meant 0..N+1 for N-bit precision.
+       * Here we allow 0..10 for 8-bit data; Al larger than 10 results in
+       * out-of-range reconstructed DC values during the first DC scan,
+       * which might cause problems for some decoders.
+       */
+#if BITS_IN_JSAMPLE == 8
+#define MAX_AH_AL 10
+#else
+#define MAX_AH_AL 13
+#endif
+      if (Ss < 0 || Ss >= DCTSIZE2 || Se < Ss || Se >= DCTSIZE2 ||
+	  Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
+	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      if (Ss == 0) {
+	if (Se != 0)		/* DC and AC together not OK */
+	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      } else {
+	if (ncomps != 1)	/* AC scans must be for only one component */
+	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      }
+      for (ci = 0; ci < ncomps; ci++) {
+	last_bitpos_ptr = & last_bitpos[scanptr->component_index[ci]][0];
+	if (Ss != 0 && last_bitpos_ptr[0] < 0) /* AC without prior DC scan */
+	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+	for (coefi = Ss; coefi <= Se; coefi++) {
+	  if (last_bitpos_ptr[coefi] < 0) {
+	    /* first scan of this coefficient */
+	    if (Ah != 0)
+	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+	  } else {
+	    /* not first scan */
+	    if (Ah != last_bitpos_ptr[coefi] || Al != Ah-1)
+	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+	  }
+	  last_bitpos_ptr[coefi] = Al;
+	}
+      }
+#endif
+    } else {
+      /* For sequential JPEG, all progression parameters must be these: */
+      if (Ss != 0 || Se != DCTSIZE2-1 || Ah != 0 || Al != 0)
+	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+      /* Make sure components are not sent twice */
+      for (ci = 0; ci < ncomps; ci++) {
+	thisi = scanptr->component_index[ci];
+	if (component_sent[thisi])
+	  ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+	component_sent[thisi] = TRUE;
+      }
+    }
+  }
+
+  /* Now verify that everything got sent. */
+  if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+    /* For progressive mode, we only check that at least some DC data
+     * got sent for each component; the spec does not require that all bits
+     * of all coefficients be transmitted.  Would it be wiser to enforce
+     * transmission of all coefficient bits??
+     */
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      if (last_bitpos[ci][0] < 0)
+	ERREXIT(cinfo, JERR_MISSING_DATA);
+    }
+#endif
+  } else {
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      if (! component_sent[ci])
+	ERREXIT(cinfo, JERR_MISSING_DATA);
+    }
+  }
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+LOCAL(void)
+select_scan_parameters (j_compress_ptr cinfo)
+/* Set up the scan parameters for the current scan */
+{
+  int ci;
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  if (cinfo->scan_info != NULL) {
+    /* Prepare for current scan --- the script is already validated */
+    my_master_ptr master = (my_master_ptr) cinfo->master;
+    const jpeg_scan_info * scanptr = cinfo->scan_info + master->scan_number;
+
+    cinfo->comps_in_scan = scanptr->comps_in_scan;
+    for (ci = 0; ci < scanptr->comps_in_scan; ci++) {
+      cinfo->cur_comp_info[ci] =
+	&cinfo->comp_info[scanptr->component_index[ci]];
+    }
+    cinfo->Ss = scanptr->Ss;
+    cinfo->Se = scanptr->Se;
+    cinfo->Ah = scanptr->Ah;
+    cinfo->Al = scanptr->Al;
+  }
+  else
+#endif
+  {
+    /* Prepare for single sequential-JPEG scan containing all components */
+    if (cinfo->num_components > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	       MAX_COMPS_IN_SCAN);
+    cinfo->comps_in_scan = cinfo->num_components;
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      cinfo->cur_comp_info[ci] = &cinfo->comp_info[ci];
+    }
+    cinfo->Ss = 0;
+    cinfo->Se = DCTSIZE2-1;
+    cinfo->Ah = 0;
+    cinfo->Al = 0;
+  }
+}
+
+
+LOCAL(void)
+per_scan_setup (j_compress_ptr cinfo)
+/* Do computations that are needed before processing a JPEG scan */
+/* cinfo->comps_in_scan and cinfo->cur_comp_info[] are already set */
+{
+  int ci, mcublks, tmp;
+  jpeg_component_info *compptr;
+  
+  if (cinfo->comps_in_scan == 1) {
+    
+    /* Noninterleaved (single-component) scan */
+    compptr = cinfo->cur_comp_info[0];
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = compptr->width_in_blocks;
+    cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
+    
+    /* For noninterleaved scan, always one block per MCU */
+    compptr->MCU_width = 1;
+    compptr->MCU_height = 1;
+    compptr->MCU_blocks = 1;
+    compptr->MCU_sample_width = DCTSIZE;
+    compptr->last_col_width = 1;
+    /* For noninterleaved scans, it is convenient to define last_row_height
+     * as the number of block rows present in the last iMCU row.
+     */
+    tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+    if (tmp == 0) tmp = compptr->v_samp_factor;
+    compptr->last_row_height = tmp;
+    
+    /* Prepare array describing MCU composition */
+    cinfo->blocks_in_MCU = 1;
+    cinfo->MCU_membership[0] = 0;
+    
+  } else {
+    
+    /* Interleaved (multi-component) scan */
+    if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
+	       MAX_COMPS_IN_SCAN);
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width,
+		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+    cinfo->MCU_rows_in_scan = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height,
+		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+    
+    cinfo->blocks_in_MCU = 0;
+    
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      compptr = cinfo->cur_comp_info[ci];
+      /* Sampling factors give # of blocks of component in each MCU */
+      compptr->MCU_width = compptr->h_samp_factor;
+      compptr->MCU_height = compptr->v_samp_factor;
+      compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height;
+      compptr->MCU_sample_width = compptr->MCU_width * DCTSIZE;
+      /* Figure number of non-dummy blocks in last MCU column & row */
+      tmp = (int) (compptr->width_in_blocks % compptr->MCU_width);
+      if (tmp == 0) tmp = compptr->MCU_width;
+      compptr->last_col_width = tmp;
+      tmp = (int) (compptr->height_in_blocks % compptr->MCU_height);
+      if (tmp == 0) tmp = compptr->MCU_height;
+      compptr->last_row_height = tmp;
+      /* Prepare array describing MCU composition */
+      mcublks = compptr->MCU_blocks;
+      if (cinfo->blocks_in_MCU + mcublks > C_MAX_BLOCKS_IN_MCU)
+	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+      while (mcublks-- > 0) {
+	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+      }
+    }
+    
+  }
+
+  /* Convert restart specified in rows to actual MCU count. */
+  /* Note that count must fit in 16 bits, so we provide limiting. */
+  if (cinfo->restart_in_rows > 0) {
+    long nominal = (long) cinfo->restart_in_rows * (long) cinfo->MCUs_per_row;
+    cinfo->restart_interval = (unsigned int) MIN(nominal, 65535L);
+  }
+}
+
+
+/*
+ * Per-pass setup.
+ * This is called at the beginning of each pass.  We determine which modules
+ * will be active during this pass and give them appropriate start_pass calls.
+ * We also set is_last_pass to indicate whether any more passes will be
+ * required.
+ */
+
+METHODDEF(void)
+prepare_for_pass (j_compress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  switch (master->pass_type) {
+  case main_pass:
+    /* Initial pass: will collect input data, and do either Huffman
+     * optimization or data output for the first scan.
+     */
+    select_scan_parameters(cinfo);
+    per_scan_setup(cinfo);
+    if (! cinfo->raw_data_in) {
+      (*cinfo->cconvert->start_pass) (cinfo);
+      (*cinfo->downsample->start_pass) (cinfo);
+      (*cinfo->prep->start_pass) (cinfo, JBUF_PASS_THRU);
+    }
+    (*cinfo->fdct->start_pass) (cinfo);
+    (*cinfo->entropy->start_pass) (cinfo, cinfo->optimize_coding);
+    (*cinfo->coef->start_pass) (cinfo,
+				(master->total_passes > 1 ?
+				 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+    (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
+    if (cinfo->optimize_coding) {
+      /* No immediate data output; postpone writing frame/scan headers */
+      master->pub.call_pass_startup = FALSE;
+    } else {
+      /* Will write frame/scan headers at first jpeg_write_scanlines call */
+      master->pub.call_pass_startup = TRUE;
+    }
+    break;
+#ifdef ENTROPY_OPT_SUPPORTED
+  case huff_opt_pass:
+    /* Do Huffman optimization for a scan after the first one. */
+    select_scan_parameters(cinfo);
+    per_scan_setup(cinfo);
+    if (cinfo->Ss != 0 || cinfo->Ah == 0 || cinfo->arith_code) {
+      (*cinfo->entropy->start_pass) (cinfo, TRUE);
+      (*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
+      master->pub.call_pass_startup = FALSE;
+      break;
+    }
+    /* Special case: Huffman DC refinement scans need no Huffman table
+     * and therefore we can skip the optimization pass for them.
+     */
+    master->pass_type = output_pass;
+    master->pass_number++;
+    /*FALLTHROUGH*/
+#endif
+  case output_pass:
+    /* Do a data-output pass. */
+    /* We need not repeat per-scan setup if prior optimization pass did it. */
+    if (! cinfo->optimize_coding) {
+      select_scan_parameters(cinfo);
+      per_scan_setup(cinfo);
+    }
+    (*cinfo->entropy->start_pass) (cinfo, FALSE);
+    (*cinfo->coef->start_pass) (cinfo, JBUF_CRANK_DEST);
+    /* We emit frame/scan headers now */
+    if (master->scan_number == 0)
+      (*cinfo->marker->write_frame_header) (cinfo);
+    (*cinfo->marker->write_scan_header) (cinfo);
+    master->pub.call_pass_startup = FALSE;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+  }
+
+  master->pub.is_last_pass = (master->pass_number == master->total_passes-1);
+
+  /* Set up progress monitor's pass info if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->completed_passes = master->pass_number;
+    cinfo->progress->total_passes = master->total_passes;
+  }
+}
+
+
+/*
+ * Special start-of-pass hook.
+ * This is called by jpeg_write_scanlines if call_pass_startup is TRUE.
+ * In single-pass processing, we need this hook because we don't want to
+ * write frame/scan headers during jpeg_start_compress; we want to let the
+ * application write COM markers etc. between jpeg_start_compress and the
+ * jpeg_write_scanlines loop.
+ * In multi-pass processing, this routine is not used.
+ */
+
+METHODDEF(void)
+pass_startup (j_compress_ptr cinfo)
+{
+  cinfo->master->call_pass_startup = FALSE; /* reset flag so call only once */
+
+  (*cinfo->marker->write_frame_header) (cinfo);
+  (*cinfo->marker->write_scan_header) (cinfo);
+}
+
+
+/*
+ * Finish up at end of pass.
+ */
+
+METHODDEF(void)
+finish_pass_master (j_compress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  /* The entropy coder always needs an end-of-pass call,
+   * either to analyze statistics or to flush its output buffer.
+   */
+  (*cinfo->entropy->finish_pass) (cinfo);
+
+  /* Update state for next pass */
+  switch (master->pass_type) {
+  case main_pass:
+    /* next pass is either output of scan 0 (after optimization)
+     * or output of scan 1 (if no optimization).
+     */
+    master->pass_type = output_pass;
+    if (! cinfo->optimize_coding)
+      master->scan_number++;
+    break;
+  case huff_opt_pass:
+    /* next pass is always output of current scan */
+    master->pass_type = output_pass;
+    break;
+  case output_pass:
+    /* next pass is either optimization or output of next scan */
+    if (cinfo->optimize_coding)
+      master->pass_type = huff_opt_pass;
+    master->scan_number++;
+    break;
+  }
+
+  master->pass_number++;
+}
+
+
+/*
+ * Initialize master compression control.
+ */
+
+GLOBAL(void)
+jinit_c_master_control (j_compress_ptr cinfo, boolean transcode_only)
+{
+  my_master_ptr master;
+
+  master = (my_master_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(my_comp_master));
+  cinfo->master = (struct jpeg_comp_master *) master;
+  master->pub.prepare_for_pass = prepare_for_pass;
+  master->pub.pass_startup = pass_startup;
+  master->pub.finish_pass = finish_pass_master;
+  master->pub.is_last_pass = FALSE;
+
+  /* Validate parameters, determine derived values */
+  initial_setup(cinfo);
+
+  if (cinfo->scan_info != NULL) {
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    validate_script(cinfo);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    cinfo->progressive_mode = FALSE;
+    cinfo->num_scans = 1;
+  }
+
+  if (cinfo->progressive_mode)	/*  TEMPORARY HACK ??? */
+    cinfo->optimize_coding = TRUE; /* assume default tables no good for progressive mode */
+
+  /* Initialize my private state */
+  if (transcode_only) {
+    /* no main pass in transcoding */
+    if (cinfo->optimize_coding)
+      master->pass_type = huff_opt_pass;
+    else
+      master->pass_type = output_pass;
+  } else {
+    /* for normal compression, first pass is always this type: */
+    master->pass_type = main_pass;
+  }
+  master->scan_number = 0;
+  master->pass_number = 0;
+  if (cinfo->optimize_coding)
+    master->total_passes = cinfo->num_scans * 2;
+  else
+    master->total_passes = cinfo->num_scans;
+}
diff --git a/jpeg/jcomapi.c b/jpeg/jcomapi.c
new file mode 100644
index 000000000..9b1fa7568
--- /dev/null
+++ b/jpeg/jcomapi.c
@@ -0,0 +1,106 @@
+/*
+ * jcomapi.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface routines that are used for both
+ * compression and decompression.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Abort processing of a JPEG compression or decompression operation,
+ * but don't destroy the object itself.
+ *
+ * For this, we merely clean up all the nonpermanent memory pools.
+ * Note that temp files (virtual arrays) are not allowed to belong to
+ * the permanent pool, so we will be able to close all temp files here.
+ * Closing a data source or destination, if necessary, is the application's
+ * responsibility.
+ */
+
+GLOBAL(void)
+jpeg_abort (j_common_ptr cinfo)
+{
+  int pool;
+
+  /* Do nothing if called on a not-initialized or destroyed JPEG object. */
+  if (cinfo->mem == NULL)
+    return;
+
+  /* Releasing pools in reverse order might help avoid fragmentation
+   * with some (brain-damaged) malloc libraries.
+   */
+  for (pool = JPOOL_NUMPOOLS-1; pool > JPOOL_PERMANENT; pool--) {
+    (*cinfo->mem->free_pool) (cinfo, pool);
+  }
+
+  /* Reset overall state for possible reuse of object */
+  if (cinfo->is_decompressor) {
+    cinfo->global_state = DSTATE_START;
+    /* Try to keep application from accessing now-deleted marker list.
+     * A bit kludgy to do it here, but this is the most central place.
+     */
+    ((j_decompress_ptr) cinfo)->marker_list = NULL;
+  } else {
+    cinfo->global_state = CSTATE_START;
+  }
+}
+
+
+/*
+ * Destruction of a JPEG object.
+ *
+ * Everything gets deallocated except the master jpeg_compress_struct itself
+ * and the error manager struct.  Both of these are supplied by the application
+ * and must be freed, if necessary, by the application.  (Often they are on
+ * the stack and so don't need to be freed anyway.)
+ * Closing a data source or destination, if necessary, is the application's
+ * responsibility.
+ */
+
+GLOBAL(void)
+jpeg_destroy (j_common_ptr cinfo)
+{
+  /* We need only tell the memory manager to release everything. */
+  /* NB: mem pointer is NULL if memory mgr failed to initialize. */
+  if (cinfo->mem != NULL)
+    (*cinfo->mem->self_destruct) (cinfo);
+  cinfo->mem = NULL;		/* be safe if jpeg_destroy is called twice */
+  cinfo->global_state = 0;	/* mark it destroyed */
+}
+
+
+/*
+ * Convenience routines for allocating quantization and Huffman tables.
+ * (Would jutils.c be a more reasonable place to put these?)
+ */
+
+GLOBAL(JQUANT_TBL *)
+jpeg_alloc_quant_table (j_common_ptr cinfo)
+{
+  JQUANT_TBL *tbl;
+
+  tbl = (JQUANT_TBL *)
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JQUANT_TBL));
+  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+  return tbl;
+}
+
+
+GLOBAL(JHUFF_TBL *)
+jpeg_alloc_huff_table (j_common_ptr cinfo)
+{
+  JHUFF_TBL *tbl;
+
+  tbl = (JHUFF_TBL *)
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JHUFF_TBL));
+  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+  return tbl;
+}
diff --git a/jpeg/jconfig.doc b/jpeg/jconfig.doc
new file mode 100644
index 000000000..c18d1c064
--- /dev/null
+++ b/jpeg/jconfig.doc
@@ -0,0 +1,155 @@
+/*
+ * jconfig.doc
+ *
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file documents the configuration options that are required to
+ * customize the JPEG software for a particular system.
+ *
+ * The actual configuration options for a particular installation are stored
+ * in jconfig.h.  On many machines, jconfig.h can be generated automatically
+ * or copied from one of the "canned" jconfig files that we supply.  But if
+ * you need to generate a jconfig.h file by hand, this file tells you how.
+ *
+ * DO NOT EDIT THIS FILE --- IT WON'T ACCOMPLISH ANYTHING.
+ * EDIT A COPY NAMED JCONFIG.H.
+ */
+
+
+/*
+ * These symbols indicate the properties of your machine or compiler.
+ * #define the symbol if yes, #undef it if no.
+ */
+
+/* Does your compiler support function prototypes?
+ * (If not, you also need to use ansi2knr, see install.doc)
+ */
+#define HAVE_PROTOTYPES
+
+/* Does your compiler support the declaration "unsigned char" ?
+ * How about "unsigned short" ?
+ */
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+
+/* Define "void" as "char" if your compiler doesn't know about type void.
+ * NOTE: be sure to define void such that "void *" represents the most general
+ * pointer type, e.g., that returned by malloc().
+ */
+/* #define void char */
+
+/* Define "const" as empty if your compiler doesn't know the "const" keyword.
+ */
+/* #define const */
+
+/* Define this if an ordinary "char" type is unsigned.
+ * If you're not sure, leaving it undefined will work at some cost in speed.
+ * If you defined HAVE_UNSIGNED_CHAR then the speed difference is minimal.
+ */
+#undef CHAR_IS_UNSIGNED
+
+/* Define this if your system has an ANSI-conforming <stddef.h> file.
+ */
+#define HAVE_STDDEF_H
+
+/* Define this if your system has an ANSI-conforming <stdlib.h> file.
+ */
+#define HAVE_STDLIB_H
+
+/* Define this if your system does not have an ANSI/SysV <string.h>,
+ * but does have a BSD-style <strings.h>.
+ */
+#undef NEED_BSD_STRINGS
+
+/* Define this if your system does not provide typedef size_t in any of the
+ * ANSI-standard places (stddef.h, stdlib.h, or stdio.h), but places it in
+ * <sys/types.h> instead.
+ */
+#undef NEED_SYS_TYPES_H
+
+/* For 80x86 machines, you need to define NEED_FAR_POINTERS,
+ * unless you are using a large-data memory model or 80386 flat-memory mode.
+ * On less brain-damaged CPUs this symbol must not be defined.
+ * (Defining this symbol causes large data structures to be referenced through
+ * "far" pointers and to be allocated with a special version of malloc.)
+ */
+#undef NEED_FAR_POINTERS
+
+/* Define this if your linker needs global names to be unique in less
+ * than the first 15 characters.
+ */
+#undef NEED_SHORT_EXTERNAL_NAMES
+
+/* Although a real ANSI C compiler can deal perfectly well with pointers to
+ * unspecified structures (see "incomplete types" in the spec), a few pre-ANSI
+ * and pseudo-ANSI compilers get confused.  To keep one of these bozos happy,
+ * define INCOMPLETE_TYPES_BROKEN.  This is not recommended unless you
+ * actually get "missing structure definition" warnings or errors while
+ * compiling the JPEG code.
+ */
+#undef INCOMPLETE_TYPES_BROKEN
+
+
+/*
+ * The following options affect code selection within the JPEG library,
+ * but they don't need to be visible to applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+
+/* Define this if your compiler implements ">>" on signed values as a logical
+ * (unsigned) shift; leave it undefined if ">>" is a signed (arithmetic) shift,
+ * which is the normal and rational definition.
+ */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+
+#endif /* JPEG_INTERNALS */
+
+
+/*
+ * The remaining options do not affect the JPEG library proper,
+ * but only the sample applications cjpeg/djpeg (see cjpeg.c, djpeg.c).
+ * Other applications can ignore these.
+ */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+/* These defines indicate which image (non-JPEG) file formats are allowed. */
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+/* Define this if you want to name both input and output files on the command
+ * line, rather than using stdout and optionally stdin.  You MUST do this if
+ * your system can't cope with binary I/O to stdin/stdout.  See comments at
+ * head of cjpeg.c or djpeg.c.
+ */
+#undef TWO_FILE_COMMANDLINE
+
+/* Define this if your system needs explicit cleanup of temporary files.
+ * This is crucial under MS-DOS, where the temporary "files" may be areas
+ * of extended memory; on most other systems it's not as important.
+ */
+#undef NEED_SIGNAL_CATCHER
+
+/* By default, we open image files with fopen(...,"rb") or fopen(...,"wb").
+ * This is necessary on systems that distinguish text files from binary files,
+ * and is harmless on most systems that don't.  If you have one of the rare
+ * systems that complains about the "b" spec, define this symbol.
+ */
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg.
+ */
+#undef PROGRESS_REPORT
+
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/jpeg/jconfig.h b/jpeg/jconfig.h
new file mode 100644
index 000000000..b1fa66895
--- /dev/null
+++ b/jpeg/jconfig.h
@@ -0,0 +1,52 @@
+/* jconfig.cfg --- source file edited by configure script */
+/* see jconfig.doc for explanations */
+
+#include <config.h>
+
+#define HAVE_PROTOTYPES
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+#ifdef __CHAR_UNSIGNED__
+#  define CHAR_IS_UNSIGNED
+#endif /* __CHAR_UNSIGNED__ */
+/* Define this if you get warnings about undefined structures. */
+#undef INCOMPLETE_TYPES_BROKEN
+
+#if defined(WIN32) || defined(__EMX__)
+/* Define "boolean" as unsigned char, not int, per Windows custom */
+#  ifndef __RPCNDR_H__		/* don't conflict if rpcndr.h already read */
+typedef unsigned char boolean;
+#  endif
+#  define HAVE_BOOLEAN		/* prevent jmorecfg.h from redefining it */
+#endif /* WIN32 || __EMX__ */
+
+#ifdef JPEG_INTERNALS
+
+#undef RIGHT_SHIFT_IS_UNSIGNED
+#undef INLINE
+/* These are for configuring the JPEG memory manager. */
+#undef DEFAULT_MAX_MEM
+#undef NO_MKTEMP
+
+#endif /* JPEG_INTERNALS */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+#define BMP_SUPPORTED		/* BMP image file format */
+#define GIF_SUPPORTED		/* GIF image file format */
+#define PPM_SUPPORTED		/* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED		/* Utah RLE image file format */
+#define TARGA_SUPPORTED		/* Targa image file format */
+
+#undef TWO_FILE_COMMANDLINE
+#undef NEED_SIGNAL_CATCHER
+#undef DONT_USE_B_MODE
+
+#if defined(WIN32) || defined(__EMX__)
+#  define USE_SETMODE
+#endif /* WIN32 || __EMX__ */
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg. */
+#undef PROGRESS_REPORT
+
+#endif /* JPEG_CJPEG_DJPEG */
diff --git a/jpeg/jcparam.c b/jpeg/jcparam.c
new file mode 100644
index 000000000..6fc48f536
--- /dev/null
+++ b/jpeg/jcparam.c
@@ -0,0 +1,610 @@
+/*
+ * jcparam.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains optional default-setting code for the JPEG compressor.
+ * Applications do not have to use this file, but those that don't use it
+ * must know a lot more about the innards of the JPEG code.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Quantization table setup routines
+ */
+
+GLOBAL(void)
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+		      const unsigned int *basic_table,
+		      int scale_factor, boolean force_baseline)
+/* Define a quantization table equal to the basic_table times
+ * a scale factor (given as a percentage).
+ * If force_baseline is TRUE, the computed quantization table entries
+ * are limited to 1..255 for JPEG baseline compatibility.
+ */
+{
+  JQUANT_TBL ** qtblptr;
+  int i;
+  long temp;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (which_tbl < 0 || which_tbl >= NUM_QUANT_TBLS)
+    ERREXIT1(cinfo, JERR_DQT_INDEX, which_tbl);
+
+  qtblptr = & cinfo->quant_tbl_ptrs[which_tbl];
+
+  if (*qtblptr == NULL)
+    *qtblptr = jpeg_alloc_quant_table((j_common_ptr) cinfo);
+
+  for (i = 0; i < DCTSIZE2; i++) {
+    temp = ((long) basic_table[i] * scale_factor + 50L) / 100L;
+    /* limit the values to the valid range */
+    if (temp <= 0L) temp = 1L;
+    if (temp > 32767L) temp = 32767L; /* max quantizer needed for 12 bits */
+    if (force_baseline && temp > 255L)
+      temp = 255L;		/* limit to baseline range if requested */
+    (*qtblptr)->quantval[i] = (UINT16) temp;
+  }
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*qtblptr)->sent_table = FALSE;
+}
+
+
+GLOBAL(void)
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+			 boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables
+ * and a straight percentage-scaling quality scale.  In most cases it's better
+ * to use jpeg_set_quality (below); this entry point is provided for
+ * applications that insist on a linear percentage scaling.
+ */
+{
+  /* These are the sample quantization tables given in JPEG spec section K.1.
+   * The spec says that the values given produce "good" quality, and
+   * when divided by 2, "very good" quality.
+   */
+  static const unsigned int std_luminance_quant_tbl[DCTSIZE2] = {
+    16,  11,  10,  16,  24,  40,  51,  61,
+    12,  12,  14,  19,  26,  58,  60,  55,
+    14,  13,  16,  24,  40,  57,  69,  56,
+    14,  17,  22,  29,  51,  87,  80,  62,
+    18,  22,  37,  56,  68, 109, 103,  77,
+    24,  35,  55,  64,  81, 104, 113,  92,
+    49,  64,  78,  87, 103, 121, 120, 101,
+    72,  92,  95,  98, 112, 100, 103,  99
+  };
+  static const unsigned int std_chrominance_quant_tbl[DCTSIZE2] = {
+    17,  18,  24,  47,  99,  99,  99,  99,
+    18,  21,  26,  66,  99,  99,  99,  99,
+    24,  26,  56,  99,  99,  99,  99,  99,
+    47,  66,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99,
+    99,  99,  99,  99,  99,  99,  99,  99
+  };
+
+  /* Set up two quantization tables using the specified scaling */
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
+		       scale_factor, force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
+		       scale_factor, force_baseline);
+}
+
+
+GLOBAL(int)
+jpeg_quality_scaling (int quality)
+/* Convert a user-specified quality rating to a percentage scaling factor
+ * for an underlying quantization table, using our recommended scaling curve.
+ * The input 'quality' factor should be 0 (terrible) to 100 (very good).
+ */
+{
+  /* Safety limit on quality factor.  Convert 0 to 1 to avoid zero divide. */
+  if (quality <= 0) quality = 1;
+  if (quality > 100) quality = 100;
+
+  /* The basic table is used as-is (scaling 100) for a quality of 50.
+   * Qualities 50..100 are converted to scaling percentage 200 - 2*Q;
+   * note that at Q=100 the scaling is 0, which will cause jpeg_add_quant_table
+   * to make all the table entries 1 (hence, minimum quantization loss).
+   * Qualities 1..50 are converted to scaling percentage 5000/Q.
+   */
+  if (quality < 50)
+    quality = 5000 / quality;
+  else
+    quality = 200 - quality*2;
+
+  return quality;
+}
+
+
+GLOBAL(void)
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+/* Set or change the 'quality' (quantization) setting, using default tables.
+ * This is the standard quality-adjusting entry point for typical user
+ * interfaces; only those who want detailed control over quantization tables
+ * would use the preceding three routines directly.
+ */
+{
+  /* Convert user 0-100 rating to percentage scaling */
+  quality = jpeg_quality_scaling(quality);
+
+  /* Set up standard quality tables */
+  jpeg_set_linear_quality(cinfo, quality, force_baseline);
+}
+
+
+/*
+ * Huffman table setup routines
+ */
+
+LOCAL(void)
+add_huff_table (j_compress_ptr cinfo,
+		JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
+/* Define a Huffman table */
+{
+  int nsymbols, len;
+
+  if (*htblptr == NULL)
+    *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+
+  /* Copy the number-of-symbols-of-each-code-length counts */
+  MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
+
+  /* Validate the counts.  We do this here mainly so we can copy the right
+   * number of symbols from the val[] array, without risking marching off
+   * the end of memory.  jchuff.c will do a more thorough test later.
+   */
+  nsymbols = 0;
+  for (len = 1; len <= 16; len++)
+    nsymbols += bits[len];
+  if (nsymbols < 1 || nsymbols > 256)
+    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+  MEMCOPY((*htblptr)->huffval, val, nsymbols * SIZEOF(UINT8));
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*htblptr)->sent_table = FALSE;
+}
+
+
+LOCAL(void)
+std_huff_tables (j_compress_ptr cinfo)
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+{
+  static const UINT8 bits_dc_luminance[17] =
+    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_luminance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+  
+  static const UINT8 bits_dc_chrominance[17] =
+    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_chrominance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+  
+  static const UINT8 bits_ac_luminance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
+  static const UINT8 val_ac_luminance[] =
+    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+  
+  static const UINT8 bits_ac_chrominance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
+  static const UINT8 val_ac_chrominance[] =
+    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+  
+  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[0],
+		 bits_dc_luminance, val_dc_luminance);
+  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[0],
+		 bits_ac_luminance, val_ac_luminance);
+  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[1],
+		 bits_dc_chrominance, val_dc_chrominance);
+  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[1],
+		 bits_ac_chrominance, val_ac_chrominance);
+}
+
+
+/*
+ * Default parameter setup for compression.
+ *
+ * Applications that don't choose to use this routine must do their
+ * own setup of all these parameters.  Alternately, you can call this
+ * to establish defaults and then alter parameters selectively.  This
+ * is the recommended approach since, if we add any new parameters,
+ * your code will still work (they'll be set to reasonable defaults).
+ */
+
+GLOBAL(void)
+jpeg_set_defaults (j_compress_ptr cinfo)
+{
+  int i;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Allocate comp_info array large enough for maximum component count.
+   * Array is made permanent in case application wants to compress
+   * multiple images at same param settings.
+   */
+  if (cinfo->comp_info == NULL)
+    cinfo->comp_info = (jpeg_component_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  MAX_COMPONENTS * SIZEOF(jpeg_component_info));
+
+  /* Initialize everything not dependent on the color space */
+
+  cinfo->data_precision = BITS_IN_JSAMPLE;
+  /* Set up two quantization tables using default quality of 75 */
+  jpeg_set_quality(cinfo, 75, TRUE);
+  /* Set up two Huffman tables */
+  std_huff_tables(cinfo);
+
+  /* Initialize default arithmetic coding conditioning */
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    cinfo->arith_dc_L[i] = 0;
+    cinfo->arith_dc_U[i] = 1;
+    cinfo->arith_ac_K[i] = 5;
+  }
+
+  /* Default is no multiple-scan output */
+  cinfo->scan_info = NULL;
+  cinfo->num_scans = 0;
+
+  /* Expect normal source image, not raw downsampled data */
+  cinfo->raw_data_in = FALSE;
+
+  /* Use Huffman coding, not arithmetic coding, by default */
+  cinfo->arith_code = FALSE;
+
+  /* By default, don't do extra passes to optimize entropy coding */
+  cinfo->optimize_coding = FALSE;
+  /* The standard Huffman tables are only valid for 8-bit data precision.
+   * If the precision is higher, force optimization on so that usable
+   * tables will be computed.  This test can be removed if default tables
+   * are supplied that are valid for the desired precision.
+   */
+  if (cinfo->data_precision > 8)
+    cinfo->optimize_coding = TRUE;
+
+  /* By default, use the simpler non-cosited sampling alignment */
+  cinfo->CCIR601_sampling = FALSE;
+
+  /* No input smoothing */
+  cinfo->smoothing_factor = 0;
+
+  /* DCT algorithm preference */
+  cinfo->dct_method = JDCT_DEFAULT;
+
+  /* No restart markers */
+  cinfo->restart_interval = 0;
+  cinfo->restart_in_rows = 0;
+
+  /* Fill in default JFIF marker parameters.  Note that whether the marker
+   * will actually be written is determined by jpeg_set_colorspace.
+   *
+   * By default, the library emits JFIF version code 1.01.
+   * An application that wants to emit JFIF 1.02 extension markers should set
+   * JFIF_minor_version to 2.  We could probably get away with just defaulting
+   * to 1.02, but there may still be some decoders in use that will complain
+   * about that; saying 1.01 should minimize compatibility problems.
+   */
+  cinfo->JFIF_major_version = 1; /* Default JFIF version = 1.01 */
+  cinfo->JFIF_minor_version = 1;
+  cinfo->density_unit = 0;	/* Pixel size is unknown by default */
+  cinfo->X_density = 1;		/* Pixel aspect ratio is square by default */
+  cinfo->Y_density = 1;
+
+  /* Choose JPEG colorspace based on input space, set defaults accordingly */
+
+  jpeg_default_colorspace(cinfo);
+}
+
+
+/*
+ * Select an appropriate JPEG colorspace for in_color_space.
+ */
+
+GLOBAL(void)
+jpeg_default_colorspace (j_compress_ptr cinfo)
+{
+  switch (cinfo->in_color_space) {
+  case JCS_GRAYSCALE:
+    jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+    break;
+  case JCS_RGB:
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+    break;
+  case JCS_YCbCr:
+    jpeg_set_colorspace(cinfo, JCS_YCbCr);
+    break;
+  case JCS_CMYK:
+    jpeg_set_colorspace(cinfo, JCS_CMYK); /* By default, no translation */
+    break;
+  case JCS_YCCK:
+    jpeg_set_colorspace(cinfo, JCS_YCCK);
+    break;
+  case JCS_UNKNOWN:
+    jpeg_set_colorspace(cinfo, JCS_UNKNOWN);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
+  }
+}
+
+
+/*
+ * Set the JPEG colorspace, and choose colorspace-dependent default values.
+ */
+
+GLOBAL(void)
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+{
+  jpeg_component_info * compptr;
+  int ci;
+
+#define SET_COMP(index,id,hsamp,vsamp,quant,dctbl,actbl)  \
+  (compptr = &cinfo->comp_info[index], \
+   compptr->component_id = (id), \
+   compptr->h_samp_factor = (hsamp), \
+   compptr->v_samp_factor = (vsamp), \
+   compptr->quant_tbl_no = (quant), \
+   compptr->dc_tbl_no = (dctbl), \
+   compptr->ac_tbl_no = (actbl) )
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* For all colorspaces, we use Q and Huff tables 0 for luminance components,
+   * tables 1 for chrominance components.
+   */
+
+  cinfo->jpeg_color_space = colorspace;
+
+  cinfo->write_JFIF_header = FALSE; /* No marker for non-JFIF colorspaces */
+  cinfo->write_Adobe_marker = FALSE; /* write no Adobe marker by default */
+
+  switch (colorspace) {
+  case JCS_GRAYSCALE:
+    cinfo->write_JFIF_header = TRUE; /* Write a JFIF marker */
+    cinfo->num_components = 1;
+    /* JFIF specifies component ID 1 */
+    SET_COMP(0, 1, 1,1, 0, 0,0);
+    break;
+  case JCS_RGB:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag RGB */
+    cinfo->num_components = 3;
+    SET_COMP(0, 0x52 /* 'R' */, 1,1, 0, 0,0);
+    SET_COMP(1, 0x47 /* 'G' */, 1,1, 0, 0,0);
+    SET_COMP(2, 0x42 /* 'B' */, 1,1, 0, 0,0);
+    break;
+  case JCS_YCbCr:
+    cinfo->write_JFIF_header = TRUE; /* Write a JFIF marker */
+    cinfo->num_components = 3;
+    /* JFIF specifies component IDs 1,2,3 */
+    /* We default to 2x2 subsamples of chrominance */
+    SET_COMP(0, 1, 2,2, 0, 0,0);
+    SET_COMP(1, 2, 1,1, 1, 1,1);
+    SET_COMP(2, 3, 1,1, 1, 1,1);
+    break;
+  case JCS_CMYK:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag CMYK */
+    cinfo->num_components = 4;
+    SET_COMP(0, 0x43 /* 'C' */, 1,1, 0, 0,0);
+    SET_COMP(1, 0x4D /* 'M' */, 1,1, 0, 0,0);
+    SET_COMP(2, 0x59 /* 'Y' */, 1,1, 0, 0,0);
+    SET_COMP(3, 0x4B /* 'K' */, 1,1, 0, 0,0);
+    break;
+  case JCS_YCCK:
+    cinfo->write_Adobe_marker = TRUE; /* write Adobe marker to flag YCCK */
+    cinfo->num_components = 4;
+    SET_COMP(0, 1, 2,2, 0, 0,0);
+    SET_COMP(1, 2, 1,1, 1, 1,1);
+    SET_COMP(2, 3, 1,1, 1, 1,1);
+    SET_COMP(3, 4, 2,2, 0, 0,0);
+    break;
+  case JCS_UNKNOWN:
+    cinfo->num_components = cinfo->input_components;
+    if (cinfo->num_components < 1 || cinfo->num_components > MAX_COMPONENTS)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	       MAX_COMPONENTS);
+    for (ci = 0; ci < cinfo->num_components; ci++) {
+      SET_COMP(ci, ci, 1,1, 0, 0,0);
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+  }
+}
+
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+
+LOCAL(jpeg_scan_info *)
+fill_a_scan (jpeg_scan_info * scanptr, int ci,
+	     int Ss, int Se, int Ah, int Al)
+/* Support routine: generate one scan for specified component */
+{
+  scanptr->comps_in_scan = 1;
+  scanptr->component_index[0] = ci;
+  scanptr->Ss = Ss;
+  scanptr->Se = Se;
+  scanptr->Ah = Ah;
+  scanptr->Al = Al;
+  scanptr++;
+  return scanptr;
+}
+
+LOCAL(jpeg_scan_info *)
+fill_scans (jpeg_scan_info * scanptr, int ncomps,
+	    int Ss, int Se, int Ah, int Al)
+/* Support routine: generate one scan for each component */
+{
+  int ci;
+
+  for (ci = 0; ci < ncomps; ci++) {
+    scanptr->comps_in_scan = 1;
+    scanptr->component_index[0] = ci;
+    scanptr->Ss = Ss;
+    scanptr->Se = Se;
+    scanptr->Ah = Ah;
+    scanptr->Al = Al;
+    scanptr++;
+  }
+  return scanptr;
+}
+
+LOCAL(jpeg_scan_info *)
+fill_dc_scans (jpeg_scan_info * scanptr, int ncomps, int Ah, int Al)
+/* Support routine: generate interleaved DC scan if possible, else N scans */
+{
+  int ci;
+
+  if (ncomps <= MAX_COMPS_IN_SCAN) {
+    /* Single interleaved DC scan */
+    scanptr->comps_in_scan = ncomps;
+    for (ci = 0; ci < ncomps; ci++)
+      scanptr->component_index[ci] = ci;
+    scanptr->Ss = scanptr->Se = 0;
+    scanptr->Ah = Ah;
+    scanptr->Al = Al;
+    scanptr++;
+  } else {
+    /* Noninterleaved DC scan for each component */
+    scanptr = fill_scans(scanptr, ncomps, 0, 0, Ah, Al);
+  }
+  return scanptr;
+}
+
+
+/*
+ * Create a recommended progressive-JPEG script.
+ * cinfo->num_components and cinfo->jpeg_color_space must be correct.
+ */
+
+GLOBAL(void)
+jpeg_simple_progression (j_compress_ptr cinfo)
+{
+  int ncomps = cinfo->num_components;
+  int nscans;
+  jpeg_scan_info * scanptr;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Figure space needed for script.  Calculation must match code below! */
+  if (ncomps == 3 && cinfo->jpeg_color_space == JCS_YCbCr) {
+    /* Custom script for YCbCr color images. */
+    nscans = 10;
+  } else {
+    /* All-purpose script for other color spaces. */
+    if (ncomps > MAX_COMPS_IN_SCAN)
+      nscans = 6 * ncomps;	/* 2 DC + 4 AC scans per component */
+    else
+      nscans = 2 + 4 * ncomps;	/* 2 DC scans; 4 AC scans per component */
+  }
+
+  /* Allocate space for script.
+   * We need to put it in the permanent pool in case the application performs
+   * multiple compressions without changing the settings.  To avoid a memory
+   * leak if jpeg_simple_progression is called repeatedly for the same JPEG
+   * object, we try to re-use previously allocated space, and we allocate
+   * enough space to handle YCbCr even if initially asked for grayscale.
+   */
+  if (cinfo->script_space == NULL || cinfo->script_space_size < nscans) {
+    cinfo->script_space_size = MAX(nscans, 10);
+    cinfo->script_space = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+			cinfo->script_space_size * SIZEOF(jpeg_scan_info));
+  }
+  scanptr = cinfo->script_space;
+  cinfo->scan_info = scanptr;
+  cinfo->num_scans = nscans;
+
+  if (ncomps == 3 && cinfo->jpeg_color_space == JCS_YCbCr) {
+    /* Custom script for YCbCr color images. */
+    /* Initial DC scan */
+    scanptr = fill_dc_scans(scanptr, ncomps, 0, 1);
+    /* Initial AC scan: get some luma data out in a hurry */
+    scanptr = fill_a_scan(scanptr, 0, 1, 5, 0, 2);
+    /* Chroma data is too small to be worth expending many scans on */
+    scanptr = fill_a_scan(scanptr, 2, 1, 63, 0, 1);
+    scanptr = fill_a_scan(scanptr, 1, 1, 63, 0, 1);
+    /* Complete spectral selection for luma AC */
+    scanptr = fill_a_scan(scanptr, 0, 6, 63, 0, 2);
+    /* Refine next bit of luma AC */
+    scanptr = fill_a_scan(scanptr, 0, 1, 63, 2, 1);
+    /* Finish DC successive approximation */
+    scanptr = fill_dc_scans(scanptr, ncomps, 1, 0);
+    /* Finish AC successive approximation */
+    scanptr = fill_a_scan(scanptr, 2, 1, 63, 1, 0);
+    scanptr = fill_a_scan(scanptr, 1, 1, 63, 1, 0);
+    /* Luma bottom bit comes last since it's usually largest scan */
+    scanptr = fill_a_scan(scanptr, 0, 1, 63, 1, 0);
+  } else {
+    /* All-purpose script for other color spaces. */
+    /* Successive approximation first pass */
+    scanptr = fill_dc_scans(scanptr, ncomps, 0, 1);
+    scanptr = fill_scans(scanptr, ncomps, 1, 5, 0, 2);
+    scanptr = fill_scans(scanptr, ncomps, 6, 63, 0, 2);
+    /* Successive approximation second pass */
+    scanptr = fill_scans(scanptr, ncomps, 1, 63, 2, 1);
+    /* Successive approximation final pass */
+    scanptr = fill_dc_scans(scanptr, ncomps, 1, 0);
+    scanptr = fill_scans(scanptr, ncomps, 1, 63, 1, 0);
+  }
+}
+
+#endif /* C_PROGRESSIVE_SUPPORTED */
diff --git a/jpeg/jcphuff.c b/jpeg/jcphuff.c
new file mode 100644
index 000000000..07f9178b0
--- /dev/null
+++ b/jpeg/jcphuff.c
@@ -0,0 +1,833 @@
+/*
+ * jcphuff.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy encoding routines for progressive JPEG.
+ *
+ * We do not support output suspension in this module, since the library
+ * currently does not allow multiple-scan files to be written with output
+ * suspension.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jchuff.h"		/* Declarations shared with jchuff.c */
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+
+/* Expanded entropy encoder object for progressive Huffman encoding. */
+
+typedef struct {
+  struct jpeg_entropy_encoder pub; /* public fields */
+
+  /* Mode flag: TRUE for optimization, FALSE for actual data output */
+  boolean gather_statistics;
+
+  /* Bit-level coding status.
+   * next_output_byte/free_in_buffer are local copies of cinfo->dest fields.
+   */
+  JOCTET * next_output_byte;	/* => next byte to write in buffer */
+  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+  INT32 put_buffer;		/* current bit-accumulation buffer */
+  int put_bits;			/* # of bits now in it */
+  j_compress_ptr cinfo;		/* link to cinfo (needed for dump_buffer) */
+
+  /* Coding status for DC components */
+  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+
+  /* Coding status for AC components */
+  int ac_tbl_no;		/* the table number of the single component */
+  unsigned int EOBRUN;		/* run length of EOBs */
+  unsigned int BE;		/* # of buffered correction bits before MCU */
+  char * bit_buffer;		/* buffer for correction bits (1 per char) */
+  /* packing correction bits tightly would save some space but cost time... */
+
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  int next_restart_num;		/* next restart number to write (0-7) */
+
+  /* Pointers to derived tables (these workspaces have image lifespan).
+   * Since any one scan codes only DC or only AC, we only need one set
+   * of tables, not one for DC and one for AC.
+   */
+  c_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+
+  /* Statistics tables for optimization; again, one set is enough */
+  long * count_ptrs[NUM_HUFF_TBLS];
+} phuff_entropy_encoder;
+
+typedef phuff_entropy_encoder * phuff_entropy_ptr;
+
+/* MAX_CORR_BITS is the number of bits the AC refinement correction-bit
+ * buffer can hold.  Larger sizes may slightly improve compression, but
+ * 1000 is already well into the realm of overkill.
+ * The minimum safe size is 64 bits.
+ */
+
+#define MAX_CORR_BITS  1000	/* Max # of correction bits I can buffer */
+
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
+ * We assume that int right shift is unsigned if INT32 right shift is,
+ * which should be safe.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS	int ishift_temp;
+#define IRIGHT_SHIFT(x,shft)  \
+	((ishift_temp = (x)) < 0 ? \
+	 (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+	 (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#endif
+
+/* Forward declarations */
+METHODDEF(boolean) encode_mcu_DC_first JPP((j_compress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) encode_mcu_AC_first JPP((j_compress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) encode_mcu_DC_refine JPP((j_compress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+METHODDEF(boolean) encode_mcu_AC_refine JPP((j_compress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+METHODDEF(void) finish_pass_phuff JPP((j_compress_ptr cinfo));
+METHODDEF(void) finish_pass_gather_phuff JPP((j_compress_ptr cinfo));
+
+
+/*
+ * Initialize for a Huffman-compressed scan using progressive JPEG.
+ */
+
+METHODDEF(void)
+start_pass_phuff (j_compress_ptr cinfo, boolean gather_statistics)
+{  
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  boolean is_DC_band;
+  int ci, tbl;
+  jpeg_component_info * compptr;
+
+  entropy->cinfo = cinfo;
+  entropy->gather_statistics = gather_statistics;
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* We assume jcmaster.c already validated the scan parameters. */
+
+  /* Select execution routines */
+  if (cinfo->Ah == 0) {
+    if (is_DC_band)
+      entropy->pub.encode_mcu = encode_mcu_DC_first;
+    else
+      entropy->pub.encode_mcu = encode_mcu_AC_first;
+  } else {
+    if (is_DC_band)
+      entropy->pub.encode_mcu = encode_mcu_DC_refine;
+    else {
+      entropy->pub.encode_mcu = encode_mcu_AC_refine;
+      /* AC refinement needs a correction bit buffer */
+      if (entropy->bit_buffer == NULL)
+	entropy->bit_buffer = (char *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      MAX_CORR_BITS * SIZEOF(char));
+    }
+  }
+  if (gather_statistics)
+    entropy->pub.finish_pass = finish_pass_gather_phuff;
+  else
+    entropy->pub.finish_pass = finish_pass_phuff;
+
+  /* Only DC coefficients may be interleaved, so cinfo->comps_in_scan = 1
+   * for AC coefficients.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    /* Initialize DC predictions to 0 */
+    entropy->last_dc_val[ci] = 0;
+    /* Get table index */
+    if (is_DC_band) {
+      if (cinfo->Ah != 0)	/* DC refinement needs no table */
+	continue;
+      tbl = compptr->dc_tbl_no;
+    } else {
+      entropy->ac_tbl_no = tbl = compptr->ac_tbl_no;
+    }
+    if (gather_statistics) {
+      /* Check for invalid table index */
+      /* (make_c_derived_tbl does this in the other path) */
+      if (tbl < 0 || tbl >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tbl);
+      /* Allocate and zero the statistics tables */
+      /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
+      if (entropy->count_ptrs[tbl] == NULL)
+	entropy->count_ptrs[tbl] = (long *)
+	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				      257 * SIZEOF(long));
+      MEMZERO(entropy->count_ptrs[tbl], 257 * SIZEOF(long));
+    } else {
+      /* Compute derived values for Huffman table */
+      /* We may do this more than once for a table, but it's not expensive */
+      jpeg_make_c_derived_tbl(cinfo, is_DC_band, tbl,
+			      & entropy->derived_tbls[tbl]);
+    }
+  }
+
+  /* Initialize AC stuff */
+  entropy->EOBRUN = 0;
+  entropy->BE = 0;
+
+  /* Initialize bit buffer to empty */
+  entropy->put_buffer = 0;
+  entropy->put_bits = 0;
+
+  /* Initialize restart stuff */
+  entropy->restarts_to_go = cinfo->restart_interval;
+  entropy->next_restart_num = 0;
+}
+
+
+/* Outputting bytes to the file.
+ * NB: these must be called only when actually outputting,
+ * that is, entropy->gather_statistics == FALSE.
+ */
+
+/* Emit a byte */
+#define emit_byte(entropy,val)  \
+	{ *(entropy)->next_output_byte++ = (JOCTET) (val);  \
+	  if (--(entropy)->free_in_buffer == 0)  \
+	    dump_buffer(entropy); }
+
+
+LOCAL(void)
+dump_buffer (phuff_entropy_ptr entropy)
+/* Empty the output buffer; we do not support suspension in this module. */
+{
+  struct jpeg_destination_mgr * dest = entropy->cinfo->dest;
+
+  if (! (*dest->empty_output_buffer) (entropy->cinfo))
+    ERREXIT(entropy->cinfo, JERR_CANT_SUSPEND);
+  /* After a successful buffer dump, must reset buffer pointers */
+  entropy->next_output_byte = dest->next_output_byte;
+  entropy->free_in_buffer = dest->free_in_buffer;
+}
+
+
+/* Outputting bits to the file */
+
+/* Only the right 24 bits of put_buffer are used; the valid bits are
+ * left-justified in this part.  At most 16 bits can be passed to emit_bits
+ * in one call, and we never retain more than 7 bits in put_buffer
+ * between calls, so 24 bits are sufficient.
+ */
+
+INLINE
+LOCAL(void)
+emit_bits (phuff_entropy_ptr entropy, unsigned int code, int size)
+/* Emit some bits, unless we are in gather mode */
+{
+  /* This routine is heavily used, so it's worth coding tightly. */
+  register INT32 put_buffer = (INT32) code;
+  register int put_bits = entropy->put_bits;
+
+  /* if size is 0, caller used an invalid Huffman table entry */
+  if (size == 0)
+    ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
+
+  if (entropy->gather_statistics)
+    return;			/* do nothing if we're only getting stats */
+
+  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
+  
+  put_bits += size;		/* new number of bits in buffer */
+  
+  put_buffer <<= 24 - put_bits; /* align incoming bits */
+
+  put_buffer |= entropy->put_buffer; /* and merge with old buffer contents */
+
+  while (put_bits >= 8) {
+    int c = (int) ((put_buffer >> 16) & 0xFF);
+    
+    emit_byte(entropy, c);
+    if (c == 0xFF) {		/* need to stuff a zero byte? */
+      emit_byte(entropy, 0);
+    }
+    put_buffer <<= 8;
+    put_bits -= 8;
+  }
+
+  entropy->put_buffer = put_buffer; /* update variables */
+  entropy->put_bits = put_bits;
+}
+
+
+LOCAL(void)
+flush_bits (phuff_entropy_ptr entropy)
+{
+  emit_bits(entropy, 0x7F, 7); /* fill any partial byte with ones */
+  entropy->put_buffer = 0;     /* and reset bit-buffer to empty */
+  entropy->put_bits = 0;
+}
+
+
+/*
+ * Emit (or just count) a Huffman symbol.
+ */
+
+INLINE
+LOCAL(void)
+emit_symbol (phuff_entropy_ptr entropy, int tbl_no, int symbol)
+{
+  if (entropy->gather_statistics)
+    entropy->count_ptrs[tbl_no][symbol]++;
+  else {
+    c_derived_tbl * tbl = entropy->derived_tbls[tbl_no];
+    emit_bits(entropy, tbl->ehufco[symbol], tbl->ehufsi[symbol]);
+  }
+}
+
+
+/*
+ * Emit bits from a correction bit buffer.
+ */
+
+LOCAL(void)
+emit_buffered_bits (phuff_entropy_ptr entropy, char * bufstart,
+		    unsigned int nbits)
+{
+  if (entropy->gather_statistics)
+    return;			/* no real work */
+
+  while (nbits > 0) {
+    emit_bits(entropy, (unsigned int) (*bufstart), 1);
+    bufstart++;
+    nbits--;
+  }
+}
+
+
+/*
+ * Emit any pending EOBRUN symbol.
+ */
+
+LOCAL(void)
+emit_eobrun (phuff_entropy_ptr entropy)
+{
+  register int temp, nbits;
+
+  if (entropy->EOBRUN > 0) {	/* if there is any pending EOBRUN */
+    temp = entropy->EOBRUN;
+    nbits = 0;
+    while ((temp >>= 1))
+      nbits++;
+    /* safety check: shouldn't happen given limited correction-bit buffer */
+    if (nbits > 14)
+      ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
+
+    emit_symbol(entropy, entropy->ac_tbl_no, nbits << 4);
+    if (nbits)
+      emit_bits(entropy, entropy->EOBRUN, nbits);
+
+    entropy->EOBRUN = 0;
+
+    /* Emit any buffered correction bits */
+    emit_buffered_bits(entropy, entropy->bit_buffer, entropy->BE);
+    entropy->BE = 0;
+  }
+}
+
+
+/*
+ * Emit a restart marker & resynchronize predictions.
+ */
+
+LOCAL(void)
+emit_restart (phuff_entropy_ptr entropy, int restart_num)
+{
+  int ci;
+
+  emit_eobrun(entropy);
+
+  if (! entropy->gather_statistics) {
+    flush_bits(entropy);
+    emit_byte(entropy, 0xFF);
+    emit_byte(entropy, JPEG_RST0 + restart_num);
+  }
+
+  if (entropy->cinfo->Ss == 0) {
+    /* Re-initialize DC predictions to 0 */
+    for (ci = 0; ci < entropy->cinfo->comps_in_scan; ci++)
+      entropy->last_dc_val[ci] = 0;
+  } else {
+    /* Re-initialize all AC-related fields to 0 */
+    entropy->EOBRUN = 0;
+    entropy->BE = 0;
+  }
+}
+
+
+/*
+ * MCU encoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_first (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp, temp2;
+  register int nbits;
+  int blkn, ci;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+  jpeg_component_info * compptr;
+  ISHIFT_TEMPS
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data[blkn];
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+
+    /* Compute the DC value after the required point transform by Al.
+     * This is simply an arithmetic right shift.
+     */
+    temp2 = IRIGHT_SHIFT((int) ((*block)[0]), Al);
+
+    /* DC differences are figured on the point-transformed values. */
+    temp = temp2 - entropy->last_dc_val[ci];
+    entropy->last_dc_val[ci] = temp2;
+
+    /* Encode the DC coefficient difference per section G.1.2.1 */
+    temp2 = temp;
+    if (temp < 0) {
+      temp = -temp;		/* temp is abs value of input */
+      /* For a negative input, want temp2 = bitwise complement of abs(input) */
+      /* This code assumes we are on a two's complement machine */
+      temp2--;
+    }
+    
+    /* Find the number of bits needed for the magnitude of the coefficient */
+    nbits = 0;
+    while (temp) {
+      nbits++;
+      temp >>= 1;
+    }
+    /* Check for out-of-range coefficient values.
+     * Since we're encoding a difference, the range limit is twice as much.
+     */
+    if (nbits > MAX_COEF_BITS+1)
+      ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+    
+    /* Count/emit the Huffman-coded symbol for the number of bits */
+    emit_symbol(entropy, compptr->dc_tbl_no, nbits);
+    
+    /* Emit that number of bits of the value, if positive, */
+    /* or the complement of its magnitude, if negative. */
+    if (nbits)			/* emit_bits rejects calls with size 0 */
+      emit_bits(entropy, (unsigned int) temp2, nbits);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_first (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp, temp2;
+  register int nbits;
+  register int r, k;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data block */
+  block = MCU_data[0];
+
+  /* Encode the AC coefficients per section G.1.2.2, fig. G.3 */
+  
+  r = 0;			/* r = run length of zeros */
+   
+  for (k = cinfo->Ss; k <= Se; k++) {
+    if ((temp = (*block)[jpeg_natural_order[k]]) == 0) {
+      r++;
+      continue;
+    }
+    /* We must apply the point transform by Al.  For AC coefficients this
+     * is an integer division with rounding towards 0.  To do this portably
+     * in C, we shift after obtaining the absolute value; so the code is
+     * interwoven with finding the abs value (temp) and output bits (temp2).
+     */
+    if (temp < 0) {
+      temp = -temp;		/* temp is abs value of input */
+      temp >>= Al;		/* apply the point transform */
+      /* For a negative coef, want temp2 = bitwise complement of abs(coef) */
+      temp2 = ~temp;
+    } else {
+      temp >>= Al;		/* apply the point transform */
+      temp2 = temp;
+    }
+    /* Watch out for case that nonzero coef is zero after point transform */
+    if (temp == 0) {
+      r++;
+      continue;
+    }
+
+    /* Emit any pending EOBRUN */
+    if (entropy->EOBRUN > 0)
+      emit_eobrun(entropy);
+    /* if run length > 15, must emit special run-length-16 codes (0xF0) */
+    while (r > 15) {
+      emit_symbol(entropy, entropy->ac_tbl_no, 0xF0);
+      r -= 16;
+    }
+
+    /* Find the number of bits needed for the magnitude of the coefficient */
+    nbits = 1;			/* there must be at least one 1 bit */
+    while ((temp >>= 1))
+      nbits++;
+    /* Check for out-of-range coefficient values */
+    if (nbits > MAX_COEF_BITS)
+      ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
+    /* Count/emit Huffman symbol for run length / number of bits */
+    emit_symbol(entropy, entropy->ac_tbl_no, (r << 4) + nbits);
+
+    /* Emit that number of bits of the value, if positive, */
+    /* or the complement of its magnitude, if negative. */
+    emit_bits(entropy, (unsigned int) temp2, nbits);
+
+    r = 0;			/* reset zero run length */
+  }
+
+  if (r > 0) {			/* If there are trailing zeroes, */
+    entropy->EOBRUN++;		/* count an EOB */
+    if (entropy->EOBRUN == 0x7FFF)
+      emit_eobrun(entropy);	/* force it out to avoid overflow */
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for DC successive approximation refinement scan.
+ * Note: we assume such scans can be multi-component, although the spec
+ * is not very clear on the point.
+ */
+
+METHODDEF(boolean)
+encode_mcu_DC_refine (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp;
+  int blkn;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data blocks */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data[blkn];
+
+    /* We simply emit the Al'th bit of the DC coefficient value. */
+    temp = (*block)[0];
+    emit_bits(entropy, (unsigned int) (temp >> Al), 1);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * MCU encoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+encode_mcu_AC_refine (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  register int temp;
+  register int r, k;
+  int EOB;
+  char *BR_buffer;
+  unsigned int BR;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  JBLOCKROW block;
+  int absvalues[DCTSIZE2];
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Emit restart marker if needed */
+  if (cinfo->restart_interval)
+    if (entropy->restarts_to_go == 0)
+      emit_restart(entropy, entropy->next_restart_num);
+
+  /* Encode the MCU data block */
+  block = MCU_data[0];
+
+  /* It is convenient to make a pre-pass to determine the transformed
+   * coefficients' absolute values and the EOB position.
+   */
+  EOB = 0;
+  for (k = cinfo->Ss; k <= Se; k++) {
+    temp = (*block)[jpeg_natural_order[k]];
+    /* We must apply the point transform by Al.  For AC coefficients this
+     * is an integer division with rounding towards 0.  To do this portably
+     * in C, we shift after obtaining the absolute value.
+     */
+    if (temp < 0)
+      temp = -temp;		/* temp is abs value of input */
+    temp >>= Al;		/* apply the point transform */
+    absvalues[k] = temp;	/* save abs value for main pass */
+    if (temp == 1)
+      EOB = k;			/* EOB = index of last newly-nonzero coef */
+  }
+
+  /* Encode the AC coefficients per section G.1.2.3, fig. G.7 */
+  
+  r = 0;			/* r = run length of zeros */
+  BR = 0;			/* BR = count of buffered bits added now */
+  BR_buffer = entropy->bit_buffer + entropy->BE; /* Append bits to buffer */
+
+  for (k = cinfo->Ss; k <= Se; k++) {
+    if ((temp = absvalues[k]) == 0) {
+      r++;
+      continue;
+    }
+
+    /* Emit any required ZRLs, but not if they can be folded into EOB */
+    while (r > 15 && k <= EOB) {
+      /* emit any pending EOBRUN and the BE correction bits */
+      emit_eobrun(entropy);
+      /* Emit ZRL */
+      emit_symbol(entropy, entropy->ac_tbl_no, 0xF0);
+      r -= 16;
+      /* Emit buffered correction bits that must be associated with ZRL */
+      emit_buffered_bits(entropy, BR_buffer, BR);
+      BR_buffer = entropy->bit_buffer; /* BE bits are gone now */
+      BR = 0;
+    }
+
+    /* If the coef was previously nonzero, it only needs a correction bit.
+     * NOTE: a straight translation of the spec's figure G.7 would suggest
+     * that we also need to test r > 15.  But if r > 15, we can only get here
+     * if k > EOB, which implies that this coefficient is not 1.
+     */
+    if (temp > 1) {
+      /* The correction bit is the next bit of the absolute value. */
+      BR_buffer[BR++] = (char) (temp & 1);
+      continue;
+    }
+
+    /* Emit any pending EOBRUN and the BE correction bits */
+    emit_eobrun(entropy);
+
+    /* Count/emit Huffman symbol for run length / number of bits */
+    emit_symbol(entropy, entropy->ac_tbl_no, (r << 4) + 1);
+
+    /* Emit output bit for newly-nonzero coef */
+    temp = ((*block)[jpeg_natural_order[k]] < 0) ? 0 : 1;
+    emit_bits(entropy, (unsigned int) temp, 1);
+
+    /* Emit buffered correction bits that must be associated with this code */
+    emit_buffered_bits(entropy, BR_buffer, BR);
+    BR_buffer = entropy->bit_buffer; /* BE bits are gone now */
+    BR = 0;
+    r = 0;			/* reset zero run length */
+  }
+
+  if (r > 0 || BR > 0) {	/* If there are trailing zeroes, */
+    entropy->EOBRUN++;		/* count an EOB */
+    entropy->BE += BR;		/* concat my correction bits to older ones */
+    /* We force out the EOB if we risk either:
+     * 1. overflow of the EOB counter;
+     * 2. overflow of the correction bit buffer during the next MCU.
+     */
+    if (entropy->EOBRUN == 0x7FFF || entropy->BE > (MAX_CORR_BITS-DCTSIZE2+1))
+      emit_eobrun(entropy);
+  }
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+
+  /* Update restart-interval state too */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0) {
+      entropy->restarts_to_go = cinfo->restart_interval;
+      entropy->next_restart_num++;
+      entropy->next_restart_num &= 7;
+    }
+    entropy->restarts_to_go--;
+  }
+
+  return TRUE;
+}
+
+
+/*
+ * Finish up at the end of a Huffman-compressed progressive scan.
+ */
+
+METHODDEF(void)
+finish_pass_phuff (j_compress_ptr cinfo)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+
+  entropy->next_output_byte = cinfo->dest->next_output_byte;
+  entropy->free_in_buffer = cinfo->dest->free_in_buffer;
+
+  /* Flush out any buffered data */
+  emit_eobrun(entropy);
+  flush_bits(entropy);
+
+  cinfo->dest->next_output_byte = entropy->next_output_byte;
+  cinfo->dest->free_in_buffer = entropy->free_in_buffer;
+}
+
+
+/*
+ * Finish up a statistics-gathering pass and create the new Huffman tables.
+ */
+
+METHODDEF(void)
+finish_pass_gather_phuff (j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  boolean is_DC_band;
+  int ci, tbl;
+  jpeg_component_info * compptr;
+  JHUFF_TBL **htblptr;
+  boolean did[NUM_HUFF_TBLS];
+
+  /* Flush out buffered data (all we care about is counting the EOB symbol) */
+  emit_eobrun(entropy);
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* It's important not to apply jpeg_gen_optimal_table more than once
+   * per table, because it clobbers the input frequency counts!
+   */
+  MEMZERO(did, SIZEOF(did));
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    if (is_DC_band) {
+      if (cinfo->Ah != 0)	/* DC refinement needs no table */
+	continue;
+      tbl = compptr->dc_tbl_no;
+    } else {
+      tbl = compptr->ac_tbl_no;
+    }
+    if (! did[tbl]) {
+      if (is_DC_band)
+        htblptr = & cinfo->dc_huff_tbl_ptrs[tbl];
+      else
+        htblptr = & cinfo->ac_huff_tbl_ptrs[tbl];
+      if (*htblptr == NULL)
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+      jpeg_gen_optimal_table(cinfo, *htblptr, entropy->count_ptrs[tbl]);
+      did[tbl] = TRUE;
+    }
+  }
+}
+
+
+/*
+ * Module initialization routine for progressive Huffman entropy encoding.
+ */
+
+GLOBAL(void)
+jinit_phuff_encoder (j_compress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy;
+  int i;
+
+  entropy = (phuff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(phuff_entropy_encoder));
+  cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
+  entropy->pub.start_pass = start_pass_phuff;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->derived_tbls[i] = NULL;
+    entropy->count_ptrs[i] = NULL;
+  }
+  entropy->bit_buffer = NULL;	/* needed only in AC refinement scan */
+}
+
+#endif /* C_PROGRESSIVE_SUPPORTED */
diff --git a/jpeg/jcprepct.c b/jpeg/jcprepct.c
new file mode 100644
index 000000000..fa93333db
--- /dev/null
+++ b/jpeg/jcprepct.c
@@ -0,0 +1,354 @@
+/*
+ * jcprepct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the compression preprocessing controller.
+ * This controller manages the color conversion, downsampling,
+ * and edge expansion steps.
+ *
+ * Most of the complexity here is associated with buffering input rows
+ * as required by the downsampler.  See the comments at the head of
+ * jcsample.c for the downsampler's needs.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* At present, jcsample.c can request context rows only for smoothing.
+ * In the future, we might also need context rows for CCIR601 sampling
+ * or other more-complex downsampling procedures.  The code to support
+ * context rows should be compiled only if needed.
+ */
+#ifdef INPUT_SMOOTHING_SUPPORTED
+#define CONTEXT_ROWS_SUPPORTED
+#endif
+
+
+/*
+ * For the simple (no-context-row) case, we just need to buffer one
+ * row group's worth of pixels for the downsampling step.  At the bottom of
+ * the image, we pad to a full row group by replicating the last pixel row.
+ * The downsampler's last output row is then replicated if needed to pad
+ * out to a full iMCU row.
+ *
+ * When providing context rows, we must buffer three row groups' worth of
+ * pixels.  Three row groups are physically allocated, but the row pointer
+ * arrays are made five row groups high, with the extra pointers above and
+ * below "wrapping around" to point to the last and first real row groups.
+ * This allows the downsampler to access the proper context rows.
+ * At the top and bottom of the image, we create dummy context rows by
+ * copying the first or last real pixel row.  This copying could be avoided
+ * by pointer hacking as is done in jdmainct.c, but it doesn't seem worth the
+ * trouble on the compression side.
+ */
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_prep_controller pub; /* public fields */
+
+  /* Downsampling input buffer.  This buffer holds color-converted data
+   * until we have enough to do a downsample step.
+   */
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+
+  JDIMENSION rows_to_go;	/* counts rows remaining in source image */
+  int next_buf_row;		/* index of next row to store in color_buf */
+
+#ifdef CONTEXT_ROWS_SUPPORTED	/* only needed for context case */
+  int this_row_group;		/* starting row index of group to process */
+  int next_buf_stop;		/* downsample when we reach this index */
+#endif
+} my_prep_controller;
+
+typedef my_prep_controller * my_prep_ptr;
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_prep (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+
+  if (pass_mode != JBUF_PASS_THRU)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  /* Initialize total-height counter for detecting bottom of image */
+  prep->rows_to_go = cinfo->image_height;
+  /* Mark the conversion buffer empty */
+  prep->next_buf_row = 0;
+#ifdef CONTEXT_ROWS_SUPPORTED
+  /* Preset additional state variables for context mode.
+   * These aren't used in non-context mode, so we needn't test which mode.
+   */
+  prep->this_row_group = 0;
+  /* Set next_buf_stop to stop after two row groups have been read in. */
+  prep->next_buf_stop = 2 * cinfo->max_v_samp_factor;
+#endif
+}
+
+
+/*
+ * Expand an image vertically from height input_rows to height output_rows,
+ * by duplicating the bottom row.
+ */
+
+LOCAL(void)
+expand_bottom_edge (JSAMPARRAY image_data, JDIMENSION num_cols,
+		    int input_rows, int output_rows)
+{
+  register int row;
+
+  for (row = input_rows; row < output_rows; row++) {
+    jcopy_sample_rows(image_data, input_rows-1, image_data, row,
+		      1, num_cols);
+  }
+}
+
+
+/*
+ * Process some data in the simple no-context case.
+ *
+ * Preprocessor output data is counted in "row groups".  A row group
+ * is defined to be v_samp_factor sample rows of each component.
+ * Downsampling will produce this much data from each max_v_samp_factor
+ * input rows.
+ */
+
+METHODDEF(void)
+pre_process_data (j_compress_ptr cinfo,
+		  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+		  JDIMENSION in_rows_avail,
+		  JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+		  JDIMENSION out_row_groups_avail)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+  int numrows, ci;
+  JDIMENSION inrows;
+  jpeg_component_info * compptr;
+
+  while (*in_row_ctr < in_rows_avail &&
+	 *out_row_group_ctr < out_row_groups_avail) {
+    /* Do color conversion to fill the conversion buffer. */
+    inrows = in_rows_avail - *in_row_ctr;
+    numrows = cinfo->max_v_samp_factor - prep->next_buf_row;
+    numrows = (int) MIN((JDIMENSION) numrows, inrows);
+    (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
+				       prep->color_buf,
+				       (JDIMENSION) prep->next_buf_row,
+				       numrows);
+    *in_row_ctr += numrows;
+    prep->next_buf_row += numrows;
+    prep->rows_to_go -= numrows;
+    /* If at bottom of image, pad to fill the conversion buffer. */
+    if (prep->rows_to_go == 0 &&
+	prep->next_buf_row < cinfo->max_v_samp_factor) {
+      for (ci = 0; ci < cinfo->num_components; ci++) {
+	expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+			   prep->next_buf_row, cinfo->max_v_samp_factor);
+      }
+      prep->next_buf_row = cinfo->max_v_samp_factor;
+    }
+    /* If we've filled the conversion buffer, empty it. */
+    if (prep->next_buf_row == cinfo->max_v_samp_factor) {
+      (*cinfo->downsample->downsample) (cinfo,
+					prep->color_buf, (JDIMENSION) 0,
+					output_buf, *out_row_group_ctr);
+      prep->next_buf_row = 0;
+      (*out_row_group_ctr)++;
+    }
+    /* If at bottom of image, pad the output to a full iMCU height.
+     * Note we assume the caller is providing a one-iMCU-height output buffer!
+     */
+    if (prep->rows_to_go == 0 &&
+	*out_row_group_ctr < out_row_groups_avail) {
+      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	   ci++, compptr++) {
+	expand_bottom_edge(output_buf[ci],
+			   compptr->width_in_blocks * DCTSIZE,
+			   (int) (*out_row_group_ctr * compptr->v_samp_factor),
+			   (int) (out_row_groups_avail * compptr->v_samp_factor));
+      }
+      *out_row_group_ctr = out_row_groups_avail;
+      break;			/* can exit outer loop without test */
+    }
+  }
+}
+
+
+#ifdef CONTEXT_ROWS_SUPPORTED
+
+/*
+ * Process some data in the context case.
+ */
+
+METHODDEF(void)
+pre_process_context (j_compress_ptr cinfo,
+		     JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+		     JDIMENSION in_rows_avail,
+		     JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+		     JDIMENSION out_row_groups_avail)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+  int numrows, ci;
+  int buf_height = cinfo->max_v_samp_factor * 3;
+  JDIMENSION inrows;
+
+  while (*out_row_group_ctr < out_row_groups_avail) {
+    if (*in_row_ctr < in_rows_avail) {
+      /* Do color conversion to fill the conversion buffer. */
+      inrows = in_rows_avail - *in_row_ctr;
+      numrows = prep->next_buf_stop - prep->next_buf_row;
+      numrows = (int) MIN((JDIMENSION) numrows, inrows);
+      (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
+					 prep->color_buf,
+					 (JDIMENSION) prep->next_buf_row,
+					 numrows);
+      /* Pad at top of image, if first time through */
+      if (prep->rows_to_go == cinfo->image_height) {
+	for (ci = 0; ci < cinfo->num_components; ci++) {
+	  int row;
+	  for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
+	    jcopy_sample_rows(prep->color_buf[ci], 0,
+			      prep->color_buf[ci], -row,
+			      1, cinfo->image_width);
+	  }
+	}
+      }
+      *in_row_ctr += numrows;
+      prep->next_buf_row += numrows;
+      prep->rows_to_go -= numrows;
+    } else {
+      /* Return for more data, unless we are at the bottom of the image. */
+      if (prep->rows_to_go != 0)
+	break;
+      /* When at bottom of image, pad to fill the conversion buffer. */
+      if (prep->next_buf_row < prep->next_buf_stop) {
+	for (ci = 0; ci < cinfo->num_components; ci++) {
+	  expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+			     prep->next_buf_row, prep->next_buf_stop);
+	}
+	prep->next_buf_row = prep->next_buf_stop;
+      }
+    }
+    /* If we've gotten enough data, downsample a row group. */
+    if (prep->next_buf_row == prep->next_buf_stop) {
+      (*cinfo->downsample->downsample) (cinfo,
+					prep->color_buf,
+					(JDIMENSION) prep->this_row_group,
+					output_buf, *out_row_group_ctr);
+      (*out_row_group_ctr)++;
+      /* Advance pointers with wraparound as necessary. */
+      prep->this_row_group += cinfo->max_v_samp_factor;
+      if (prep->this_row_group >= buf_height)
+	prep->this_row_group = 0;
+      if (prep->next_buf_row >= buf_height)
+	prep->next_buf_row = 0;
+      prep->next_buf_stop = prep->next_buf_row + cinfo->max_v_samp_factor;
+    }
+  }
+}
+
+
+/*
+ * Create the wrapped-around downsampling input buffer needed for context mode.
+ */
+
+LOCAL(void)
+create_context_buffer (j_compress_ptr cinfo)
+{
+  my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
+  int rgroup_height = cinfo->max_v_samp_factor;
+  int ci, i;
+  jpeg_component_info * compptr;
+  JSAMPARRAY true_buffer, fake_buffer;
+
+  /* Grab enough space for fake row pointers for all the components;
+   * we need five row groups' worth of pointers for each component.
+   */
+  fake_buffer = (JSAMPARRAY)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(cinfo->num_components * 5 * rgroup_height) *
+				SIZEOF(JSAMPROW));
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Allocate the actual buffer space (3 row groups) for this component.
+     * We make the buffer wide enough to allow the downsampler to edge-expand
+     * horizontally within the buffer, if it so chooses.
+     */
+    true_buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
+		      cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+       (JDIMENSION) (3 * rgroup_height));
+    /* Copy true buffer row pointers into the middle of the fake row array */
+    MEMCOPY(fake_buffer + rgroup_height, true_buffer,
+	    3 * rgroup_height * SIZEOF(JSAMPROW));
+    /* Fill in the above and below wraparound pointers */
+    for (i = 0; i < rgroup_height; i++) {
+      fake_buffer[i] = true_buffer[2 * rgroup_height + i];
+      fake_buffer[4 * rgroup_height + i] = true_buffer[i];
+    }
+    prep->color_buf[ci] = fake_buffer + rgroup_height;
+    fake_buffer += 5 * rgroup_height; /* point to space for next component */
+  }
+}
+
+#endif /* CONTEXT_ROWS_SUPPORTED */
+
+
+/*
+ * Initialize preprocessing controller.
+ */
+
+GLOBAL(void)
+jinit_c_prep_controller (j_compress_ptr cinfo, boolean need_full_buffer)
+{
+  my_prep_ptr prep;
+  int ci;
+  jpeg_component_info * compptr;
+
+  if (need_full_buffer)		/* safety check */
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  prep = (my_prep_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_prep_controller));
+  cinfo->prep = (struct jpeg_c_prep_controller *) prep;
+  prep->pub.start_pass = start_pass_prep;
+
+  /* Allocate the color conversion buffer.
+   * We make the buffer wide enough to allow the downsampler to edge-expand
+   * horizontally within the buffer, if it so chooses.
+   */
+  if (cinfo->downsample->need_context_rows) {
+    /* Set up to provide context rows */
+#ifdef CONTEXT_ROWS_SUPPORTED
+    prep->pub.pre_process_data = pre_process_context;
+    create_context_buffer(cinfo);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    /* No context, just make it tall enough for one row group */
+    prep->pub.pre_process_data = pre_process_data;
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      prep->color_buf[ci] = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
+			cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+	 (JDIMENSION) cinfo->max_v_samp_factor);
+    }
+  }
+}
diff --git a/jpeg/jcsample.c b/jpeg/jcsample.c
new file mode 100644
index 000000000..212ec8757
--- /dev/null
+++ b/jpeg/jcsample.c
@@ -0,0 +1,519 @@
+/*
+ * jcsample.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains downsampling routines.
+ *
+ * Downsampling input data is counted in "row groups".  A row group
+ * is defined to be max_v_samp_factor pixel rows of each component,
+ * from which the downsampler produces v_samp_factor sample rows.
+ * A single row group is processed in each call to the downsampler module.
+ *
+ * The downsampler is responsible for edge-expansion of its output data
+ * to fill an integral number of DCT blocks horizontally.  The source buffer
+ * may be modified if it is helpful for this purpose (the source buffer is
+ * allocated wide enough to correspond to the desired output width).
+ * The caller (the prep controller) is responsible for vertical padding.
+ *
+ * The downsampler may request "context rows" by setting need_context_rows
+ * during startup.  In this case, the input arrays will contain at least
+ * one row group's worth of pixels above and below the passed-in data;
+ * the caller will create dummy rows at image top and bottom by replicating
+ * the first or last real pixel row.
+ *
+ * An excellent reference for image resampling is
+ *   Digital Image Warping, George Wolberg, 1990.
+ *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
+ *
+ * The downsampling algorithm used here is a simple average of the source
+ * pixels covered by the output pixel.  The hi-falutin sampling literature
+ * refers to this as a "box filter".  In general the characteristics of a box
+ * filter are not very good, but for the specific cases we normally use (1:1
+ * and 2:1 ratios) the box is equivalent to a "triangle filter" which is not
+ * nearly so bad.  If you intend to use other sampling ratios, you'd be well
+ * advised to improve this code.
+ *
+ * A simple input-smoothing capability is provided.  This is mainly intended
+ * for cleaning up color-dithered GIF input files (if you find it inadequate,
+ * we suggest using an external filtering program such as pnmconvol).  When
+ * enabled, each input pixel P is replaced by a weighted sum of itself and its
+ * eight neighbors.  P's weight is 1-8*SF and each neighbor's weight is SF,
+ * where SF = (smoothing_factor / 1024).
+ * Currently, smoothing is only supported for 2h2v sampling factors.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Pointer to routine to downsample a single component */
+typedef JMETHOD(void, downsample1_ptr,
+		(j_compress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY output_data));
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_downsampler pub;	/* public fields */
+
+  /* Downsampling method pointers, one per component */
+  downsample1_ptr methods[MAX_COMPONENTS];
+} my_downsampler;
+
+typedef my_downsampler * my_downsample_ptr;
+
+
+/*
+ * Initialize for a downsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_downsample (j_compress_ptr cinfo)
+{
+  /* no work for now */
+}
+
+
+/*
+ * Expand a component horizontally from width input_cols to width output_cols,
+ * by duplicating the rightmost samples.
+ */
+
+LOCAL(void)
+expand_right_edge (JSAMPARRAY image_data, int num_rows,
+		   JDIMENSION input_cols, JDIMENSION output_cols)
+{
+  register JSAMPROW ptr;
+  register JSAMPLE pixval;
+  register int count;
+  int row;
+  int numcols = (int) (output_cols - input_cols);
+
+  if (numcols > 0) {
+    for (row = 0; row < num_rows; row++) {
+      ptr = image_data[row] + input_cols;
+      pixval = ptr[-1];		/* don't need GETJSAMPLE() here */
+      for (count = numcols; count > 0; count--)
+	*ptr++ = pixval;
+    }
+  }
+}
+
+
+/*
+ * Do downsampling for a whole row group (all components).
+ *
+ * In this version we simply downsample each component independently.
+ */
+
+METHODDEF(void)
+sep_downsample (j_compress_ptr cinfo,
+		JSAMPIMAGE input_buf, JDIMENSION in_row_index,
+		JSAMPIMAGE output_buf, JDIMENSION out_row_group_index)
+{
+  my_downsample_ptr downsample = (my_downsample_ptr) cinfo->downsample;
+  int ci;
+  jpeg_component_info * compptr;
+  JSAMPARRAY in_ptr, out_ptr;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    in_ptr = input_buf[ci] + in_row_index;
+    out_ptr = output_buf[ci] + (out_row_group_index * compptr->v_samp_factor);
+    (*downsample->methods[ci]) (cinfo, compptr, in_ptr, out_ptr);
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * One row group is processed per call.
+ * This version handles arbitrary integral sampling ratios, without smoothing.
+ * Note that this version is not actually used for customary sampling ratios.
+ */
+
+METHODDEF(void)
+int_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, h_expand, v_expand, numpix, numpix2, h, v;
+  JDIMENSION outcol, outcol_h;	/* outcol_h == outcol*h_expand */
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  JSAMPROW inptr, outptr;
+  INT32 outvalue;
+
+  h_expand = cinfo->max_h_samp_factor / compptr->h_samp_factor;
+  v_expand = cinfo->max_v_samp_factor / compptr->v_samp_factor;
+  numpix = h_expand * v_expand;
+  numpix2 = numpix/2;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, output_cols * h_expand);
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    for (outcol = 0, outcol_h = 0; outcol < output_cols;
+	 outcol++, outcol_h += h_expand) {
+      outvalue = 0;
+      for (v = 0; v < v_expand; v++) {
+	inptr = input_data[inrow+v] + outcol_h;
+	for (h = 0; h < h_expand; h++) {
+	  outvalue += (INT32) GETJSAMPLE(*inptr++);
+	}
+      }
+      *outptr++ = (JSAMPLE) ((outvalue + numpix2) / numpix);
+    }
+    inrow += v_expand;
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the special case of a full-size component,
+ * without smoothing.
+ */
+
+METHODDEF(void)
+fullsize_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		     JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  /* Copy the data */
+  jcopy_sample_rows(input_data, 0, output_data, 0,
+		    cinfo->max_v_samp_factor, cinfo->image_width);
+  /* Edge-expand */
+  expand_right_edge(output_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, compptr->width_in_blocks * DCTSIZE);
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 1:1 vertical,
+ * without smoothing.
+ *
+ * A note about the "bias" calculations: when rounding fractional values to
+ * integer, we do not want to always round 0.5 up to the next integer.
+ * If we did that, we'd introduce a noticeable bias towards larger values.
+ * Instead, this code is arranged so that 0.5 will be rounded up or down at
+ * alternate pixel locations (a simple ordered dither pattern).
+ */
+
+METHODDEF(void)
+h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow;
+  JDIMENSION outcol;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr, outptr;
+  register int bias;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, output_cols * 2);
+
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr = input_data[outrow];
+    bias = 0;			/* bias = 0,1,0,1,... for successive samples */
+    for (outcol = 0; outcol < output_cols; outcol++) {
+      *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1])
+			      + bias) >> 1);
+      bias ^= 1;		/* 0=>1, 1=>0 */
+      inptr += 2;
+    }
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * without smoothing.
+ */
+
+METHODDEF(void)
+h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow;
+  JDIMENSION outcol;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr0, inptr1, outptr;
+  register int bias;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data, cinfo->max_v_samp_factor,
+		    cinfo->image_width, output_cols * 2);
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow+1];
+    bias = 1;			/* bias = 1,2,1,2,... for successive samples */
+    for (outcol = 0; outcol < output_cols; outcol++) {
+      *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+			      GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1])
+			      + bias) >> 2);
+      bias ^= 3;		/* 1=>2, 2=>1 */
+      inptr0 += 2; inptr1 += 2;
+    }
+    inrow += 2;
+  }
+}
+
+
+#ifdef INPUT_SMOOTHING_SUPPORTED
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+ * with smoothing.  One row of context is required.
+ */
+
+METHODDEF(void)
+h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+			JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow;
+  JDIMENSION colctr;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr0, inptr1, above_ptr, below_ptr, outptr;
+  INT32 membersum, neighsum, memberscale, neighscale;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
+		    cinfo->image_width, output_cols * 2);
+
+  /* We don't bother to form the individual "smoothed" input pixel values;
+   * we can directly compute the output which is the average of the four
+   * smoothed values.  Each of the four member pixels contributes a fraction
+   * (1-8*SF) to its own smoothed image and a fraction SF to each of the three
+   * other smoothed pixels, therefore a total fraction (1-5*SF)/4 to the final
+   * output.  The four corner-adjacent neighbor pixels contribute a fraction
+   * SF to just one smoothed pixel, or SF/4 to the final output; while the
+   * eight edge-adjacent neighbors contribute SF to each of two smoothed
+   * pixels, or SF/2 overall.  In order to use integer arithmetic, these
+   * factors are scaled by 2^16 = 65536.
+   * Also recall that SF = smoothing_factor / 1024.
+   */
+
+  memberscale = 16384 - cinfo->smoothing_factor * 80; /* scaled (1-5*SF)/4 */
+  neighscale = cinfo->smoothing_factor * 16; /* scaled SF/4 */
+
+  inrow = 0;
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow+1];
+    above_ptr = input_data[inrow-1];
+    below_ptr = input_data[inrow+2];
+
+    /* Special case for first column: pretend column -1 is same as column 0 */
+    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+	       GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
+	       GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
+    neighsum += neighsum;
+    neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
+		GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+    inptr0 += 2; inptr1 += 2; above_ptr += 2; below_ptr += 2;
+
+    for (colctr = output_cols - 2; colctr > 0; colctr--) {
+      /* sum of pixels directly mapped to this output element */
+      membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+		  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+      /* sum of edge-neighbor pixels */
+      neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+		 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+		 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
+		 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
+      /* The edge-neighbors count twice as much as corner-neighbors */
+      neighsum += neighsum;
+      /* Add in the corner-neighbors */
+      neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
+		  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
+      /* form final output scaled up by 2^16 */
+      membersum = membersum * memberscale + neighsum * neighscale;
+      /* round, descale and output it */
+      *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+      inptr0 += 2; inptr1 += 2; above_ptr += 2; below_ptr += 2;
+    }
+
+    /* Special case for last column */
+    membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
+		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+    neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
+	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+	       GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
+	       GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
+    neighsum += neighsum;
+    neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[1]) +
+		GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr = (JSAMPLE) ((membersum + 32768) >> 16);
+
+    inrow += 2;
+  }
+}
+
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the special case of a full-size component,
+ * with smoothing.  One row of context is required.
+ */
+
+METHODDEF(void)
+fullsize_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+			    JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow;
+  JDIMENSION colctr;
+  JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
+  register JSAMPROW inptr, above_ptr, below_ptr, outptr;
+  INT32 membersum, neighsum, memberscale, neighscale;
+  int colsum, lastcolsum, nextcolsum;
+
+  /* Expand input data enough to let all the output samples be generated
+   * by the standard loop.  Special-casing padded output would be more
+   * efficient.
+   */
+  expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
+		    cinfo->image_width, output_cols);
+
+  /* Each of the eight neighbor pixels contributes a fraction SF to the
+   * smoothed pixel, while the main pixel contributes (1-8*SF).  In order
+   * to use integer arithmetic, these factors are multiplied by 2^16 = 65536.
+   * Also recall that SF = smoothing_factor / 1024.
+   */
+
+  memberscale = 65536L - cinfo->smoothing_factor * 512L; /* scaled 1-8*SF */
+  neighscale = cinfo->smoothing_factor * 64; /* scaled SF */
+
+  for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr = input_data[outrow];
+    above_ptr = input_data[outrow-1];
+    below_ptr = input_data[outrow+1];
+
+    /* Special case for first column */
+    colsum = GETJSAMPLE(*above_ptr++) + GETJSAMPLE(*below_ptr++) +
+	     GETJSAMPLE(*inptr);
+    membersum = GETJSAMPLE(*inptr++);
+    nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
+		 GETJSAMPLE(*inptr);
+    neighsum = colsum + (colsum - membersum) + nextcolsum;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+    lastcolsum = colsum; colsum = nextcolsum;
+
+    for (colctr = output_cols - 2; colctr > 0; colctr--) {
+      membersum = GETJSAMPLE(*inptr++);
+      above_ptr++; below_ptr++;
+      nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
+		   GETJSAMPLE(*inptr);
+      neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
+      membersum = membersum * memberscale + neighsum * neighscale;
+      *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
+      lastcolsum = colsum; colsum = nextcolsum;
+    }
+
+    /* Special case for last column */
+    membersum = GETJSAMPLE(*inptr);
+    neighsum = lastcolsum + (colsum - membersum) + colsum;
+    membersum = membersum * memberscale + neighsum * neighscale;
+    *outptr = (JSAMPLE) ((membersum + 32768) >> 16);
+
+  }
+}
+
+#endif /* INPUT_SMOOTHING_SUPPORTED */
+
+
+/*
+ * Module initialization routine for downsampling.
+ * Note that we must select a routine for each component.
+ */
+
+GLOBAL(void)
+jinit_downsampler (j_compress_ptr cinfo)
+{
+  my_downsample_ptr downsample;
+  int ci;
+  jpeg_component_info * compptr;
+  boolean smoothok = TRUE;
+
+  downsample = (my_downsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_downsampler));
+  cinfo->downsample = (struct jpeg_downsampler *) downsample;
+  downsample->pub.start_pass = start_pass_downsample;
+  downsample->pub.downsample = sep_downsample;
+  downsample->pub.need_context_rows = FALSE;
+
+  if (cinfo->CCIR601_sampling)
+    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
+
+  /* Verify we can handle the sampling factors, and set up method pointers */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor == cinfo->max_h_samp_factor &&
+	compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+#ifdef INPUT_SMOOTHING_SUPPORTED
+      if (cinfo->smoothing_factor) {
+	downsample->methods[ci] = fullsize_smooth_downsample;
+	downsample->pub.need_context_rows = TRUE;
+      } else
+#endif
+	downsample->methods[ci] = fullsize_downsample;
+    } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
+	       compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+      smoothok = FALSE;
+      downsample->methods[ci] = h2v1_downsample;
+    } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
+	       compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
+#ifdef INPUT_SMOOTHING_SUPPORTED
+      if (cinfo->smoothing_factor) {
+	downsample->methods[ci] = h2v2_smooth_downsample;
+	downsample->pub.need_context_rows = TRUE;
+      } else
+#endif
+	downsample->methods[ci] = h2v2_downsample;
+    } else if ((cinfo->max_h_samp_factor % compptr->h_samp_factor) == 0 &&
+	       (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
+      smoothok = FALSE;
+      downsample->methods[ci] = int_downsample;
+    } else
+      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
+  }
+
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  if (cinfo->smoothing_factor && !smoothok)
+    TRACEMS(cinfo, 0, JTRC_SMOOTH_NOTIMPL);
+#endif
+}
diff --git a/jpeg/jctrans.c b/jpeg/jctrans.c
new file mode 100644
index 000000000..0e6d70769
--- /dev/null
+++ b/jpeg/jctrans.c
@@ -0,0 +1,388 @@
+/*
+ * jctrans.c
+ *
+ * Copyright (C) 1995-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains library routines for transcoding compression,
+ * that is, writing raw DCT coefficient arrays to an output JPEG file.
+ * The routines in jcapimin.c will also be needed by a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Forward declarations */
+LOCAL(void) transencode_master_selection
+	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+LOCAL(void) transencode_coef_controller
+	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+
+
+/*
+ * Compression initialization for writing raw-coefficient data.
+ * Before calling this, all parameters and a data destination must be set up.
+ * Call jpeg_finish_compress() to actually write the data.
+ *
+ * The number of passed virtual arrays must match cinfo->num_components.
+ * Note that the virtual arrays need not be filled or even realized at
+ * the time write_coefficients is called; indeed, if the virtual arrays
+ * were requested from this compression object's memory manager, they
+ * typically will be realized during this routine and filled afterwards.
+ */
+
+GLOBAL(void)
+jpeg_write_coefficients (j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays)
+{
+  if (cinfo->global_state != CSTATE_START)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Mark all tables to be written */
+  jpeg_suppress_tables(cinfo, FALSE);
+  /* (Re)initialize error mgr and destination modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->dest->init_destination) (cinfo);
+  /* Perform master selection of active modules */
+  transencode_master_selection(cinfo, coef_arrays);
+  /* Wait for jpeg_finish_compress() call */
+  cinfo->next_scanline = 0;	/* so jpeg_write_marker works */
+  cinfo->global_state = CSTATE_WRCOEFS;
+}
+
+
+/*
+ * Initialize the compression object with default parameters,
+ * then copy from the source object all parameters needed for lossless
+ * transcoding.  Parameters that can be varied without loss (such as
+ * scan script and Huffman optimization) are left in their default states.
+ */
+
+GLOBAL(void)
+jpeg_copy_critical_parameters (j_decompress_ptr srcinfo,
+			       j_compress_ptr dstinfo)
+{
+  JQUANT_TBL ** qtblptr;
+  jpeg_component_info *incomp, *outcomp;
+  JQUANT_TBL *c_quant, *slot_quant;
+  int tblno, ci, coefi;
+
+  /* Safety check to ensure start_compress not called yet. */
+  if (dstinfo->global_state != CSTATE_START)
+    ERREXIT1(dstinfo, JERR_BAD_STATE, dstinfo->global_state);
+  /* Copy fundamental image dimensions */
+  dstinfo->image_width = srcinfo->image_width;
+  dstinfo->image_height = srcinfo->image_height;
+  dstinfo->input_components = srcinfo->num_components;
+  dstinfo->in_color_space = srcinfo->jpeg_color_space;
+  /* Initialize all parameters to default values */
+  jpeg_set_defaults(dstinfo);
+  /* jpeg_set_defaults may choose wrong colorspace, eg YCbCr if input is RGB.
+   * Fix it to get the right header markers for the image colorspace.
+   */
+  jpeg_set_colorspace(dstinfo, srcinfo->jpeg_color_space);
+  dstinfo->data_precision = srcinfo->data_precision;
+  dstinfo->CCIR601_sampling = srcinfo->CCIR601_sampling;
+  /* Copy the source's quantization tables. */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    if (srcinfo->quant_tbl_ptrs[tblno] != NULL) {
+      qtblptr = & dstinfo->quant_tbl_ptrs[tblno];
+      if (*qtblptr == NULL)
+	*qtblptr = jpeg_alloc_quant_table((j_common_ptr) dstinfo);
+      MEMCOPY((*qtblptr)->quantval,
+	      srcinfo->quant_tbl_ptrs[tblno]->quantval,
+	      SIZEOF((*qtblptr)->quantval));
+      (*qtblptr)->sent_table = FALSE;
+    }
+  }
+  /* Copy the source's per-component info.
+   * Note we assume jpeg_set_defaults has allocated the dest comp_info array.
+   */
+  dstinfo->num_components = srcinfo->num_components;
+  if (dstinfo->num_components < 1 || dstinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(dstinfo, JERR_COMPONENT_COUNT, dstinfo->num_components,
+	     MAX_COMPONENTS);
+  for (ci = 0, incomp = srcinfo->comp_info, outcomp = dstinfo->comp_info;
+       ci < dstinfo->num_components; ci++, incomp++, outcomp++) {
+    outcomp->component_id = incomp->component_id;
+    outcomp->h_samp_factor = incomp->h_samp_factor;
+    outcomp->v_samp_factor = incomp->v_samp_factor;
+    outcomp->quant_tbl_no = incomp->quant_tbl_no;
+    /* Make sure saved quantization table for component matches the qtable
+     * slot.  If not, the input file re-used this qtable slot.
+     * IJG encoder currently cannot duplicate this.
+     */
+    tblno = outcomp->quant_tbl_no;
+    if (tblno < 0 || tblno >= NUM_QUANT_TBLS ||
+	srcinfo->quant_tbl_ptrs[tblno] == NULL)
+      ERREXIT1(dstinfo, JERR_NO_QUANT_TABLE, tblno);
+    slot_quant = srcinfo->quant_tbl_ptrs[tblno];
+    c_quant = incomp->quant_table;
+    if (c_quant != NULL) {
+      for (coefi = 0; coefi < DCTSIZE2; coefi++) {
+	if (c_quant->quantval[coefi] != slot_quant->quantval[coefi])
+	  ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
+      }
+    }
+    /* Note: we do not copy the source's Huffman table assignments;
+     * instead we rely on jpeg_set_colorspace to have made a suitable choice.
+     */
+  }
+  /* Also copy JFIF version and resolution information, if available.
+   * Strictly speaking this isn't "critical" info, but it's nearly
+   * always appropriate to copy it if available.  In particular,
+   * if the application chooses to copy JFIF 1.02 extension markers from
+   * the source file, we need to copy the version to make sure we don't
+   * emit a file that has 1.02 extensions but a claimed version of 1.01.
+   * We will *not*, however, copy version info from mislabeled "2.01" files.
+   */
+  if (srcinfo->saw_JFIF_marker) {
+    if (srcinfo->JFIF_major_version == 1) {
+      dstinfo->JFIF_major_version = srcinfo->JFIF_major_version;
+      dstinfo->JFIF_minor_version = srcinfo->JFIF_minor_version;
+    }
+    dstinfo->density_unit = srcinfo->density_unit;
+    dstinfo->X_density = srcinfo->X_density;
+    dstinfo->Y_density = srcinfo->Y_density;
+  }
+}
+
+
+/*
+ * Master selection of compression modules for transcoding.
+ * This substitutes for jcinit.c's initialization of the full compressor.
+ */
+
+LOCAL(void)
+transencode_master_selection (j_compress_ptr cinfo,
+			      jvirt_barray_ptr * coef_arrays)
+{
+  /* Although we don't actually use input_components for transcoding,
+   * jcmaster.c's initial_setup will complain if input_components is 0.
+   */
+  cinfo->input_components = 1;
+  /* Initialize master control (includes parameter checking/processing) */
+  jinit_c_master_control(cinfo, TRUE /* transcode only */);
+
+  /* Entropy encoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef C_PROGRESSIVE_SUPPORTED
+      jinit_phuff_encoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_encoder(cinfo);
+  }
+
+  /* We need a special coefficient buffer controller. */
+  transencode_coef_controller(cinfo, coef_arrays);
+
+  jinit_marker_writer(cinfo);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Write the datastream header (SOI, JFIF) immediately.
+   * Frame and scan headers are postponed till later.
+   * This lets application insert special markers after the SOI.
+   */
+  (*cinfo->marker->write_file_header) (cinfo);
+}
+
+
+/*
+ * The rest of this file is a special implementation of the coefficient
+ * buffer controller.  This is similar to jccoefct.c, but it handles only
+ * output from presupplied virtual arrays.  Furthermore, we generate any
+ * dummy padding blocks on-the-fly rather than expecting them to be present
+ * in the arrays.
+ */
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_c_coef_controller pub; /* public fields */
+
+  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
+  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
+  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+
+  /* Virtual block array for each component. */
+  jvirt_barray_ptr * whole_image;
+
+  /* Workspace for constructing dummy blocks at right/bottom edges. */
+  JBLOCKROW dummy_buffer[C_MAX_BLOCKS_IN_MCU];
+} my_coef_controller;
+
+typedef my_coef_controller * my_coef_ptr;
+
+
+LOCAL(void)
+start_iMCU_row (j_compress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (coef->iMCU_row_num < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->mcu_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_coef (j_compress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  if (pass_mode != JBUF_CRANK_DEST)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  coef->iMCU_row_num = 0;
+  start_iMCU_row(cinfo);
+}
+
+
+/*
+ * Process some data.
+ * We process the equivalent of one fully interleaved MCU row ("iMCU" row)
+ * per call, ie, v_samp_factor block rows for each component in the scan.
+ * The data is obtained from the virtual arrays and fed to the entropy coder.
+ * Returns TRUE if the iMCU row is completed, FALSE if suspended.
+ *
+ * NB: input_buf is ignored; it is likely to be a NULL pointer.
+ */
+
+METHODDEF(boolean)
+compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, ci, xindex, yindex, yoffset, blockcnt;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW MCU_buffer[C_MAX_BLOCKS_IN_MCU];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       coef->iMCU_row_num * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, FALSE);
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	start_col = MCU_col_num * compptr->MCU_width;
+	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+						: compptr->last_col_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  if (coef->iMCU_row_num < last_iMCU_row ||
+	      yindex+yoffset < compptr->last_row_height) {
+	    /* Fill in pointers to real blocks in this row */
+	    buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+	    for (xindex = 0; xindex < blockcnt; xindex++)
+	      MCU_buffer[blkn++] = buffer_ptr++;
+	  } else {
+	    /* At bottom of image, need a whole row of dummy blocks */
+	    xindex = 0;
+	  }
+	  /* Fill in any dummy blocks needed in this row.
+	   * Dummy blocks are filled in the same way as in jccoefct.c:
+	   * all zeroes in the AC entries, DC entries equal to previous
+	   * block's DC value.  The init routine has already zeroed the
+	   * AC entries, so we need only set the DC entries correctly.
+	   */
+	  for (; xindex < compptr->MCU_width; xindex++) {
+	    MCU_buffer[blkn] = coef->dummy_buffer[blkn];
+	    MCU_buffer[blkn][0][0] = MCU_buffer[blkn-1][0][0];
+	    blkn++;
+	  }
+	}
+      }
+      /* Try to write the MCU. */
+      if (! (*cinfo->entropy->encode_mcu) (cinfo, MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->mcu_ctr = MCU_col_num;
+	return FALSE;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->mcu_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  coef->iMCU_row_num++;
+  start_iMCU_row(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Initialize coefficient buffer controller.
+ *
+ * Each passed coefficient array must be the right size for that
+ * coefficient: width_in_blocks wide and height_in_blocks high,
+ * with unitheight at least v_samp_factor.
+ */
+
+LOCAL(void)
+transencode_coef_controller (j_compress_ptr cinfo,
+			     jvirt_barray_ptr * coef_arrays)
+{
+  my_coef_ptr coef;
+  JBLOCKROW buffer;
+  int i;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_coef_controller));
+  cinfo->coef = (struct jpeg_c_coef_controller *) coef;
+  coef->pub.start_pass = start_pass_coef;
+  coef->pub.compress_data = compress_output;
+
+  /* Save pointer to virtual arrays */
+  coef->whole_image = coef_arrays;
+
+  /* Allocate and pre-zero space for dummy DCT blocks. */
+  buffer = (JBLOCKROW)
+    (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+  jzero_far((void FAR *) buffer, C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+  for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
+    coef->dummy_buffer[i] = buffer + i;
+  }
+}
diff --git a/jpeg/jdapimin.c b/jpeg/jdapimin.c
new file mode 100644
index 000000000..cadb59fce
--- /dev/null
+++ b/jpeg/jdapimin.c
@@ -0,0 +1,395 @@
+/*
+ * jdapimin.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the decompression half
+ * of the JPEG library.  These are the "minimum" API routines that may be
+ * needed in either the normal full-decompression case or the
+ * transcoding-only case.
+ *
+ * Most of the routines intended to be called directly by an application
+ * are in this file or in jdapistd.c.  But also see jcomapi.c for routines
+ * shared by compression and decompression, and jdtrans.c for the transcoding
+ * case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * Initialization of a JPEG decompression object.
+ * The error manager must already be set up (in case memory manager fails).
+ */
+
+GLOBAL(void)
+jpeg_CreateDecompress (j_decompress_ptr cinfo, int version, size_t structsize)
+{
+  int i;
+
+  /* Guard against version mismatches between library and caller. */
+  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  if (version != JPEG_LIB_VERSION)
+    ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
+  if (structsize != SIZEOF(struct jpeg_decompress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
+	     (int) SIZEOF(struct jpeg_decompress_struct), (int) structsize);
+
+  /* For debugging purposes, we zero the whole master structure.
+   * But the application has already set the err pointer, and may have set
+   * client_data, so we have to save and restore those fields.
+   * Note: if application hasn't set client_data, tools like Purify may
+   * complain here.
+   */
+  {
+    struct jpeg_error_mgr * err = cinfo->err;
+    void * client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, SIZEOF(struct jpeg_decompress_struct));
+    cinfo->err = err;
+    cinfo->client_data = client_data;
+  }
+  cinfo->is_decompressor = TRUE;
+
+  /* Initialize a memory manager instance for this object */
+  jinit_memory_mgr((j_common_ptr) cinfo);
+
+  /* Zero out pointers to permanent structures. */
+  cinfo->progress = NULL;
+  cinfo->src = NULL;
+
+  for (i = 0; i < NUM_QUANT_TBLS; i++)
+    cinfo->quant_tbl_ptrs[i] = NULL;
+
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    cinfo->dc_huff_tbl_ptrs[i] = NULL;
+    cinfo->ac_huff_tbl_ptrs[i] = NULL;
+  }
+
+  /* Initialize marker processor so application can override methods
+   * for COM, APPn markers before calling jpeg_read_header.
+   */
+  cinfo->marker_list = NULL;
+  jinit_marker_reader(cinfo);
+
+  /* And initialize the overall input controller. */
+  jinit_input_controller(cinfo);
+
+  /* OK, I'm ready */
+  cinfo->global_state = DSTATE_START;
+}
+
+
+/*
+ * Destruction of a JPEG decompression object
+ */
+
+GLOBAL(void)
+jpeg_destroy_decompress (j_decompress_ptr cinfo)
+{
+  jpeg_destroy((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Abort processing of a JPEG decompression operation,
+ * but don't destroy the object itself.
+ */
+
+GLOBAL(void)
+jpeg_abort_decompress (j_decompress_ptr cinfo)
+{
+  jpeg_abort((j_common_ptr) cinfo); /* use common routine */
+}
+
+
+/*
+ * Set default decompression parameters.
+ */
+
+LOCAL(void)
+default_decompress_parms (j_decompress_ptr cinfo)
+{
+  /* Guess the input colorspace, and set output colorspace accordingly. */
+  /* (Wish JPEG committee had provided a real way to specify this...) */
+  /* Note application may override our guesses. */
+  switch (cinfo->num_components) {
+  case 1:
+    cinfo->jpeg_color_space = JCS_GRAYSCALE;
+    cinfo->out_color_space = JCS_GRAYSCALE;
+    break;
+    
+  case 3:
+    if (cinfo->saw_JFIF_marker) {
+      cinfo->jpeg_color_space = JCS_YCbCr; /* JFIF implies YCbCr */
+    } else if (cinfo->saw_Adobe_marker) {
+      switch (cinfo->Adobe_transform) {
+      case 0:
+	cinfo->jpeg_color_space = JCS_RGB;
+	break;
+      case 1:
+	cinfo->jpeg_color_space = JCS_YCbCr;
+	break;
+      default:
+	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+	break;
+      }
+    } else {
+      /* Saw no special markers, try to guess from the component IDs */
+      int cid0 = cinfo->comp_info[0].component_id;
+      int cid1 = cinfo->comp_info[1].component_id;
+      int cid2 = cinfo->comp_info[2].component_id;
+
+      if (cid0 == 1 && cid1 == 2 && cid2 == 3)
+	cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
+      else if (cid0 == 82 && cid1 == 71 && cid2 == 66)
+	cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
+      else {
+	TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
+	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+      }
+    }
+    /* Always guess RGB is proper output colorspace. */
+    cinfo->out_color_space = JCS_RGB;
+    break;
+    
+  case 4:
+    if (cinfo->saw_Adobe_marker) {
+      switch (cinfo->Adobe_transform) {
+      case 0:
+	cinfo->jpeg_color_space = JCS_CMYK;
+	break;
+      case 2:
+	cinfo->jpeg_color_space = JCS_YCCK;
+	break;
+      default:
+	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+	cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
+	break;
+      }
+    } else {
+      /* No special markers, assume straight CMYK. */
+      cinfo->jpeg_color_space = JCS_CMYK;
+    }
+    cinfo->out_color_space = JCS_CMYK;
+    break;
+    
+  default:
+    cinfo->jpeg_color_space = JCS_UNKNOWN;
+    cinfo->out_color_space = JCS_UNKNOWN;
+    break;
+  }
+
+  /* Set defaults for other decompression parameters. */
+  cinfo->scale_num = 1;		/* 1:1 scaling */
+  cinfo->scale_denom = 1;
+  cinfo->output_gamma = 1.0;
+  cinfo->buffered_image = FALSE;
+  cinfo->raw_data_out = FALSE;
+  cinfo->dct_method = JDCT_DEFAULT;
+  cinfo->do_fancy_upsampling = TRUE;
+  cinfo->do_block_smoothing = TRUE;
+  cinfo->quantize_colors = FALSE;
+  /* We set these in case application only sets quantize_colors. */
+  cinfo->dither_mode = JDITHER_FS;
+#ifdef QUANT_2PASS_SUPPORTED
+  cinfo->two_pass_quantize = TRUE;
+#else
+  cinfo->two_pass_quantize = FALSE;
+#endif
+  cinfo->desired_number_of_colors = 256;
+  cinfo->colormap = NULL;
+  /* Initialize for no mode change in buffered-image mode. */
+  cinfo->enable_1pass_quant = FALSE;
+  cinfo->enable_external_quant = FALSE;
+  cinfo->enable_2pass_quant = FALSE;
+}
+
+
+/*
+ * Decompression startup: read start of JPEG datastream to see what's there.
+ * Need only initialize JPEG object and supply a data source before calling.
+ *
+ * This routine will read as far as the first SOS marker (ie, actual start of
+ * compressed data), and will save all tables and parameters in the JPEG
+ * object.  It will also initialize the decompression parameters to default
+ * values, and finally return JPEG_HEADER_OK.  On return, the application may
+ * adjust the decompression parameters and then call jpeg_start_decompress.
+ * (Or, if the application only wanted to determine the image parameters,
+ * the data need not be decompressed.  In that case, call jpeg_abort or
+ * jpeg_destroy to release any temporary space.)
+ * If an abbreviated (tables only) datastream is presented, the routine will
+ * return JPEG_HEADER_TABLES_ONLY upon reaching EOI.  The application may then
+ * re-use the JPEG object to read the abbreviated image datastream(s).
+ * It is unnecessary (but OK) to call jpeg_abort in this case.
+ * The JPEG_SUSPENDED return code only occurs if the data source module
+ * requests suspension of the decompressor.  In this case the application
+ * should load more source data and then re-call jpeg_read_header to resume
+ * processing.
+ * If a non-suspending data source is used and require_image is TRUE, then the
+ * return code need not be inspected since only JPEG_HEADER_OK is possible.
+ *
+ * This routine is now just a front end to jpeg_consume_input, with some
+ * extra error checking.
+ */
+
+GLOBAL(int)
+jpeg_read_header (j_decompress_ptr cinfo, boolean require_image)
+{
+  int retcode;
+
+  if (cinfo->global_state != DSTATE_START &&
+      cinfo->global_state != DSTATE_INHEADER)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  retcode = jpeg_consume_input(cinfo);
+
+  switch (retcode) {
+  case JPEG_REACHED_SOS:
+    retcode = JPEG_HEADER_OK;
+    break;
+  case JPEG_REACHED_EOI:
+    if (require_image)		/* Complain if application wanted an image */
+      ERREXIT(cinfo, JERR_NO_IMAGE);
+    /* Reset to start state; it would be safer to require the application to
+     * call jpeg_abort, but we can't change it now for compatibility reasons.
+     * A side effect is to free any temporary memory (there shouldn't be any).
+     */
+    jpeg_abort((j_common_ptr) cinfo); /* sets state = DSTATE_START */
+    retcode = JPEG_HEADER_TABLES_ONLY;
+    break;
+  case JPEG_SUSPENDED:
+    /* no work */
+    break;
+  }
+
+  return retcode;
+}
+
+
+/*
+ * Consume data in advance of what the decompressor requires.
+ * This can be called at any time once the decompressor object has
+ * been created and a data source has been set up.
+ *
+ * This routine is essentially a state machine that handles a couple
+ * of critical state-transition actions, namely initial setup and
+ * transition from header scanning to ready-for-start_decompress.
+ * All the actual input is done via the input controller's consume_input
+ * method.
+ */
+
+GLOBAL(int)
+jpeg_consume_input (j_decompress_ptr cinfo)
+{
+  int retcode = JPEG_SUSPENDED;
+
+  /* NB: every possible DSTATE value should be listed in this switch */
+  switch (cinfo->global_state) {
+  case DSTATE_START:
+    /* Start-of-datastream actions: reset appropriate modules */
+    (*cinfo->inputctl->reset_input_controller) (cinfo);
+    /* Initialize application's data source module */
+    (*cinfo->src->init_source) (cinfo);
+    cinfo->global_state = DSTATE_INHEADER;
+    /*FALLTHROUGH*/
+  case DSTATE_INHEADER:
+    retcode = (*cinfo->inputctl->consume_input) (cinfo);
+    if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */
+      /* Set up default parameters based on header data */
+      default_decompress_parms(cinfo);
+      /* Set global state: ready for start_decompress */
+      cinfo->global_state = DSTATE_READY;
+    }
+    break;
+  case DSTATE_READY:
+    /* Can't advance past first SOS until start_decompress is called */
+    retcode = JPEG_REACHED_SOS;
+    break;
+  case DSTATE_PRELOAD:
+  case DSTATE_PRESCAN:
+  case DSTATE_SCANNING:
+  case DSTATE_RAW_OK:
+  case DSTATE_BUFIMAGE:
+  case DSTATE_BUFPOST:
+  case DSTATE_STOPPING:
+    retcode = (*cinfo->inputctl->consume_input) (cinfo);
+    break;
+  default:
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  return retcode;
+}
+
+
+/*
+ * Have we finished reading the input file?
+ */
+
+GLOBAL(boolean)
+jpeg_input_complete (j_decompress_ptr cinfo)
+{
+  /* Check for valid jpeg object */
+  if (cinfo->global_state < DSTATE_START ||
+      cinfo->global_state > DSTATE_STOPPING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return cinfo->inputctl->eoi_reached;
+}
+
+
+/*
+ * Is there more than one scan?
+ */
+
+GLOBAL(boolean)
+jpeg_has_multiple_scans (j_decompress_ptr cinfo)
+{
+  /* Only valid after jpeg_read_header completes */
+  if (cinfo->global_state < DSTATE_READY ||
+      cinfo->global_state > DSTATE_STOPPING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return cinfo->inputctl->has_multiple_scans;
+}
+
+
+/*
+ * Finish JPEG decompression.
+ *
+ * This will normally just verify the file trailer and release temp storage.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_finish_decompress (j_decompress_ptr cinfo)
+{
+  if ((cinfo->global_state == DSTATE_SCANNING ||
+       cinfo->global_state == DSTATE_RAW_OK) && ! cinfo->buffered_image) {
+    /* Terminate final pass of non-buffered mode */
+    if (cinfo->output_scanline < cinfo->output_height)
+      ERREXIT(cinfo, JERR_TOO_LITTLE_DATA);
+    (*cinfo->master->finish_output_pass) (cinfo);
+    cinfo->global_state = DSTATE_STOPPING;
+  } else if (cinfo->global_state == DSTATE_BUFIMAGE) {
+    /* Finishing after a buffered-image operation */
+    cinfo->global_state = DSTATE_STOPPING;
+  } else if (cinfo->global_state != DSTATE_STOPPING) {
+    /* STOPPING = repeat call after a suspension, anything else is error */
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  /* Read until EOI */
+  while (! cinfo->inputctl->eoi_reached) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return FALSE;		/* Suspend, come back later */
+  }
+  /* Do final cleanup */
+  (*cinfo->src->term_source) (cinfo);
+  /* We can use jpeg_abort to release memory and reset global_state */
+  jpeg_abort((j_common_ptr) cinfo);
+  return TRUE;
+}
diff --git a/jpeg/jdapistd.c b/jpeg/jdapistd.c
new file mode 100644
index 000000000..c8e3fa0c3
--- /dev/null
+++ b/jpeg/jdapistd.c
@@ -0,0 +1,275 @@
+/*
+ * jdapistd.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains application interface code for the decompression half
+ * of the JPEG library.  These are the "standard" API routines that are
+ * used in the normal full-decompression case.  They are not used by a
+ * transcoding-only application.  Note that if an application links in
+ * jpeg_start_decompress, it will end up linking in the entire decompressor.
+ * We thus must separate this file from jdapimin.c to avoid linking the
+ * whole decompression library into a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Forward declarations */
+LOCAL(boolean) output_pass_setup JPP((j_decompress_ptr cinfo));
+
+
+/*
+ * Decompression initialization.
+ * jpeg_read_header must be completed before calling this.
+ *
+ * If a multipass operating mode was selected, this will do all but the
+ * last pass, and thus may take a great deal of time.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_start_decompress (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize master control, select active modules */
+    jinit_master_decompress(cinfo);
+    if (cinfo->buffered_image) {
+      /* No more work here; expecting jpeg_start_output next */
+      cinfo->global_state = DSTATE_BUFIMAGE;
+      return TRUE;
+    }
+    cinfo->global_state = DSTATE_PRELOAD;
+  }
+  if (cinfo->global_state == DSTATE_PRELOAD) {
+    /* If file has multiple scans, absorb them all into the coef buffer */
+    if (cinfo->inputctl->has_multiple_scans) {
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+      for (;;) {
+	int retcode;
+	/* Call progress monitor hook if present */
+	if (cinfo->progress != NULL)
+	  (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+	/* Absorb some more input */
+	retcode = (*cinfo->inputctl->consume_input) (cinfo);
+	if (retcode == JPEG_SUSPENDED)
+	  return FALSE;
+	if (retcode == JPEG_REACHED_EOI)
+	  break;
+	/* Advance progress counter if appropriate */
+	if (cinfo->progress != NULL &&
+	    (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+	  if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+	    /* jdmaster underestimated number of scans; ratchet up one scan */
+	    cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+	  }
+	}
+      }
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+    }
+    cinfo->output_scan_number = cinfo->input_scan_number;
+  } else if (cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Perform any dummy output passes, and set up for the final pass */
+  return output_pass_setup(cinfo);
+}
+
+
+/*
+ * Set up for an output pass, and perform any dummy pass(es) needed.
+ * Common subroutine for jpeg_start_decompress and jpeg_start_output.
+ * Entry: global_state = DSTATE_PRESCAN only if previously suspended.
+ * Exit: If done, returns TRUE and sets global_state for proper output mode.
+ *       If suspended, returns FALSE and sets global_state = DSTATE_PRESCAN.
+ */
+
+LOCAL(boolean)
+output_pass_setup (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state != DSTATE_PRESCAN) {
+    /* First call: do pass setup */
+    (*cinfo->master->prepare_for_output_pass) (cinfo);
+    cinfo->output_scanline = 0;
+    cinfo->global_state = DSTATE_PRESCAN;
+  }
+  /* Loop over any required dummy passes */
+  while (cinfo->master->is_dummy_pass) {
+#ifdef QUANT_2PASS_SUPPORTED
+    /* Crank through the dummy pass */
+    while (cinfo->output_scanline < cinfo->output_height) {
+      JDIMENSION last_scanline;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL) {
+	cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+	cinfo->progress->pass_limit = (long) cinfo->output_height;
+	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      }
+      /* Process some data */
+      last_scanline = cinfo->output_scanline;
+      (*cinfo->main->process_data) (cinfo, (JSAMPARRAY) NULL,
+				    &cinfo->output_scanline, (JDIMENSION) 0);
+      if (cinfo->output_scanline == last_scanline)
+	return FALSE;		/* No progress made, must suspend */
+    }
+    /* Finish up dummy pass, and set up for another one */
+    (*cinfo->master->finish_output_pass) (cinfo);
+    (*cinfo->master->prepare_for_output_pass) (cinfo);
+    cinfo->output_scanline = 0;
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* QUANT_2PASS_SUPPORTED */
+  }
+  /* Ready for application to drive output pass through
+   * jpeg_read_scanlines or jpeg_read_raw_data.
+   */
+  cinfo->global_state = cinfo->raw_data_out ? DSTATE_RAW_OK : DSTATE_SCANNING;
+  return TRUE;
+}
+
+
+/*
+ * Read some scanlines of data from the JPEG decompressor.
+ *
+ * The return value will be the number of lines actually read.
+ * This may be less than the number requested in several cases,
+ * including bottom of image, data source suspension, and operating
+ * modes that emit multiple scanlines at a time.
+ *
+ * Note: we warn about excess calls to jpeg_read_scanlines() since
+ * this likely signals an application programmer error.  However,
+ * an oversize buffer (max_lines > scanlines remaining) is not an error.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_scanlines (j_decompress_ptr cinfo, JSAMPARRAY scanlines,
+		     JDIMENSION max_lines)
+{
+  JDIMENSION row_ctr;
+
+  if (cinfo->global_state != DSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->output_scanline >= cinfo->output_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->output_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Process some data */
+  row_ctr = 0;
+  (*cinfo->main->process_data) (cinfo, scanlines, &row_ctr, max_lines);
+  cinfo->output_scanline += row_ctr;
+  return row_ctr;
+}
+
+
+/*
+ * Alternate entry point to read raw data.
+ * Processes exactly one iMCU row per call, unless suspended.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data,
+		    JDIMENSION max_lines)
+{
+  JDIMENSION lines_per_iMCU_row;
+
+  if (cinfo->global_state != DSTATE_RAW_OK)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  if (cinfo->output_scanline >= cinfo->output_height) {
+    WARNMS(cinfo, JWRN_TOO_MUCH_DATA);
+    return 0;
+  }
+
+  /* Call progress monitor hook if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+    cinfo->progress->pass_limit = (long) cinfo->output_height;
+    (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+  }
+
+  /* Verify that at least one iMCU row can be returned. */
+  lines_per_iMCU_row = cinfo->max_v_samp_factor * cinfo->min_DCT_scaled_size;
+  if (max_lines < lines_per_iMCU_row)
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* Decompress directly into user's buffer. */
+  if (! (*cinfo->coef->decompress_data) (cinfo, data))
+    return 0;			/* suspension forced, can do nothing more */
+
+  /* OK, we processed one iMCU row. */
+  cinfo->output_scanline += lines_per_iMCU_row;
+  return lines_per_iMCU_row;
+}
+
+
+/* Additional entry points for buffered-image mode. */
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Initialize for an output pass in buffered-image mode.
+ */
+
+GLOBAL(boolean)
+jpeg_start_output (j_decompress_ptr cinfo, int scan_number)
+{
+  if (cinfo->global_state != DSTATE_BUFIMAGE &&
+      cinfo->global_state != DSTATE_PRESCAN)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  /* Limit scan number to valid range */
+  if (scan_number <= 0)
+    scan_number = 1;
+  if (cinfo->inputctl->eoi_reached &&
+      scan_number > cinfo->input_scan_number)
+    scan_number = cinfo->input_scan_number;
+  cinfo->output_scan_number = scan_number;
+  /* Perform any dummy output passes, and set up for the real pass */
+  return output_pass_setup(cinfo);
+}
+
+
+/*
+ * Finish up after an output pass in buffered-image mode.
+ *
+ * Returns FALSE if suspended.  The return value need be inspected only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(boolean)
+jpeg_finish_output (j_decompress_ptr cinfo)
+{
+  if ((cinfo->global_state == DSTATE_SCANNING ||
+       cinfo->global_state == DSTATE_RAW_OK) && cinfo->buffered_image) {
+    /* Terminate this pass. */
+    /* We do not require the whole pass to have been completed. */
+    (*cinfo->master->finish_output_pass) (cinfo);
+    cinfo->global_state = DSTATE_BUFPOST;
+  } else if (cinfo->global_state != DSTATE_BUFPOST) {
+    /* BUFPOST = repeat call after a suspension, anything else is error */
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  }
+  /* Read markers looking for SOS or EOI */
+  while (cinfo->input_scan_number <= cinfo->output_scan_number &&
+	 ! cinfo->inputctl->eoi_reached) {
+    if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
+      return FALSE;		/* Suspend, come back later */
+  }
+  cinfo->global_state = DSTATE_BUFIMAGE;
+  return TRUE;
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
diff --git a/jpeg/jdatadst.c b/jpeg/jdatadst.c
new file mode 100644
index 000000000..a8f6fb0e0
--- /dev/null
+++ b/jpeg/jdatadst.c
@@ -0,0 +1,151 @@
+/*
+ * jdatadst.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains compression data destination routines for the case of
+ * emitting JPEG data to a file (or any stdio stream).  While these routines
+ * are sufficient for most applications, some will want to use a different
+ * destination manager.
+ * IMPORTANT: we assume that fwrite() will correctly transcribe an array of
+ * JOCTETs into 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data destination object for stdio output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  FILE * outfile;		/* target stream */
+  JOCTET * buffer;		/* start of buffer */
+} my_destination_mgr;
+
+typedef my_destination_mgr * my_dest_ptr;
+
+#define OUTPUT_BUF_SIZE  4096	/* choose an efficiently fwrite'able size */
+
+
+/*
+ * Initialize destination --- called by jpeg_start_compress
+ * before any data is actually written.
+ */
+
+METHODDEF(void)
+init_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  /* Allocate the output buffer --- it will be released when done with image */
+  dest->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  OUTPUT_BUF_SIZE * SIZEOF(JOCTET));
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+}
+
+
+/*
+ * Empty the output buffer --- called whenever buffer fills up.
+ *
+ * In typical applications, this should write the entire output buffer
+ * (ignoring the current state of next_output_byte & free_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been dumped.
+ *
+ * In applications that need to be able to suspend compression due to output
+ * overrun, a FALSE return indicates that the buffer cannot be emptied now.
+ * In this situation, the compressor will return to its caller (possibly with
+ * an indication that it has not accepted all the supplied scanlines).  The
+ * application should resume compression after it has made more room in the
+ * output buffer.  Note that there are substantial restrictions on the use of
+ * suspension --- see the documentation.
+ *
+ * When suspending, the compressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_output_byte & free_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point will be regenerated after resumption, so do not
+ * write it out when emptying the buffer externally.
+ */
+
+METHODDEF(boolean)
+empty_output_buffer (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  if (JFWRITE(dest->outfile, dest->buffer, OUTPUT_BUF_SIZE) !=
+      (size_t) OUTPUT_BUF_SIZE)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+
+  return TRUE;
+}
+
+
+/*
+ * Terminate destination --- called by jpeg_finish_compress
+ * after all data has been written.  Usually needs to flush buffer.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+  size_t datacount = OUTPUT_BUF_SIZE - dest->pub.free_in_buffer;
+
+  /* Write any data remaining in the buffer */
+  if (datacount > 0) {
+    if (JFWRITE(dest->outfile, dest->buffer, datacount) != datacount)
+      ERREXIT(cinfo, JERR_FILE_WRITE);
+  }
+  fflush(dest->outfile);
+  /* Make sure we wrote the output file OK */
+  if (ferror(dest->outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Prepare for output to a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing compression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_dest (j_compress_ptr cinfo, FILE * outfile)
+{
+  my_dest_ptr dest;
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same file without re-executing jpeg_stdio_dest.
+   * This makes it dangerous to use this manager and a different destination
+   * manager serially with the same JPEG object, because their private object
+   * sizes may be different.  Caveat programmer.
+   */
+  if (cinfo->dest == NULL) {	/* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_destination_mgr));
+  }
+
+  dest = (my_dest_ptr) cinfo->dest;
+  dest->pub.init_destination = init_destination;
+  dest->pub.empty_output_buffer = empty_output_buffer;
+  dest->pub.term_destination = term_destination;
+  dest->outfile = outfile;
+}
diff --git a/jpeg/jdatasrc.c b/jpeg/jdatasrc.c
new file mode 100644
index 000000000..edc752bf5
--- /dev/null
+++ b/jpeg/jdatasrc.c
@@ -0,0 +1,212 @@
+/*
+ * jdatasrc.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains decompression data source routines for the case of
+ * reading JPEG data from a file (or any stdio stream).  While these routines
+ * are sufficient for most applications, some will want to use a different
+ * source manager.
+ * IMPORTANT: we assume that fread() will correctly transcribe an array of
+ * JOCTETs from 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data source object for stdio input */
+
+typedef struct {
+  struct jpeg_source_mgr pub;	/* public fields */
+
+  FILE * infile;		/* source stream */
+  JOCTET * buffer;		/* start of buffer */
+  boolean start_of_file;	/* have we gotten any data yet? */
+} my_source_mgr;
+
+typedef my_source_mgr * my_src_ptr;
+
+#define INPUT_BUF_SIZE  4096	/* choose an efficiently fread'able size */
+
+
+/*
+ * Initialize source --- called by jpeg_read_header
+ * before any data is actually read.
+ */
+
+METHODDEF(void)
+init_source (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+
+  /* We reset the empty-input-file flag for each image,
+   * but we don't clear the input buffer.
+   * This is correct behavior for reading a series of images from one source.
+   */
+  src->start_of_file = TRUE;
+}
+
+
+/*
+ * Fill the input buffer --- called whenever buffer is emptied.
+ *
+ * In typical applications, this should read fresh data into the buffer
+ * (ignoring the current state of next_input_byte & bytes_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been reloaded.  It is not necessary to
+ * fill the buffer entirely, only to obtain at least one more byte.
+ *
+ * There is no such thing as an EOF return.  If the end of the file has been
+ * reached, the routine has a choice of ERREXIT() or inserting fake data into
+ * the buffer.  In most cases, generating a warning message and inserting a
+ * fake EOI marker is the best course of action --- this will allow the
+ * decompressor to output however much of the image is there.  However,
+ * the resulting error message is misleading if the real problem is an empty
+ * input file, so we handle that case specially.
+ *
+ * In applications that need to be able to suspend compression due to input
+ * not being available yet, a FALSE return indicates that no more data can be
+ * obtained right now, but more may be forthcoming later.  In this situation,
+ * the decompressor will return to its caller (with an indication of the
+ * number of scanlines it has read, if any).  The application should resume
+ * decompression after it has loaded more data into the input buffer.  Note
+ * that there are substantial restrictions on the use of suspension --- see
+ * the documentation.
+ *
+ * When suspending, the decompressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_input_byte & bytes_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point must be rescanned after resumption, so move it to
+ * the front of the buffer rather than discarding it.
+ */
+
+METHODDEF(boolean)
+fill_input_buffer (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+  size_t nbytes;
+
+  nbytes = JFREAD(src->infile, src->buffer, INPUT_BUF_SIZE);
+
+  if (nbytes <= 0) {
+    if (src->start_of_file)	/* Treat empty input file as fatal error */
+      ERREXIT(cinfo, JERR_INPUT_EMPTY);
+    WARNMS(cinfo, JWRN_JPEG_EOF);
+    /* Insert a fake EOI marker */
+    src->buffer[0] = (JOCTET) 0xFF;
+    src->buffer[1] = (JOCTET) JPEG_EOI;
+    nbytes = 2;
+  }
+
+  src->pub.next_input_byte = src->buffer;
+  src->pub.bytes_in_buffer = nbytes;
+  src->start_of_file = FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Skip data --- used to skip over a potentially large amount of
+ * uninteresting data (such as an APPn marker).
+ *
+ * Writers of suspendable-input applications must note that skip_input_data
+ * is not granted the right to give a suspension return.  If the skip extends
+ * beyond the data currently in the buffer, the buffer can be marked empty so
+ * that the next read will cause a fill_input_buffer call that can suspend.
+ * Arranging for additional bytes to be discarded before reloading the input
+ * buffer is the application writer's problem.
+ */
+
+METHODDEF(void)
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+
+  /* Just a dumb implementation for now.  Could use fseek() except
+   * it doesn't work on pipes.  Not clear that being smart is worth
+   * any trouble anyway --- large skips are infrequent.
+   */
+  if (num_bytes > 0) {
+    while (num_bytes > (long) src->pub.bytes_in_buffer) {
+      num_bytes -= (long) src->pub.bytes_in_buffer;
+      (void) fill_input_buffer(cinfo);
+      /* note we assume that fill_input_buffer will never return FALSE,
+       * so suspension need not be handled.
+       */
+    }
+    src->pub.next_input_byte += (size_t) num_bytes;
+    src->pub.bytes_in_buffer -= (size_t) num_bytes;
+  }
+}
+
+
+/*
+ * An additional method that can be provided by data source modules is the
+ * resync_to_restart method for error recovery in the presence of RST markers.
+ * For the moment, this source module just uses the default resync method
+ * provided by the JPEG library.  That method assumes that no backtracking
+ * is possible.
+ */
+
+
+/*
+ * Terminate source --- called by jpeg_finish_decompress
+ * after all data has been read.  Often a no-op.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_source (j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Prepare for input from a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing decompression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_src (j_decompress_ptr cinfo, FILE * infile)
+{
+  my_src_ptr src;
+
+  /* The source object and input buffer are made permanent so that a series
+   * of JPEG images can be read from the same file by calling jpeg_stdio_src
+   * only before the first one.  (If we discarded the buffer at the end of
+   * one image, we'd likely lose the start of the next one.)
+   * This makes it unsafe to use this manager and a different source
+   * manager serially with the same JPEG object.  Caveat programmer.
+   */
+  if (cinfo->src == NULL) {	/* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  SIZEOF(my_source_mgr));
+    src = (my_src_ptr) cinfo->src;
+    src->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				  INPUT_BUF_SIZE * SIZEOF(JOCTET));
+  }
+
+  src = (my_src_ptr) cinfo->src;
+  src->pub.init_source = init_source;
+  src->pub.fill_input_buffer = fill_input_buffer;
+  src->pub.skip_input_data = skip_input_data;
+  src->pub.resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->pub.term_source = term_source;
+  src->infile = infile;
+  src->pub.bytes_in_buffer = 0; /* forces fill_input_buffer on first read */
+  src->pub.next_input_byte = NULL; /* until buffer loaded */
+}
diff --git a/jpeg/jdcoefct.c b/jpeg/jdcoefct.c
new file mode 100644
index 000000000..4938d20fc
--- /dev/null
+++ b/jpeg/jdcoefct.c
@@ -0,0 +1,736 @@
+/*
+ * jdcoefct.c
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the coefficient buffer controller for decompression.
+ * This controller is the top level of the JPEG decompressor proper.
+ * The coefficient buffer lies between entropy decoding and inverse-DCT steps.
+ *
+ * In buffered-image mode, this controller is the interface between
+ * input-oriented processing and output-oriented processing.
+ * Also, the input side (only) is used when reading a file for transcoding.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+/* Block smoothing is only applicable for progressive JPEG, so: */
+#ifndef D_PROGRESSIVE_SUPPORTED
+#undef BLOCK_SMOOTHING_SUPPORTED
+#endif
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_coef_controller pub; /* public fields */
+
+  /* These variables keep track of the current location of the input side. */
+  /* cinfo->input_iMCU_row is also used for this. */
+  JDIMENSION MCU_ctr;		/* counts MCUs processed in current row */
+  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+
+  /* The output side's location is represented by cinfo->output_iMCU_row. */
+
+  /* In single-pass modes, it's sufficient to buffer just one MCU.
+   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
+   * and let the entropy decoder write into that workspace each time.
+   * (On 80x86, the workspace is FAR even though it's not really very big;
+   * this is to keep the module interfaces unchanged when a large coefficient
+   * buffer is necessary.)
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays; it is used only by the input side.
+   */
+  JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU];
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
+#endif
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  /* When doing block smoothing, we latch coefficient Al values here */
+  int * coef_bits_latch;
+#define SAVED_COEFS  6		/* we save coef_bits[0..5] */
+#endif
+} my_coef_controller;
+
+typedef my_coef_controller * my_coef_ptr;
+
+/* Forward declarations */
+METHODDEF(int) decompress_onepass
+	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+METHODDEF(int) decompress_data
+	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+#endif
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+LOCAL(boolean) smoothing_ok JPP((j_decompress_ptr cinfo));
+METHODDEF(int) decompress_smooth_data
+	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+#endif
+
+
+LOCAL(void)
+start_iMCU_row (j_decompress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row (input side) */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->MCU_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
+
+
+/*
+ * Initialize for an input processing pass.
+ */
+
+METHODDEF(void)
+start_input_pass (j_decompress_ptr cinfo)
+{
+  cinfo->input_iMCU_row = 0;
+  start_iMCU_row(cinfo);
+}
+
+
+/*
+ * Initialize for an output processing pass.
+ */
+
+METHODDEF(void)
+start_output_pass (j_decompress_ptr cinfo)
+{
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* If multipass, check to see whether to use block smoothing on this pass */
+  if (coef->pub.coef_arrays != NULL) {
+    if (cinfo->do_block_smoothing && smoothing_ok(cinfo))
+      coef->pub.decompress_data = decompress_smooth_data;
+    else
+      coef->pub.decompress_data = decompress_data;
+  }
+#endif
+  cinfo->output_iMCU_row = 0;
+}
+
+
+/*
+ * Decompress and return some data in the single-pass case.
+ * Always attempts to emit one fully interleaved MCU row ("iMCU" row).
+ * Input and output must run in lockstep since we have only a one-MCU buffer.
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ *
+ * NB: output_buf contains a plane for each component in image,
+ * which we index according to the component's SOF position.
+ */
+
+METHODDEF(int)
+decompress_onepass (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  int blkn, ci, xindex, yindex, yoffset, useful_width;
+  JSAMPARRAY output_ptr;
+  JDIMENSION start_col, output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+
+  /* Loop to process as much as one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col;
+	 MCU_col_num++) {
+      /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
+      jzero_far((void FAR *) coef->MCU_buffer[0],
+		(size_t) (cinfo->blocks_in_MCU * SIZEOF(JBLOCK)));
+      if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->MCU_ctr = MCU_col_num;
+	return JPEG_SUSPENDED;
+      }
+      /* Determine where data should go in output_buf and do the IDCT thing.
+       * We skip dummy blocks at the right and bottom edges (but blkn gets
+       * incremented past them!).  Note the inner loop relies on having
+       * allocated the MCU_buffer[] blocks sequentially.
+       */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	/* Don't bother to IDCT an uninteresting component. */
+	if (! compptr->component_needed) {
+	  blkn += compptr->MCU_blocks;
+	  continue;
+	}
+	inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index];
+	useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+						    : compptr->last_col_width;
+	output_ptr = output_buf[compptr->component_index] +
+	  yoffset * compptr->DCT_scaled_size;
+	start_col = MCU_col_num * compptr->MCU_sample_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  if (cinfo->input_iMCU_row < last_iMCU_row ||
+	      yoffset+yindex < compptr->last_row_height) {
+	    output_col = start_col;
+	    for (xindex = 0; xindex < useful_width; xindex++) {
+	      (*inverse_DCT) (cinfo, compptr,
+			      (JCOEFPTR) coef->MCU_buffer[blkn+xindex],
+			      output_ptr, output_col);
+	      output_col += compptr->DCT_scaled_size;
+	    }
+	  }
+	  blkn += compptr->MCU_width;
+	  output_ptr += compptr->DCT_scaled_size;
+	}
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  cinfo->output_iMCU_row++;
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+
+/*
+ * Dummy consume-input routine for single-pass operation.
+ */
+
+METHODDEF(int)
+dummy_consume_data (j_decompress_ptr cinfo)
+{
+  return JPEG_SUSPENDED;	/* Always indicate nothing was done */
+}
+
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Consume input data and store it in the full-image coefficient buffer.
+ * We read as much as one fully interleaved MCU row ("iMCU" row) per call,
+ * ie, v_samp_factor block rows for each component in the scan.
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ */
+
+METHODDEF(int)
+consume_data (j_decompress_ptr cinfo)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  int blkn, ci, xindex, yindex, yoffset;
+  JDIMENSION start_col;
+  JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
+  JBLOCKROW buffer_ptr;
+  jpeg_component_info *compptr;
+
+  /* Align the virtual buffers for the components used in this scan. */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    buffer[ci] = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[compptr->component_index],
+       cinfo->input_iMCU_row * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, TRUE);
+    /* Note: entropy decoder expects buffer to be zeroed,
+     * but this is handled automatically by the memory manager
+     * because we requested a pre-zeroed array.
+     */
+  }
+
+  /* Loop to process one whole iMCU row */
+  for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
+       yoffset++) {
+    for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row;
+	 MCU_col_num++) {
+      /* Construct list of pointers to DCT blocks belonging to this MCU */
+      blkn = 0;			/* index of current DCT block within MCU */
+      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+	compptr = cinfo->cur_comp_info[ci];
+	start_col = MCU_col_num * compptr->MCU_width;
+	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+	    coef->MCU_buffer[blkn++] = buffer_ptr++;
+	  }
+	}
+      }
+      /* Try to fetch the MCU. */
+      if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
+	/* Suspension forced; update state counters and exit */
+	coef->MCU_vert_offset = yoffset;
+	coef->MCU_ctr = MCU_col_num;
+	return JPEG_SUSPENDED;
+      }
+    }
+    /* Completed an MCU row, but perhaps not an iMCU row */
+    coef->MCU_ctr = 0;
+  }
+  /* Completed the iMCU row, advance counters for next one */
+  if (++(cinfo->input_iMCU_row) < cinfo->total_iMCU_rows) {
+    start_iMCU_row(cinfo);
+    return JPEG_ROW_COMPLETED;
+  }
+  /* Completed the scan */
+  (*cinfo->inputctl->finish_input_pass) (cinfo);
+  return JPEG_SCAN_COMPLETED;
+}
+
+
+/*
+ * Decompress and return some data in the multi-pass case.
+ * Always attempts to emit one fully interleaved MCU row ("iMCU" row).
+ * Return value is JPEG_ROW_COMPLETED, JPEG_SCAN_COMPLETED, or JPEG_SUSPENDED.
+ *
+ * NB: output_buf contains a plane for each component in image.
+ */
+
+METHODDEF(int)
+decompress_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION block_num;
+  int ci, block_row, block_rows;
+  JBLOCKARRAY buffer;
+  JBLOCKROW buffer_ptr;
+  JSAMPARRAY output_ptr;
+  JDIMENSION output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+
+  /* Force some input to be done if we are getting ahead of the input. */
+  while (cinfo->input_scan_number < cinfo->output_scan_number ||
+	 (cinfo->input_scan_number == cinfo->output_scan_number &&
+	  cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
+    if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
+      return JPEG_SUSPENDED;
+  }
+
+  /* OK, output from the virtual arrays. */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Don't bother to IDCT an uninteresting component. */
+    if (! compptr->component_needed)
+      continue;
+    /* Align the virtual buffer for this component. */
+    buffer = (*cinfo->mem->access_virt_barray)
+      ((j_common_ptr) cinfo, coef->whole_image[ci],
+       cinfo->output_iMCU_row * compptr->v_samp_factor,
+       (JDIMENSION) compptr->v_samp_factor, FALSE);
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (cinfo->output_iMCU_row < last_iMCU_row)
+      block_rows = compptr->v_samp_factor;
+    else {
+      /* NB: can't use last_row_height here; it is input-side-dependent! */
+      block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+    }
+    inverse_DCT = cinfo->idct->inverse_DCT[ci];
+    output_ptr = output_buf[ci];
+    /* Loop over all DCT blocks to be processed. */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      buffer_ptr = buffer[block_row];
+      output_col = 0;
+      for (block_num = 0; block_num < compptr->width_in_blocks; block_num++) {
+	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr,
+			output_ptr, output_col);
+	buffer_ptr++;
+	output_col += compptr->DCT_scaled_size;
+      }
+      output_ptr += compptr->DCT_scaled_size;
+    }
+  }
+
+  if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows)
+    return JPEG_ROW_COMPLETED;
+  return JPEG_SCAN_COMPLETED;
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+
+/*
+ * This code applies interblock smoothing as described by section K.8
+ * of the JPEG standard: the first 5 AC coefficients are estimated from
+ * the DC values of a DCT block and its 8 neighboring blocks.
+ * We apply smoothing only for progressive JPEG decoding, and only if
+ * the coefficients it can estimate are not yet known to full precision.
+ */
+
+/* Natural-order array positions of the first 5 zigzag-order coefficients */
+#define Q01_POS  1
+#define Q10_POS  8
+#define Q20_POS  16
+#define Q11_POS  9
+#define Q02_POS  2
+
+/*
+ * Determine whether block smoothing is applicable and safe.
+ * We also latch the current states of the coef_bits[] entries for the
+ * AC coefficients; otherwise, if the input side of the decompressor
+ * advances into a new scan, we might think the coefficients are known
+ * more accurately than they really are.
+ */
+
+LOCAL(boolean)
+smoothing_ok (j_decompress_ptr cinfo)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  boolean smoothing_useful = FALSE;
+  int ci, coefi;
+  jpeg_component_info *compptr;
+  JQUANT_TBL * qtable;
+  int * coef_bits;
+  int * coef_bits_latch;
+
+  if (! cinfo->progressive_mode || cinfo->coef_bits == NULL)
+    return FALSE;
+
+  /* Allocate latch area if not already done */
+  if (coef->coef_bits_latch == NULL)
+    coef->coef_bits_latch = (int *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  cinfo->num_components *
+				  (SAVED_COEFS * SIZEOF(int)));
+  coef_bits_latch = coef->coef_bits_latch;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* All components' quantization values must already be latched. */
+    if ((qtable = compptr->quant_table) == NULL)
+      return FALSE;
+    /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */
+    if (qtable->quantval[0] == 0 ||
+	qtable->quantval[Q01_POS] == 0 ||
+	qtable->quantval[Q10_POS] == 0 ||
+	qtable->quantval[Q20_POS] == 0 ||
+	qtable->quantval[Q11_POS] == 0 ||
+	qtable->quantval[Q02_POS] == 0)
+      return FALSE;
+    /* DC values must be at least partly known for all components. */
+    coef_bits = cinfo->coef_bits[ci];
+    if (coef_bits[0] < 0)
+      return FALSE;
+    /* Block smoothing is helpful if some AC coefficients remain inaccurate. */
+    for (coefi = 1; coefi <= 5; coefi++) {
+      coef_bits_latch[coefi] = coef_bits[coefi];
+      if (coef_bits[coefi] != 0)
+	smoothing_useful = TRUE;
+    }
+    coef_bits_latch += SAVED_COEFS;
+  }
+
+  return smoothing_useful;
+}
+
+
+/*
+ * Variant of decompress_data for use when doing block smoothing.
+ */
+
+METHODDEF(int)
+decompress_smooth_data (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
+  JDIMENSION block_num, last_block_column;
+  int ci, block_row, block_rows, access_rows;
+  JBLOCKARRAY buffer;
+  JBLOCKROW buffer_ptr, prev_block_row, next_block_row;
+  JSAMPARRAY output_ptr;
+  JDIMENSION output_col;
+  jpeg_component_info *compptr;
+  inverse_DCT_method_ptr inverse_DCT;
+  boolean first_row, last_row;
+  JBLOCK workspace;
+  int *coef_bits;
+  JQUANT_TBL *quanttbl;
+  INT32 Q00,Q01,Q02,Q10,Q11,Q20, num;
+  int DC1,DC2,DC3,DC4,DC5,DC6,DC7,DC8,DC9;
+  int Al, pred;
+
+  /* Force some input to be done if we are getting ahead of the input. */
+  while (cinfo->input_scan_number <= cinfo->output_scan_number &&
+	 ! cinfo->inputctl->eoi_reached) {
+    if (cinfo->input_scan_number == cinfo->output_scan_number) {
+      /* If input is working on current scan, we ordinarily want it to
+       * have completed the current row.  But if input scan is DC,
+       * we want it to keep one row ahead so that next block row's DC
+       * values are up to date.
+       */
+      JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
+      if (cinfo->input_iMCU_row > cinfo->output_iMCU_row+delta)
+	break;
+    }
+    if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
+      return JPEG_SUSPENDED;
+  }
+
+  /* OK, output from the virtual arrays. */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Don't bother to IDCT an uninteresting component. */
+    if (! compptr->component_needed)
+      continue;
+    /* Count non-dummy DCT block rows in this iMCU row. */
+    if (cinfo->output_iMCU_row < last_iMCU_row) {
+      block_rows = compptr->v_samp_factor;
+      access_rows = block_rows * 2; /* this and next iMCU row */
+      last_row = FALSE;
+    } else {
+      /* NB: can't use last_row_height here; it is input-side-dependent! */
+      block_rows = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+      if (block_rows == 0) block_rows = compptr->v_samp_factor;
+      access_rows = block_rows; /* this iMCU row only */
+      last_row = TRUE;
+    }
+    /* Align the virtual buffer for this component. */
+    if (cinfo->output_iMCU_row > 0) {
+      access_rows += compptr->v_samp_factor; /* prior iMCU row too */
+      buffer = (*cinfo->mem->access_virt_barray)
+	((j_common_ptr) cinfo, coef->whole_image[ci],
+	 (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
+	 (JDIMENSION) access_rows, FALSE);
+      buffer += compptr->v_samp_factor;	/* point to current iMCU row */
+      first_row = FALSE;
+    } else {
+      buffer = (*cinfo->mem->access_virt_barray)
+	((j_common_ptr) cinfo, coef->whole_image[ci],
+	 (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE);
+      first_row = TRUE;
+    }
+    /* Fetch component-dependent info */
+    coef_bits = coef->coef_bits_latch + (ci * SAVED_COEFS);
+    quanttbl = compptr->quant_table;
+    Q00 = quanttbl->quantval[0];
+    Q01 = quanttbl->quantval[Q01_POS];
+    Q10 = quanttbl->quantval[Q10_POS];
+    Q20 = quanttbl->quantval[Q20_POS];
+    Q11 = quanttbl->quantval[Q11_POS];
+    Q02 = quanttbl->quantval[Q02_POS];
+    inverse_DCT = cinfo->idct->inverse_DCT[ci];
+    output_ptr = output_buf[ci];
+    /* Loop over all DCT blocks to be processed. */
+    for (block_row = 0; block_row < block_rows; block_row++) {
+      buffer_ptr = buffer[block_row];
+      if (first_row && block_row == 0)
+	prev_block_row = buffer_ptr;
+      else
+	prev_block_row = buffer[block_row-1];
+      if (last_row && block_row == block_rows-1)
+	next_block_row = buffer_ptr;
+      else
+	next_block_row = buffer[block_row+1];
+      /* We fetch the surrounding DC values using a sliding-register approach.
+       * Initialize all nine here so as to do the right thing on narrow pics.
+       */
+      DC1 = DC2 = DC3 = (int) prev_block_row[0][0];
+      DC4 = DC5 = DC6 = (int) buffer_ptr[0][0];
+      DC7 = DC8 = DC9 = (int) next_block_row[0][0];
+      output_col = 0;
+      last_block_column = compptr->width_in_blocks - 1;
+      for (block_num = 0; block_num <= last_block_column; block_num++) {
+	/* Fetch current DCT block into workspace so we can modify it. */
+	jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1);
+	/* Update DC values */
+	if (block_num < last_block_column) {
+	  DC3 = (int) prev_block_row[1][0];
+	  DC6 = (int) buffer_ptr[1][0];
+	  DC9 = (int) next_block_row[1][0];
+	}
+	/* Compute coefficient estimates per K.8.
+	 * An estimate is applied only if coefficient is still zero,
+	 * and is not known to be fully accurate.
+	 */
+	/* AC01 */
+	if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) {
+	  num = 36 * Q00 * (DC4 - DC6);
+	  if (num >= 0) {
+	    pred = (int) (((Q01<<7) + num) / (Q01<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q01<<7) - num) / (Q01<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[1] = (JCOEF) pred;
+	}
+	/* AC10 */
+	if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) {
+	  num = 36 * Q00 * (DC2 - DC8);
+	  if (num >= 0) {
+	    pred = (int) (((Q10<<7) + num) / (Q10<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q10<<7) - num) / (Q10<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[8] = (JCOEF) pred;
+	}
+	/* AC20 */
+	if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) {
+	  num = 9 * Q00 * (DC2 + DC8 - 2*DC5);
+	  if (num >= 0) {
+	    pred = (int) (((Q20<<7) + num) / (Q20<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q20<<7) - num) / (Q20<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[16] = (JCOEF) pred;
+	}
+	/* AC11 */
+	if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) {
+	  num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
+	  if (num >= 0) {
+	    pred = (int) (((Q11<<7) + num) / (Q11<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q11<<7) - num) / (Q11<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[9] = (JCOEF) pred;
+	}
+	/* AC02 */
+	if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) {
+	  num = 9 * Q00 * (DC4 + DC6 - 2*DC5);
+	  if (num >= 0) {
+	    pred = (int) (((Q02<<7) + num) / (Q02<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	  } else {
+	    pred = (int) (((Q02<<7) - num) / (Q02<<8));
+	    if (Al > 0 && pred >= (1<<Al))
+	      pred = (1<<Al)-1;
+	    pred = -pred;
+	  }
+	  workspace[2] = (JCOEF) pred;
+	}
+	/* OK, do the IDCT */
+	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace,
+			output_ptr, output_col);
+	/* Advance for next column */
+	DC1 = DC2; DC2 = DC3;
+	DC4 = DC5; DC5 = DC6;
+	DC7 = DC8; DC8 = DC9;
+	buffer_ptr++, prev_block_row++, next_block_row++;
+	output_col += compptr->DCT_scaled_size;
+      }
+      output_ptr += compptr->DCT_scaled_size;
+    }
+  }
+
+  if (++(cinfo->output_iMCU_row) < cinfo->total_iMCU_rows)
+    return JPEG_ROW_COMPLETED;
+  return JPEG_SCAN_COMPLETED;
+}
+
+#endif /* BLOCK_SMOOTHING_SUPPORTED */
+
+
+/*
+ * Initialize coefficient buffer controller.
+ */
+
+GLOBAL(void)
+jinit_d_coef_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_coef_ptr coef;
+
+  coef = (my_coef_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_coef_controller));
+  cinfo->coef = (struct jpeg_d_coef_controller *) coef;
+  coef->pub.start_input_pass = start_input_pass;
+  coef->pub.start_output_pass = start_output_pass;
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  coef->coef_bits_latch = NULL;
+#endif
+
+  /* Create the coefficient buffer. */
+  if (need_full_buffer) {
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+    /* Allocate a full-image virtual array for each component, */
+    /* padded to a multiple of samp_factor DCT blocks in each direction. */
+    /* Note we ask for a pre-zeroed array. */
+    int ci, access_rows;
+    jpeg_component_info *compptr;
+
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      access_rows = compptr->v_samp_factor;
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+      /* If block smoothing could be used, need a bigger window */
+      if (cinfo->progressive_mode)
+	access_rows *= 3;
+#endif
+      coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
+	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+				(long) compptr->h_samp_factor),
+	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+				(long) compptr->v_samp_factor),
+	 (JDIMENSION) access_rows);
+    }
+    coef->pub.consume_data = consume_data;
+    coef->pub.decompress_data = decompress_data;
+    coef->pub.coef_arrays = coef->whole_image; /* link to virtual arrays */
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+  } else {
+    /* We only need a single-MCU buffer. */
+    JBLOCKROW buffer;
+    int i;
+
+    buffer = (JBLOCKROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+    for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) {
+      coef->MCU_buffer[i] = buffer + i;
+    }
+    coef->pub.consume_data = dummy_consume_data;
+    coef->pub.decompress_data = decompress_onepass;
+    coef->pub.coef_arrays = NULL; /* flag for no virtual arrays */
+  }
+}
diff --git a/jpeg/jdcolor.c b/jpeg/jdcolor.c
new file mode 100644
index 000000000..6c04dfe8a
--- /dev/null
+++ b/jpeg/jdcolor.c
@@ -0,0 +1,396 @@
+/*
+ * jdcolor.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_deconverter pub; /* public fields */
+
+  /* Private state for YCC->RGB conversion */
+  int * Cr_r_tab;		/* => table for Cr to R conversion */
+  int * Cb_b_tab;		/* => table for Cb to B conversion */
+  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
+  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+} my_color_deconverter;
+
+typedef my_color_deconverter * my_cconvert_ptr;
+
+
+/**************** YCbCr -> RGB conversion: most common case **************/
+
+/*
+ * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
+ * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
+ * The conversion equations to be implemented are therefore
+ *	R = Y                + 1.40200 * Cr
+ *	G = Y - 0.34414 * Cb - 0.71414 * Cr
+ *	B = Y + 1.77200 * Cb
+ * where Cb and Cr represent the incoming values less CENTERJSAMPLE.
+ * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
+ *
+ * To avoid floating-point arithmetic, we represent the fractional constants
+ * as integers scaled up by 2^16 (about 4 digits precision); we have to divide
+ * the products by 2^16, with appropriate rounding, to get the correct answer.
+ * Notice that Y, being an integral input, does not contribute any fraction
+ * so it need not participate in the rounding.
+ *
+ * For even more speed, we avoid doing any multiplications in the inner loop
+ * by precalculating the constants times Cb and Cr for all possible values.
+ * For 8-bit JSAMPLEs this is very reasonable (only 256 entries per table);
+ * for 12-bit samples it is still acceptable.  It's not very reasonable for
+ * 16-bit samples, but if you want lossless storage you shouldn't be changing
+ * colorspace anyway.
+ * The Cr=>R and Cb=>B values can be rounded to integers in advance; the
+ * values for the G calculation are left scaled up, since we must add them
+ * together before rounding.
+ */
+
+#define SCALEBITS	16	/* speediest right-shift on some machines */
+#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
+#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+
+
+/*
+ * Initialize tables for YCC->RGB colorspace conversion.
+ */
+
+LOCAL(void)
+build_ycc_rgb_table (j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  int i;
+  INT32 x;
+  SHIFT_TEMPS
+
+  cconvert->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  cconvert->Cr_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+  cconvert->Cb_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+
+  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
+    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
+    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
+    /* Cr=>R value is nearest int to 1.40200 * x */
+    cconvert->Cr_r_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+    /* Cb=>B value is nearest int to 1.77200 * x */
+    cconvert->Cb_b_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+    /* Cr=>G value is scaled-up -0.71414 * x */
+    cconvert->Cr_g_tab[i] = (- FIX(0.71414)) * x;
+    /* Cb=>G value is scaled-up -0.34414 * x */
+    /* We also add in ONE_HALF so that need not do it in inner loop */
+    cconvert->Cb_g_tab[i] = (- FIX(0.34414)) * x + ONE_HALF;
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ *
+ * Note that we change from noninterleaved, one-plane-per-component format
+ * to interleaved-pixel format.  The output buffer is therefore three times
+ * as wide as the input buffer.
+ * A starting row offset is provided only for the input buffer.  The caller
+ * can easily adjust the passed output_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+METHODDEF(void)
+ycc_rgb_convert (j_decompress_ptr cinfo,
+		 JSAMPIMAGE input_buf, JDIMENSION input_row,
+		 JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
+      outptr[RGB_GREEN] = range_limit[y +
+			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+						 SCALEBITS))];
+      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/**************** Cases other than YCbCr -> RGB **************/
+
+
+/*
+ * Color conversion for no colorspace change: just copy the data,
+ * converting from separate-planes to interleaved representation.
+ */
+
+METHODDEF(void)
+null_convert (j_decompress_ptr cinfo,
+	      JSAMPIMAGE input_buf, JDIMENSION input_row,
+	      JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION count;
+  register int num_components = cinfo->num_components;
+  JDIMENSION num_cols = cinfo->output_width;
+  int ci;
+
+  while (--num_rows >= 0) {
+    for (ci = 0; ci < num_components; ci++) {
+      inptr = input_buf[ci][input_row];
+      outptr = output_buf[0] + ci;
+      for (count = num_cols; count > 0; count--) {
+	*outptr = *inptr++;	/* needn't bother with GETJSAMPLE() here */
+	outptr += num_components;
+      }
+    }
+    input_row++;
+    output_buf++;
+  }
+}
+
+
+/*
+ * Color conversion for grayscale: just copy the data.
+ * This also works for YCbCr -> grayscale conversion, in which
+ * we just copy the Y (luminance) component and ignore chrominance.
+ */
+
+METHODDEF(void)
+grayscale_convert (j_decompress_ptr cinfo,
+		   JSAMPIMAGE input_buf, JDIMENSION input_row,
+		   JSAMPARRAY output_buf, int num_rows)
+{
+  jcopy_sample_rows(input_buf[0], (int) input_row, output_buf, 0,
+		    num_rows, cinfo->output_width);
+}
+
+
+/*
+ * Convert grayscale to RGB: just duplicate the graylevel three times.
+ * This is provided to support applications that don't want to cope
+ * with grayscale as a separate case.
+ */
+
+METHODDEF(void)
+gray_rgb_convert (j_decompress_ptr cinfo,
+		  JSAMPIMAGE input_buf, JDIMENSION input_row,
+		  JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Adobe-style YCCK->CMYK conversion.
+ * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same
+ * conversion as above, while passing K (black) unchanged.
+ * We assume build_ycc_rgb_table has been called.
+ */
+
+METHODDEF(void)
+ycck_cmyk_convert (j_decompress_ptr cinfo,
+		   JSAMPIMAGE input_buf, JDIMENSION input_row,
+		   JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2, inptr3;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    inptr3 = input_buf[3][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];	/* red */
+      outptr[1] = range_limit[MAXJSAMPLE - (y +			/* green */
+			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+						 SCALEBITS)))];
+      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];	/* blue */
+      /* K passes through unchanged */
+      outptr[3] = inptr3[col];	/* don't need GETJSAMPLE here */
+      outptr += 4;
+    }
+  }
+}
+
+
+/*
+ * Empty method for start_pass.
+ */
+
+METHODDEF(void)
+start_pass_dcolor (j_decompress_ptr cinfo)
+{
+  /* no work needed */
+}
+
+
+/*
+ * Module initialization routine for output colorspace conversion.
+ */
+
+GLOBAL(void)
+jinit_color_deconverter (j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert;
+  int ci;
+
+  cconvert = (my_cconvert_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_color_deconverter));
+  cinfo->cconvert = (struct jpeg_color_deconverter *) cconvert;
+  cconvert->pub.start_pass = start_pass_dcolor;
+
+  /* Make sure num_components agrees with jpeg_color_space */
+  switch (cinfo->jpeg_color_space) {
+  case JCS_GRAYSCALE:
+    if (cinfo->num_components != 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  case JCS_RGB:
+  case JCS_YCbCr:
+    if (cinfo->num_components != 3)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  case JCS_CMYK:
+  case JCS_YCCK:
+    if (cinfo->num_components != 4)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+
+  default:			/* JCS_UNKNOWN can be anything */
+    if (cinfo->num_components < 1)
+      ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
+    break;
+  }
+
+  /* Set out_color_components and conversion method based on requested space.
+   * Also clear the component_needed flags for any unused components,
+   * so that earlier pipeline stages can avoid useless computation.
+   */
+
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    cinfo->out_color_components = 1;
+    if (cinfo->jpeg_color_space == JCS_GRAYSCALE ||
+	cinfo->jpeg_color_space == JCS_YCbCr) {
+      cconvert->pub.color_convert = grayscale_convert;
+      /* For color->grayscale conversion, only the Y (0) component is needed */
+      for (ci = 1; ci < cinfo->num_components; ci++)
+	cinfo->comp_info[ci].component_needed = FALSE;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_RGB:
+    cinfo->out_color_components = RGB_PIXELSIZE;
+    if (cinfo->jpeg_color_space == JCS_YCbCr) {
+      cconvert->pub.color_convert = ycc_rgb_convert;
+      build_ycc_rgb_table(cinfo);
+    } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+      cconvert->pub.color_convert = gray_rgb_convert;
+    } else if (cinfo->jpeg_color_space == JCS_RGB && RGB_PIXELSIZE == 3) {
+      cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  case JCS_CMYK:
+    cinfo->out_color_components = 4;
+    if (cinfo->jpeg_color_space == JCS_YCCK) {
+      cconvert->pub.color_convert = ycck_cmyk_convert;
+      build_ycc_rgb_table(cinfo);
+    } else if (cinfo->jpeg_color_space == JCS_CMYK) {
+      cconvert->pub.color_convert = null_convert;
+    } else
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+
+  default:
+    /* Permit null conversion to same output space */
+    if (cinfo->out_color_space == cinfo->jpeg_color_space) {
+      cinfo->out_color_components = cinfo->num_components;
+      cconvert->pub.color_convert = null_convert;
+    } else			/* unsupported non-null conversion */
+      ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    break;
+  }
+
+  if (cinfo->quantize_colors)
+    cinfo->output_components = 1; /* single colormapped output component */
+  else
+    cinfo->output_components = cinfo->out_color_components;
+}
diff --git a/jpeg/jdct.h b/jpeg/jdct.h
new file mode 100644
index 000000000..04192a266
--- /dev/null
+++ b/jpeg/jdct.h
@@ -0,0 +1,176 @@
+/*
+ * jdct.h
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This include file contains common declarations for the forward and
+ * inverse DCT modules.  These declarations are private to the DCT managers
+ * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
+ * The individual DCT algorithms are kept in separate files to ease 
+ * machine-dependent tuning (e.g., assembly coding).
+ */
+
+
+/*
+ * A forward DCT routine is given a pointer to a work area of type DCTELEM[];
+ * the DCT is to be performed in-place in that buffer.  Type DCTELEM is int
+ * for 8-bit samples, INT32 for 12-bit samples.  (NOTE: Floating-point DCT
+ * implementations use an array of type FAST_FLOAT, instead.)
+ * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE).
+ * The DCT outputs are returned scaled up by a factor of 8; they therefore
+ * have a range of +-8K for 8-bit data, +-128K for 12-bit data.  This
+ * convention improves accuracy in integer implementations and saves some
+ * work in floating-point ones.
+ * Quantization of the output coefficients is done by jcdctmgr.c.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef int DCTELEM;		/* 16 or 32 bits is fine */
+#else
+typedef INT32 DCTELEM;		/* must have 32 bits */
+#endif
+
+typedef JMETHOD(void, forward_DCT_method_ptr, (DCTELEM * data));
+typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data));
+
+
+/*
+ * An inverse DCT routine is given a pointer to the input JBLOCK and a pointer
+ * to an output sample array.  The routine must dequantize the input data as
+ * well as perform the IDCT; for dequantization, it uses the multiplier table
+ * pointed to by compptr->dct_table.  The output data is to be placed into the
+ * sample array starting at a specified column.  (Any row offset needed will
+ * be applied to the array pointer before it is passed to the IDCT code.)
+ * Note that the number of samples emitted by the IDCT routine is
+ * DCT_scaled_size * DCT_scaled_size.
+ */
+
+/* typedef inverse_DCT_method_ptr is declared in jpegint.h */
+
+/*
+ * Each IDCT routine has its own ideas about the best dct_table element type.
+ */
+
+typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
+#if BITS_IN_JSAMPLE == 8
+typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
+#define IFAST_SCALE_BITS  2	/* fractional bits in scale factors */
+#else
+typedef INT32 IFAST_MULT_TYPE;	/* need 32 bits for scaled quantizers */
+#define IFAST_SCALE_BITS  13	/* fractional bits in scale factors */
+#endif
+typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
+
+
+/*
+ * Each IDCT routine is responsible for range-limiting its results and
+ * converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
+ * be quite far out of range if the input data is corrupt, so a bulletproof
+ * range-limiting step is required.  We use a mask-and-table-lookup method
+ * to do the combined operations quickly.  See the comments with
+ * prepare_range_limit_table (in jdmaster.c) for more info.
+ */
+
+#define IDCT_range_limit(cinfo)  ((cinfo)->sample_range_limit + CENTERJSAMPLE)
+
+#define RANGE_MASK  (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_fdct_islow		jFDislow
+#define jpeg_fdct_ifast		jFDifast
+#define jpeg_fdct_float		jFDfloat
+#define jpeg_idct_islow		jRDislow
+#define jpeg_idct_ifast		jRDifast
+#define jpeg_idct_float		jRDfloat
+#define jpeg_idct_4x4		jRD4x4
+#define jpeg_idct_2x2		jRD2x2
+#define jpeg_idct_1x1		jRD1x1
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+/* Extern declarations for the forward and inverse DCT routines. */
+
+EXTERN(void) jpeg_fdct_islow JPP((DCTELEM * data));
+EXTERN(void) jpeg_fdct_ifast JPP((DCTELEM * data));
+EXTERN(void) jpeg_fdct_float JPP((FAST_FLOAT * data));
+
+EXTERN(void) jpeg_idct_islow
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_ifast
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_float
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_4x4
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_2x2
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+EXTERN(void) jpeg_idct_1x1
+    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+
+
+/*
+ * Macros for handling fixed-point arithmetic; these are used by many
+ * but not all of the DCT/IDCT modules.
+ *
+ * All values are expected to be of type INT32.
+ * Fractional constants are scaled left by CONST_BITS bits.
+ * CONST_BITS is defined within each module using these macros,
+ * and may differ from one module to the next.
+ */
+
+#define ONE	((INT32) 1)
+#define CONST_SCALE (ONE << CONST_BITS)
+
+/* Convert a positive real constant to an integer scaled by CONST_SCALE.
+ * Caution: some C compilers fail to reduce "FIX(constant)" at compile time,
+ * thus causing a lot of useless floating-point operations at run time.
+ */
+
+#define FIX(x)	((INT32) ((x) * CONST_SCALE + 0.5))
+
+/* Descale and correctly round an INT32 value that's scaled by N bits.
+ * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
+ * the fudge factor is correct for either sign of X.
+ */
+
+#define DESCALE(x,n)  RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * This macro is used only when the two inputs will actually be no more than
+ * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
+ * full 32x32 multiply.  This provides a useful speedup on many machines.
+ * Unfortunately there is no way to specify a 16x16->32 multiply portably
+ * in C, but some C compilers will do the right thing if you provide the
+ * correct combination of casts.
+ */
+
+#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT16) (const)))
+#endif
+#ifdef SHORTxLCONST_32		/* known to work with Microsoft C 6.0 */
+#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT32) (const)))
+#endif
+
+#ifndef MULTIPLY16C16		/* default definition */
+#define MULTIPLY16C16(var,const)  ((var) * (const))
+#endif
+
+/* Same except both inputs are variables. */
+
+#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#define MULTIPLY16V16(var1,var2)  (((INT16) (var1)) * ((INT16) (var2)))
+#endif
+
+#ifndef MULTIPLY16V16		/* default definition */
+#define MULTIPLY16V16(var1,var2)  ((var1) * (var2))
+#endif
diff --git a/jpeg/jddctmgr.c b/jpeg/jddctmgr.c
new file mode 100644
index 000000000..bbf8d0e92
--- /dev/null
+++ b/jpeg/jddctmgr.c
@@ -0,0 +1,269 @@
+/*
+ * jddctmgr.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the inverse-DCT management logic.
+ * This code selects a particular IDCT implementation to be used,
+ * and it performs related housekeeping chores.  No code in this file
+ * is executed per IDCT step, only during output pass setup.
+ *
+ * Note that the IDCT routines are responsible for performing coefficient
+ * dequantization as well as the IDCT proper.  This module sets up the
+ * dequantization multiplier table needed by the IDCT routine.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+
+/*
+ * The decompressor input side (jdinput.c) saves away the appropriate
+ * quantization table for each component at the start of the first scan
+ * involving that component.  (This is necessary in order to correctly
+ * decode files that reuse Q-table slots.)
+ * When we are ready to make an output pass, the saved Q-table is converted
+ * to a multiplier table that will actually be used by the IDCT routine.
+ * The multiplier table contents are IDCT-method-dependent.  To support
+ * application changes in IDCT method between scans, we can remake the
+ * multiplier tables if necessary.
+ * In buffered-image mode, the first output pass may occur before any data
+ * has been seen for some components, and thus before their Q-tables have
+ * been saved away.  To handle this case, multiplier tables are preset
+ * to zeroes; the result of the IDCT will be a neutral gray level.
+ */
+
+
+/* Private subobject for this module */
+
+typedef struct {
+  struct jpeg_inverse_dct pub;	/* public fields */
+
+  /* This array contains the IDCT method code that each multiplier table
+   * is currently set up for, or -1 if it's not yet set up.
+   * The actual multiplier tables are pointed to by dct_table in the
+   * per-component comp_info structures.
+   */
+  int cur_method[MAX_COMPONENTS];
+} my_idct_controller;
+
+typedef my_idct_controller * my_idct_ptr;
+
+
+/* Allocated multiplier tables: big enough for any supported variant */
+
+typedef union {
+  ISLOW_MULT_TYPE islow_array[DCTSIZE2];
+#ifdef DCT_IFAST_SUPPORTED
+  IFAST_MULT_TYPE ifast_array[DCTSIZE2];
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  FLOAT_MULT_TYPE float_array[DCTSIZE2];
+#endif
+} multiplier_table;
+
+
+/* The current scaled-IDCT routines require ISLOW-style multiplier tables,
+ * so be sure to compile that code if either ISLOW or SCALING is requested.
+ */
+#ifdef DCT_ISLOW_SUPPORTED
+#define PROVIDE_ISLOW_TABLES
+#else
+#ifdef IDCT_SCALING_SUPPORTED
+#define PROVIDE_ISLOW_TABLES
+#endif
+#endif
+
+
+/*
+ * Prepare for an output pass.
+ * Here we select the proper IDCT routine for each component and build
+ * a matching multiplier table.
+ */
+
+METHODDEF(void)
+start_pass (j_decompress_ptr cinfo)
+{
+  my_idct_ptr idct = (my_idct_ptr) cinfo->idct;
+  int ci, i;
+  jpeg_component_info *compptr;
+  int method = 0;
+  inverse_DCT_method_ptr method_ptr = NULL;
+  JQUANT_TBL * qtbl;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Select the proper IDCT routine for this component's scaling */
+    switch (compptr->DCT_scaled_size) {
+#ifdef IDCT_SCALING_SUPPORTED
+    case 1:
+      method_ptr = jpeg_idct_1x1;
+      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      break;
+    case 2:
+      method_ptr = jpeg_idct_2x2;
+      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      break;
+    case 4:
+      method_ptr = jpeg_idct_4x4;
+      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      break;
+#endif
+    case DCTSIZE:
+      switch (cinfo->dct_method) {
+#ifdef DCT_ISLOW_SUPPORTED
+      case JDCT_ISLOW:
+	method_ptr = jpeg_idct_islow;
+	method = JDCT_ISLOW;
+	break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+      case JDCT_IFAST:
+	method_ptr = jpeg_idct_ifast;
+	method = JDCT_IFAST;
+	break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+      case JDCT_FLOAT:
+	method_ptr = jpeg_idct_float;
+	method = JDCT_FLOAT;
+	break;
+#endif
+      default:
+	ERREXIT(cinfo, JERR_NOT_COMPILED);
+	break;
+      }
+      break;
+    default:
+      ERREXIT1(cinfo, JERR_BAD_DCTSIZE, compptr->DCT_scaled_size);
+      break;
+    }
+    idct->pub.inverse_DCT[ci] = method_ptr;
+    /* Create multiplier table from quant table.
+     * However, we can skip this if the component is uninteresting
+     * or if we already built the table.  Also, if no quant table
+     * has yet been saved for the component, we leave the
+     * multiplier table all-zero; we'll be reading zeroes from the
+     * coefficient controller's buffer anyway.
+     */
+    if (! compptr->component_needed || idct->cur_method[ci] == method)
+      continue;
+    qtbl = compptr->quant_table;
+    if (qtbl == NULL)		/* happens if no data yet for component */
+      continue;
+    idct->cur_method[ci] = method;
+    switch (method) {
+#ifdef PROVIDE_ISLOW_TABLES
+    case JDCT_ISLOW:
+      {
+	/* For LL&M IDCT method, multipliers are equal to raw quantization
+	 * coefficients, but are stored as ints to ensure access efficiency.
+	 */
+	ISLOW_MULT_TYPE * ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table;
+	for (i = 0; i < DCTSIZE2; i++) {
+	  ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i];
+	}
+      }
+      break;
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+    case JDCT_IFAST:
+      {
+	/* For AA&N IDCT method, multipliers are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 * For integer operation, the multiplier table is to be scaled by
+	 * IFAST_SCALE_BITS.
+	 */
+	IFAST_MULT_TYPE * ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table;
+#define CONST_BITS 14
+	static const INT16 aanscales[DCTSIZE2] = {
+	  /* precomputed values scaled up by 14 bits */
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+	};
+	SHIFT_TEMPS
+
+	for (i = 0; i < DCTSIZE2; i++) {
+	  ifmtbl[i] = (IFAST_MULT_TYPE)
+	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
+				  (INT32) aanscales[i]),
+		    CONST_BITS-IFAST_SCALE_BITS);
+	}
+      }
+      break;
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+    case JDCT_FLOAT:
+      {
+	/* For float AA&N IDCT method, multipliers are equal to quantization
+	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
+	 *   scalefactor[0] = 1
+	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+	 */
+	FLOAT_MULT_TYPE * fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table;
+	int row, col;
+	static const double aanscalefactor[DCTSIZE] = {
+	  1.0, 1.387039845, 1.306562965, 1.175875602,
+	  1.0, 0.785694958, 0.541196100, 0.275899379
+	};
+
+	i = 0;
+	for (row = 0; row < DCTSIZE; row++) {
+	  for (col = 0; col < DCTSIZE; col++) {
+	    fmtbl[i] = (FLOAT_MULT_TYPE)
+	      ((double) qtbl->quantval[i] *
+	       aanscalefactor[row] * aanscalefactor[col]);
+	    i++;
+	  }
+	}
+      }
+      break;
+#endif
+    default:
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+      break;
+    }
+  }
+}
+
+
+/*
+ * Initialize IDCT manager.
+ */
+
+GLOBAL(void)
+jinit_inverse_dct (j_decompress_ptr cinfo)
+{
+  my_idct_ptr idct;
+  int ci;
+  jpeg_component_info *compptr;
+
+  idct = (my_idct_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_idct_controller));
+  cinfo->idct = (struct jpeg_inverse_dct *) idct;
+  idct->pub.start_pass = start_pass;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Allocate and pre-zero a multiplier table for each component */
+    compptr->dct_table =
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(multiplier_table));
+    MEMZERO(compptr->dct_table, SIZEOF(multiplier_table));
+    /* Mark multiplier table not yet set up for any method */
+    idct->cur_method[ci] = -1;
+  }
+}
diff --git a/jpeg/jdhuff.c b/jpeg/jdhuff.c
new file mode 100644
index 000000000..b5ba39f73
--- /dev/null
+++ b/jpeg/jdhuff.c
@@ -0,0 +1,651 @@
+/*
+ * jdhuff.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy decoding routines.
+ *
+ * Much of the complexity here has to do with supporting input suspension.
+ * If the data source module demands suspension, we want to be able to back
+ * up to the start of the current MCU.  To do this, we copy state variables
+ * into local working storage, and update them back to the permanent
+ * storage only upon successful completion of an MCU.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdhuff.h"		/* Declarations shared with jdphuff.c */
+
+
+/*
+ * Expanded entropy decoder object for Huffman decoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
+} savable_state;
+
+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest,src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest,src)  \
+	((dest).last_dc_val[0] = (src).last_dc_val[0], \
+	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
+	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
+	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  /* These fields are loaded into local variables at start of each MCU.
+   * In case of suspension, we exit WITHOUT updating them.
+   */
+  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
+  savable_state saved;		/* Other state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  d_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+
+  /* Precalculated info set up by start_pass for use in decode_mcu: */
+
+  /* Pointers to derived tables to be used for each block within an MCU */
+  d_derived_tbl * dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  d_derived_tbl * ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  /* Whether we care about the DC and AC coefficient values for each block */
+  boolean dc_needed[D_MAX_BLOCKS_IN_MCU];
+  boolean ac_needed[D_MAX_BLOCKS_IN_MCU];
+} huff_entropy_decoder;
+
+typedef huff_entropy_decoder * huff_entropy_ptr;
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass_huff_decoder (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci, blkn, dctbl, actbl;
+  jpeg_component_info * compptr;
+
+  /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
+   * This ought to be an error condition, but we make it a warning because
+   * there are some baseline files out there with all zeroes in these bytes.
+   */
+  if (cinfo->Ss != 0 || cinfo->Se != DCTSIZE2-1 ||
+      cinfo->Ah != 0 || cinfo->Al != 0)
+    WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    dctbl = compptr->dc_tbl_no;
+    actbl = compptr->ac_tbl_no;
+    /* Compute derived values for Huffman tables */
+    /* We may do this more than once for a table, but it's not expensive */
+    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl,
+			    & entropy->dc_derived_tbls[dctbl]);
+    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl,
+			    & entropy->ac_derived_tbls[actbl]);
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_val[ci] = 0;
+  }
+
+  /* Precalculate decoding info for each block in an MCU of this scan */
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    ci = cinfo->MCU_membership[blkn];
+    compptr = cinfo->cur_comp_info[ci];
+    /* Precalculate which table to use for each block */
+    entropy->dc_cur_tbls[blkn] = entropy->dc_derived_tbls[compptr->dc_tbl_no];
+    entropy->ac_cur_tbls[blkn] = entropy->ac_derived_tbls[compptr->ac_tbl_no];
+    /* Decide whether we really care about the coefficient values */
+    if (compptr->component_needed) {
+      entropy->dc_needed[blkn] = TRUE;
+      /* we don't need the ACs if producing a 1/8th-size image */
+      entropy->ac_needed[blkn] = (compptr->DCT_scaled_size > 1);
+    } else {
+      entropy->dc_needed[blkn] = entropy->ac_needed[blkn] = FALSE;
+    }
+  }
+
+  /* Initialize bitread state variables */
+  entropy->bitstate.bits_left = 0;
+  entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Compute the derived values for a Huffman table.
+ * This routine also performs some validation checks on the table.
+ *
+ * Note this is also used by jdphuff.c.
+ */
+
+GLOBAL(void)
+jpeg_make_d_derived_tbl (j_decompress_ptr cinfo, boolean isDC, int tblno,
+			 d_derived_tbl ** pdtbl)
+{
+  JHUFF_TBL *htbl;
+  d_derived_tbl *dtbl;
+  int p, i, l, si, numsymbols;
+  int lookbits, ctr;
+  char huffsize[257];
+  unsigned int huffcode[257];
+  unsigned int code;
+
+  /* Note that huffsize[] and huffcode[] are filled in code-length order,
+   * paralleling the order of the symbols themselves in htbl->huffval[].
+   */
+
+  /* Find the input Huffman table */
+  if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+  htbl =
+    isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
+  if (htbl == NULL)
+    ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
+
+  /* Allocate a workspace if we haven't already done so. */
+  if (*pdtbl == NULL)
+    *pdtbl = (d_derived_tbl *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(d_derived_tbl));
+  dtbl = *pdtbl;
+  dtbl->pub = htbl;		/* fill in back link */
+  
+  /* Figure C.1: make table of Huffman code length for each symbol */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    i = (int) htbl->bits[l];
+    if (i < 0 || p + i > 256)	/* protect against table overrun */
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    while (i--)
+      huffsize[p++] = (char) l;
+  }
+  huffsize[p] = 0;
+  numsymbols = p;
+  
+  /* Figure C.2: generate the codes themselves */
+  /* We also validate that the counts represent a legal Huffman code tree. */
+  
+  code = 0;
+  si = huffsize[0];
+  p = 0;
+  while (huffsize[p]) {
+    while (((int) huffsize[p]) == si) {
+      huffcode[p++] = code;
+      code++;
+    }
+    /* code is now 1 more than the last code used for codelength si; but
+     * it must still fit in si bits, since no code is allowed to be all ones.
+     */
+    if (((INT32) code) >= (((INT32) 1) << si))
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    code <<= 1;
+    si++;
+  }
+
+  /* Figure F.15: generate decoding tables for bit-sequential decoding */
+
+  p = 0;
+  for (l = 1; l <= 16; l++) {
+    if (htbl->bits[l]) {
+      /* valoffset[l] = huffval[] index of 1st symbol of code length l,
+       * minus the minimum code of length l
+       */
+      dtbl->valoffset[l] = (INT32) p - (INT32) huffcode[p];
+      p += htbl->bits[l];
+      dtbl->maxcode[l] = huffcode[p-1]; /* maximum code of length l */
+    } else {
+      dtbl->maxcode[l] = -1;	/* -1 if no codes of this length */
+    }
+  }
+  dtbl->maxcode[17] = 0xFFFFFL; /* ensures jpeg_huff_decode terminates */
+
+  /* Compute lookahead tables to speed up decoding.
+   * First we set all the table entries to 0, indicating "too long";
+   * then we iterate through the Huffman codes that are short enough and
+   * fill in all the entries that correspond to bit sequences starting
+   * with that code.
+   */
+
+  MEMZERO(dtbl->look_nbits, SIZEOF(dtbl->look_nbits));
+
+  p = 0;
+  for (l = 1; l <= HUFF_LOOKAHEAD; l++) {
+    for (i = 1; i <= (int) htbl->bits[l]; i++, p++) {
+      /* l = current code's length, p = its index in huffcode[] & huffval[]. */
+      /* Generate left-justified code followed by all possible bit sequences */
+      lookbits = huffcode[p] << (HUFF_LOOKAHEAD-l);
+      for (ctr = 1 << (HUFF_LOOKAHEAD-l); ctr > 0; ctr--) {
+	dtbl->look_nbits[lookbits] = l;
+	dtbl->look_sym[lookbits] = htbl->huffval[p];
+	lookbits++;
+      }
+    }
+  }
+
+  /* Validate symbols as being reasonable.
+   * For AC tables, we make no check, but accept all byte values 0..255.
+   * For DC tables, we require the symbols to be in range 0..15.
+   * (Tighter bounds could be applied depending on the data depth and mode,
+   * but this is sufficient to ensure safe decoding.)
+   */
+  if (isDC) {
+    for (i = 0; i < numsymbols; i++) {
+      int sym = htbl->huffval[i];
+      if (sym < 0 || sym > 15)
+	ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+    }
+  }
+}
+
+
+/*
+ * Out-of-line code for bit fetching (shared with jdphuff.c).
+ * See jdhuff.h for info about usage.
+ * Note: current values of get_buffer and bits_left are passed as parameters,
+ * but are returned in the corresponding fields of the state struct.
+ *
+ * On most machines MIN_GET_BITS should be 25 to allow the full 32-bit width
+ * of get_buffer to be used.  (On machines with wider words, an even larger
+ * buffer could be used.)  However, on some machines 32-bit shifts are
+ * quite slow and take time proportional to the number of places shifted.
+ * (This is true with most PC compilers, for instance.)  In this case it may
+ * be a win to set MIN_GET_BITS to the minimum value of 15.  This reduces the
+ * average shift distance at the cost of more calls to jpeg_fill_bit_buffer.
+ */
+
+#ifdef SLOW_SHIFT_32
+#define MIN_GET_BITS  15	/* minimum allowable value */
+#else
+#define MIN_GET_BITS  (BIT_BUF_SIZE-7)
+#endif
+
+
+GLOBAL(boolean)
+jpeg_fill_bit_buffer (bitread_working_state * state,
+		      register bit_buf_type get_buffer, register int bits_left,
+		      int nbits)
+/* Load up the bit buffer to a depth of at least nbits */
+{
+  /* Copy heavily used state fields into locals (hopefully registers) */
+  register const JOCTET * next_input_byte = state->next_input_byte;
+  register size_t bytes_in_buffer = state->bytes_in_buffer;
+  j_decompress_ptr cinfo = state->cinfo;
+
+  /* Attempt to load at least MIN_GET_BITS bits into get_buffer. */
+  /* (It is assumed that no request will be for more than that many bits.) */
+  /* We fail to do so only if we hit a marker or are forced to suspend. */
+
+  if (cinfo->unread_marker == 0) {	/* cannot advance past a marker */
+    while (bits_left < MIN_GET_BITS) {
+      register int c;
+
+      /* Attempt to read a byte */
+      if (bytes_in_buffer == 0) {
+	if (! (*cinfo->src->fill_input_buffer) (cinfo))
+	  return FALSE;
+	next_input_byte = cinfo->src->next_input_byte;
+	bytes_in_buffer = cinfo->src->bytes_in_buffer;
+      }
+      bytes_in_buffer--;
+      c = GETJOCTET(*next_input_byte++);
+
+      /* If it's 0xFF, check and discard stuffed zero byte */
+      if (c == 0xFF) {
+	/* Loop here to discard any padding FF's on terminating marker,
+	 * so that we can save a valid unread_marker value.  NOTE: we will
+	 * accept multiple FF's followed by a 0 as meaning a single FF data
+	 * byte.  This data pattern is not valid according to the standard.
+	 */
+	do {
+	  if (bytes_in_buffer == 0) {
+	    if (! (*cinfo->src->fill_input_buffer) (cinfo))
+	      return FALSE;
+	    next_input_byte = cinfo->src->next_input_byte;
+	    bytes_in_buffer = cinfo->src->bytes_in_buffer;
+	  }
+	  bytes_in_buffer--;
+	  c = GETJOCTET(*next_input_byte++);
+	} while (c == 0xFF);
+
+	if (c == 0) {
+	  /* Found FF/00, which represents an FF data byte */
+	  c = 0xFF;
+	} else {
+	  /* Oops, it's actually a marker indicating end of compressed data.
+	   * Save the marker code for later use.
+	   * Fine point: it might appear that we should save the marker into
+	   * bitread working state, not straight into permanent state.  But
+	   * once we have hit a marker, we cannot need to suspend within the
+	   * current MCU, because we will read no more bytes from the data
+	   * source.  So it is OK to update permanent state right away.
+	   */
+	  cinfo->unread_marker = c;
+	  /* See if we need to insert some fake zero bits. */
+	  goto no_more_bytes;
+	}
+      }
+
+      /* OK, load c into get_buffer */
+      get_buffer = (get_buffer << 8) | c;
+      bits_left += 8;
+    } /* end while */
+  } else {
+  no_more_bytes:
+    /* We get here if we've read the marker that terminates the compressed
+     * data segment.  There should be enough bits in the buffer register
+     * to satisfy the request; if so, no problem.
+     */
+    if (nbits > bits_left) {
+      /* Uh-oh.  Report corrupted data to user and stuff zeroes into
+       * the data stream, so that we can produce some kind of image.
+       * We use a nonvolatile flag to ensure that only one warning message
+       * appears per data segment.
+       */
+      if (! cinfo->entropy->insufficient_data) {
+	WARNMS(cinfo, JWRN_HIT_MARKER);
+	cinfo->entropy->insufficient_data = TRUE;
+      }
+      /* Fill the buffer with zero bits */
+      get_buffer <<= MIN_GET_BITS - bits_left;
+      bits_left = MIN_GET_BITS;
+    }
+  }
+
+  /* Unload the local registers */
+  state->next_input_byte = next_input_byte;
+  state->bytes_in_buffer = bytes_in_buffer;
+  state->get_buffer = get_buffer;
+  state->bits_left = bits_left;
+
+  return TRUE;
+}
+
+
+/*
+ * Out-of-line code for Huffman code decoding.
+ * See jdhuff.h for info about usage.
+ */
+
+GLOBAL(int)
+jpeg_huff_decode (bitread_working_state * state,
+		  register bit_buf_type get_buffer, register int bits_left,
+		  d_derived_tbl * htbl, int min_bits)
+{
+  register int l = min_bits;
+  register INT32 code;
+
+  /* HUFF_DECODE has determined that the code is at least min_bits */
+  /* bits long, so fetch that many bits in one swoop. */
+
+  CHECK_BIT_BUFFER(*state, l, return -1);
+  code = GET_BITS(l);
+
+  /* Collect the rest of the Huffman code one bit at a time. */
+  /* This is per Figure F.16 in the JPEG spec. */
+
+  while (code > htbl->maxcode[l]) {
+    code <<= 1;
+    CHECK_BIT_BUFFER(*state, 1, return -1);
+    code |= GET_BITS(1);
+    l++;
+  }
+
+  /* Unload the local registers */
+  state->get_buffer = get_buffer;
+  state->bits_left = bits_left;
+
+  /* With garbage input we may reach the sentinel value l = 17. */
+
+  if (l > 16) {
+    WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
+    return 0;			/* fake a zero as the safest result */
+  }
+
+  return htbl->pub->huffval[ (int) (code + htbl->valoffset[l]) ];
+}
+
+
+/*
+ * Figure F.12: extend sign bit.
+ * On some machines, a shift and add will be faster than a table lookup.
+ */
+
+#ifdef AVOID_TABLES
+
+#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))
+
+#else
+
+#define HUFF_EXTEND(x,s)  ((x) < extend_test[s] ? (x) + extend_offset[s] : (x))
+
+static const int extend_test[16] =   /* entry n is 2**(n-1) */
+  { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+    0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 };
+
+static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */
+  { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1,
+    ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1,
+    ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1,
+    ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 };
+
+#endif /* AVOID_TABLES */
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ * Returns FALSE if must suspend.
+ */
+
+LOCAL(boolean)
+process_restart (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int ci;
+
+  /* Throw away any unused bits remaining in bit buffer; */
+  /* include any full bytes in next_marker's count of discarded bytes */
+  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
+  entropy->bitstate.bits_left = 0;
+
+  /* Advance past the RSTn marker */
+  if (! (*cinfo->marker->read_restart_marker) (cinfo))
+    return FALSE;
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    entropy->saved.last_dc_val[ci] = 0;
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+
+  /* Reset out-of-data flag, unless read_restart_marker left us smack up
+   * against a marker.  In that case we will end up treating the next data
+   * segment as empty, and we can avoid producing bogus output pixels by
+   * leaving the flag set.
+   */
+  if (cinfo->unread_marker == 0)
+    entropy->pub.insufficient_data = FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Decode and return one MCU's worth of Huffman-compressed coefficients.
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_data[i].  WE ASSUME THIS AREA HAS BEEN ZEROED BY THE CALLER.
+ * (Wholesale zeroing is usually a little faster than retail...)
+ *
+ * Returns FALSE if data source requested suspension.  In that case no
+ * changes have been made to permanent state.  (Exception: some output
+ * coefficients may already have been assigned.  This is harmless for
+ * this module, since we'll just re-assign them on the next call.)
+ */
+
+METHODDEF(boolean)
+decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{
+  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
+  int blkn;
+  BITREAD_STATE_VARS;
+  savable_state state;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(state, entropy->saved);
+
+    /* Outer loop handles each block in the MCU */
+
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      JBLOCKROW block = MCU_data[blkn];
+      d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
+      d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+      register int s, k, r;
+
+      /* Decode a single block's worth of coefficients */
+
+      /* Section F.2.2.1: decode the DC coefficient difference */
+      HUFF_DECODE(s, br_state, dctbl, return FALSE, label1);
+      if (s) {
+	CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	r = GET_BITS(s);
+	s = HUFF_EXTEND(r, s);
+      }
+
+      if (entropy->dc_needed[blkn]) {
+	/* Convert DC difference to actual value, update last_dc_val */
+	int ci = cinfo->MCU_membership[blkn];
+	s += state.last_dc_val[ci];
+	state.last_dc_val[ci] = s;
+	/* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
+	(*block)[0] = (JCOEF) s;
+      }
+
+      if (entropy->ac_needed[blkn]) {
+
+	/* Section F.2.2.2: decode the AC coefficients */
+	/* Since zeroes are skipped, output area must be cleared beforehand */
+	for (k = 1; k < DCTSIZE2; k++) {
+	  HUFF_DECODE(s, br_state, actbl, return FALSE, label2);
+      
+	  r = s >> 4;
+	  s &= 15;
+      
+	  if (s) {
+	    k += r;
+	    CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	    r = GET_BITS(s);
+	    s = HUFF_EXTEND(r, s);
+	    /* Output coefficient in natural (dezigzagged) order.
+	     * Note: the extra entries in jpeg_natural_order[] will save us
+	     * if k >= DCTSIZE2, which could happen if the data is corrupted.
+	     */
+	    (*block)[jpeg_natural_order[k]] = (JCOEF) s;
+	  } else {
+	    if (r != 15)
+	      break;
+	    k += 15;
+	  }
+	}
+
+      } else {
+
+	/* Section F.2.2.2: decode the AC coefficients */
+	/* In this path we just discard the values */
+	for (k = 1; k < DCTSIZE2; k++) {
+	  HUFF_DECODE(s, br_state, actbl, return FALSE, label3);
+      
+	  r = s >> 4;
+	  s &= 15;
+      
+	  if (s) {
+	    k += r;
+	    CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	    DROP_BITS(s);
+	  } else {
+	    if (r != 15)
+	      break;
+	    k += 15;
+	  }
+	}
+
+      }
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(entropy->saved, state);
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * Module initialization routine for Huffman entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_huff_decoder (j_decompress_ptr cinfo)
+{
+  huff_entropy_ptr entropy;
+  int i;
+
+  entropy = (huff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(huff_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
+  entropy->pub.start_pass = start_pass_huff_decoder;
+  entropy->pub.decode_mcu = decode_mcu;
+
+  /* Mark tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL;
+  }
+}
diff --git a/jpeg/jdhuff.h b/jpeg/jdhuff.h
new file mode 100644
index 000000000..ae19b6caf
--- /dev/null
+++ b/jpeg/jdhuff.h
@@ -0,0 +1,201 @@
+/*
+ * jdhuff.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains declarations for Huffman entropy decoding routines
+ * that are shared between the sequential decoder (jdhuff.c) and the
+ * progressive decoder (jdphuff.c).  No other modules need to see these.
+ */
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_make_d_derived_tbl	jMkDDerived
+#define jpeg_fill_bit_buffer	jFilBitBuf
+#define jpeg_huff_decode	jHufDecode
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/* Derived data constructed for each Huffman table */
+
+#define HUFF_LOOKAHEAD	8	/* # of bits of lookahead */
+
+typedef struct {
+  /* Basic tables: (element [0] of each array is unused) */
+  INT32 maxcode[18];		/* largest code of length k (-1 if none) */
+  /* (maxcode[17] is a sentinel to ensure jpeg_huff_decode terminates) */
+  INT32 valoffset[17];		/* huffval[] offset for codes of length k */
+  /* valoffset[k] = huffval[] index of 1st symbol of code length k, less
+   * the smallest code of length k; so given a code of length k, the
+   * corresponding symbol is huffval[code + valoffset[k]]
+   */
+
+  /* Link to public Huffman table (needed only in jpeg_huff_decode) */
+  JHUFF_TBL *pub;
+
+  /* Lookahead tables: indexed by the next HUFF_LOOKAHEAD bits of
+   * the input data stream.  If the next Huffman code is no more
+   * than HUFF_LOOKAHEAD bits long, we can obtain its length and
+   * the corresponding symbol directly from these tables.
+   */
+  int look_nbits[1<<HUFF_LOOKAHEAD]; /* # bits, or 0 if too long */
+  UINT8 look_sym[1<<HUFF_LOOKAHEAD]; /* symbol, or unused */
+} d_derived_tbl;
+
+/* Expand a Huffman table definition into the derived format */
+EXTERN(void) jpeg_make_d_derived_tbl
+	JPP((j_decompress_ptr cinfo, boolean isDC, int tblno,
+	     d_derived_tbl ** pdtbl));
+
+
+/*
+ * Fetching the next N bits from the input stream is a time-critical operation
+ * for the Huffman decoders.  We implement it with a combination of inline
+ * macros and out-of-line subroutines.  Note that N (the number of bits
+ * demanded at one time) never exceeds 15 for JPEG use.
+ *
+ * We read source bytes into get_buffer and dole out bits as needed.
+ * If get_buffer already contains enough bits, they are fetched in-line
+ * by the macros CHECK_BIT_BUFFER and GET_BITS.  When there aren't enough
+ * bits, jpeg_fill_bit_buffer is called; it will attempt to fill get_buffer
+ * as full as possible (not just to the number of bits needed; this
+ * prefetching reduces the overhead cost of calling jpeg_fill_bit_buffer).
+ * Note that jpeg_fill_bit_buffer may return FALSE to indicate suspension.
+ * On TRUE return, jpeg_fill_bit_buffer guarantees that get_buffer contains
+ * at least the requested number of bits --- dummy zeroes are inserted if
+ * necessary.
+ */
+
+typedef INT32 bit_buf_type;	/* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  32	/* size of buffer in bits */
+
+/* If long is > 32 bits on your machine, and shifting/masking longs is
+ * reasonably fast, making bit_buf_type be long and setting BIT_BUF_SIZE
+ * appropriately should be a win.  Unfortunately we can't define the size
+ * with something like  #define BIT_BUF_SIZE (sizeof(bit_buf_type)*8)
+ * because not all machines measure sizeof in 8-bit bytes.
+ */
+
+typedef struct {		/* Bitreading state saved across MCUs */
+  bit_buf_type get_buffer;	/* current bit-extraction buffer */
+  int bits_left;		/* # of unused bits in it */
+} bitread_perm_state;
+
+typedef struct {		/* Bitreading working state within an MCU */
+  /* Current data source location */
+  /* We need a copy, rather than munging the original, in case of suspension */
+  const JOCTET * next_input_byte; /* => next byte to read from source */
+  size_t bytes_in_buffer;	/* # of bytes remaining in source buffer */
+  /* Bit input buffer --- note these values are kept in register variables,
+   * not in this struct, inside the inner loops.
+   */
+  bit_buf_type get_buffer;	/* current bit-extraction buffer */
+  int bits_left;		/* # of unused bits in it */
+  /* Pointer needed by jpeg_fill_bit_buffer. */
+  j_decompress_ptr cinfo;	/* back link to decompress master record */
+} bitread_working_state;
+
+/* Macros to declare and load/save bitread local variables. */
+#define BITREAD_STATE_VARS  \
+	register bit_buf_type get_buffer;  \
+	register int bits_left;  \
+	bitread_working_state br_state
+
+#define BITREAD_LOAD_STATE(cinfop,permstate)  \
+	br_state.cinfo = cinfop; \
+	br_state.next_input_byte = cinfop->src->next_input_byte; \
+	br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
+	get_buffer = permstate.get_buffer; \
+	bits_left = permstate.bits_left;
+
+#define BITREAD_SAVE_STATE(cinfop,permstate)  \
+	cinfop->src->next_input_byte = br_state.next_input_byte; \
+	cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
+	permstate.get_buffer = get_buffer; \
+	permstate.bits_left = bits_left
+
+/*
+ * These macros provide the in-line portion of bit fetching.
+ * Use CHECK_BIT_BUFFER to ensure there are N bits in get_buffer
+ * before using GET_BITS, PEEK_BITS, or DROP_BITS.
+ * The variables get_buffer and bits_left are assumed to be locals,
+ * but the state struct might not be (jpeg_huff_decode needs this).
+ *	CHECK_BIT_BUFFER(state,n,action);
+ *		Ensure there are N bits in get_buffer; if suspend, take action.
+ *      val = GET_BITS(n);
+ *		Fetch next N bits.
+ *      val = PEEK_BITS(n);
+ *		Fetch next N bits without removing them from the buffer.
+ *	DROP_BITS(n);
+ *		Discard next N bits.
+ * The value N should be a simple variable, not an expression, because it
+ * is evaluated multiple times.
+ */
+
+#define CHECK_BIT_BUFFER(state,nbits,action) \
+	{ if (bits_left < (nbits)) {  \
+	    if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
+	      { action; }  \
+	    get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }
+
+#define GET_BITS(nbits) \
+	(((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))
+
+#define PEEK_BITS(nbits) \
+	(((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))
+
+#define DROP_BITS(nbits) \
+	(bits_left -= (nbits))
+
+/* Load up the bit buffer to a depth of at least nbits */
+EXTERN(boolean) jpeg_fill_bit_buffer
+	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
+	     register int bits_left, int nbits));
+
+
+/*
+ * Code for extracting next Huffman-coded symbol from input bit stream.
+ * Again, this is time-critical and we make the main paths be macros.
+ *
+ * We use a lookahead table to process codes of up to HUFF_LOOKAHEAD bits
+ * without looping.  Usually, more than 95% of the Huffman codes will be 8
+ * or fewer bits long.  The few overlength codes are handled with a loop,
+ * which need not be inline code.
+ *
+ * Notes about the HUFF_DECODE macro:
+ * 1. Near the end of the data segment, we may fail to get enough bits
+ *    for a lookahead.  In that case, we do it the hard way.
+ * 2. If the lookahead table contains no entry, the next code must be
+ *    more than HUFF_LOOKAHEAD bits long.
+ * 3. jpeg_huff_decode returns -1 if forced to suspend.
+ */
+
+#define HUFF_DECODE(result,state,htbl,failaction,slowlabel) \
+{ register int nb, look; \
+  if (bits_left < HUFF_LOOKAHEAD) { \
+    if (! jpeg_fill_bit_buffer(&state,get_buffer,bits_left, 0)) {failaction;} \
+    get_buffer = state.get_buffer; bits_left = state.bits_left; \
+    if (bits_left < HUFF_LOOKAHEAD) { \
+      nb = 1; goto slowlabel; \
+    } \
+  } \
+  look = PEEK_BITS(HUFF_LOOKAHEAD); \
+  if ((nb = htbl->look_nbits[look]) != 0) { \
+    DROP_BITS(nb); \
+    result = htbl->look_sym[look]; \
+  } else { \
+    nb = HUFF_LOOKAHEAD+1; \
+slowlabel: \
+    if ((result=jpeg_huff_decode(&state,get_buffer,bits_left,htbl,nb)) < 0) \
+	{ failaction; } \
+    get_buffer = state.get_buffer; bits_left = state.bits_left; \
+  } \
+}
+
+/* Out-of-line case for Huffman code fetching */
+EXTERN(int) jpeg_huff_decode
+	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
+	     register int bits_left, d_derived_tbl * htbl, int min_bits));
diff --git a/jpeg/jdinput.c b/jpeg/jdinput.c
new file mode 100644
index 000000000..0c2ac8f12
--- /dev/null
+++ b/jpeg/jdinput.c
@@ -0,0 +1,381 @@
+/*
+ * jdinput.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains input control logic for the JPEG decompressor.
+ * These routines are concerned with controlling the decompressor's input
+ * processing (marker reading and coefficient decoding).  The actual input
+ * reading is done in jdmarker.c, jdhuff.c, and jdphuff.c.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_input_controller pub; /* public fields */
+
+  boolean inheaders;		/* TRUE until first SOS is reached */
+} my_input_controller;
+
+typedef my_input_controller * my_inputctl_ptr;
+
+
+/* Forward declarations */
+METHODDEF(int) consume_markers JPP((j_decompress_ptr cinfo));
+
+
+/*
+ * Routines to calculate various quantities related to the size of the image.
+ */
+
+LOCAL(void)
+initial_setup (j_decompress_ptr cinfo)
+/* Called once, when first SOS marker is reached */
+{
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Make sure image isn't bigger than I can handle */
+  if ((long) cinfo->image_height > (long) JPEG_MAX_DIMENSION ||
+      (long) cinfo->image_width > (long) JPEG_MAX_DIMENSION)
+    ERREXIT1(cinfo, JERR_IMAGE_TOO_BIG, (unsigned int) JPEG_MAX_DIMENSION);
+
+  /* For now, precision must match compiled-in value... */
+  if (cinfo->data_precision != BITS_IN_JSAMPLE)
+    ERREXIT1(cinfo, JERR_BAD_PRECISION, cinfo->data_precision);
+
+  /* Check that number of components won't exceed internal array sizes */
+  if (cinfo->num_components > MAX_COMPONENTS)
+    ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
+	     MAX_COMPONENTS);
+
+  /* Compute maximum sampling factors; check factor validity */
+  cinfo->max_h_samp_factor = 1;
+  cinfo->max_v_samp_factor = 1;
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
+	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+      ERREXIT(cinfo, JERR_BAD_SAMPLING);
+    cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
+				   compptr->h_samp_factor);
+    cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
+				   compptr->v_samp_factor);
+  }
+
+  /* We initialize DCT_scaled_size and min_DCT_scaled_size to DCTSIZE.
+   * In the full decompressor, this will be overridden by jdmaster.c;
+   * but in the transcoder, jdmaster.c is not used, so we must do it here.
+   */
+  cinfo->min_DCT_scaled_size = DCTSIZE;
+
+  /* Compute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->DCT_scaled_size = DCTSIZE;
+    /* Size in DCT blocks */
+    compptr->width_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->height_in_blocks = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+    /* downsampled_width and downsampled_height will also be overridden by
+     * jdmaster.c if we are doing full decompression.  The transcoder library
+     * doesn't use these values, but the calling application might.
+     */
+    /* Size in samples */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
+		    (long) cinfo->max_h_samp_factor);
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
+		    (long) cinfo->max_v_samp_factor);
+    /* Mark component needed, until color conversion says otherwise */
+    compptr->component_needed = TRUE;
+    /* Mark no quantization table yet saved for component */
+    compptr->quant_table = NULL;
+  }
+
+  /* Compute number of fully interleaved MCU rows. */
+  cinfo->total_iMCU_rows = (JDIMENSION)
+    jdiv_round_up((long) cinfo->image_height,
+		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+
+  /* Decide whether file contains multiple scans */
+  if (cinfo->comps_in_scan < cinfo->num_components || cinfo->progressive_mode)
+    cinfo->inputctl->has_multiple_scans = TRUE;
+  else
+    cinfo->inputctl->has_multiple_scans = FALSE;
+}
+
+
+LOCAL(void)
+per_scan_setup (j_decompress_ptr cinfo)
+/* Do computations that are needed before processing a JPEG scan */
+/* cinfo->comps_in_scan and cinfo->cur_comp_info[] were set from SOS marker */
+{
+  int ci, mcublks, tmp;
+  jpeg_component_info *compptr;
+  
+  if (cinfo->comps_in_scan == 1) {
+    
+    /* Noninterleaved (single-component) scan */
+    compptr = cinfo->cur_comp_info[0];
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = compptr->width_in_blocks;
+    cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
+    
+    /* For noninterleaved scan, always one block per MCU */
+    compptr->MCU_width = 1;
+    compptr->MCU_height = 1;
+    compptr->MCU_blocks = 1;
+    compptr->MCU_sample_width = compptr->DCT_scaled_size;
+    compptr->last_col_width = 1;
+    /* For noninterleaved scans, it is convenient to define last_row_height
+     * as the number of block rows present in the last iMCU row.
+     */
+    tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
+    if (tmp == 0) tmp = compptr->v_samp_factor;
+    compptr->last_row_height = tmp;
+    
+    /* Prepare array describing MCU composition */
+    cinfo->blocks_in_MCU = 1;
+    cinfo->MCU_membership[0] = 0;
+    
+  } else {
+    
+    /* Interleaved (multi-component) scan */
+    if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
+      ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
+	       MAX_COMPS_IN_SCAN);
+    
+    /* Overall image size in MCUs */
+    cinfo->MCUs_per_row = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width,
+		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+    cinfo->MCU_rows_in_scan = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height,
+		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+    
+    cinfo->blocks_in_MCU = 0;
+    
+    for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+      compptr = cinfo->cur_comp_info[ci];
+      /* Sampling factors give # of blocks of component in each MCU */
+      compptr->MCU_width = compptr->h_samp_factor;
+      compptr->MCU_height = compptr->v_samp_factor;
+      compptr->MCU_blocks = compptr->MCU_width * compptr->MCU_height;
+      compptr->MCU_sample_width = compptr->MCU_width * compptr->DCT_scaled_size;
+      /* Figure number of non-dummy blocks in last MCU column & row */
+      tmp = (int) (compptr->width_in_blocks % compptr->MCU_width);
+      if (tmp == 0) tmp = compptr->MCU_width;
+      compptr->last_col_width = tmp;
+      tmp = (int) (compptr->height_in_blocks % compptr->MCU_height);
+      if (tmp == 0) tmp = compptr->MCU_height;
+      compptr->last_row_height = tmp;
+      /* Prepare array describing MCU composition */
+      mcublks = compptr->MCU_blocks;
+      if (cinfo->blocks_in_MCU + mcublks > D_MAX_BLOCKS_IN_MCU)
+	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+      while (mcublks-- > 0) {
+	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+      }
+    }
+    
+  }
+}
+
+
+/*
+ * Save away a copy of the Q-table referenced by each component present
+ * in the current scan, unless already saved during a prior scan.
+ *
+ * In a multiple-scan JPEG file, the encoder could assign different components
+ * the same Q-table slot number, but change table definitions between scans
+ * so that each component uses a different Q-table.  (The IJG encoder is not
+ * currently capable of doing this, but other encoders might.)  Since we want
+ * to be able to dequantize all the components at the end of the file, this
+ * means that we have to save away the table actually used for each component.
+ * We do this by copying the table at the start of the first scan containing
+ * the component.
+ * The JPEG spec prohibits the encoder from changing the contents of a Q-table
+ * slot between scans of a component using that slot.  If the encoder does so
+ * anyway, this decoder will simply use the Q-table values that were current
+ * at the start of the first scan for the component.
+ *
+ * The decompressor output side looks only at the saved quant tables,
+ * not at the current Q-table slots.
+ */
+
+LOCAL(void)
+latch_quant_tables (j_decompress_ptr cinfo)
+{
+  int ci, qtblno;
+  jpeg_component_info *compptr;
+  JQUANT_TBL * qtbl;
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    /* No work if we already saved Q-table for this component */
+    if (compptr->quant_table != NULL)
+      continue;
+    /* Make sure specified quantization table is present */
+    qtblno = compptr->quant_tbl_no;
+    if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
+	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+      ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
+    /* OK, save away the quantization table */
+    qtbl = (JQUANT_TBL *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(JQUANT_TBL));
+    MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], SIZEOF(JQUANT_TBL));
+    compptr->quant_table = qtbl;
+  }
+}
+
+
+/*
+ * Initialize the input modules to read a scan of compressed data.
+ * The first call to this is done by jdmaster.c after initializing
+ * the entire decompressor (during jpeg_start_decompress).
+ * Subsequent calls come from consume_markers, below.
+ */
+
+METHODDEF(void)
+start_input_pass (j_decompress_ptr cinfo)
+{
+  per_scan_setup(cinfo);
+  latch_quant_tables(cinfo);
+  (*cinfo->entropy->start_pass) (cinfo);
+  (*cinfo->coef->start_input_pass) (cinfo);
+  cinfo->inputctl->consume_input = cinfo->coef->consume_data;
+}
+
+
+/*
+ * Finish up after inputting a compressed-data scan.
+ * This is called by the coefficient controller after it's read all
+ * the expected data of the scan.
+ */
+
+METHODDEF(void)
+finish_input_pass (j_decompress_ptr cinfo)
+{
+  cinfo->inputctl->consume_input = consume_markers;
+}
+
+
+/*
+ * Read JPEG markers before, between, or after compressed-data scans.
+ * Change state as necessary when a new scan is reached.
+ * Return value is JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+ *
+ * The consume_input method pointer points either here or to the
+ * coefficient controller's consume_data routine, depending on whether
+ * we are reading a compressed data segment or inter-segment markers.
+ */
+
+METHODDEF(int)
+consume_markers (j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl = (my_inputctl_ptr) cinfo->inputctl;
+  int val;
+
+  if (inputctl->pub.eoi_reached) /* After hitting EOI, read no further */
+    return JPEG_REACHED_EOI;
+
+  val = (*cinfo->marker->read_markers) (cinfo);
+
+  switch (val) {
+  case JPEG_REACHED_SOS:	/* Found SOS */
+    if (inputctl->inheaders) {	/* 1st SOS */
+      initial_setup(cinfo);
+      inputctl->inheaders = FALSE;
+      /* Note: start_input_pass must be called by jdmaster.c
+       * before any more input can be consumed.  jdapimin.c is
+       * responsible for enforcing this sequencing.
+       */
+    } else {			/* 2nd or later SOS marker */
+      if (! inputctl->pub.has_multiple_scans)
+	ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
+      start_input_pass(cinfo);
+    }
+    break;
+  case JPEG_REACHED_EOI:	/* Found EOI */
+    inputctl->pub.eoi_reached = TRUE;
+    if (inputctl->inheaders) {	/* Tables-only datastream, apparently */
+      if (cinfo->marker->saw_SOF)
+	ERREXIT(cinfo, JERR_SOF_NO_SOS);
+    } else {
+      /* Prevent infinite loop in coef ctlr's decompress_data routine
+       * if user set output_scan_number larger than number of scans.
+       */
+      if (cinfo->output_scan_number > cinfo->input_scan_number)
+	cinfo->output_scan_number = cinfo->input_scan_number;
+    }
+    break;
+  case JPEG_SUSPENDED:
+    break;
+  }
+
+  return val;
+}
+
+
+/*
+ * Reset state to begin a fresh datastream.
+ */
+
+METHODDEF(void)
+reset_input_controller (j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl = (my_inputctl_ptr) cinfo->inputctl;
+
+  inputctl->pub.consume_input = consume_markers;
+  inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */
+  inputctl->pub.eoi_reached = FALSE;
+  inputctl->inheaders = TRUE;
+  /* Reset other modules */
+  (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
+  (*cinfo->marker->reset_marker_reader) (cinfo);
+  /* Reset progression state -- would be cleaner if entropy decoder did this */
+  cinfo->coef_bits = NULL;
+}
+
+
+/*
+ * Initialize the input controller module.
+ * This is called only once, when the decompression object is created.
+ */
+
+GLOBAL(void)
+jinit_input_controller (j_decompress_ptr cinfo)
+{
+  my_inputctl_ptr inputctl;
+
+  /* Create subobject in permanent pool */
+  inputctl = (my_inputctl_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				SIZEOF(my_input_controller));
+  cinfo->inputctl = (struct jpeg_input_controller *) inputctl;
+  /* Initialize method pointers */
+  inputctl->pub.consume_input = consume_markers;
+  inputctl->pub.reset_input_controller = reset_input_controller;
+  inputctl->pub.start_input_pass = start_input_pass;
+  inputctl->pub.finish_input_pass = finish_input_pass;
+  /* Initialize state: can't use reset_input_controller since we don't
+   * want to try to reset other modules yet.
+   */
+  inputctl->pub.has_multiple_scans = FALSE; /* "unknown" would be better */
+  inputctl->pub.eoi_reached = FALSE;
+  inputctl->inheaders = TRUE;
+}
diff --git a/jpeg/jdmainct.c b/jpeg/jdmainct.c
new file mode 100644
index 000000000..da19c7e5e
--- /dev/null
+++ b/jpeg/jdmainct.c
@@ -0,0 +1,512 @@
+/*
+ * jdmainct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the main buffer controller for decompression.
+ * The main buffer lies between the JPEG decompressor proper and the
+ * post-processor; it holds downsampled data in the JPEG colorspace.
+ *
+ * Note that this code is bypassed in raw-data mode, since the application
+ * supplies the equivalent of the main buffer in that case.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * In the current system design, the main buffer need never be a full-image
+ * buffer; any full-height buffers will be found inside the coefficient or
+ * postprocessing controllers.  Nonetheless, the main controller is not
+ * trivial.  Its responsibility is to provide context rows for upsampling/
+ * rescaling, and doing this in an efficient fashion is a bit tricky.
+ *
+ * Postprocessor input data is counted in "row groups".  A row group
+ * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
+ * sample rows of each component.  (We require DCT_scaled_size values to be
+ * chosen such that these numbers are integers.  In practice DCT_scaled_size
+ * values will likely be powers of two, so we actually have the stronger
+ * condition that DCT_scaled_size / min_DCT_scaled_size is an integer.)
+ * Upsampling will typically produce max_v_samp_factor pixel rows from each
+ * row group (times any additional scale factor that the upsampler is
+ * applying).
+ *
+ * The coefficient controller will deliver data to us one iMCU row at a time;
+ * each iMCU row contains v_samp_factor * DCT_scaled_size sample rows, or
+ * exactly min_DCT_scaled_size row groups.  (This amount of data corresponds
+ * to one row of MCUs when the image is fully interleaved.)  Note that the
+ * number of sample rows varies across components, but the number of row
+ * groups does not.  Some garbage sample rows may be included in the last iMCU
+ * row at the bottom of the image.
+ *
+ * Depending on the vertical scaling algorithm used, the upsampler may need
+ * access to the sample row(s) above and below its current input row group.
+ * The upsampler is required to set need_context_rows TRUE at global selection
+ * time if so.  When need_context_rows is FALSE, this controller can simply
+ * obtain one iMCU row at a time from the coefficient controller and dole it
+ * out as row groups to the postprocessor.
+ *
+ * When need_context_rows is TRUE, this controller guarantees that the buffer
+ * passed to postprocessing contains at least one row group's worth of samples
+ * above and below the row group(s) being processed.  Note that the context
+ * rows "above" the first passed row group appear at negative row offsets in
+ * the passed buffer.  At the top and bottom of the image, the required
+ * context rows are manufactured by duplicating the first or last real sample
+ * row; this avoids having special cases in the upsampling inner loops.
+ *
+ * The amount of context is fixed at one row group just because that's a
+ * convenient number for this controller to work with.  The existing
+ * upsamplers really only need one sample row of context.  An upsampler
+ * supporting arbitrary output rescaling might wish for more than one row
+ * group of context when shrinking the image; tough, we don't handle that.
+ * (This is justified by the assumption that downsizing will be handled mostly
+ * by adjusting the DCT_scaled_size values, so that the actual scale factor at
+ * the upsample step needn't be much less than one.)
+ *
+ * To provide the desired context, we have to retain the last two row groups
+ * of one iMCU row while reading in the next iMCU row.  (The last row group
+ * can't be processed until we have another row group for its below-context,
+ * and so we have to save the next-to-last group too for its above-context.)
+ * We could do this most simply by copying data around in our buffer, but
+ * that'd be very slow.  We can avoid copying any data by creating a rather
+ * strange pointer structure.  Here's how it works.  We allocate a workspace
+ * consisting of M+2 row groups (where M = min_DCT_scaled_size is the number
+ * of row groups per iMCU row).  We create two sets of redundant pointers to
+ * the workspace.  Labeling the physical row groups 0 to M+1, the synthesized
+ * pointer lists look like this:
+ *                   M+1                          M-1
+ * master pointer --> 0         master pointer --> 0
+ *                    1                            1
+ *                   ...                          ...
+ *                   M-3                          M-3
+ *                   M-2                           M
+ *                   M-1                          M+1
+ *                    M                           M-2
+ *                   M+1                          M-1
+ *                    0                            0
+ * We read alternate iMCU rows using each master pointer; thus the last two
+ * row groups of the previous iMCU row remain un-overwritten in the workspace.
+ * The pointer lists are set up so that the required context rows appear to
+ * be adjacent to the proper places when we pass the pointer lists to the
+ * upsampler.
+ *
+ * The above pictures describe the normal state of the pointer lists.
+ * At top and bottom of the image, we diddle the pointer lists to duplicate
+ * the first or last sample row as necessary (this is cheaper than copying
+ * sample rows around).
+ *
+ * This scheme breaks down if M < 2, ie, min_DCT_scaled_size is 1.  In that
+ * situation each iMCU row provides only one row group so the buffering logic
+ * must be different (eg, we must read two iMCU rows before we can emit the
+ * first row group).  For now, we simply do not support providing context
+ * rows when min_DCT_scaled_size is 1.  That combination seems unlikely to
+ * be worth providing --- if someone wants a 1/8th-size preview, they probably
+ * want it quick and dirty, so a context-free upsampler is sufficient.
+ */
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_main_controller pub; /* public fields */
+
+  /* Pointer to allocated workspace (M or M+2 row groups). */
+  JSAMPARRAY buffer[MAX_COMPONENTS];
+
+  boolean buffer_full;		/* Have we gotten an iMCU row from decoder? */
+  JDIMENSION rowgroup_ctr;	/* counts row groups output to postprocessor */
+
+  /* Remaining fields are only used in the context case. */
+
+  /* These are the master pointers to the funny-order pointer lists. */
+  JSAMPIMAGE xbuffer[2];	/* pointers to weird pointer lists */
+
+  int whichptr;			/* indicates which pointer set is now in use */
+  int context_state;		/* process_data state machine status */
+  JDIMENSION rowgroups_avail;	/* row groups available to postprocessor */
+  JDIMENSION iMCU_row_ctr;	/* counts iMCU rows to detect image top/bot */
+} my_main_controller;
+
+typedef my_main_controller * my_main_ptr;
+
+/* context_state values: */
+#define CTX_PREPARE_FOR_IMCU	0	/* need to prepare for MCU row */
+#define CTX_PROCESS_IMCU	1	/* feeding iMCU to postprocessor */
+#define CTX_POSTPONED_ROW	2	/* feeding postponed row group */
+
+
+/* Forward declarations */
+METHODDEF(void) process_data_simple_main
+	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+METHODDEF(void) process_data_context_main
+	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+#ifdef QUANT_2PASS_SUPPORTED
+METHODDEF(void) process_data_crank_post
+	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+#endif
+
+
+LOCAL(void)
+alloc_funny_pointers (j_decompress_ptr cinfo)
+/* Allocate space for the funny pointer lists.
+ * This is done only once, not once per pass.
+ */
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+  int ci, rgroup;
+  int M = cinfo->min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf;
+
+  /* Get top-level space for component array pointers.
+   * We alloc both arrays with one call to save a few cycles.
+   */
+  jmain->xbuffer[0] = (JSAMPIMAGE)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				cinfo->num_components * 2 * SIZEOF(JSAMPARRAY));
+  jmain->xbuffer[1] = jmain->xbuffer[0] + cinfo->num_components;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    /* Get space for pointer lists --- M+4 row groups in each list.
+     * We alloc both pointer lists with one call to save a few cycles.
+     */
+    xbuf = (JSAMPARRAY)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  2 * (rgroup * (M + 4)) * SIZEOF(JSAMPROW));
+    xbuf += rgroup;		/* want one row group at negative offsets */
+    jmain->xbuffer[0][ci] = xbuf;
+    xbuf += rgroup * (M + 4);
+    jmain->xbuffer[1][ci] = xbuf;
+  }
+}
+
+
+LOCAL(void)
+make_funny_pointers (j_decompress_ptr cinfo)
+/* Create the funny pointer lists discussed in the comments above.
+ * The actual workspace is already allocated (in jmain->buffer),
+ * and the space for the pointer lists is allocated too.
+ * This routine just fills in the curiously ordered lists.
+ * This will be repeated at the beginning of each pass.
+ */
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY buf, xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = jmain->xbuffer[0][ci];
+    xbuf1 = jmain->xbuffer[1][ci];
+    /* First copy the workspace pointers as-is */
+    buf = jmain->buffer[ci];
+    for (i = 0; i < rgroup * (M + 2); i++) {
+      xbuf0[i] = xbuf1[i] = buf[i];
+    }
+    /* In the second list, put the last four row groups in swapped order */
+    for (i = 0; i < rgroup * 2; i++) {
+      xbuf1[rgroup*(M-2) + i] = buf[rgroup*M + i];
+      xbuf1[rgroup*M + i] = buf[rgroup*(M-2) + i];
+    }
+    /* The wraparound pointers at top and bottom will be filled later
+     * (see set_wraparound_pointers, below).  Initially we want the "above"
+     * pointers to duplicate the first actual data line.  This only needs
+     * to happen in xbuffer[0].
+     */
+    for (i = 0; i < rgroup; i++) {
+      xbuf0[i - rgroup] = xbuf0[0];
+    }
+  }
+}
+
+
+LOCAL(void)
+set_wraparound_pointers (j_decompress_ptr cinfo)
+/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
+ * This changes the pointer list state from top-of-image to the normal state.
+ */
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = jmain->xbuffer[0][ci];
+    xbuf1 = jmain->xbuffer[1][ci];
+    for (i = 0; i < rgroup; i++) {
+      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
+      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
+      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
+      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
+    }
+  }
+}
+
+
+LOCAL(void)
+set_bottom_pointers (j_decompress_ptr cinfo)
+/* Change the pointer lists to duplicate the last sample row at the bottom
+ * of the image.  whichptr indicates which xbuffer holds the final iMCU row.
+ * Also sets rowgroups_avail to indicate number of nondummy row groups in row.
+ */
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup, iMCUheight, rows_left;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Count sample rows in one iMCU row and in one row group */
+    iMCUheight = compptr->v_samp_factor * compptr->DCT_scaled_size;
+    rgroup = iMCUheight / cinfo->min_DCT_scaled_size;
+    /* Count nondummy sample rows remaining for this component */
+    rows_left = (int) (compptr->downsampled_height % (JDIMENSION) iMCUheight);
+    if (rows_left == 0) rows_left = iMCUheight;
+    /* Count nondummy row groups.  Should get same answer for each component,
+     * so we need only do it once.
+     */
+    if (ci == 0) {
+      jmain->rowgroups_avail = (JDIMENSION) ((rows_left-1) / rgroup + 1);
+    }
+    /* Duplicate the last real sample row rgroup*2 times; this pads out the
+     * last partial rowgroup and ensures at least one full rowgroup of context.
+     */
+    xbuf = jmain->xbuffer[jmain->whichptr][ci];
+    for (i = 0; i < rgroup * 2; i++) {
+      xbuf[rows_left + i] = xbuf[rows_left-1];
+    }
+  }
+}
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_main (j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (cinfo->upsample->need_context_rows) {
+      jmain->pub.process_data = process_data_context_main;
+      make_funny_pointers(cinfo); /* Create the xbuffer[] lists */
+      jmain->whichptr = 0;	/* Read first iMCU row into xbuffer[0] */
+      jmain->context_state = CTX_PREPARE_FOR_IMCU;
+      jmain->iMCU_row_ctr = 0;
+    } else {
+      /* Simple case with no context needed */
+      jmain->pub.process_data = process_data_simple_main;
+    }
+    jmain->buffer_full = FALSE;	/* Mark buffer empty */
+    jmain->rowgroup_ctr = 0;
+    break;
+#ifdef QUANT_2PASS_SUPPORTED
+  case JBUF_CRANK_DEST:
+    /* For last pass of 2-pass quantization, just crank the postprocessor */
+    jmain->pub.process_data = process_data_crank_post;
+    break;
+#endif
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This handles the simple case where no context is required.
+ */
+
+METHODDEF(void)
+process_data_simple_main (j_decompress_ptr cinfo,
+			  JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			  JDIMENSION out_rows_avail)
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+  JDIMENSION rowgroups_avail;
+
+  /* Read input data if we haven't filled the main buffer yet */
+  if (! jmain->buffer_full) {
+    if (! (*cinfo->coef->decompress_data) (cinfo, jmain->buffer))
+      return;			/* suspension forced, can do nothing more */
+    jmain->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
+  }
+
+  /* There are always min_DCT_scaled_size row groups in an iMCU row. */
+  rowgroups_avail = (JDIMENSION) cinfo->min_DCT_scaled_size;
+  /* Note: at the bottom of the image, we may pass extra garbage row groups
+   * to the postprocessor.  The postprocessor has to check for bottom
+   * of image anyway (at row resolution), so no point in us doing it too.
+   */
+
+  /* Feed the postprocessor */
+  (*cinfo->post->post_process_data) (cinfo, jmain->buffer,
+				     &jmain->rowgroup_ctr, rowgroups_avail,
+				     output_buf, out_row_ctr, out_rows_avail);
+
+  /* Has postprocessor consumed all the data yet? If so, mark buffer empty */
+  if (jmain->rowgroup_ctr >= rowgroups_avail) {
+    jmain->buffer_full = FALSE;
+    jmain->rowgroup_ctr = 0;
+  }
+}
+
+
+/*
+ * Process some data.
+ * This handles the case where context rows must be provided.
+ */
+
+METHODDEF(void)
+process_data_context_main (j_decompress_ptr cinfo,
+			   JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			   JDIMENSION out_rows_avail)
+{
+  my_main_ptr jmain = (my_main_ptr) cinfo->main;
+
+  /* Read input data if we haven't filled the main buffer yet */
+  if (! jmain->buffer_full) {
+    if (! (*cinfo->coef->decompress_data) (cinfo,
+					   jmain->xbuffer[jmain->whichptr]))
+      return;			/* suspension forced, can do nothing more */
+    jmain->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
+    jmain->iMCU_row_ctr++;	/* count rows received */
+  }
+
+  /* Postprocessor typically will not swallow all the input data it is handed
+   * in one call (due to filling the output buffer first).  Must be prepared
+   * to exit and restart.  This switch lets us keep track of how far we got.
+   * Note that each case falls through to the next on successful completion.
+   */
+  switch (jmain->context_state) {
+  case CTX_POSTPONED_ROW:
+    /* Call postprocessor using previously set pointers for postponed row */
+    (*cinfo->post->post_process_data) (cinfo, jmain->xbuffer[jmain->whichptr],
+			&jmain->rowgroup_ctr, jmain->rowgroups_avail,
+			output_buf, out_row_ctr, out_rows_avail);
+    if (jmain->rowgroup_ctr < jmain->rowgroups_avail)
+      return;			/* Need to suspend */
+    jmain->context_state = CTX_PREPARE_FOR_IMCU;
+    if (*out_row_ctr >= out_rows_avail)
+      return;			/* Postprocessor exactly filled output buf */
+    /*FALLTHROUGH*/
+  case CTX_PREPARE_FOR_IMCU:
+    /* Prepare to process first M-1 row groups of this iMCU row */
+    jmain->rowgroup_ctr = 0;
+    jmain->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_scaled_size - 1);
+    /* Check for bottom of image: if so, tweak pointers to "duplicate"
+     * the last sample row, and adjust rowgroups_avail to ignore padding rows.
+     */
+    if (jmain->iMCU_row_ctr == cinfo->total_iMCU_rows)
+      set_bottom_pointers(cinfo);
+    jmain->context_state = CTX_PROCESS_IMCU;
+    /*FALLTHROUGH*/
+  case CTX_PROCESS_IMCU:
+    /* Call postprocessor using previously set pointers */
+    (*cinfo->post->post_process_data) (cinfo, jmain->xbuffer[jmain->whichptr],
+			&jmain->rowgroup_ctr, jmain->rowgroups_avail,
+			output_buf, out_row_ctr, out_rows_avail);
+    if (jmain->rowgroup_ctr < jmain->rowgroups_avail)
+      return;			/* Need to suspend */
+    /* After the first iMCU, change wraparound pointers to normal state */
+    if (jmain->iMCU_row_ctr == 1)
+      set_wraparound_pointers(cinfo);
+    /* Prepare to load new iMCU row using other xbuffer list */
+    jmain->whichptr ^= 1;	/* 0=>1 or 1=>0 */
+    jmain->buffer_full = FALSE;
+    /* Still need to process last row group of this iMCU row, */
+    /* which is saved at index M+1 of the other xbuffer */
+    jmain->rowgroup_ctr = (JDIMENSION) (cinfo->min_DCT_scaled_size + 1);
+    jmain->rowgroups_avail = (JDIMENSION) (cinfo->min_DCT_scaled_size + 2);
+    jmain->context_state = CTX_POSTPONED_ROW;
+  }
+}
+
+
+/*
+ * Process some data.
+ * Final pass of two-pass quantization: just call the postprocessor.
+ * Source data will be the postprocessor controller's internal buffer.
+ */
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+METHODDEF(void)
+process_data_crank_post (j_decompress_ptr cinfo,
+			 JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			 JDIMENSION out_rows_avail)
+{
+  (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL,
+				     (JDIMENSION *) NULL, (JDIMENSION) 0,
+				     output_buf, out_row_ctr, out_rows_avail);
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
+
+
+/*
+ * Initialize main buffer controller.
+ */
+
+GLOBAL(void)
+jinit_d_main_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_main_ptr jmain;
+  int ci, rgroup, ngroups;
+  jpeg_component_info *compptr;
+
+  jmain = (my_main_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_main_controller));
+  cinfo->main = (struct jpeg_d_main_controller *) jmain;
+  jmain->pub.start_pass = start_pass_main;
+
+  if (need_full_buffer)		/* shouldn't happen */
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  /* Allocate the workspace.
+   * ngroups is the number of row groups we need.
+   */
+  if (cinfo->upsample->need_context_rows) {
+    if (cinfo->min_DCT_scaled_size < 2) /* unsupported, see comments above */
+      ERREXIT(cinfo, JERR_NOTIMPL);
+    alloc_funny_pointers(cinfo); /* Alloc space for xbuffer[] lists */
+    ngroups = cinfo->min_DCT_scaled_size + 2;
+  } else {
+    ngroups = cinfo->min_DCT_scaled_size;
+  }
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+      cinfo->min_DCT_scaled_size; /* height of a row group of component */
+    jmain->buffer[ci] = (*cinfo->mem->alloc_sarray)
+			((j_common_ptr) cinfo, JPOOL_IMAGE,
+			 compptr->width_in_blocks * compptr->DCT_scaled_size,
+			 (JDIMENSION) (rgroup * ngroups));
+  }
+}
diff --git a/jpeg/jdmarker.c b/jpeg/jdmarker.c
new file mode 100644
index 000000000..f4cca8cc8
--- /dev/null
+++ b/jpeg/jdmarker.c
@@ -0,0 +1,1360 @@
+/*
+ * jdmarker.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains routines to decode JPEG datastream markers.
+ * Most of the complexity arises from our desire to support input
+ * suspension: if not all of the data for a marker is available,
+ * we must exit back to the application.  On resumption, we reprocess
+ * the marker.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+typedef enum {			/* JPEG marker codes */
+  M_SOF0  = 0xc0,
+  M_SOF1  = 0xc1,
+  M_SOF2  = 0xc2,
+  M_SOF3  = 0xc3,
+  
+  M_SOF5  = 0xc5,
+  M_SOF6  = 0xc6,
+  M_SOF7  = 0xc7,
+  
+  M_JPG   = 0xc8,
+  M_SOF9  = 0xc9,
+  M_SOF10 = 0xca,
+  M_SOF11 = 0xcb,
+  
+  M_SOF13 = 0xcd,
+  M_SOF14 = 0xce,
+  M_SOF15 = 0xcf,
+  
+  M_DHT   = 0xc4,
+  
+  M_DAC   = 0xcc,
+  
+  M_RST0  = 0xd0,
+  M_RST1  = 0xd1,
+  M_RST2  = 0xd2,
+  M_RST3  = 0xd3,
+  M_RST4  = 0xd4,
+  M_RST5  = 0xd5,
+  M_RST6  = 0xd6,
+  M_RST7  = 0xd7,
+  
+  M_SOI   = 0xd8,
+  M_EOI   = 0xd9,
+  M_SOS   = 0xda,
+  M_DQT   = 0xdb,
+  M_DNL   = 0xdc,
+  M_DRI   = 0xdd,
+  M_DHP   = 0xde,
+  M_EXP   = 0xdf,
+  
+  M_APP0  = 0xe0,
+  M_APP1  = 0xe1,
+  M_APP2  = 0xe2,
+  M_APP3  = 0xe3,
+  M_APP4  = 0xe4,
+  M_APP5  = 0xe5,
+  M_APP6  = 0xe6,
+  M_APP7  = 0xe7,
+  M_APP8  = 0xe8,
+  M_APP9  = 0xe9,
+  M_APP10 = 0xea,
+  M_APP11 = 0xeb,
+  M_APP12 = 0xec,
+  M_APP13 = 0xed,
+  M_APP14 = 0xee,
+  M_APP15 = 0xef,
+  
+  M_JPG0  = 0xf0,
+  M_JPG13 = 0xfd,
+  M_COM   = 0xfe,
+  
+  M_TEM   = 0x01,
+  
+  M_ERROR = 0x100
+} JPEG_MARKER;
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_marker_reader pub; /* public fields */
+
+  /* Application-overridable marker processing methods */
+  jpeg_marker_parser_method process_COM;
+  jpeg_marker_parser_method process_APPn[16];
+
+  /* Limit on marker data length to save for each marker type */
+  unsigned int length_limit_COM;
+  unsigned int length_limit_APPn[16];
+
+  /* Status of COM/APPn marker saving */
+  jpeg_saved_marker_ptr cur_marker;	/* NULL if not processing a marker */
+  unsigned int bytes_read;		/* data bytes read so far in marker */
+  /* Note: cur_marker is not linked into marker_list until it's all read. */
+} my_marker_reader;
+
+typedef my_marker_reader * my_marker_ptr;
+
+
+/*
+ * Macros for fetching data from the data source module.
+ *
+ * At all times, cinfo->src->next_input_byte and ->bytes_in_buffer reflect
+ * the current restart point; we update them only when we have reached a
+ * suitable place to restart if a suspension occurs.
+ */
+
+/* Declare and initialize local copies of input pointer/count */
+#define INPUT_VARS(cinfo)  \
+	struct jpeg_source_mgr * datasrc = (cinfo)->src;  \
+	const JOCTET * next_input_byte = datasrc->next_input_byte;  \
+	size_t bytes_in_buffer = datasrc->bytes_in_buffer
+
+/* Unload the local copies --- do this only at a restart boundary */
+#define INPUT_SYNC(cinfo)  \
+	( datasrc->next_input_byte = next_input_byte,  \
+	  datasrc->bytes_in_buffer = bytes_in_buffer )
+
+/* Reload the local copies --- used only in MAKE_BYTE_AVAIL */
+#define INPUT_RELOAD(cinfo)  \
+	( next_input_byte = datasrc->next_input_byte,  \
+	  bytes_in_buffer = datasrc->bytes_in_buffer )
+
+/* Internal macro for INPUT_BYTE and INPUT_2BYTES: make a byte available.
+ * Note we do *not* do INPUT_SYNC before calling fill_input_buffer,
+ * but we must reload the local copies after a successful fill.
+ */
+#define MAKE_BYTE_AVAIL(cinfo,action)  \
+	if (bytes_in_buffer == 0) {  \
+	  if (! (*datasrc->fill_input_buffer) (cinfo))  \
+	    { action; }  \
+	  INPUT_RELOAD(cinfo);  \
+	}
+
+/* Read a byte into variable V.
+ * If must suspend, take the specified action (typically "return FALSE").
+ */
+#define INPUT_BYTE(cinfo,V,action)  \
+	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+		  bytes_in_buffer--; \
+		  V = GETJOCTET(*next_input_byte++); )
+
+/* As above, but read two bytes interpreted as an unsigned 16-bit integer.
+ * V should be declared unsigned int or perhaps INT32.
+ */
+#define INPUT_2BYTES(cinfo,V,action)  \
+	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+		  bytes_in_buffer--; \
+		  V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \
+		  MAKE_BYTE_AVAIL(cinfo,action); \
+		  bytes_in_buffer--; \
+		  V += GETJOCTET(*next_input_byte++); )
+
+
+/*
+ * Routines to process JPEG markers.
+ *
+ * Entry condition: JPEG marker itself has been read and its code saved
+ *   in cinfo->unread_marker; input restart point is just after the marker.
+ *
+ * Exit: if return TRUE, have read and processed any parameters, and have
+ *   updated the restart point to point after the parameters.
+ *   If return FALSE, was forced to suspend before reaching end of
+ *   marker parameters; restart point has not been moved.  Same routine
+ *   will be called again after application supplies more input data.
+ *
+ * This approach to suspension assumes that all of a marker's parameters
+ * can fit into a single input bufferload.  This should hold for "normal"
+ * markers.  Some COM/APPn markers might have large parameter segments
+ * that might not fit.  If we are simply dropping such a marker, we use
+ * skip_input_data to get past it, and thereby put the problem on the
+ * source manager's shoulders.  If we are saving the marker's contents
+ * into memory, we use a slightly different convention: when forced to
+ * suspend, the marker processor updates the restart point to the end of
+ * what it's consumed (ie, the end of the buffer) before returning FALSE.
+ * On resumption, cinfo->unread_marker still contains the marker code,
+ * but the data source will point to the next chunk of marker data.
+ * The marker processor must retain internal state to deal with this.
+ *
+ * Note that we don't bother to avoid duplicate trace messages if a
+ * suspension occurs within marker parameters.  Other side effects
+ * require more care.
+ */
+
+
+LOCAL(boolean)
+get_soi (j_decompress_ptr cinfo)
+/* Process an SOI marker */
+{
+  int i;
+  
+  TRACEMS(cinfo, 1, JTRC_SOI);
+
+  if (cinfo->marker->saw_SOI)
+    ERREXIT(cinfo, JERR_SOI_DUPLICATE);
+
+  /* Reset all parameters that are defined to be reset by SOI */
+
+  for (i = 0; i < NUM_ARITH_TBLS; i++) {
+    cinfo->arith_dc_L[i] = 0;
+    cinfo->arith_dc_U[i] = 1;
+    cinfo->arith_ac_K[i] = 5;
+  }
+  cinfo->restart_interval = 0;
+
+  /* Set initial assumptions for colorspace etc */
+
+  cinfo->jpeg_color_space = JCS_UNKNOWN;
+  cinfo->CCIR601_sampling = FALSE; /* Assume non-CCIR sampling??? */
+
+  cinfo->saw_JFIF_marker = FALSE;
+  cinfo->JFIF_major_version = 1; /* set default JFIF APP0 values */
+  cinfo->JFIF_minor_version = 1;
+  cinfo->density_unit = 0;
+  cinfo->X_density = 1;
+  cinfo->Y_density = 1;
+  cinfo->saw_Adobe_marker = FALSE;
+  cinfo->Adobe_transform = 0;
+
+  cinfo->marker->saw_SOI = TRUE;
+
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_sof (j_decompress_ptr cinfo, boolean is_prog, boolean is_arith)
+/* Process a SOFn marker */
+{
+  INT32 length;
+  int c, ci;
+  jpeg_component_info * compptr;
+  INPUT_VARS(cinfo);
+
+  cinfo->progressive_mode = is_prog;
+  cinfo->arith_code = is_arith;
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  INPUT_BYTE(cinfo, cinfo->data_precision, return FALSE);
+  INPUT_2BYTES(cinfo, cinfo->image_height, return FALSE);
+  INPUT_2BYTES(cinfo, cinfo->image_width, return FALSE);
+  INPUT_BYTE(cinfo, cinfo->num_components, return FALSE);
+
+  length -= 8;
+
+  TRACEMS4(cinfo, 1, JTRC_SOF, cinfo->unread_marker,
+	   (int) cinfo->image_width, (int) cinfo->image_height,
+	   cinfo->num_components);
+
+  if (cinfo->marker->saw_SOF)
+    ERREXIT(cinfo, JERR_SOF_DUPLICATE);
+
+  /* We don't support files in which the image height is initially specified */
+  /* as 0 and is later redefined by DNL.  As long as we have to check that,  */
+  /* might as well have a general sanity check. */
+  if (cinfo->image_height <= 0 || cinfo->image_width <= 0
+      || cinfo->num_components <= 0)
+    ERREXIT(cinfo, JERR_EMPTY_IMAGE);
+
+  if (length != (cinfo->num_components * 3))
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  if (cinfo->comp_info == NULL)	/* do only once, even if suspend */
+    cinfo->comp_info = (jpeg_component_info *) (*cinfo->mem->alloc_small)
+			((j_common_ptr) cinfo, JPOOL_IMAGE,
+			 cinfo->num_components * SIZEOF(jpeg_component_info));
+  
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->component_index = ci;
+    INPUT_BYTE(cinfo, compptr->component_id, return FALSE);
+    INPUT_BYTE(cinfo, c, return FALSE);
+    compptr->h_samp_factor = (c >> 4) & 15;
+    compptr->v_samp_factor = (c     ) & 15;
+    INPUT_BYTE(cinfo, compptr->quant_tbl_no, return FALSE);
+
+    TRACEMS4(cinfo, 1, JTRC_SOF_COMPONENT,
+	     compptr->component_id, compptr->h_samp_factor,
+	     compptr->v_samp_factor, compptr->quant_tbl_no);
+  }
+
+  cinfo->marker->saw_SOF = TRUE;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_sos (j_decompress_ptr cinfo)
+/* Process a SOS marker */
+{
+  INT32 length;
+  int i, ci, n, c, cc;
+  jpeg_component_info * compptr;
+  INPUT_VARS(cinfo);
+
+  if (! cinfo->marker->saw_SOF)
+    ERREXIT(cinfo, JERR_SOS_NO_SOF);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+
+  INPUT_BYTE(cinfo, n, return FALSE); /* Number of components */
+
+  TRACEMS1(cinfo, 1, JTRC_SOS, n);
+
+  if (length != (n * 2 + 6) || n < 1 || n > MAX_COMPS_IN_SCAN)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  cinfo->comps_in_scan = n;
+
+  /* Collect the component-spec parameters */
+
+  for (i = 0; i < n; i++) {
+    INPUT_BYTE(cinfo, cc, return FALSE);
+    INPUT_BYTE(cinfo, c, return FALSE);
+    
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      if (cc == compptr->component_id)
+	goto id_found;
+    }
+
+    ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
+
+  id_found:
+
+    cinfo->cur_comp_info[i] = compptr;
+    compptr->dc_tbl_no = (c >> 4) & 15;
+    compptr->ac_tbl_no = (c     ) & 15;
+    
+    TRACEMS3(cinfo, 1, JTRC_SOS_COMPONENT, cc,
+	     compptr->dc_tbl_no, compptr->ac_tbl_no);
+  }
+
+  /* Collect the additional scan parameters Ss, Se, Ah/Al. */
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Ss = c;
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Se = c;
+  INPUT_BYTE(cinfo, c, return FALSE);
+  cinfo->Ah = (c >> 4) & 15;
+  cinfo->Al = (c     ) & 15;
+
+  TRACEMS4(cinfo, 1, JTRC_SOS_PARAMS, cinfo->Ss, cinfo->Se,
+	   cinfo->Ah, cinfo->Al);
+
+  /* Prepare to scan data & restart markers */
+  cinfo->marker->next_restart_num = 0;
+
+  /* Count another SOS marker */
+  cinfo->input_scan_number++;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+#ifdef D_ARITH_CODING_SUPPORTED
+
+LOCAL(boolean)
+get_dac (j_decompress_ptr cinfo)
+/* Process a DAC marker */
+{
+  INT32 length;
+  int index, val;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+  
+  while (length > 0) {
+    INPUT_BYTE(cinfo, index, return FALSE);
+    INPUT_BYTE(cinfo, val, return FALSE);
+
+    length -= 2;
+
+    TRACEMS2(cinfo, 1, JTRC_DAC, index, val);
+
+    if (index < 0 || index >= (2*NUM_ARITH_TBLS))
+      ERREXIT1(cinfo, JERR_DAC_INDEX, index);
+
+    if (index >= NUM_ARITH_TBLS) { /* define AC table */
+      cinfo->arith_ac_K[index-NUM_ARITH_TBLS] = (UINT8) val;
+    } else {			/* define DC table */
+      cinfo->arith_dc_L[index] = (UINT8) (val & 0x0F);
+      cinfo->arith_dc_U[index] = (UINT8) (val >> 4);
+      if (cinfo->arith_dc_L[index] > cinfo->arith_dc_U[index])
+	ERREXIT1(cinfo, JERR_DAC_VALUE, val);
+    }
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+#else /* ! D_ARITH_CODING_SUPPORTED */
+
+#define get_dac(cinfo)  skip_variable(cinfo)
+
+#endif /* D_ARITH_CODING_SUPPORTED */
+
+
+LOCAL(boolean)
+get_dht (j_decompress_ptr cinfo)
+/* Process a DHT marker */
+{
+  INT32 length;
+  UINT8 bits[17];
+  UINT8 huffval[256];
+  int i, index, count;
+  JHUFF_TBL **htblptr;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+  
+  while (length > 16) {
+    INPUT_BYTE(cinfo, index, return FALSE);
+
+    TRACEMS1(cinfo, 1, JTRC_DHT, index);
+      
+    bits[0] = 0;
+    count = 0;
+    for (i = 1; i <= 16; i++) {
+      INPUT_BYTE(cinfo, bits[i], return FALSE);
+      count += bits[i];
+    }
+
+    length -= 1 + 16;
+
+    TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
+	     bits[1], bits[2], bits[3], bits[4],
+	     bits[5], bits[6], bits[7], bits[8]);
+    TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
+	     bits[9], bits[10], bits[11], bits[12],
+	     bits[13], bits[14], bits[15], bits[16]);
+
+    /* Here we just do minimal validation of the counts to avoid walking
+     * off the end of our table space.  jdhuff.c will check more carefully.
+     */
+    if (count > 256 || ((INT32) count) > length)
+      ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+    for (i = 0; i < count; i++)
+      INPUT_BYTE(cinfo, huffval[i], return FALSE);
+
+    length -= count;
+
+    if (index & 0x10) {		/* AC table definition */
+      index -= 0x10;
+      htblptr = &cinfo->ac_huff_tbl_ptrs[index];
+    } else {			/* DC table definition */
+      htblptr = &cinfo->dc_huff_tbl_ptrs[index];
+    }
+
+    if (index < 0 || index >= NUM_HUFF_TBLS)
+      ERREXIT1(cinfo, JERR_DHT_INDEX, index);
+
+    if (*htblptr == NULL)
+      *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+  
+    MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
+    MEMCOPY((*htblptr)->huffval, huffval, SIZEOF((*htblptr)->huffval));
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_dqt (j_decompress_ptr cinfo)
+/* Process a DQT marker */
+{
+  INT32 length;
+  int n, i, prec;
+  unsigned int tmp;
+  JQUANT_TBL *quant_ptr;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  while (length > 0) {
+    INPUT_BYTE(cinfo, n, return FALSE);
+    prec = n >> 4;
+    n &= 0x0F;
+
+    TRACEMS2(cinfo, 1, JTRC_DQT, n, prec);
+
+    if (n >= NUM_QUANT_TBLS)
+      ERREXIT1(cinfo, JERR_DQT_INDEX, n);
+      
+    if (cinfo->quant_tbl_ptrs[n] == NULL)
+      cinfo->quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) cinfo);
+    quant_ptr = cinfo->quant_tbl_ptrs[n];
+
+    for (i = 0; i < DCTSIZE2; i++) {
+      if (prec)
+	INPUT_2BYTES(cinfo, tmp, return FALSE);
+      else
+	INPUT_BYTE(cinfo, tmp, return FALSE);
+      /* We convert the zigzag-order table to natural array order. */
+      quant_ptr->quantval[jpeg_natural_order[i]] = (UINT16) tmp;
+    }
+
+    if (cinfo->err->trace_level >= 2) {
+      for (i = 0; i < DCTSIZE2; i += 8) {
+	TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
+		 quant_ptr->quantval[i],   quant_ptr->quantval[i+1],
+		 quant_ptr->quantval[i+2], quant_ptr->quantval[i+3],
+		 quant_ptr->quantval[i+4], quant_ptr->quantval[i+5],
+		 quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]);
+      }
+    }
+
+    length -= DCTSIZE2+1;
+    if (prec) length -= DCTSIZE2;
+  }
+
+  if (length != 0)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+get_dri (j_decompress_ptr cinfo)
+/* Process a DRI marker */
+{
+  INT32 length;
+  unsigned int tmp;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  
+  if (length != 4)
+    ERREXIT(cinfo, JERR_BAD_LENGTH);
+
+  INPUT_2BYTES(cinfo, tmp, return FALSE);
+
+  TRACEMS1(cinfo, 1, JTRC_DRI, tmp);
+
+  cinfo->restart_interval = tmp;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Routines for processing APPn and COM markers.
+ * These are either saved in memory or discarded, per application request.
+ * APP0 and APP14 are specially checked to see if they are
+ * JFIF and Adobe markers, respectively.
+ */
+
+#define APP0_DATA_LEN	14	/* Length of interesting data in APP0 */
+#define APP14_DATA_LEN	12	/* Length of interesting data in APP14 */
+#define APPN_DATA_LEN	14	/* Must be the largest of the above!! */
+
+
+LOCAL(void)
+examine_app0 (j_decompress_ptr cinfo, JOCTET FAR * data,
+	      unsigned int datalen, INT32 remaining)
+/* Examine first few bytes from an APP0.
+ * Take appropriate action if it is a JFIF marker.
+ * datalen is # of bytes at data[], remaining is length of rest of marker data.
+ */
+{
+  INT32 totallen = (INT32) datalen + remaining;
+
+  if (datalen >= APP0_DATA_LEN &&
+      GETJOCTET(data[0]) == 0x4A &&
+      GETJOCTET(data[1]) == 0x46 &&
+      GETJOCTET(data[2]) == 0x49 &&
+      GETJOCTET(data[3]) == 0x46 &&
+      GETJOCTET(data[4]) == 0) {
+    /* Found JFIF APP0 marker: save info */
+    cinfo->saw_JFIF_marker = TRUE;
+    cinfo->JFIF_major_version = GETJOCTET(data[5]);
+    cinfo->JFIF_minor_version = GETJOCTET(data[6]);
+    cinfo->density_unit = GETJOCTET(data[7]);
+    cinfo->X_density = (GETJOCTET(data[8]) << 8) + GETJOCTET(data[9]);
+    cinfo->Y_density = (GETJOCTET(data[10]) << 8) + GETJOCTET(data[11]);
+    /* Check version.
+     * Major version must be 1, anything else signals an incompatible change.
+     * (We used to treat this as an error, but now it's a nonfatal warning,
+     * because some bozo at Hijaak couldn't read the spec.)
+     * Minor version should be 0..2, but process anyway if newer.
+     */
+    if (cinfo->JFIF_major_version != 1)
+      WARNMS2(cinfo, JWRN_JFIF_MAJOR,
+	      cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
+    /* Generate trace messages */
+    TRACEMS5(cinfo, 1, JTRC_JFIF,
+	     cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
+	     cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
+    /* Validate thumbnail dimensions and issue appropriate messages */
+    if (GETJOCTET(data[12]) | GETJOCTET(data[13]))
+      TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
+	       GETJOCTET(data[12]), GETJOCTET(data[13]));
+    totallen -= APP0_DATA_LEN;
+    if (totallen !=
+	((INT32)GETJOCTET(data[12]) * (INT32)GETJOCTET(data[13]) * (INT32) 3))
+      TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int) totallen);
+  } else if (datalen >= 6 &&
+      GETJOCTET(data[0]) == 0x4A &&
+      GETJOCTET(data[1]) == 0x46 &&
+      GETJOCTET(data[2]) == 0x58 &&
+      GETJOCTET(data[3]) == 0x58 &&
+      GETJOCTET(data[4]) == 0) {
+    /* Found JFIF "JFXX" extension APP0 marker */
+    /* The library doesn't actually do anything with these,
+     * but we try to produce a helpful trace message.
+     */
+    switch (GETJOCTET(data[5])) {
+    case 0x10:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_JPEG, (int) totallen);
+      break;
+    case 0x11:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_PALETTE, (int) totallen);
+      break;
+    case 0x13:
+      TRACEMS1(cinfo, 1, JTRC_THUMB_RGB, (int) totallen);
+      break;
+    default:
+      TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION,
+	       GETJOCTET(data[5]), (int) totallen);
+      break;
+    }
+  } else {
+    /* Start of APP0 does not match "JFIF" or "JFXX", or too short */
+    TRACEMS1(cinfo, 1, JTRC_APP0, (int) totallen);
+  }
+}
+
+
+LOCAL(void)
+examine_app14 (j_decompress_ptr cinfo, JOCTET FAR * data,
+	       unsigned int datalen, INT32 remaining)
+/* Examine first few bytes from an APP14.
+ * Take appropriate action if it is an Adobe marker.
+ * datalen is # of bytes at data[], remaining is length of rest of marker data.
+ */
+{
+  unsigned int version, flags0, flags1, transform;
+
+  if (datalen >= APP14_DATA_LEN &&
+      GETJOCTET(data[0]) == 0x41 &&
+      GETJOCTET(data[1]) == 0x64 &&
+      GETJOCTET(data[2]) == 0x6F &&
+      GETJOCTET(data[3]) == 0x62 &&
+      GETJOCTET(data[4]) == 0x65) {
+    /* Found Adobe APP14 marker */
+    version = (GETJOCTET(data[5]) << 8) + GETJOCTET(data[6]);
+    flags0 = (GETJOCTET(data[7]) << 8) + GETJOCTET(data[8]);
+    flags1 = (GETJOCTET(data[9]) << 8) + GETJOCTET(data[10]);
+    transform = GETJOCTET(data[11]);
+    TRACEMS4(cinfo, 1, JTRC_ADOBE, version, flags0, flags1, transform);
+    cinfo->saw_Adobe_marker = TRUE;
+    cinfo->Adobe_transform = (UINT8) transform;
+  } else {
+    /* Start of APP14 does not match "Adobe", or too short */
+    TRACEMS1(cinfo, 1, JTRC_APP14, (int) (datalen + remaining));
+  }
+}
+
+
+METHODDEF(boolean)
+get_interesting_appn (j_decompress_ptr cinfo)
+/* Process an APP0 or APP14 marker without saving it */
+{
+  INT32 length;
+  JOCTET b[APPN_DATA_LEN];
+  unsigned int i, numtoread;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+
+  /* get the interesting part of the marker data */
+  if (length >= APPN_DATA_LEN)
+    numtoread = APPN_DATA_LEN;
+  else if (length > 0)
+    numtoread = (unsigned int) length;
+  else
+    numtoread = 0;
+  for (i = 0; i < numtoread; i++)
+    INPUT_BYTE(cinfo, b[i], return FALSE);
+  length -= numtoread;
+
+  /* process it */
+  switch (cinfo->unread_marker) {
+  case M_APP0:
+    examine_app0(cinfo, (JOCTET FAR *) b, numtoread, length);
+    break;
+  case M_APP14:
+    examine_app14(cinfo, (JOCTET FAR *) b, numtoread, length);
+    break;
+  default:
+    /* can't get here unless jpeg_save_markers chooses wrong processor */
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker);
+    break;
+  }
+
+  /* skip any remaining data -- could be lots */
+  INPUT_SYNC(cinfo);
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long) length);
+
+  return TRUE;
+}
+
+
+#ifdef SAVE_MARKERS_SUPPORTED
+
+METHODDEF(boolean)
+save_marker (j_decompress_ptr cinfo)
+/* Save an APPn or COM marker into the marker list */
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+  jpeg_saved_marker_ptr cur_marker = marker->cur_marker;
+  unsigned int bytes_read, data_length;
+  JOCTET FAR * data;
+  INT32 length = 0;
+  INPUT_VARS(cinfo);
+
+  if (cur_marker == NULL) {
+    /* begin reading a marker */
+    INPUT_2BYTES(cinfo, length, return FALSE);
+    length -= 2;
+    if (length >= 0) {		/* watch out for bogus length word */
+      /* figure out how much we want to save */
+      unsigned int limit;
+      if (cinfo->unread_marker == (int) M_COM)
+	limit = marker->length_limit_COM;
+      else
+	limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0];
+      if ((unsigned int) length < limit)
+	limit = (unsigned int) length;
+      /* allocate and initialize the marker item */
+      cur_marker = (jpeg_saved_marker_ptr)
+	(*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				    SIZEOF(struct jpeg_marker_struct) + limit);
+      cur_marker->next = NULL;
+      cur_marker->marker = (UINT8) cinfo->unread_marker;
+      cur_marker->original_length = (unsigned int) length;
+      cur_marker->data_length = limit;
+      /* data area is just beyond the jpeg_marker_struct */
+      data = cur_marker->data = (JOCTET FAR *) (cur_marker + 1);
+      marker->cur_marker = cur_marker;
+      marker->bytes_read = 0;
+      bytes_read = 0;
+      data_length = limit;
+    } else {
+      /* deal with bogus length word */
+      bytes_read = data_length = 0;
+      data = NULL;
+    }
+  } else {
+    /* resume reading a marker */
+    bytes_read = marker->bytes_read;
+    data_length = cur_marker->data_length;
+    data = cur_marker->data + bytes_read;
+  }
+
+  while (bytes_read < data_length) {
+    INPUT_SYNC(cinfo);		/* move the restart point to here */
+    marker->bytes_read = bytes_read;
+    /* If there's not at least one byte in buffer, suspend */
+    MAKE_BYTE_AVAIL(cinfo, return FALSE);
+    /* Copy bytes with reasonable rapidity */
+    while (bytes_read < data_length && bytes_in_buffer > 0) {
+      *data++ = *next_input_byte++;
+      bytes_in_buffer--;
+      bytes_read++;
+    }
+  }
+
+  /* Done reading what we want to read */
+  if (cur_marker != NULL) {	/* will be NULL if bogus length word */
+    /* Add new marker to end of list */
+    if (cinfo->marker_list == NULL) {
+      cinfo->marker_list = cur_marker;
+    } else {
+      jpeg_saved_marker_ptr prev = cinfo->marker_list;
+      while (prev->next != NULL)
+	prev = prev->next;
+      prev->next = cur_marker;
+    }
+    /* Reset pointer & calc remaining data length */
+    data = cur_marker->data;
+    length = cur_marker->original_length - data_length;
+  }
+  /* Reset to initial state for next marker */
+  marker->cur_marker = NULL;
+
+  /* Process the marker if interesting; else just make a generic trace msg */
+  switch (cinfo->unread_marker) {
+  case M_APP0:
+    examine_app0(cinfo, data, data_length, length);
+    break;
+  case M_APP14:
+    examine_app14(cinfo, data, data_length, length);
+    break;
+  default:
+    TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker,
+	     (int) (data_length + length));
+    break;
+  }
+
+  /* skip any remaining data -- could be lots */
+  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long) length);
+
+  return TRUE;
+}
+
+#endif /* SAVE_MARKERS_SUPPORTED */
+
+
+METHODDEF(boolean)
+skip_variable (j_decompress_ptr cinfo)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  INT32 length;
+  INPUT_VARS(cinfo);
+
+  INPUT_2BYTES(cinfo, length, return FALSE);
+  length -= 2;
+  
+  TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, (int) length);
+
+  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  if (length > 0)
+    (*cinfo->src->skip_input_data) (cinfo, (long) length);
+
+  return TRUE;
+}
+
+
+/*
+ * Find the next JPEG marker, save it in cinfo->unread_marker.
+ * Returns FALSE if had to suspend before reaching a marker;
+ * in that case cinfo->unread_marker is unchanged.
+ *
+ * Note that the result might not be a valid marker code,
+ * but it will never be 0 or FF.
+ */
+
+LOCAL(boolean)
+next_marker (j_decompress_ptr cinfo)
+{
+  int c;
+  INPUT_VARS(cinfo);
+
+  for (;;) {
+    INPUT_BYTE(cinfo, c, return FALSE);
+    /* Skip any non-FF bytes.
+     * This may look a bit inefficient, but it will not occur in a valid file.
+     * We sync after each discarded byte so that a suspending data source
+     * can discard the byte from its buffer.
+     */
+    while (c != 0xFF) {
+      cinfo->marker->discarded_bytes++;
+      INPUT_SYNC(cinfo);
+      INPUT_BYTE(cinfo, c, return FALSE);
+    }
+    /* This loop swallows any duplicate FF bytes.  Extra FFs are legal as
+     * pad bytes, so don't count them in discarded_bytes.  We assume there
+     * will not be so many consecutive FF bytes as to overflow a suspending
+     * data source's input buffer.
+     */
+    do {
+      INPUT_BYTE(cinfo, c, return FALSE);
+    } while (c == 0xFF);
+    if (c != 0)
+      break;			/* found a valid marker, exit loop */
+    /* Reach here if we found a stuffed-zero data sequence (FF/00).
+     * Discard it and loop back to try again.
+     */
+    cinfo->marker->discarded_bytes += 2;
+    INPUT_SYNC(cinfo);
+  }
+
+  if (cinfo->marker->discarded_bytes != 0) {
+    WARNMS2(cinfo, JWRN_EXTRANEOUS_DATA, cinfo->marker->discarded_bytes, c);
+    cinfo->marker->discarded_bytes = 0;
+  }
+
+  cinfo->unread_marker = c;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+LOCAL(boolean)
+first_marker (j_decompress_ptr cinfo)
+/* Like next_marker, but used to obtain the initial SOI marker. */
+/* For this marker, we do not allow preceding garbage or fill; otherwise,
+ * we might well scan an entire input file before realizing it ain't JPEG.
+ * If an application wants to process non-JFIF files, it must seek to the
+ * SOI before calling the JPEG library.
+ */
+{
+  int c, c2;
+  INPUT_VARS(cinfo);
+
+  INPUT_BYTE(cinfo, c, return FALSE);
+  INPUT_BYTE(cinfo, c2, return FALSE);
+  if (c != 0xFF || c2 != (int) M_SOI)
+    ERREXIT2(cinfo, JERR_NO_SOI, c, c2);
+
+  cinfo->unread_marker = c2;
+
+  INPUT_SYNC(cinfo);
+  return TRUE;
+}
+
+
+/*
+ * Read markers until SOS or EOI.
+ *
+ * Returns same codes as are defined for jpeg_consume_input:
+ * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+ */
+
+METHODDEF(int)
+read_markers (j_decompress_ptr cinfo)
+{
+  /* Outer loop repeats once for each marker. */
+  for (;;) {
+    /* Collect the marker proper, unless we already did. */
+    /* NB: first_marker() enforces the requirement that SOI appear first. */
+    if (cinfo->unread_marker == 0) {
+      if (! cinfo->marker->saw_SOI) {
+	if (! first_marker(cinfo))
+	  return JPEG_SUSPENDED;
+      } else {
+	if (! next_marker(cinfo))
+	  return JPEG_SUSPENDED;
+      }
+    }
+    /* At this point cinfo->unread_marker contains the marker code and the
+     * input point is just past the marker proper, but before any parameters.
+     * A suspension will cause us to return with this state still true.
+     */
+    switch (cinfo->unread_marker) {
+    case M_SOI:
+      if (! get_soi(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF0:		/* Baseline */
+    case M_SOF1:		/* Extended sequential, Huffman */
+      if (! get_sof(cinfo, FALSE, FALSE))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF2:		/* Progressive, Huffman */
+      if (! get_sof(cinfo, TRUE, FALSE))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF9:		/* Extended sequential, arithmetic */
+      if (! get_sof(cinfo, FALSE, TRUE))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_SOF10:		/* Progressive, arithmetic */
+      if (! get_sof(cinfo, TRUE, TRUE))
+	return JPEG_SUSPENDED;
+      break;
+
+    /* Currently unsupported SOFn types */
+    case M_SOF3:		/* Lossless, Huffman */
+    case M_SOF5:		/* Differential sequential, Huffman */
+    case M_SOF6:		/* Differential progressive, Huffman */
+    case M_SOF7:		/* Differential lossless, Huffman */
+    case M_JPG:			/* Reserved for JPEG extensions */
+    case M_SOF11:		/* Lossless, arithmetic */
+    case M_SOF13:		/* Differential sequential, arithmetic */
+    case M_SOF14:		/* Differential progressive, arithmetic */
+    case M_SOF15:		/* Differential lossless, arithmetic */
+      ERREXIT1(cinfo, JERR_SOF_UNSUPPORTED, cinfo->unread_marker);
+      break;
+
+    case M_SOS:
+      if (! get_sos(cinfo))
+	return JPEG_SUSPENDED;
+      cinfo->unread_marker = 0;	/* processed the marker */
+      return JPEG_REACHED_SOS;
+    
+    case M_EOI:
+      TRACEMS(cinfo, 1, JTRC_EOI);
+      cinfo->unread_marker = 0;	/* processed the marker */
+      return JPEG_REACHED_EOI;
+      
+    case M_DAC:
+      if (! get_dac(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_DHT:
+      if (! get_dht(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_DQT:
+      if (! get_dqt(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_DRI:
+      if (! get_dri(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_APP0:
+    case M_APP1:
+    case M_APP2:
+    case M_APP3:
+    case M_APP4:
+    case M_APP5:
+    case M_APP6:
+    case M_APP7:
+    case M_APP8:
+    case M_APP9:
+    case M_APP10:
+    case M_APP11:
+    case M_APP12:
+    case M_APP13:
+    case M_APP14:
+    case M_APP15:
+      if (! (*((my_marker_ptr) cinfo->marker)->process_APPn[
+		cinfo->unread_marker - (int) M_APP0]) (cinfo))
+	return JPEG_SUSPENDED;
+      break;
+      
+    case M_COM:
+      if (! (*((my_marker_ptr) cinfo->marker)->process_COM) (cinfo))
+	return JPEG_SUSPENDED;
+      break;
+
+    case M_RST0:		/* these are all parameterless */
+    case M_RST1:
+    case M_RST2:
+    case M_RST3:
+    case M_RST4:
+    case M_RST5:
+    case M_RST6:
+    case M_RST7:
+    case M_TEM:
+      TRACEMS1(cinfo, 1, JTRC_PARMLESS_MARKER, cinfo->unread_marker);
+      break;
+
+    case M_DNL:			/* Ignore DNL ... perhaps the wrong thing */
+      if (! skip_variable(cinfo))
+	return JPEG_SUSPENDED;
+      break;
+
+    default:			/* must be DHP, EXP, JPGn, or RESn */
+      /* For now, we treat the reserved markers as fatal errors since they are
+       * likely to be used to signal incompatible JPEG Part 3 extensions.
+       * Once the JPEG 3 version-number marker is well defined, this code
+       * ought to change!
+       */
+      ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, cinfo->unread_marker);
+      break;
+    }
+    /* Successfully processed marker, so reset state variable */
+    cinfo->unread_marker = 0;
+  } /* end loop */
+}
+
+
+/*
+ * Read a restart marker, which is expected to appear next in the datastream;
+ * if the marker is not there, take appropriate recovery action.
+ * Returns FALSE if suspension is required.
+ *
+ * This is called by the entropy decoder after it has read an appropriate
+ * number of MCUs.  cinfo->unread_marker may be nonzero if the entropy decoder
+ * has already read a marker from the data source.  Under normal conditions
+ * cinfo->unread_marker will be reset to 0 before returning; if not reset,
+ * it holds a marker which the decoder will be unable to read past.
+ */
+
+METHODDEF(boolean)
+read_restart_marker (j_decompress_ptr cinfo)
+{
+  /* Obtain a marker unless we already did. */
+  /* Note that next_marker will complain if it skips any data. */
+  if (cinfo->unread_marker == 0) {
+    if (! next_marker(cinfo))
+      return FALSE;
+  }
+
+  if (cinfo->unread_marker ==
+      ((int) M_RST0 + cinfo->marker->next_restart_num)) {
+    /* Normal case --- swallow the marker and let entropy decoder continue */
+    TRACEMS1(cinfo, 3, JTRC_RST, cinfo->marker->next_restart_num);
+    cinfo->unread_marker = 0;
+  } else {
+    /* Uh-oh, the restart markers have been messed up. */
+    /* Let the data source manager determine how to resync. */
+    if (! (*cinfo->src->resync_to_restart) (cinfo,
+					    cinfo->marker->next_restart_num))
+      return FALSE;
+  }
+
+  /* Update next-restart state */
+  cinfo->marker->next_restart_num = (cinfo->marker->next_restart_num + 1) & 7;
+
+  return TRUE;
+}
+
+
+/*
+ * This is the default resync_to_restart method for data source managers
+ * to use if they don't have any better approach.  Some data source managers
+ * may be able to back up, or may have additional knowledge about the data
+ * which permits a more intelligent recovery strategy; such managers would
+ * presumably supply their own resync method.
+ *
+ * read_restart_marker calls resync_to_restart if it finds a marker other than
+ * the restart marker it was expecting.  (This code is *not* used unless
+ * a nonzero restart interval has been declared.)  cinfo->unread_marker is
+ * the marker code actually found (might be anything, except 0 or FF).
+ * The desired restart marker number (0..7) is passed as a parameter.
+ * This routine is supposed to apply whatever error recovery strategy seems
+ * appropriate in order to position the input stream to the next data segment.
+ * Note that cinfo->unread_marker is treated as a marker appearing before
+ * the current data-source input point; usually it should be reset to zero
+ * before returning.
+ * Returns FALSE if suspension is required.
+ *
+ * This implementation is substantially constrained by wanting to treat the
+ * input as a data stream; this means we can't back up.  Therefore, we have
+ * only the following actions to work with:
+ *   1. Simply discard the marker and let the entropy decoder resume at next
+ *      byte of file.
+ *   2. Read forward until we find another marker, discarding intervening
+ *      data.  (In theory we could look ahead within the current bufferload,
+ *      without having to discard data if we don't find the desired marker.
+ *      This idea is not implemented here, in part because it makes behavior
+ *      dependent on buffer size and chance buffer-boundary positions.)
+ *   3. Leave the marker unread (by failing to zero cinfo->unread_marker).
+ *      This will cause the entropy decoder to process an empty data segment,
+ *      inserting dummy zeroes, and then we will reprocess the marker.
+ *
+ * #2 is appropriate if we think the desired marker lies ahead, while #3 is
+ * appropriate if the found marker is a future restart marker (indicating
+ * that we have missed the desired restart marker, probably because it got
+ * corrupted).
+ * We apply #2 or #3 if the found marker is a restart marker no more than
+ * two counts behind or ahead of the expected one.  We also apply #2 if the
+ * found marker is not a legal JPEG marker code (it's certainly bogus data).
+ * If the found marker is a restart marker more than 2 counts away, we do #1
+ * (too much risk that the marker is erroneous; with luck we will be able to
+ * resync at some future point).
+ * For any valid non-restart JPEG marker, we apply #3.  This keeps us from
+ * overrunning the end of a scan.  An implementation limited to single-scan
+ * files might find it better to apply #2 for markers other than EOI, since
+ * any other marker would have to be bogus data in that case.
+ */
+
+GLOBAL(boolean)
+jpeg_resync_to_restart (j_decompress_ptr cinfo, int desired)
+{
+  int marker = cinfo->unread_marker;
+  int action = 1;
+  
+  /* Always put up a warning. */
+  WARNMS2(cinfo, JWRN_MUST_RESYNC, marker, desired);
+  
+  /* Outer loop handles repeated decision after scanning forward. */
+  for (;;) {
+    if (marker < (int) M_SOF0)
+      action = 2;		/* invalid marker */
+    else if (marker < (int) M_RST0 || marker > (int) M_RST7)
+      action = 3;		/* valid non-restart marker */
+    else {
+      if (marker == ((int) M_RST0 + ((desired+1) & 7)) ||
+	  marker == ((int) M_RST0 + ((desired+2) & 7)))
+	action = 3;		/* one of the next two expected restarts */
+      else if (marker == ((int) M_RST0 + ((desired-1) & 7)) ||
+	       marker == ((int) M_RST0 + ((desired-2) & 7)))
+	action = 2;		/* a prior restart, so advance */
+      else
+	action = 1;		/* desired restart or too far away */
+    }
+    TRACEMS2(cinfo, 4, JTRC_RECOVERY_ACTION, marker, action);
+    switch (action) {
+    case 1:
+      /* Discard marker and let entropy decoder resume processing. */
+      cinfo->unread_marker = 0;
+      return TRUE;
+    case 2:
+      /* Scan to the next marker, and repeat the decision loop. */
+      if (! next_marker(cinfo))
+	return FALSE;
+      marker = cinfo->unread_marker;
+      break;
+    case 3:
+      /* Return without advancing past this marker. */
+      /* Entropy decoder will be forced to process an empty segment. */
+      return TRUE;
+    }
+  } /* end loop */
+}
+
+
+/*
+ * Reset marker processing state to begin a fresh datastream.
+ */
+
+METHODDEF(void)
+reset_marker_reader (j_decompress_ptr cinfo)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+
+  cinfo->comp_info = NULL;		/* until allocated by get_sof */
+  cinfo->input_scan_number = 0;		/* no SOS seen yet */
+  cinfo->unread_marker = 0;		/* no pending marker */
+  marker->pub.saw_SOI = FALSE;		/* set internal state too */
+  marker->pub.saw_SOF = FALSE;
+  marker->pub.discarded_bytes = 0;
+  marker->cur_marker = NULL;
+}
+
+
+/*
+ * Initialize the marker reader module.
+ * This is called only once, when the decompression object is created.
+ */
+
+GLOBAL(void)
+jinit_marker_reader (j_decompress_ptr cinfo)
+{
+  my_marker_ptr marker;
+  int i;
+
+  /* Create subobject in permanent pool */
+  marker = (my_marker_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+				SIZEOF(my_marker_reader));
+  cinfo->marker = (struct jpeg_marker_reader *) marker;
+  /* Initialize public method pointers */
+  marker->pub.reset_marker_reader = reset_marker_reader;
+  marker->pub.read_markers = read_markers;
+  marker->pub.read_restart_marker = read_restart_marker;
+  /* Initialize COM/APPn processing.
+   * By default, we examine and then discard APP0 and APP14,
+   * but simply discard COM and all other APPn.
+   */
+  marker->process_COM = skip_variable;
+  marker->length_limit_COM = 0;
+  for (i = 0; i < 16; i++) {
+    marker->process_APPn[i] = skip_variable;
+    marker->length_limit_APPn[i] = 0;
+  }
+  marker->process_APPn[0] = get_interesting_appn;
+  marker->process_APPn[14] = get_interesting_appn;
+  /* Reset marker processing state */
+  reset_marker_reader(cinfo);
+}
+
+
+/*
+ * Control saving of COM and APPn markers into marker_list.
+ */
+
+#ifdef SAVE_MARKERS_SUPPORTED
+
+GLOBAL(void)
+jpeg_save_markers (j_decompress_ptr cinfo, int marker_code,
+		   unsigned int length_limit)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+  long maxlength;
+  jpeg_marker_parser_method processor;
+
+  /* Length limit mustn't be larger than what we can allocate
+   * (should only be a concern in a 16-bit environment).
+   */
+  maxlength = cinfo->mem->max_alloc_chunk - SIZEOF(struct jpeg_marker_struct);
+  if (((long) length_limit) > maxlength)
+    length_limit = (unsigned int) maxlength;
+
+  /* Choose processor routine to use.
+   * APP0/APP14 have special requirements.
+   */
+  if (length_limit) {
+    processor = save_marker;
+    /* If saving APP0/APP14, save at least enough for our internal use. */
+    if (marker_code == (int) M_APP0 && length_limit < APP0_DATA_LEN)
+      length_limit = APP0_DATA_LEN;
+    else if (marker_code == (int) M_APP14 && length_limit < APP14_DATA_LEN)
+      length_limit = APP14_DATA_LEN;
+  } else {
+    processor = skip_variable;
+    /* If discarding APP0/APP14, use our regular on-the-fly processor. */
+    if (marker_code == (int) M_APP0 || marker_code == (int) M_APP14)
+      processor = get_interesting_appn;
+  }
+
+  if (marker_code == (int) M_COM) {
+    marker->process_COM = processor;
+    marker->length_limit_COM = length_limit;
+  } else if (marker_code >= (int) M_APP0 && marker_code <= (int) M_APP15) {
+    marker->process_APPn[marker_code - (int) M_APP0] = processor;
+    marker->length_limit_APPn[marker_code - (int) M_APP0] = length_limit;
+  } else
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code);
+}
+
+#endif /* SAVE_MARKERS_SUPPORTED */
+
+
+/*
+ * Install a special processing method for COM or APPn markers.
+ */
+
+GLOBAL(void)
+jpeg_set_marker_processor (j_decompress_ptr cinfo, int marker_code,
+			   jpeg_marker_parser_method routine)
+{
+  my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
+
+  if (marker_code == (int) M_COM)
+    marker->process_COM = routine;
+  else if (marker_code >= (int) M_APP0 && marker_code <= (int) M_APP15)
+    marker->process_APPn[marker_code - (int) M_APP0] = routine;
+  else
+    ERREXIT1(cinfo, JERR_UNKNOWN_MARKER, marker_code);
+}
diff --git a/jpeg/jdmaster.c b/jpeg/jdmaster.c
new file mode 100644
index 000000000..2802c5b7b
--- /dev/null
+++ b/jpeg/jdmaster.c
@@ -0,0 +1,557 @@
+/*
+ * jdmaster.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains master control logic for the JPEG decompressor.
+ * These routines are concerned with selecting the modules to be executed
+ * and with determining the number of passes and the work to be done in each
+ * pass.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_decomp_master pub; /* public fields */
+
+  int pass_number;		/* # of passes completed */
+
+  boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */
+
+  /* Saved references to initialized quantizer modules,
+   * in case we need to switch modes.
+   */
+  struct jpeg_color_quantizer * quantizer_1pass;
+  struct jpeg_color_quantizer * quantizer_2pass;
+} my_decomp_master;
+
+typedef my_decomp_master * my_master_ptr;
+
+
+/*
+ * Determine whether merged upsample/color conversion should be used.
+ * CRUCIAL: this must match the actual capabilities of jdmerge.c!
+ */
+
+LOCAL(boolean)
+use_merged_upsample (j_decompress_ptr cinfo)
+{
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+  /* Merging is the equivalent of plain box-filter upsampling */
+  if (cinfo->do_fancy_upsampling || cinfo->CCIR601_sampling)
+    return FALSE;
+  /* jdmerge.c only supports YCC=>RGB color conversion */
+  if (cinfo->jpeg_color_space != JCS_YCbCr || cinfo->num_components != 3 ||
+      cinfo->out_color_space != JCS_RGB ||
+      cinfo->out_color_components != RGB_PIXELSIZE)
+    return FALSE;
+  /* and it only handles 2h1v or 2h2v sampling ratios */
+  if (cinfo->comp_info[0].h_samp_factor != 2 ||
+      cinfo->comp_info[1].h_samp_factor != 1 ||
+      cinfo->comp_info[2].h_samp_factor != 1 ||
+      cinfo->comp_info[0].v_samp_factor >  2 ||
+      cinfo->comp_info[1].v_samp_factor != 1 ||
+      cinfo->comp_info[2].v_samp_factor != 1)
+    return FALSE;
+  /* furthermore, it doesn't work if we've scaled the IDCTs differently */
+  if (cinfo->comp_info[0].DCT_scaled_size != cinfo->min_DCT_scaled_size ||
+      cinfo->comp_info[1].DCT_scaled_size != cinfo->min_DCT_scaled_size ||
+      cinfo->comp_info[2].DCT_scaled_size != cinfo->min_DCT_scaled_size)
+    return FALSE;
+  /* ??? also need to test for upsample-time rescaling, when & if supported */
+  return TRUE;			/* by golly, it'll work... */
+#else
+  return FALSE;
+#endif
+}
+
+
+/*
+ * Compute output image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
+ * Also note that it may be called before the master module is initialized!
+ */
+
+GLOBAL(void)
+jpeg_calc_output_dimensions (j_decompress_ptr cinfo)
+/* Do computations that are needed before master selection phase */
+{
+#ifdef IDCT_SCALING_SUPPORTED
+  int ci;
+  jpeg_component_info *compptr;
+#endif
+
+  /* Prevent application from calling me at wrong times */
+  if (cinfo->global_state != DSTATE_READY)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+  /* Compute actual output image dimensions and DCT scaling choices. */
+  if (cinfo->scale_num * 8 <= cinfo->scale_denom) {
+    /* Provide 1/8 scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, 8L);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, 8L);
+    cinfo->min_DCT_scaled_size = 1;
+  } else if (cinfo->scale_num * 4 <= cinfo->scale_denom) {
+    /* Provide 1/4 scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, 4L);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, 4L);
+    cinfo->min_DCT_scaled_size = 2;
+  } else if (cinfo->scale_num * 2 <= cinfo->scale_denom) {
+    /* Provide 1/2 scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, 2L);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, 2L);
+    cinfo->min_DCT_scaled_size = 4;
+  } else {
+    /* Provide 1/1 scaling */
+    cinfo->output_width = cinfo->image_width;
+    cinfo->output_height = cinfo->image_height;
+    cinfo->min_DCT_scaled_size = DCTSIZE;
+  }
+  /* In selecting the actual DCT scaling for each component, we try to
+   * scale up the chroma components via IDCT scaling rather than upsampling.
+   * This saves time if the upsampler gets to use 1:1 scaling.
+   * Note this code assumes that the supported DCT scalings are powers of 2.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    int ssize = cinfo->min_DCT_scaled_size;
+    while (ssize < DCTSIZE &&
+	   (compptr->h_samp_factor * ssize * 2 <=
+	    cinfo->max_h_samp_factor * cinfo->min_DCT_scaled_size) &&
+	   (compptr->v_samp_factor * ssize * 2 <=
+	    cinfo->max_v_samp_factor * cinfo->min_DCT_scaled_size)) {
+      ssize = ssize * 2;
+    }
+    compptr->DCT_scaled_size = ssize;
+  }
+
+  /* Recompute downsampled dimensions of components;
+   * application needs to know these if using raw downsampled data.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Size in samples, after IDCT scaling */
+    compptr->downsampled_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width *
+		    (long) (compptr->h_samp_factor * compptr->DCT_scaled_size),
+		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+    compptr->downsampled_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height *
+		    (long) (compptr->v_samp_factor * compptr->DCT_scaled_size),
+		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+  }
+
+#else /* !IDCT_SCALING_SUPPORTED */
+
+  /* Hardwire it to "no scaling" */
+  cinfo->output_width = cinfo->image_width;
+  cinfo->output_height = cinfo->image_height;
+  /* jdinput.c has already initialized DCT_scaled_size to DCTSIZE,
+   * and has computed unscaled downsampled_width and downsampled_height.
+   */
+
+#endif /* IDCT_SCALING_SUPPORTED */
+
+  /* Report number of components in selected colorspace. */
+  /* Probably this should be in the color conversion module... */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    cinfo->out_color_components = 1;
+    break;
+  case JCS_RGB:
+#if RGB_PIXELSIZE != 3
+    cinfo->out_color_components = RGB_PIXELSIZE;
+    break;
+#endif /* else share code with YCbCr */
+  case JCS_YCbCr:
+    cinfo->out_color_components = 3;
+    break;
+  case JCS_CMYK:
+  case JCS_YCCK:
+    cinfo->out_color_components = 4;
+    break;
+  default:			/* else must be same colorspace as in file */
+    cinfo->out_color_components = cinfo->num_components;
+    break;
+  }
+  cinfo->output_components = (cinfo->quantize_colors ? 1 :
+			      cinfo->out_color_components);
+
+  /* See if upsampler will want to emit more than one row at a time */
+  if (use_merged_upsample(cinfo))
+    cinfo->rec_outbuf_height = cinfo->max_v_samp_factor;
+  else
+    cinfo->rec_outbuf_height = 1;
+}
+
+
+/*
+ * Several decompression processes need to range-limit values to the range
+ * 0..MAXJSAMPLE; the input value may fall somewhat outside this range
+ * due to noise introduced by quantization, roundoff error, etc.  These
+ * processes are inner loops and need to be as fast as possible.  On most
+ * machines, particularly CPUs with pipelines or instruction prefetch,
+ * a (subscript-check-less) C table lookup
+ *		x = sample_range_limit[x];
+ * is faster than explicit tests
+ *		if (x < 0)  x = 0;
+ *		else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
+ * These processes all use a common table prepared by the routine below.
+ *
+ * For most steps we can mathematically guarantee that the initial value
+ * of x is within MAXJSAMPLE+1 of the legal range, so a table running from
+ * -(MAXJSAMPLE+1) to 2*MAXJSAMPLE+1 is sufficient.  But for the initial
+ * limiting step (just after the IDCT), a wildly out-of-range value is 
+ * possible if the input data is corrupt.  To avoid any chance of indexing
+ * off the end of memory and getting a bad-pointer trap, we perform the
+ * post-IDCT limiting thus:
+ *		x = range_limit[x & MASK];
+ * where MASK is 2 bits wider than legal sample data, ie 10 bits for 8-bit
+ * samples.  Under normal circumstances this is more than enough range and
+ * a correct output will be generated; with bogus input data the mask will
+ * cause wraparound, and we will safely generate a bogus-but-in-range output.
+ * For the post-IDCT step, we want to convert the data from signed to unsigned
+ * representation by adding CENTERJSAMPLE at the same time that we limit it.
+ * So the post-IDCT limiting table ends up looking like this:
+ *   CENTERJSAMPLE,CENTERJSAMPLE+1,...,MAXJSAMPLE,
+ *   MAXJSAMPLE (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times),
+ *   0          (repeat 2*(MAXJSAMPLE+1)-CENTERJSAMPLE times),
+ *   0,1,...,CENTERJSAMPLE-1
+ * Negative inputs select values from the upper half of the table after
+ * masking.
+ *
+ * We can save some space by overlapping the start of the post-IDCT table
+ * with the simpler range limiting table.  The post-IDCT table begins at
+ * sample_range_limit + CENTERJSAMPLE.
+ *
+ * Note that the table is allocated in near data space on PCs; it's small
+ * enough and used often enough to justify this.
+ */
+
+LOCAL(void)
+prepare_range_limit_table (j_decompress_ptr cinfo)
+/* Allocate and fill in the sample_range_limit table */
+{
+  JSAMPLE * table;
+  int i;
+
+  table = (JSAMPLE *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+		(5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * SIZEOF(JSAMPLE));
+  table += (MAXJSAMPLE+1);	/* allow negative subscripts of simple table */
+  cinfo->sample_range_limit = table;
+  /* First segment of "simple" table: limit[x] = 0 for x < 0 */
+  MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * SIZEOF(JSAMPLE));
+  /* Main part of "simple" table: limit[x] = x */
+  for (i = 0; i <= MAXJSAMPLE; i++)
+    table[i] = (JSAMPLE) i;
+  table += CENTERJSAMPLE;	/* Point to where post-IDCT table starts */
+  /* End of simple table, rest of first half of post-IDCT table */
+  for (i = CENTERJSAMPLE; i < 2*(MAXJSAMPLE+1); i++)
+    table[i] = MAXJSAMPLE;
+  /* Second half of post-IDCT table */
+  MEMZERO(table + (2 * (MAXJSAMPLE+1)),
+	  (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * SIZEOF(JSAMPLE));
+  MEMCOPY(table + (4 * (MAXJSAMPLE+1) - CENTERJSAMPLE),
+	  cinfo->sample_range_limit, CENTERJSAMPLE * SIZEOF(JSAMPLE));
+}
+
+
+/*
+ * Master selection of decompression modules.
+ * This is done once at jpeg_start_decompress time.  We determine
+ * which modules will be used and give them appropriate initialization calls.
+ * We also initialize the decompressor input side to begin consuming data.
+ *
+ * Since jpeg_read_header has finished, we know what is in the SOF
+ * and (first) SOS markers.  We also have all the application parameter
+ * settings.
+ */
+
+LOCAL(void)
+master_selection (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+  boolean use_c_buffer;
+  long samplesperrow;
+  JDIMENSION jd_samplesperrow;
+
+  /* Initialize dimensions and other stuff */
+  jpeg_calc_output_dimensions(cinfo);
+  prepare_range_limit_table(cinfo);
+
+  /* Width of an output scanline must be representable as JDIMENSION. */
+  samplesperrow = (long) cinfo->output_width * (long) cinfo->out_color_components;
+  jd_samplesperrow = (JDIMENSION) samplesperrow;
+  if ((long) jd_samplesperrow != samplesperrow)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* Initialize my private state */
+  master->pass_number = 0;
+  master->using_merged_upsample = use_merged_upsample(cinfo);
+
+  /* Color quantizer selection */
+  master->quantizer_1pass = NULL;
+  master->quantizer_2pass = NULL;
+  /* No mode changes if not using buffered-image mode. */
+  if (! cinfo->quantize_colors || ! cinfo->buffered_image) {
+    cinfo->enable_1pass_quant = FALSE;
+    cinfo->enable_external_quant = FALSE;
+    cinfo->enable_2pass_quant = FALSE;
+  }
+  if (cinfo->quantize_colors) {
+    if (cinfo->raw_data_out)
+      ERREXIT(cinfo, JERR_NOTIMPL);
+    /* 2-pass quantizer only works in 3-component color space. */
+    if (cinfo->out_color_components != 3) {
+      cinfo->enable_1pass_quant = TRUE;
+      cinfo->enable_external_quant = FALSE;
+      cinfo->enable_2pass_quant = FALSE;
+      cinfo->colormap = NULL;
+    } else if (cinfo->colormap != NULL) {
+      cinfo->enable_external_quant = TRUE;
+    } else if (cinfo->two_pass_quantize) {
+      cinfo->enable_2pass_quant = TRUE;
+    } else {
+      cinfo->enable_1pass_quant = TRUE;
+    }
+
+    if (cinfo->enable_1pass_quant) {
+#ifdef QUANT_1PASS_SUPPORTED
+      jinit_1pass_quantizer(cinfo);
+      master->quantizer_1pass = cinfo->cquantize;
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    }
+
+    /* We use the 2-pass code to map to external colormaps. */
+    if (cinfo->enable_2pass_quant || cinfo->enable_external_quant) {
+#ifdef QUANT_2PASS_SUPPORTED
+      jinit_2pass_quantizer(cinfo);
+      master->quantizer_2pass = cinfo->cquantize;
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    }
+    /* If both quantizers are initialized, the 2-pass one is left active;
+     * this is necessary for starting with quantization to an external map.
+     */
+  }
+
+  /* Post-processing: in particular, color conversion first */
+  if (! cinfo->raw_data_out) {
+    if (master->using_merged_upsample) {
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+      jinit_merged_upsampler(cinfo); /* does color conversion too */
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else {
+      jinit_color_deconverter(cinfo);
+      jinit_upsampler(cinfo);
+    }
+    jinit_d_post_controller(cinfo, cinfo->enable_2pass_quant);
+  }
+  /* Inverse DCT */
+  jinit_inverse_dct(cinfo);
+  /* Entropy decoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef D_PROGRESSIVE_SUPPORTED
+      jinit_phuff_decoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_decoder(cinfo);
+  }
+
+  /* Initialize principal buffer controllers. */
+  use_c_buffer = cinfo->inputctl->has_multiple_scans || cinfo->buffered_image;
+  jinit_d_coef_controller(cinfo, use_c_buffer);
+
+  if (! cinfo->raw_data_out)
+    jinit_d_main_controller(cinfo, FALSE /* never need full buffer here */);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Initialize input side of decompressor to consume first scan. */
+  (*cinfo->inputctl->start_input_pass) (cinfo);
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* If jpeg_start_decompress will read the whole file, initialize
+   * progress monitoring appropriately.  The input step is counted
+   * as one pass.
+   */
+  if (cinfo->progress != NULL && ! cinfo->buffered_image &&
+      cinfo->inputctl->has_multiple_scans) {
+    int nscans;
+    /* Estimate number of scans to set pass_limit. */
+    if (cinfo->progressive_mode) {
+      /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */
+      nscans = 2 + 3 * cinfo->num_components;
+    } else {
+      /* For a nonprogressive multiscan file, estimate 1 scan per component. */
+      nscans = cinfo->num_components;
+    }
+    cinfo->progress->pass_counter = 0L;
+    cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows * nscans;
+    cinfo->progress->completed_passes = 0;
+    cinfo->progress->total_passes = (cinfo->enable_2pass_quant ? 3 : 2);
+    /* Count the input pass as done */
+    master->pass_number++;
+  }
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+}
+
+
+/*
+ * Per-pass setup.
+ * This is called at the beginning of each output pass.  We determine which
+ * modules will be active during this pass and give them appropriate
+ * start_pass calls.  We also set is_dummy_pass to indicate whether this
+ * is a "real" output pass or a dummy pass for color quantization.
+ * (In the latter case, jdapistd.c will crank the pass to completion.)
+ */
+
+METHODDEF(void)
+prepare_for_output_pass (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  if (master->pub.is_dummy_pass) {
+#ifdef QUANT_2PASS_SUPPORTED
+    /* Final pass of 2-pass quantization */
+    master->pub.is_dummy_pass = FALSE;
+    (*cinfo->cquantize->start_pass) (cinfo, FALSE);
+    (*cinfo->post->start_pass) (cinfo, JBUF_CRANK_DEST);
+    (*cinfo->main->start_pass) (cinfo, JBUF_CRANK_DEST);
+#else
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif /* QUANT_2PASS_SUPPORTED */
+  } else {
+    if (cinfo->quantize_colors && cinfo->colormap == NULL) {
+      /* Select new quantization method */
+      if (cinfo->two_pass_quantize && cinfo->enable_2pass_quant) {
+	cinfo->cquantize = master->quantizer_2pass;
+	master->pub.is_dummy_pass = TRUE;
+      } else if (cinfo->enable_1pass_quant) {
+	cinfo->cquantize = master->quantizer_1pass;
+      } else {
+	ERREXIT(cinfo, JERR_MODE_CHANGE);
+      }
+    }
+    (*cinfo->idct->start_pass) (cinfo);
+    (*cinfo->coef->start_output_pass) (cinfo);
+    if (! cinfo->raw_data_out) {
+      if (! master->using_merged_upsample)
+	(*cinfo->cconvert->start_pass) (cinfo);
+      (*cinfo->upsample->start_pass) (cinfo);
+      if (cinfo->quantize_colors)
+	(*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
+      (*cinfo->post->start_pass) (cinfo,
+	    (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+      (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
+    }
+  }
+
+  /* Set up progress monitor's pass info if present */
+  if (cinfo->progress != NULL) {
+    cinfo->progress->completed_passes = master->pass_number;
+    cinfo->progress->total_passes = master->pass_number +
+				    (master->pub.is_dummy_pass ? 2 : 1);
+    /* In buffered-image mode, we assume one more output pass if EOI not
+     * yet reached, but no more passes if EOI has been reached.
+     */
+    if (cinfo->buffered_image && ! cinfo->inputctl->eoi_reached) {
+      cinfo->progress->total_passes += (cinfo->enable_2pass_quant ? 2 : 1);
+    }
+  }
+}
+
+
+/*
+ * Finish up at end of an output pass.
+ */
+
+METHODDEF(void)
+finish_output_pass (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  if (cinfo->quantize_colors)
+    (*cinfo->cquantize->finish_pass) (cinfo);
+  master->pass_number++;
+}
+
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+
+/*
+ * Switch to a new external colormap between output passes.
+ */
+
+GLOBAL(void)
+jpeg_new_colormap (j_decompress_ptr cinfo)
+{
+  my_master_ptr master = (my_master_ptr) cinfo->master;
+
+  /* Prevent application from calling me at wrong times */
+  if (cinfo->global_state != DSTATE_BUFIMAGE)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (cinfo->quantize_colors && cinfo->enable_external_quant &&
+      cinfo->colormap != NULL) {
+    /* Select 2-pass quantizer for external colormap use */
+    cinfo->cquantize = master->quantizer_2pass;
+    /* Notify quantizer of colormap change */
+    (*cinfo->cquantize->new_color_map) (cinfo);
+    master->pub.is_dummy_pass = FALSE; /* just in case */
+  } else
+    ERREXIT(cinfo, JERR_MODE_CHANGE);
+}
+
+#endif /* D_MULTISCAN_FILES_SUPPORTED */
+
+
+/*
+ * Initialize master decompression control and select active modules.
+ * This is performed at the start of jpeg_start_decompress.
+ */
+
+GLOBAL(void)
+jinit_master_decompress (j_decompress_ptr cinfo)
+{
+  my_master_ptr master;
+
+  master = (my_master_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				  SIZEOF(my_decomp_master));
+  cinfo->master = (struct jpeg_decomp_master *) master;
+  master->pub.prepare_for_output_pass = prepare_for_output_pass;
+  master->pub.finish_output_pass = finish_output_pass;
+
+  master->pub.is_dummy_pass = FALSE;
+
+  master_selection(cinfo);
+}
diff --git a/jpeg/jdmerge.c b/jpeg/jdmerge.c
new file mode 100644
index 000000000..37444468c
--- /dev/null
+++ b/jpeg/jdmerge.c
@@ -0,0 +1,400 @@
+/*
+ * jdmerge.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ *
+ * This file combines functions from jdsample.c and jdcolor.c;
+ * read those files first to understand what's going on.
+ *
+ * When the chroma components are to be upsampled by simple replication
+ * (ie, box filtering), we can save some work in color conversion by
+ * calculating all the output pixels corresponding to a pair of chroma
+ * samples at one time.  In the conversion equations
+ *	R = Y           + K1 * Cr
+ *	G = Y + K2 * Cb + K3 * Cr
+ *	B = Y + K4 * Cb
+ * only the Y term varies among the group of pixels corresponding to a pair
+ * of chroma samples, so the rest of the terms can be calculated just once.
+ * At typical sampling ratios, this eliminates half or three-quarters of the
+ * multiplications needed for color conversion.
+ *
+ * This file currently provides implementations for the following cases:
+ *	YCbCr => RGB color conversion only.
+ *	Sampling ratios of 2h1v or 2h2v.
+ *	No scaling needed at upsample time.
+ *	Corner-aligned (non-CCIR601) sampling alignment.
+ * Other special cases could be added, but in most applications these are
+ * the only common cases.  (For uncommon cases we fall back on the more
+ * general code in jdsample.c and jdcolor.c.)
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef UPSAMPLE_MERGING_SUPPORTED
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;	/* public fields */
+
+  /* Pointer to routine to do actual upsampling/conversion of one row group */
+  JMETHOD(void, upmethod, (j_decompress_ptr cinfo,
+			   JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+			   JSAMPARRAY output_buf));
+
+  /* Private state for YCC->RGB conversion */
+  int * Cr_r_tab;		/* => table for Cr to R conversion */
+  int * Cb_b_tab;		/* => table for Cb to B conversion */
+  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
+  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+
+  /* For 2:1 vertical sampling, we produce two output rows at a time.
+   * We need a "spare" row buffer to hold the second output row if the
+   * application provides just a one-row buffer; we also use the spare
+   * to discard the dummy last row if the image height is odd.
+   */
+  JSAMPROW spare_row;
+  boolean spare_full;		/* T if spare buffer is occupied */
+
+  JDIMENSION out_row_width;	/* samples per output row */
+  JDIMENSION rows_to_go;	/* counts rows remaining in image */
+} my_upsampler;
+
+typedef my_upsampler * my_upsample_ptr;
+
+#define SCALEBITS	16	/* speediest right-shift on some machines */
+#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
+#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+
+
+/*
+ * Initialize tables for YCC->RGB colorspace conversion.
+ * This is taken directly from jdcolor.c; see that file for more info.
+ */
+
+LOCAL(void)
+build_ycc_rgb_table (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  int i;
+  INT32 x;
+  SHIFT_TEMPS
+
+  upsample->Cr_r_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  upsample->Cb_b_tab = (int *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(int));
+  upsample->Cr_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+  upsample->Cb_g_tab = (INT32 *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				(MAXJSAMPLE+1) * SIZEOF(INT32));
+
+  for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
+    /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
+    /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
+    /* Cr=>R value is nearest int to 1.40200 * x */
+    upsample->Cr_r_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+    /* Cb=>B value is nearest int to 1.77200 * x */
+    upsample->Cb_b_tab[i] = (int)
+		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+    /* Cr=>G value is scaled-up -0.71414 * x */
+    upsample->Cr_g_tab[i] = (- FIX(0.71414)) * x;
+    /* Cb=>G value is scaled-up -0.34414 * x */
+    /* We also add in ONE_HALF so that need not do it in inner loop */
+    upsample->Cb_g_tab[i] = (- FIX(0.34414)) * x + ONE_HALF;
+  }
+}
+
+
+/*
+ * Initialize for an upsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_merged_upsample (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  /* Mark the spare buffer empty */
+  upsample->spare_full = FALSE;
+  /* Initialize total-height counter for detecting bottom of image */
+  upsample->rows_to_go = cinfo->output_height;
+}
+
+
+/*
+ * Control routine to do upsampling (and color conversion).
+ *
+ * The control routine just handles the row buffering considerations.
+ */
+
+METHODDEF(void)
+merged_2v_upsample (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+/* 2:1 vertical sampling case: may need a spare row. */
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPROW work_ptrs[2];
+  JDIMENSION num_rows;		/* number of rows returned to caller */
+
+  if (upsample->spare_full) {
+    /* If we have a spare row saved from a previous cycle, just return it. */
+    jcopy_sample_rows(& upsample->spare_row, 0, output_buf + *out_row_ctr, 0,
+		      1, upsample->out_row_width);
+    num_rows = 1;
+    upsample->spare_full = FALSE;
+  } else {
+    /* Figure number of rows to return to caller. */
+    num_rows = 2;
+    /* Not more than the distance to the end of the image. */
+    if (num_rows > upsample->rows_to_go)
+      num_rows = upsample->rows_to_go;
+    /* And not more than what the client can accept: */
+    out_rows_avail -= *out_row_ctr;
+    if (num_rows > out_rows_avail)
+      num_rows = out_rows_avail;
+    /* Create output pointer array for upsampler. */
+    work_ptrs[0] = output_buf[*out_row_ctr];
+    if (num_rows > 1) {
+      work_ptrs[1] = output_buf[*out_row_ctr + 1];
+    } else {
+      work_ptrs[1] = upsample->spare_row;
+      upsample->spare_full = TRUE;
+    }
+    /* Now do the upsampling. */
+    (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr, work_ptrs);
+  }
+
+  /* Adjust counts */
+  *out_row_ctr += num_rows;
+  upsample->rows_to_go -= num_rows;
+  /* When the buffer is emptied, declare this input row group consumed */
+  if (! upsample->spare_full)
+    (*in_row_group_ctr)++;
+}
+
+
+METHODDEF(void)
+merged_1v_upsample (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+/* 1:1 vertical sampling case: much easier, never need a spare row. */
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  /* Just do the upsampling. */
+  (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr,
+			 output_buf + *out_row_ctr);
+  /* Adjust counts */
+  (*out_row_ctr)++;
+  (*in_row_group_ctr)++;
+}
+
+
+/*
+ * These are the routines invoked by the control routines to do
+ * the actual upsampling/conversion.  One row group is processed per call.
+ *
+ * Note: since we may be writing directly into application-supplied buffers,
+ * we have to be honest about the output width; we can't assume the buffer
+ * has been rounded up to an even width.
+ */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+METHODDEF(void)
+h2v1_merged_upsample (j_decompress_ptr cinfo,
+		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+		      JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+    outptr += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+    outptr += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+  }
+}
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+METHODDEF(void)
+h2v2_merged_upsample (j_decompress_ptr cinfo,
+		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+		      JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+    outptr1 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+    outptr1 += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+    y  = GETJSAMPLE(*inptr01);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+  }
+}
+
+
+/*
+ * Module initialization routine for merged upsampling/color conversion.
+ *
+ * NB: this is called under the conditions determined by use_merged_upsample()
+ * in jdmaster.c.  That routine MUST correspond to the actual capabilities
+ * of this module; no safety checks are made here.
+ */
+
+GLOBAL(void)
+jinit_merged_upsampler (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample;
+
+  upsample = (my_upsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_upsampler));
+  cinfo->upsample = (struct jpeg_upsampler *) upsample;
+  upsample->pub.start_pass = start_pass_merged_upsample;
+  upsample->pub.need_context_rows = FALSE;
+
+  upsample->out_row_width = cinfo->output_width * cinfo->out_color_components;
+
+  if (cinfo->max_v_samp_factor == 2) {
+    upsample->pub.upsample = merged_2v_upsample;
+    upsample->upmethod = h2v2_merged_upsample;
+    /* Allocate a spare row buffer */
+    upsample->spare_row = (JSAMPROW)
+      (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+		(size_t) (upsample->out_row_width * SIZEOF(JSAMPLE)));
+  } else {
+    upsample->pub.upsample = merged_1v_upsample;
+    upsample->upmethod = h2v1_merged_upsample;
+    /* No spare row needed */
+    upsample->spare_row = NULL;
+  }
+
+  build_ycc_rgb_table(cinfo);
+}
+
+#endif /* UPSAMPLE_MERGING_SUPPORTED */
diff --git a/jpeg/jdphuff.c b/jpeg/jdphuff.c
new file mode 100644
index 000000000..226780994
--- /dev/null
+++ b/jpeg/jdphuff.c
@@ -0,0 +1,668 @@
+/*
+ * jdphuff.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains Huffman entropy decoding routines for progressive JPEG.
+ *
+ * Much of the complexity here has to do with supporting input suspension.
+ * If the data source module demands suspension, we want to be able to back
+ * up to the start of the current MCU.  To do this, we copy state variables
+ * into local working storage, and update them back to the permanent
+ * storage only upon successful completion of an MCU.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdhuff.h"		/* Declarations shared with jdhuff.c */
+
+
+#ifdef D_PROGRESSIVE_SUPPORTED
+
+/*
+ * Expanded entropy decoder object for progressive Huffman decoding.
+ *
+ * The savable_state subrecord contains fields that change within an MCU,
+ * but must not be updated permanently until we complete the MCU.
+ */
+
+typedef struct {
+  unsigned int EOBRUN;			/* remaining EOBs in EOBRUN */
+  int last_dc_val[MAX_COMPS_IN_SCAN];	/* last DC coef for each component */
+} savable_state;
+
+/* This macro is to work around compilers with missing or broken
+ * structure assignment.  You'll need to fix this code if you have
+ * such a compiler and you change MAX_COMPS_IN_SCAN.
+ */
+
+#ifndef NO_STRUCT_ASSIGN
+#define ASSIGN_STATE(dest,src)  ((dest) = (src))
+#else
+#if MAX_COMPS_IN_SCAN == 4
+#define ASSIGN_STATE(dest,src)  \
+	((dest).EOBRUN = (src).EOBRUN, \
+	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
+	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
+	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
+	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+#endif
+#endif
+
+
+typedef struct {
+  struct jpeg_entropy_decoder pub; /* public fields */
+
+  /* These fields are loaded into local variables at start of each MCU.
+   * In case of suspension, we exit WITHOUT updating them.
+   */
+  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
+  savable_state saved;		/* Other state at start of MCU */
+
+  /* These fields are NOT loaded into local working state. */
+  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+
+  /* Pointers to derived tables (these workspaces have image lifespan) */
+  d_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+
+  d_derived_tbl * ac_derived_tbl; /* active table during an AC scan */
+} phuff_entropy_decoder;
+
+typedef phuff_entropy_decoder * phuff_entropy_ptr;
+
+/* Forward declarations */
+METHODDEF(boolean) decode_mcu_DC_first JPP((j_decompress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_AC_first JPP((j_decompress_ptr cinfo,
+					    JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_DC_refine JPP((j_decompress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_AC_refine JPP((j_decompress_ptr cinfo,
+					     JBLOCKROW *MCU_data));
+
+
+/*
+ * Initialize for a Huffman-compressed scan.
+ */
+
+METHODDEF(void)
+start_pass_phuff_decoder (j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  boolean is_DC_band, bad;
+  int ci, coefi, tbl;
+  int *coef_bit_ptr;
+  jpeg_component_info * compptr;
+
+  is_DC_band = (cinfo->Ss == 0);
+
+  /* Validate scan parameters */
+  bad = FALSE;
+  if (is_DC_band) {
+    if (cinfo->Se != 0)
+      bad = TRUE;
+  } else {
+    /* need not check Ss/Se < 0 since they came from unsigned bytes */
+    if (cinfo->Ss > cinfo->Se || cinfo->Se >= DCTSIZE2)
+      bad = TRUE;
+    /* AC scans may have only one component */
+    if (cinfo->comps_in_scan != 1)
+      bad = TRUE;
+  }
+  if (cinfo->Ah != 0) {
+    /* Successive approximation refinement scan: must have Al = Ah-1. */
+    if (cinfo->Al != cinfo->Ah-1)
+      bad = TRUE;
+  }
+  if (cinfo->Al > 13)		/* need not check for < 0 */
+    bad = TRUE;
+  /* Arguably the maximum Al value should be less than 13 for 8-bit precision,
+   * but the spec doesn't say so, and we try to be liberal about what we
+   * accept.  Note: large Al values could result in out-of-range DC
+   * coefficients during early scans, leading to bizarre displays due to
+   * overflows in the IDCT math.  But we won't crash.
+   */
+  if (bad)
+    ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
+	     cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+  /* Update progression status, and verify that scan order is legal.
+   * Note that inter-scan inconsistencies are treated as warnings
+   * not fatal errors ... not clear if this is right way to behave.
+   */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    int cindex = cinfo->cur_comp_info[ci]->component_index;
+    coef_bit_ptr = & cinfo->coef_bits[cindex][0];
+    if (!is_DC_band && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
+      WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+    for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
+      int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
+      if (cinfo->Ah != expected)
+	WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+      coef_bit_ptr[coefi] = cinfo->Al;
+    }
+  }
+
+  /* Select MCU decoding routine */
+  if (cinfo->Ah == 0) {
+    if (is_DC_band)
+      entropy->pub.decode_mcu = decode_mcu_DC_first;
+    else
+      entropy->pub.decode_mcu = decode_mcu_AC_first;
+  } else {
+    if (is_DC_band)
+      entropy->pub.decode_mcu = decode_mcu_DC_refine;
+    else
+      entropy->pub.decode_mcu = decode_mcu_AC_refine;
+  }
+
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+    compptr = cinfo->cur_comp_info[ci];
+    /* Make sure requested tables are present, and compute derived tables.
+     * We may build same derived table more than once, but it's not expensive.
+     */
+    if (is_DC_band) {
+      if (cinfo->Ah == 0) {	/* DC refinement needs no table */
+	tbl = compptr->dc_tbl_no;
+	jpeg_make_d_derived_tbl(cinfo, TRUE, tbl,
+				& entropy->derived_tbls[tbl]);
+      }
+    } else {
+      tbl = compptr->ac_tbl_no;
+      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl,
+			      & entropy->derived_tbls[tbl]);
+      /* remember the single active table */
+      entropy->ac_derived_tbl = entropy->derived_tbls[tbl];
+    }
+    /* Initialize DC predictions to 0 */
+    entropy->saved.last_dc_val[ci] = 0;
+  }
+
+  /* Initialize bitread state variables */
+  entropy->bitstate.bits_left = 0;
+  entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
+  entropy->pub.insufficient_data = FALSE;
+
+  /* Initialize private state variables */
+  entropy->saved.EOBRUN = 0;
+
+  /* Initialize restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+}
+
+
+/*
+ * Figure F.12: extend sign bit.
+ * On some machines, a shift and add will be faster than a table lookup.
+ */
+
+#ifdef AVOID_TABLES
+
+#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))
+
+#else
+
+#define HUFF_EXTEND(x,s)  ((x) < extend_test[s] ? (x) + extend_offset[s] : (x))
+
+static const int extend_test[16] =   /* entry n is 2**(n-1) */
+  { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
+    0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 };
+
+static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */
+  { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1,
+    ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1,
+    ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1,
+    ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 };
+
+#endif /* AVOID_TABLES */
+
+
+/*
+ * Check for a restart marker & resynchronize decoder.
+ * Returns FALSE if must suspend.
+ */
+
+LOCAL(boolean)
+process_restart (j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int ci;
+
+  /* Throw away any unused bits remaining in bit buffer; */
+  /* include any full bytes in next_marker's count of discarded bytes */
+  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
+  entropy->bitstate.bits_left = 0;
+
+  /* Advance past the RSTn marker */
+  if (! (*cinfo->marker->read_restart_marker) (cinfo))
+    return FALSE;
+
+  /* Re-initialize DC predictions to 0 */
+  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
+    entropy->saved.last_dc_val[ci] = 0;
+  /* Re-init EOB run count, too */
+  entropy->saved.EOBRUN = 0;
+
+  /* Reset restart counter */
+  entropy->restarts_to_go = cinfo->restart_interval;
+
+  /* Reset out-of-data flag, unless read_restart_marker left us smack up
+   * against a marker.  In that case we will end up treating the next data
+   * segment as empty, and we can avoid producing bogus output pixels by
+   * leaving the flag set.
+   */
+  if (cinfo->unread_marker == 0)
+    entropy->pub.insufficient_data = FALSE;
+
+  return TRUE;
+}
+
+
+/*
+ * Huffman MCU decoding.
+ * Each of these routines decodes and returns one MCU's worth of
+ * Huffman-compressed coefficients. 
+ * The coefficients are reordered from zigzag order into natural array order,
+ * but are not dequantized.
+ *
+ * The i'th block of the MCU is stored into the block pointed to by
+ * MCU_data[i].  WE ASSUME THIS AREA IS INITIALLY ZEROED BY THE CALLER.
+ *
+ * We return FALSE if data source requested suspension.  In that case no
+ * changes have been made to permanent state.  (Exception: some output
+ * coefficients may already have been assigned.  This is harmless for
+ * spectral selection, since we'll just re-assign them on the next call.
+ * Successive approximation AC refinement has to be more careful, however.)
+ */
+
+/*
+ * MCU decoding for DC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int Al = cinfo->Al;
+  register int s, r;
+  int blkn, ci;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+  savable_state state;
+  d_derived_tbl * tbl;
+  jpeg_component_info * compptr;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(state, entropy->saved);
+
+    /* Outer loop handles each block in the MCU */
+
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      block = MCU_data[blkn];
+      ci = cinfo->MCU_membership[blkn];
+      compptr = cinfo->cur_comp_info[ci];
+      tbl = entropy->derived_tbls[compptr->dc_tbl_no];
+
+      /* Decode a single block's worth of coefficients */
+
+      /* Section F.2.2.1: decode the DC coefficient difference */
+      HUFF_DECODE(s, br_state, tbl, return FALSE, label1);
+      if (s) {
+	CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	r = GET_BITS(s);
+	s = HUFF_EXTEND(r, s);
+      }
+
+      /* Convert DC difference to actual value, update last_dc_val */
+      s += state.last_dc_val[ci];
+      state.last_dc_val[ci] = s;
+      /* Scale and output the coefficient (assumes jpeg_natural_order[0]=0) */
+      (*block)[0] = (JCOEF) (s << Al);
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    ASSIGN_STATE(entropy->saved, state);
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC initial scan (either spectral selection,
+ * or first pass of successive approximation).
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int Se = cinfo->Se;
+  int Al = cinfo->Al;
+  register int s, k, r;
+  unsigned int EOBRUN;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+  d_derived_tbl * tbl;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, just leave the MCU set to zeroes.
+   * This way, we return uniform gray for the remainder of the segment.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state.
+     * We can avoid loading/saving bitread state if in an EOB run.
+     */
+    EOBRUN = entropy->saved.EOBRUN;	/* only part of saved state we need */
+
+    /* There is always only one block per MCU */
+
+    if (EOBRUN > 0)		/* if it's a band of zeroes... */
+      EOBRUN--;			/* ...process it now (we do nothing) */
+    else {
+      BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+      block = MCU_data[0];
+      tbl = entropy->ac_derived_tbl;
+
+      for (k = cinfo->Ss; k <= Se; k++) {
+	HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
+	r = s >> 4;
+	s &= 15;
+	if (s) {
+	  k += r;
+	  CHECK_BIT_BUFFER(br_state, s, return FALSE);
+	  r = GET_BITS(s);
+	  s = HUFF_EXTEND(r, s);
+	  /* Scale and output coefficient in natural (dezigzagged) order */
+	  (*block)[jpeg_natural_order[k]] = (JCOEF) (s << Al);
+	} else {
+	  if (r == 15) {	/* ZRL */
+	    k += 15;		/* skip 15 zeroes in band */
+	  } else {		/* EOBr, run length is 2^r + appended bits */
+	    EOBRUN = 1 << r;
+	    if (r) {		/* EOBr, r > 0 */
+	      CHECK_BIT_BUFFER(br_state, r, return FALSE);
+	      r = GET_BITS(r);
+	      EOBRUN += r;
+	    }
+	    EOBRUN--;		/* this band is processed at this moment */
+	    break;		/* force end-of-band */
+	  }
+	}
+      }
+
+      BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    }
+
+    /* Completed MCU, so update state */
+    entropy->saved.EOBRUN = EOBRUN;	/* only part of saved state we need */
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for DC successive approximation refinement scan.
+ * Note: we assume such scans can be multi-component, although the spec
+ * is not very clear on the point.
+ */
+
+METHODDEF(boolean)
+decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
+  int blkn;
+  JBLOCKROW block;
+  BITREAD_STATE_VARS;
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* Not worth the cycles to check insufficient_data here,
+   * since we will not change the data anyway if we read zeroes.
+   */
+
+  /* Load up working state */
+  BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+
+  /* Outer loop handles each block in the MCU */
+
+  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+    block = MCU_data[blkn];
+
+    /* Encoded data is simply the next bit of the two's-complement DC value */
+    CHECK_BIT_BUFFER(br_state, 1, return FALSE);
+    if (GET_BITS(1))
+      (*block)[0] |= p1;
+    /* Note: since we use |=, repeating the assignment later is safe */
+  }
+
+  /* Completed MCU, so update state */
+  BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+}
+
+
+/*
+ * MCU decoding for AC successive approximation refinement scan.
+ */
+
+METHODDEF(boolean)
+decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
+{   
+  phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
+  int Se = cinfo->Se;
+  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
+  int m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */
+  register int s, k, r;
+  unsigned int EOBRUN;
+  JBLOCKROW block;
+  JCOEFPTR thiscoef;
+  BITREAD_STATE_VARS;
+  d_derived_tbl * tbl;
+  int num_newnz;
+  int newnz_pos[DCTSIZE2];
+
+  /* Process restart marker if needed; may have to suspend */
+  if (cinfo->restart_interval) {
+    if (entropy->restarts_to_go == 0)
+      if (! process_restart(cinfo))
+	return FALSE;
+  }
+
+  /* If we've run out of data, don't modify the MCU.
+   */
+  if (! entropy->pub.insufficient_data) {
+
+    /* Load up working state */
+    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
+    EOBRUN = entropy->saved.EOBRUN; /* only part of saved state we need */
+
+    /* There is always only one block per MCU */
+    block = MCU_data[0];
+    tbl = entropy->ac_derived_tbl;
+
+    /* If we are forced to suspend, we must undo the assignments to any newly
+     * nonzero coefficients in the block, because otherwise we'd get confused
+     * next time about which coefficients were already nonzero.
+     * But we need not undo addition of bits to already-nonzero coefficients;
+     * instead, we can test the current bit to see if we already did it.
+     */
+    num_newnz = 0;
+
+    /* initialize coefficient loop counter to start of band */
+    k = cinfo->Ss;
+
+    if (EOBRUN == 0) {
+      for (; k <= Se; k++) {
+	HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
+	r = s >> 4;
+	s &= 15;
+	if (s) {
+	  if (s != 1)		/* size of new coef should always be 1 */
+	    WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
+	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+	  if (GET_BITS(1))
+	    s = p1;		/* newly nonzero coef is positive */
+	  else
+	    s = m1;		/* newly nonzero coef is negative */
+	} else {
+	  if (r != 15) {
+	    EOBRUN = 1 << r;	/* EOBr, run length is 2^r + appended bits */
+	    if (r) {
+	      CHECK_BIT_BUFFER(br_state, r, goto undoit);
+	      r = GET_BITS(r);
+	      EOBRUN += r;
+	    }
+	    break;		/* rest of block is handled by EOB logic */
+	  }
+	  /* note s = 0 for processing ZRL */
+	}
+	/* Advance over already-nonzero coefs and r still-zero coefs,
+	 * appending correction bits to the nonzeroes.  A correction bit is 1
+	 * if the absolute value of the coefficient must be increased.
+	 */
+	do {
+	  thiscoef = *block + jpeg_natural_order[k];
+	  if (*thiscoef != 0) {
+	    CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+	    if (GET_BITS(1)) {
+	      if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
+		if (*thiscoef >= 0)
+		  *thiscoef += p1;
+		else
+		  *thiscoef += m1;
+	      }
+	    }
+	  } else {
+	    if (--r < 0)
+	      break;		/* reached target zero coefficient */
+	  }
+	  k++;
+	} while (k <= Se);
+	if (s) {
+	  int pos = jpeg_natural_order[k];
+	  /* Output newly nonzero coefficient */
+	  (*block)[pos] = (JCOEF) s;
+	  /* Remember its position in case we have to suspend */
+	  newnz_pos[num_newnz++] = pos;
+	}
+      }
+    }
+
+    if (EOBRUN > 0) {
+      /* Scan any remaining coefficient positions after the end-of-band
+       * (the last newly nonzero coefficient, if any).  Append a correction
+       * bit to each already-nonzero coefficient.  A correction bit is 1
+       * if the absolute value of the coefficient must be increased.
+       */
+      for (; k <= Se; k++) {
+	thiscoef = *block + jpeg_natural_order[k];
+	if (*thiscoef != 0) {
+	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+	  if (GET_BITS(1)) {
+	    if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
+	      if (*thiscoef >= 0)
+		*thiscoef += p1;
+	      else
+		*thiscoef += m1;
+	    }
+	  }
+	}
+      }
+      /* Count one block completed in EOB run */
+      EOBRUN--;
+    }
+
+    /* Completed MCU, so update state */
+    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
+    entropy->saved.EOBRUN = EOBRUN; /* only part of saved state we need */
+  }
+
+  /* Account for restart interval (no-op if not using restarts) */
+  entropy->restarts_to_go--;
+
+  return TRUE;
+
+undoit:
+  /* Re-zero any output coefficients that we made newly nonzero */
+  while (num_newnz > 0)
+    (*block)[newnz_pos[--num_newnz]] = 0;
+
+  return FALSE;
+}
+
+
+/*
+ * Module initialization routine for progressive Huffman entropy decoding.
+ */
+
+GLOBAL(void)
+jinit_phuff_decoder (j_decompress_ptr cinfo)
+{
+  phuff_entropy_ptr entropy;
+  int *coef_bit_ptr;
+  int ci, i;
+
+  entropy = (phuff_entropy_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(phuff_entropy_decoder));
+  cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
+  entropy->pub.start_pass = start_pass_phuff_decoder;
+
+  /* Mark derived tables unallocated */
+  for (i = 0; i < NUM_HUFF_TBLS; i++) {
+    entropy->derived_tbls[i] = NULL;
+  }
+
+  /* Create progression status table */
+  cinfo->coef_bits = (int (*)[DCTSIZE2])
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				cinfo->num_components*DCTSIZE2*SIZEOF(int));
+  coef_bit_ptr = & cinfo->coef_bits[0][0];
+  for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (i = 0; i < DCTSIZE2; i++)
+      *coef_bit_ptr++ = -1;
+}
+
+#endif /* D_PROGRESSIVE_SUPPORTED */
diff --git a/jpeg/jdpostct.c b/jpeg/jdpostct.c
new file mode 100644
index 000000000..571563d72
--- /dev/null
+++ b/jpeg/jdpostct.c
@@ -0,0 +1,290 @@
+/*
+ * jdpostct.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the decompression postprocessing controller.
+ * This controller manages the upsampling, color conversion, and color
+ * quantization/reduction steps; specifically, it controls the buffering
+ * between upsample/color conversion and color quantization/reduction.
+ *
+ * If no color quantization/reduction is required, then this module has no
+ * work to do, and it just hands off to the upsample/color conversion code.
+ * An integrated upsample/convert/quantize process would replace this module
+ * entirely.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_post_controller pub; /* public fields */
+
+  /* Color quantization source buffer: this holds output data from
+   * the upsample/color conversion step to be passed to the quantizer.
+   * For two-pass color quantization, we need a full-image buffer;
+   * for one-pass operation, a strip buffer is sufficient.
+   */
+  jvirt_sarray_ptr whole_image;	/* virtual array, or NULL if one-pass */
+  JSAMPARRAY buffer;		/* strip buffer, or current strip of virtual */
+  JDIMENSION strip_height;	/* buffer size in rows */
+  /* for two-pass mode only: */
+  JDIMENSION starting_row;	/* row # of first row in current strip */
+  JDIMENSION next_row;		/* index of next row to fill/empty in strip */
+} my_post_controller;
+
+typedef my_post_controller * my_post_ptr;
+
+
+/* Forward declarations */
+METHODDEF(void) post_process_1pass
+	JPP((j_decompress_ptr cinfo,
+	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	     JDIMENSION in_row_groups_avail,
+	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	     JDIMENSION out_rows_avail));
+#ifdef QUANT_2PASS_SUPPORTED
+METHODDEF(void) post_process_prepass
+	JPP((j_decompress_ptr cinfo,
+	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	     JDIMENSION in_row_groups_avail,
+	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	     JDIMENSION out_rows_avail));
+METHODDEF(void) post_process_2pass
+	JPP((j_decompress_ptr cinfo,
+	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	     JDIMENSION in_row_groups_avail,
+	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	     JDIMENSION out_rows_avail));
+#endif
+
+
+/*
+ * Initialize for a processing pass.
+ */
+
+METHODDEF(void)
+start_pass_dpost (j_decompress_ptr cinfo, J_BUF_MODE pass_mode)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+
+  switch (pass_mode) {
+  case JBUF_PASS_THRU:
+    if (cinfo->quantize_colors) {
+      /* Single-pass processing with color quantization. */
+      post->pub.post_process_data = post_process_1pass;
+      /* We could be doing buffered-image output before starting a 2-pass
+       * color quantization; in that case, jinit_d_post_controller did not
+       * allocate a strip buffer.  Use the virtual-array buffer as workspace.
+       */
+      if (post->buffer == NULL) {
+	post->buffer = (*cinfo->mem->access_virt_sarray)
+	  ((j_common_ptr) cinfo, post->whole_image,
+	   (JDIMENSION) 0, post->strip_height, TRUE);
+      }
+    } else {
+      /* For single-pass processing without color quantization,
+       * I have no work to do; just call the upsampler directly.
+       */
+      post->pub.post_process_data = cinfo->upsample->upsample;
+    }
+    break;
+#ifdef QUANT_2PASS_SUPPORTED
+  case JBUF_SAVE_AND_PASS:
+    /* First pass of 2-pass quantization */
+    if (post->whole_image == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    post->pub.post_process_data = post_process_prepass;
+    break;
+  case JBUF_CRANK_DEST:
+    /* Second pass of 2-pass quantization */
+    if (post->whole_image == NULL)
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    post->pub.post_process_data = post_process_2pass;
+    break;
+#endif /* QUANT_2PASS_SUPPORTED */
+  default:
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+    break;
+  }
+  post->starting_row = post->next_row = 0;
+}
+
+
+/*
+ * Process some data in the one-pass (strip buffer) case.
+ * This is used for color precision reduction as well as one-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_1pass (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+  JDIMENSION num_rows, max_rows;
+
+  /* Fill the buffer, but not more than what we can dump out in one go. */
+  /* Note we rely on the upsampler to detect bottom of image. */
+  max_rows = out_rows_avail - *out_row_ctr;
+  if (max_rows > post->strip_height)
+    max_rows = post->strip_height;
+  num_rows = 0;
+  (*cinfo->upsample->upsample) (cinfo,
+		input_buf, in_row_group_ctr, in_row_groups_avail,
+		post->buffer, &num_rows, max_rows);
+  /* Quantize and emit data. */
+  (*cinfo->cquantize->color_quantize) (cinfo,
+		post->buffer, output_buf + *out_row_ctr, (int) num_rows);
+  *out_row_ctr += num_rows;
+}
+
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+/*
+ * Process some data in the first pass of 2-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_prepass (j_decompress_ptr cinfo,
+		      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		      JDIMENSION in_row_groups_avail,
+		      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		      JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+  JDIMENSION old_next_row, num_rows;
+
+  /* Reposition virtual buffer if at start of strip. */
+  if (post->next_row == 0) {
+    post->buffer = (*cinfo->mem->access_virt_sarray)
+	((j_common_ptr) cinfo, post->whole_image,
+	 post->starting_row, post->strip_height, TRUE);
+  }
+
+  /* Upsample some data (up to a strip height's worth). */
+  old_next_row = post->next_row;
+  (*cinfo->upsample->upsample) (cinfo,
+		input_buf, in_row_group_ctr, in_row_groups_avail,
+		post->buffer, &post->next_row, post->strip_height);
+
+  /* Allow quantizer to scan new data.  No data is emitted, */
+  /* but we advance out_row_ctr so outer loop can tell when we're done. */
+  if (post->next_row > old_next_row) {
+    num_rows = post->next_row - old_next_row;
+    (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + old_next_row,
+					 (JSAMPARRAY) NULL, (int) num_rows);
+    *out_row_ctr += num_rows;
+  }
+
+  /* Advance if we filled the strip. */
+  if (post->next_row >= post->strip_height) {
+    post->starting_row += post->strip_height;
+    post->next_row = 0;
+  }
+}
+
+
+/*
+ * Process some data in the second pass of 2-pass quantization.
+ */
+
+METHODDEF(void)
+post_process_2pass (j_decompress_ptr cinfo,
+		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+		    JDIMENSION in_row_groups_avail,
+		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+		    JDIMENSION out_rows_avail)
+{
+  my_post_ptr post = (my_post_ptr) cinfo->post;
+  JDIMENSION num_rows, max_rows;
+
+  /* Reposition virtual buffer if at start of strip. */
+  if (post->next_row == 0) {
+    post->buffer = (*cinfo->mem->access_virt_sarray)
+	((j_common_ptr) cinfo, post->whole_image,
+	 post->starting_row, post->strip_height, FALSE);
+  }
+
+  /* Determine number of rows to emit. */
+  num_rows = post->strip_height - post->next_row; /* available in strip */
+  max_rows = out_rows_avail - *out_row_ctr; /* available in output area */
+  if (num_rows > max_rows)
+    num_rows = max_rows;
+  /* We have to check bottom of image here, can't depend on upsampler. */
+  max_rows = cinfo->output_height - post->starting_row;
+  if (num_rows > max_rows)
+    num_rows = max_rows;
+
+  /* Quantize and emit data. */
+  (*cinfo->cquantize->color_quantize) (cinfo,
+		post->buffer + post->next_row, output_buf + *out_row_ctr,
+		(int) num_rows);
+  *out_row_ctr += num_rows;
+
+  /* Advance if we filled the strip. */
+  post->next_row += num_rows;
+  if (post->next_row >= post->strip_height) {
+    post->starting_row += post->strip_height;
+    post->next_row = 0;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
+
+
+/*
+ * Initialize postprocessing controller.
+ */
+
+GLOBAL(void)
+jinit_d_post_controller (j_decompress_ptr cinfo, boolean need_full_buffer)
+{
+  my_post_ptr post;
+
+  post = (my_post_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_post_controller));
+  cinfo->post = (struct jpeg_d_post_controller *) post;
+  post->pub.start_pass = start_pass_dpost;
+  post->whole_image = NULL;	/* flag for no virtual arrays */
+  post->buffer = NULL;		/* flag for no strip buffer */
+
+  /* Create the quantization buffer, if needed */
+  if (cinfo->quantize_colors) {
+    /* The buffer strip height is max_v_samp_factor, which is typically
+     * an efficient number of rows for upsampling to return.
+     * (In the presence of output rescaling, we might want to be smarter?)
+     */
+    post->strip_height = (JDIMENSION) cinfo->max_v_samp_factor;
+    if (need_full_buffer) {
+      /* Two-pass color quantization: need full-image storage. */
+      /* We round up the number of rows to a multiple of the strip height. */
+#ifdef QUANT_2PASS_SUPPORTED
+      post->whole_image = (*cinfo->mem->request_virt_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+	 cinfo->output_width * cinfo->out_color_components,
+	 (JDIMENSION) jround_up((long) cinfo->output_height,
+				(long) post->strip_height),
+	 post->strip_height);
+#else
+      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+#endif /* QUANT_2PASS_SUPPORTED */
+    } else {
+      /* One-pass color quantization: just make a strip buffer. */
+      post->buffer = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 cinfo->output_width * cinfo->out_color_components,
+	 post->strip_height);
+    }
+  }
+}
diff --git a/jpeg/jdsample.c b/jpeg/jdsample.c
new file mode 100644
index 000000000..80ffefb2a
--- /dev/null
+++ b/jpeg/jdsample.c
@@ -0,0 +1,478 @@
+/*
+ * jdsample.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains upsampling routines.
+ *
+ * Upsampling input data is counted in "row groups".  A row group
+ * is defined to be (v_samp_factor * DCT_scaled_size / min_DCT_scaled_size)
+ * sample rows of each component.  Upsampling will normally produce
+ * max_v_samp_factor pixel rows from each row group (but this could vary
+ * if the upsampler is applying a scale factor of its own).
+ *
+ * An excellent reference for image resampling is
+ *   Digital Image Warping, George Wolberg, 1990.
+ *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Pointer to routine to upsample a single component */
+typedef JMETHOD(void, upsample1_ptr,
+		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;	/* public fields */
+
+  /* Color conversion buffer.  When using separate upsampling and color
+   * conversion steps, this buffer holds one upsampled row group until it
+   * has been color converted and output.
+   * Note: we do not allocate any storage for component(s) which are full-size,
+   * ie do not need rescaling.  The corresponding entry of color_buf[] is
+   * simply set to point to the input data array, thereby avoiding copying.
+   */
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+
+  /* Per-component upsampling method pointers */
+  upsample1_ptr methods[MAX_COMPONENTS];
+
+  int next_row_out;		/* counts rows emitted from color_buf */
+  JDIMENSION rows_to_go;	/* counts rows remaining in image */
+
+  /* Height of an input row group for each component. */
+  int rowgroup_height[MAX_COMPONENTS];
+
+  /* These arrays save pixel expansion factors so that int_expand need not
+   * recompute them each time.  They are unused for other upsampling methods.
+   */
+  UINT8 h_expand[MAX_COMPONENTS];
+  UINT8 v_expand[MAX_COMPONENTS];
+} my_upsampler;
+
+typedef my_upsampler * my_upsample_ptr;
+
+
+/*
+ * Initialize for an upsampling pass.
+ */
+
+METHODDEF(void)
+start_pass_upsample (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  /* Mark the conversion buffer empty */
+  upsample->next_row_out = cinfo->max_v_samp_factor;
+  /* Initialize total-height counter for detecting bottom of image */
+  upsample->rows_to_go = cinfo->output_height;
+}
+
+
+/*
+ * Control routine to do upsampling (and color conversion).
+ *
+ * In this version we upsample each component independently.
+ * We upsample one row group into the conversion buffer, then apply
+ * color conversion a row at a time.
+ */
+
+METHODDEF(void)
+sep_upsample (j_decompress_ptr cinfo,
+	      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+	      JDIMENSION in_row_groups_avail,
+	      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+	      JDIMENSION out_rows_avail)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  int ci;
+  jpeg_component_info * compptr;
+  JDIMENSION num_rows;
+
+  /* Fill the conversion buffer, if it's empty */
+  if (upsample->next_row_out >= cinfo->max_v_samp_factor) {
+    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+	 ci++, compptr++) {
+      /* Invoke per-component upsample method.  Notice we pass a POINTER
+       * to color_buf[ci], so that fullsize_upsample can change it.
+       */
+      (*upsample->methods[ci]) (cinfo, compptr,
+	input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
+	upsample->color_buf + ci);
+    }
+    upsample->next_row_out = 0;
+  }
+
+  /* Color-convert and emit rows */
+
+  /* How many we have in the buffer: */
+  num_rows = (JDIMENSION) (cinfo->max_v_samp_factor - upsample->next_row_out);
+  /* Not more than the distance to the end of the image.  Need this test
+   * in case the image height is not a multiple of max_v_samp_factor:
+   */
+  if (num_rows > upsample->rows_to_go) 
+    num_rows = upsample->rows_to_go;
+  /* And not more than what the client can accept: */
+  out_rows_avail -= *out_row_ctr;
+  if (num_rows > out_rows_avail)
+    num_rows = out_rows_avail;
+
+  (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf,
+				     (JDIMENSION) upsample->next_row_out,
+				     output_buf + *out_row_ctr,
+				     (int) num_rows);
+
+  /* Adjust counts */
+  *out_row_ctr += num_rows;
+  upsample->rows_to_go -= num_rows;
+  upsample->next_row_out += num_rows;
+  /* When the buffer is emptied, declare this input row group consumed */
+  if (upsample->next_row_out >= cinfo->max_v_samp_factor)
+    (*in_row_group_ctr)++;
+}
+
+
+/*
+ * These are the routines invoked by sep_upsample to upsample pixel values
+ * of a single component.  One row group is processed per call.
+ */
+
+
+/*
+ * For full-size components, we just make color_buf[ci] point at the
+ * input buffer, and thus avoid copying any data.  Note that this is
+ * safe only because sep_upsample doesn't declare the input row group
+ * "consumed" until we are done color converting and emitting it.
+ */
+
+METHODDEF(void)
+fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		   JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  *output_data_ptr = input_data;
+}
+
+
+/*
+ * This is a no-op version used for "uninteresting" components.
+ * These components will not be referenced by color conversion.
+ */
+
+METHODDEF(void)
+noop_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  *output_data_ptr = NULL;	/* safety check */
+}
+
+
+/*
+ * This version handles any integral sampling ratios.
+ * This is not used for typical JPEG files, so it need not be fast.
+ * Nor, for that matter, is it particularly accurate: the algorithm is
+ * simple replication of the input pixel onto the corresponding output
+ * pixels.  The hi-falutin sampling literature refers to this as a
+ * "box filter".  A box filter tends to introduce visible artifacts,
+ * so if you are actually going to use 3:1 or 4:1 sampling ratios
+ * you would be well advised to improve this code.
+ */
+
+METHODDEF(void)
+int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	      JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  register int h;
+  JSAMPROW outend;
+  int h_expand, v_expand;
+  int inrow, outrow;
+
+  h_expand = upsample->h_expand[compptr->component_index];
+  v_expand = upsample->v_expand[compptr->component_index];
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    /* Generate one output row with proper horizontal expansion */
+    inptr = input_data[inrow];
+    outptr = output_data[outrow];
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      for (h = h_expand; h > 0; h--) {
+	*outptr++ = invalue;
+      }
+    }
+    /* Generate any additional output rows by duplicating the first one */
+    if (v_expand > 1) {
+      jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
+			v_expand-1, cinfo->output_width);
+    }
+    inrow++;
+    outrow += v_expand;
+  }
+}
+
+
+/*
+ * Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+ * It's still a box filter.
+ */
+
+METHODDEF(void)
+h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  JSAMPROW outend;
+  int inrow;
+
+  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      *outptr++ = invalue;
+      *outptr++ = invalue;
+    }
+  }
+}
+
+
+/*
+ * Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+ * It's still a box filter.
+ */
+
+METHODDEF(void)
+h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register JSAMPLE invalue;
+  JSAMPROW outend;
+  int inrow, outrow;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    inptr = input_data[inrow];
+    outptr = output_data[outrow];
+    outend = outptr + cinfo->output_width;
+    while (outptr < outend) {
+      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      *outptr++ = invalue;
+      *outptr++ = invalue;
+    }
+    jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
+		      1, cinfo->output_width);
+    inrow++;
+    outrow += 2;
+  }
+}
+
+
+/*
+ * Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+ *
+ * The upsampling algorithm is linear interpolation between pixel centers,
+ * also known as a "triangle filter".  This is a good compromise between
+ * speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+ * of the way between input pixel centers.
+ *
+ * A note about the "bias" calculations: when rounding fractional values to
+ * integer, we do not want to always round 0.5 up to the next integer.
+ * If we did that, we'd introduce a noticeable bias towards larger values.
+ * Instead, this code is arranged so that 0.5 will be rounded up or down at
+ * alternate pixel locations (a simple ordered dither pattern).
+ */
+
+METHODDEF(void)
+h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr, outptr;
+  register int invalue;
+  register JDIMENSION colctr;
+  int inrow;
+
+  for (inrow = 0; inrow < cinfo->max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+    /* Special case for first column */
+    invalue = GETJSAMPLE(*inptr++);
+    *outptr++ = (JSAMPLE) invalue;
+    *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(*inptr) + 2) >> 2);
+
+    for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
+      /* General case: 3/4 * nearer pixel + 1/4 * further pixel */
+      invalue = GETJSAMPLE(*inptr++) * 3;
+      *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(inptr[-2]) + 1) >> 2);
+      *outptr++ = (JSAMPLE) ((invalue + GETJSAMPLE(*inptr) + 2) >> 2);
+    }
+
+    /* Special case for last column */
+    invalue = GETJSAMPLE(*inptr);
+    *outptr++ = (JSAMPLE) ((invalue * 3 + GETJSAMPLE(inptr[-1]) + 1) >> 2);
+    *outptr++ = (JSAMPLE) invalue;
+  }
+}
+
+
+/*
+ * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+ * Again a triangle filter; see comments for h2v1 case, above.
+ *
+ * It is OK for us to reference the adjacent input rows because we demanded
+ * context from the main buffer controller (see initialization code).
+ */
+
+METHODDEF(void)
+h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  register JSAMPROW inptr0, inptr1, outptr;
+#if BITS_IN_JSAMPLE == 8
+  register int thiscolsum, lastcolsum, nextcolsum;
+#else
+  register INT32 thiscolsum, lastcolsum, nextcolsum;
+#endif
+  register JDIMENSION colctr;
+  int inrow, outrow, v;
+
+  inrow = outrow = 0;
+  while (outrow < cinfo->max_v_samp_factor) {
+    for (v = 0; v < 2; v++) {
+      /* inptr0 points to nearest input row, inptr1 points to next nearest */
+      inptr0 = input_data[inrow];
+      if (v == 0)		/* next nearest is row above */
+	inptr1 = input_data[inrow-1];
+      else			/* next nearest is row below */
+	inptr1 = input_data[inrow+1];
+      outptr = output_data[outrow++];
+
+      /* Special case for first column */
+      thiscolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+      *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 8) >> 4);
+      *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
+      lastcolsum = thiscolsum; thiscolsum = nextcolsum;
+
+      for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
+	/* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
+	/* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
+	nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
+	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
+	lastcolsum = thiscolsum; thiscolsum = nextcolsum;
+      }
+
+      /* Special case for last column */
+      *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
+      *outptr++ = (JSAMPLE) ((thiscolsum * 4 + 7) >> 4);
+    }
+    inrow++;
+  }
+}
+
+
+/*
+ * Module initialization routine for upsampling.
+ */
+
+GLOBAL(void)
+jinit_upsampler (j_decompress_ptr cinfo)
+{
+  my_upsample_ptr upsample;
+  int ci;
+  jpeg_component_info * compptr;
+  boolean need_buffer, do_fancy;
+  int h_in_group, v_in_group, h_out_group, v_out_group;
+
+  upsample = (my_upsample_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_upsampler));
+  cinfo->upsample = (struct jpeg_upsampler *) upsample;
+  upsample->pub.start_pass = start_pass_upsample;
+  upsample->pub.upsample = sep_upsample;
+  upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+
+  if (cinfo->CCIR601_sampling)	/* this isn't supported */
+    ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
+
+  /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1,
+   * so don't ask for it.
+   */
+  do_fancy = cinfo->do_fancy_upsampling && cinfo->min_DCT_scaled_size > 1;
+
+  /* Verify we can handle the sampling factors, select per-component methods,
+   * and create storage as needed.
+   */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Compute size of an "input group" after IDCT scaling.  This many samples
+     * are to be converted to max_h_samp_factor * max_v_samp_factor pixels.
+     */
+    h_in_group = (compptr->h_samp_factor * compptr->DCT_scaled_size) /
+		 cinfo->min_DCT_scaled_size;
+    v_in_group = (compptr->v_samp_factor * compptr->DCT_scaled_size) /
+		 cinfo->min_DCT_scaled_size;
+    h_out_group = cinfo->max_h_samp_factor;
+    v_out_group = cinfo->max_v_samp_factor;
+    upsample->rowgroup_height[ci] = v_in_group; /* save for use later */
+    need_buffer = TRUE;
+    if (! compptr->component_needed) {
+      /* Don't bother to upsample an uninteresting component. */
+      upsample->methods[ci] = noop_upsample;
+      need_buffer = FALSE;
+    } else if (h_in_group == h_out_group && v_in_group == v_out_group) {
+      /* Fullsize components can be processed without any work. */
+      upsample->methods[ci] = fullsize_upsample;
+      need_buffer = FALSE;
+    } else if (h_in_group * 2 == h_out_group &&
+	       v_in_group == v_out_group) {
+      /* Special cases for 2h1v upsampling */
+      if (do_fancy && compptr->downsampled_width > 2)
+	upsample->methods[ci] = h2v1_fancy_upsample;
+      else
+	upsample->methods[ci] = h2v1_upsample;
+    } else if (h_in_group * 2 == h_out_group &&
+	       v_in_group * 2 == v_out_group) {
+      /* Special cases for 2h2v upsampling */
+      if (do_fancy && compptr->downsampled_width > 2) {
+	upsample->methods[ci] = h2v2_fancy_upsample;
+	upsample->pub.need_context_rows = TRUE;
+      } else
+	upsample->methods[ci] = h2v2_upsample;
+    } else if ((h_out_group % h_in_group) == 0 &&
+	       (v_out_group % v_in_group) == 0) {
+      /* Generic integral-factors upsampling method */
+      upsample->methods[ci] = int_upsample;
+      upsample->h_expand[ci] = (UINT8) (h_out_group / h_in_group);
+      upsample->v_expand[ci] = (UINT8) (v_out_group / v_in_group);
+    } else
+      ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
+    if (need_buffer) {
+      upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray)
+	((j_common_ptr) cinfo, JPOOL_IMAGE,
+	 (JDIMENSION) jround_up((long) cinfo->output_width,
+				(long) cinfo->max_h_samp_factor),
+	 (JDIMENSION) cinfo->max_v_samp_factor);
+    }
+  }
+}
diff --git a/jpeg/jdtrans.c b/jpeg/jdtrans.c
new file mode 100644
index 000000000..6c0ab715d
--- /dev/null
+++ b/jpeg/jdtrans.c
@@ -0,0 +1,143 @@
+/*
+ * jdtrans.c
+ *
+ * Copyright (C) 1995-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains library routines for transcoding decompression,
+ * that is, reading raw DCT coefficient arrays from an input JPEG file.
+ * The routines in jdapimin.c will also be needed by a transcoder.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/* Forward declarations */
+LOCAL(void) transdecode_master_selection JPP((j_decompress_ptr cinfo));
+
+
+/*
+ * Read the coefficient arrays from a JPEG file.
+ * jpeg_read_header must be completed before calling this.
+ *
+ * The entire image is read into a set of virtual coefficient-block arrays,
+ * one per component.  The return value is a pointer to the array of
+ * virtual-array descriptors.  These can be manipulated directly via the
+ * JPEG memory manager, or handed off to jpeg_write_coefficients().
+ * To release the memory occupied by the virtual arrays, call
+ * jpeg_finish_decompress() when done with the data.
+ *
+ * An alternative usage is to simply obtain access to the coefficient arrays
+ * during a buffered-image-mode decompression operation.  This is allowed
+ * after any jpeg_finish_output() call.  The arrays can be accessed until
+ * jpeg_finish_decompress() is called.  (Note that any call to the library
+ * may reposition the arrays, so don't rely on access_virt_barray() results
+ * to stay valid across library calls.)
+ *
+ * Returns NULL if suspended.  This case need be checked only if
+ * a suspending data source is used.
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jpeg_read_coefficients (j_decompress_ptr cinfo)
+{
+  if (cinfo->global_state == DSTATE_READY) {
+    /* First call: initialize active modules */
+    transdecode_master_selection(cinfo);
+    cinfo->global_state = DSTATE_RDCOEFS;
+  }
+  if (cinfo->global_state == DSTATE_RDCOEFS) {
+    /* Absorb whole file into the coef buffer */
+    for (;;) {
+      int retcode;
+      /* Call progress monitor hook if present */
+      if (cinfo->progress != NULL)
+	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+      /* Absorb some more input */
+      retcode = (*cinfo->inputctl->consume_input) (cinfo);
+      if (retcode == JPEG_SUSPENDED)
+	return NULL;
+      if (retcode == JPEG_REACHED_EOI)
+	break;
+      /* Advance progress counter if appropriate */
+      if (cinfo->progress != NULL &&
+	  (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+	if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+	  /* startup underestimated number of scans; ratchet up one scan */
+	  cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+	}
+      }
+    }
+    /* Set state so that jpeg_finish_decompress does the right thing */
+    cinfo->global_state = DSTATE_STOPPING;
+  }
+  /* At this point we should be in state DSTATE_STOPPING if being used
+   * standalone, or in state DSTATE_BUFIMAGE if being invoked to get access
+   * to the coefficients during a full buffered-image-mode decompression.
+   */
+  if ((cinfo->global_state == DSTATE_STOPPING ||
+       cinfo->global_state == DSTATE_BUFIMAGE) && cinfo->buffered_image) {
+    return cinfo->coef->coef_arrays;
+  }
+  /* Oops, improper usage */
+  ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+  return NULL;			/* keep compiler happy */
+}
+
+
+/*
+ * Master selection of decompression modules for transcoding.
+ * This substitutes for jdmaster.c's initialization of the full decompressor.
+ */
+
+LOCAL(void)
+transdecode_master_selection (j_decompress_ptr cinfo)
+{
+  /* This is effectively a buffered-image operation. */
+  cinfo->buffered_image = TRUE;
+
+  /* Entropy decoding: either Huffman or arithmetic coding. */
+  if (cinfo->arith_code) {
+    ERREXIT(cinfo, JERR_ARITH_NOTIMPL);
+  } else {
+    if (cinfo->progressive_mode) {
+#ifdef D_PROGRESSIVE_SUPPORTED
+      jinit_phuff_decoder(cinfo);
+#else
+      ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+    } else
+      jinit_huff_decoder(cinfo);
+  }
+
+  /* Always get a full-image coefficient buffer. */
+  jinit_d_coef_controller(cinfo, TRUE);
+
+  /* We can now tell the memory manager to allocate virtual arrays. */
+  (*cinfo->mem->realize_virt_arrays) ((j_common_ptr) cinfo);
+
+  /* Initialize input side of decompressor to consume first scan. */
+  (*cinfo->inputctl->start_input_pass) (cinfo);
+
+  /* Initialize progress monitoring. */
+  if (cinfo->progress != NULL) {
+    int nscans;
+    /* Estimate number of scans to set pass_limit. */
+    if (cinfo->progressive_mode) {
+      /* Arbitrarily estimate 2 interleaved DC scans + 3 AC scans/component. */
+      nscans = 2 + 3 * cinfo->num_components;
+    } else if (cinfo->inputctl->has_multiple_scans) {
+      /* For a nonprogressive multiscan file, estimate 1 scan per component. */
+      nscans = cinfo->num_components;
+    } else {
+      nscans = 1;
+    }
+    cinfo->progress->pass_counter = 0L;
+    cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows * nscans;
+    cinfo->progress->completed_passes = 0;
+    cinfo->progress->total_passes = 1;
+  }
+}
diff --git a/jpeg/jerror.c b/jpeg/jerror.c
new file mode 100644
index 000000000..a44463cf1
--- /dev/null
+++ b/jpeg/jerror.c
@@ -0,0 +1,253 @@
+/*
+ * jerror.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains simple error-reporting and trace-message routines.
+ * These are suitable for Unix-like systems and others where writing to
+ * stderr is the right thing to do.  Many applications will want to replace
+ * some or all of these routines.
+ *
+ * If you define USE_WINDOWS_MESSAGEBOX in jconfig.h or in the makefile,
+ * you get a Windows-specific hack to display error messages in a dialog box.
+ * It ain't much, but it beats dropping error messages into the bit bucket,
+ * which is what happens to output to stderr under most Windows C compilers.
+ *
+ * These routines are used by both the compression and decompression code.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jversion.h"
+#include "jerror.h"
+#include <stdlib.h>
+
+#ifdef USE_WINDOWS_MESSAGEBOX
+#include <windows.h>
+#endif
+
+#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+
+
+/*
+ * Create the message string table.
+ * We do this from the master message list in jerror.h by re-reading
+ * jerror.h with a suitable definition for macro JMESSAGE.
+ * The message table is made an external symbol just in case any applications
+ * want to refer to it directly.
+ */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_std_message_table	jMsgTable
+#endif
+
+#define JMESSAGE(code,string)	string ,
+
+const char * const jpeg_std_message_table[] = {
+#include "jerror.h"
+  NULL
+};
+
+
+/*
+ * Error exit handler: must not return to caller.
+ *
+ * Applications may override this if they want to get control back after
+ * an error.  Typically one would longjmp somewhere instead of exiting.
+ * The setjmp buffer can be made a private field within an expanded error
+ * handler object.  Note that the info needed to generate an error message
+ * is stored in the error object, so you can generate the message now or
+ * later, at your convenience.
+ * You should make sure that the JPEG object is cleaned up (with jpeg_abort
+ * or jpeg_destroy) at some point.
+ */
+
+METHODDEF(void)
+error_exit (j_common_ptr cinfo)
+{
+  /* Always display the message */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Let the memory manager delete any temp files before we die */
+  jpeg_destroy(cinfo);
+
+  exit(EXIT_FAILURE);
+}
+
+
+/*
+ * Actual output of an error or trace message.
+ * Applications may override this method to send JPEG messages somewhere
+ * other than stderr.
+ *
+ * On Windows, printing to stderr is generally completely useless,
+ * so we provide optional code to produce an error-dialog popup.
+ * Most Windows applications will still prefer to override this routine,
+ * but if they don't, it'll do something at least marginally useful.
+ *
+ * NOTE: to use the library in an environment that doesn't support the
+ * C stdio library, you may have to delete the call to fprintf() entirely,
+ * not just not use this routine.
+ */
+
+METHODDEF(void)
+output_message (j_common_ptr cinfo)
+{
+  char buffer[JMSG_LENGTH_MAX];
+
+  /* Create the message */
+  (*cinfo->err->format_message) (cinfo, buffer);
+
+#ifdef USE_WINDOWS_MESSAGEBOX
+  /* Display it in a message dialog box */
+  MessageBox(GetActiveWindow(), buffer, "JPEG Library Error",
+	     MB_OK | MB_ICONERROR);
+#else
+  /* Send it to stderr, adding a newline */
+  fprintf(stderr, "%s\n", buffer);
+#endif
+}
+
+
+/*
+ * Decide whether to emit a trace or warning message.
+ * msg_level is one of:
+ *   -1: recoverable corrupt-data warning, may want to abort.
+ *    0: important advisory messages (always display to user).
+ *    1: first level of tracing detail.
+ *    2,3,...: successively more detailed tracing messages.
+ * An application might override this method if it wanted to abort on warnings
+ * or change the policy about which messages to display.
+ */
+
+METHODDEF(void)
+emit_message (j_common_ptr cinfo, int msg_level)
+{
+  struct jpeg_error_mgr * err = cinfo->err;
+
+  if (msg_level < 0) {
+    /* It's a warning message.  Since corrupt files may generate many warnings,
+     * the policy implemented here is to show only the first warning,
+     * unless trace_level >= 3.
+     */
+    if (err->num_warnings == 0 || err->trace_level >= 3)
+      (*err->output_message) (cinfo);
+    /* Always count warnings in num_warnings. */
+    err->num_warnings++;
+  } else {
+    /* It's a trace message.  Show it if trace_level >= msg_level. */
+    if (err->trace_level >= msg_level)
+      (*err->output_message) (cinfo);
+  }
+}
+
+
+/*
+ * Format a message string for the most recent JPEG error or message.
+ * The message is stored into buffer, which should be at least JMSG_LENGTH_MAX
+ * characters.  Note that no '\n' character is added to the string.
+ * Few applications should need to override this method.
+ */
+
+METHODDEF(void)
+format_message (j_common_ptr cinfo, char * buffer)
+{
+  struct jpeg_error_mgr * err = cinfo->err;
+  int msg_code = err->msg_code;
+  const char * msgtext = NULL;
+  const char * msgptr;
+  char ch;
+  boolean isstring;
+
+  /* Look up message string in proper table */
+  if (msg_code > 0 && msg_code <= err->last_jpeg_message) {
+    msgtext = err->jpeg_message_table[msg_code];
+  } else if (err->addon_message_table != NULL &&
+	     msg_code >= err->first_addon_message &&
+	     msg_code <= err->last_addon_message) {
+    msgtext = err->addon_message_table[msg_code - err->first_addon_message];
+  }
+
+  /* Defend against bogus message number */
+  if (msgtext == NULL) {
+    err->msg_parm.i[0] = msg_code;
+    msgtext = err->jpeg_message_table[0];
+  }
+
+  /* Check for string parameter, as indicated by %s in the message text */
+  isstring = FALSE;
+  msgptr = msgtext;
+  while ((ch = *msgptr++) != '\0') {
+    if (ch == '%') {
+      if (*msgptr == 's') isstring = TRUE;
+      break;
+    }
+  }
+
+  /* Format the message into the passed buffer */
+  if (isstring)
+    sprintf(buffer, msgtext, err->msg_parm.s);
+  else
+    sprintf(buffer, msgtext,
+	    err->msg_parm.i[0], err->msg_parm.i[1],
+	    err->msg_parm.i[2], err->msg_parm.i[3],
+	    err->msg_parm.i[4], err->msg_parm.i[5],
+	    err->msg_parm.i[6], err->msg_parm.i[7]);
+}
+
+
+/*
+ * Reset error state variables at start of a new image.
+ * This is called during compression startup to reset trace/error
+ * processing to default state, without losing any application-specific
+ * method pointers.  An application might possibly want to override
+ * this method if it has additional error processing state.
+ */
+
+METHODDEF(void)
+reset_error_mgr (j_common_ptr cinfo)
+{
+  cinfo->err->num_warnings = 0;
+  /* trace_level is not reset since it is an application-supplied parameter */
+  cinfo->err->msg_code = 0;	/* may be useful as a flag for "no error" */
+}
+
+
+/*
+ * Fill in the standard error-handling methods in a jpeg_error_mgr object.
+ * Typical call is:
+ *	struct jpeg_compress_struct cinfo;
+ *	struct jpeg_error_mgr err;
+ *
+ *	cinfo.err = jpeg_std_error(&err);
+ * after which the application may override some of the methods.
+ */
+
+GLOBAL(struct jpeg_error_mgr *)
+jpeg_std_error (struct jpeg_error_mgr * err)
+{
+  err->error_exit = error_exit;
+  err->emit_message = emit_message;
+  err->output_message = output_message;
+  err->format_message = format_message;
+  err->reset_error_mgr = reset_error_mgr;
+
+  err->trace_level = 0;		/* default = no tracing */
+  err->num_warnings = 0;	/* no warnings emitted yet */
+  err->msg_code = 0;		/* may be useful as a flag for "no error" */
+
+  /* Initialize message table pointers */
+  err->jpeg_message_table = jpeg_std_message_table;
+  err->last_jpeg_message = (int) JMSG_LASTMSGCODE - 1;
+
+  err->addon_message_table = NULL;
+  err->first_addon_message = 0;	/* for safety */
+  err->last_addon_message = 0;
+
+  return err;
+}
diff --git a/jpeg/jerror.h b/jpeg/jerror.h
new file mode 100644
index 000000000..fc2fffeac
--- /dev/null
+++ b/jpeg/jerror.h
@@ -0,0 +1,291 @@
+/*
+ * jerror.h
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file defines the error and message codes for the JPEG library.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ * A set of error-reporting macros are defined too.  Some applications using
+ * the JPEG library may wish to include this file to get the error codes
+ * and/or the macros.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef JERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code,string)
+#endif /* JERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code,string)	code ,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_NOMESSAGE, "Bogus message code %d") /* Must be first entry! */
+
+/* For maintenance convenience, list is alphabetical by message code name */
+JMESSAGE(JERR_ARITH_NOTIMPL,
+	 "Sorry, there are legal restrictions on arithmetic coding")
+JMESSAGE(JERR_BAD_ALIGN_TYPE, "ALIGN_TYPE is wrong, please fix")
+JMESSAGE(JERR_BAD_ALLOC_CHUNK, "MAX_ALLOC_CHUNK is wrong, please fix")
+JMESSAGE(JERR_BAD_BUFFER_MODE, "Bogus buffer control mode")
+JMESSAGE(JERR_BAD_COMPONENT_ID, "Invalid component ID %d in SOS")
+JMESSAGE(JERR_BAD_DCT_COEF, "DCT coefficient out of range")
+JMESSAGE(JERR_BAD_DCTSIZE, "IDCT output block size %d not supported")
+JMESSAGE(JERR_BAD_HUFF_TABLE, "Bogus Huffman table definition")
+JMESSAGE(JERR_BAD_IN_COLORSPACE, "Bogus input colorspace")
+JMESSAGE(JERR_BAD_J_COLORSPACE, "Bogus JPEG colorspace")
+JMESSAGE(JERR_BAD_LENGTH, "Bogus marker length")
+JMESSAGE(JERR_BAD_LIB_VERSION,
+	 "Wrong JPEG library version: library is %d, caller expects %d")
+JMESSAGE(JERR_BAD_MCU_SIZE, "Sampling factors too large for interleaved scan")
+JMESSAGE(JERR_BAD_POOL_ID, "Invalid memory pool code %d")
+JMESSAGE(JERR_BAD_PRECISION, "Unsupported JPEG data precision %d")
+JMESSAGE(JERR_BAD_PROGRESSION,
+	 "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
+JMESSAGE(JERR_BAD_PROG_SCRIPT,
+	 "Invalid progressive parameters at scan script entry %d")
+JMESSAGE(JERR_BAD_SAMPLING, "Bogus sampling factors")
+JMESSAGE(JERR_BAD_SCAN_SCRIPT, "Invalid scan script at entry %d")
+JMESSAGE(JERR_BAD_STATE, "Improper call to JPEG library in state %d")
+JMESSAGE(JERR_BAD_STRUCT_SIZE,
+	 "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
+JMESSAGE(JERR_BAD_VIRTUAL_ACCESS, "Bogus virtual array access")
+JMESSAGE(JERR_BUFFER_SIZE, "Buffer passed to JPEG library is too small")
+JMESSAGE(JERR_CANT_SUSPEND, "Suspension not allowed here")
+JMESSAGE(JERR_CCIR601_NOTIMPL, "CCIR601 sampling not implemented yet")
+JMESSAGE(JERR_COMPONENT_COUNT, "Too many color components: %d, max %d")
+JMESSAGE(JERR_CONVERSION_NOTIMPL, "Unsupported color conversion request")
+JMESSAGE(JERR_DAC_INDEX, "Bogus DAC index %d")
+JMESSAGE(JERR_DAC_VALUE, "Bogus DAC value 0x%x")
+JMESSAGE(JERR_DHT_INDEX, "Bogus DHT index %d")
+JMESSAGE(JERR_DQT_INDEX, "Bogus DQT index %d")
+JMESSAGE(JERR_EMPTY_IMAGE, "Empty JPEG image (DNL not supported)")
+JMESSAGE(JERR_EMS_READ, "Read from EMS failed")
+JMESSAGE(JERR_EMS_WRITE, "Write to EMS failed")
+JMESSAGE(JERR_EOI_EXPECTED, "Didn't expect more than one scan")
+JMESSAGE(JERR_FILE_READ, "Input file read error")
+JMESSAGE(JERR_FILE_WRITE, "Output file write error --- out of disk space?")
+JMESSAGE(JERR_FRACT_SAMPLE_NOTIMPL, "Fractional sampling not implemented yet")
+JMESSAGE(JERR_HUFF_CLEN_OVERFLOW, "Huffman code size table overflow")
+JMESSAGE(JERR_HUFF_MISSING_CODE, "Missing Huffman code table entry")
+JMESSAGE(JERR_IMAGE_TOO_BIG, "Maximum supported image dimension is %u pixels")
+JMESSAGE(JERR_INPUT_EMPTY, "Empty input file")
+JMESSAGE(JERR_INPUT_EOF, "Premature end of input file")
+JMESSAGE(JERR_MISMATCHED_QUANT_TABLE,
+	 "Cannot transcode due to multiple use of quantization table %d")
+JMESSAGE(JERR_MISSING_DATA, "Scan script does not transmit all data")
+JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change")
+JMESSAGE(JERR_NOTIMPL, "Not implemented yet")
+JMESSAGE(JERR_NOT_COMPILED, "Requested feature was omitted at compile time")
+JMESSAGE(JERR_NO_BACKING_STORE, "Backing store not supported")
+JMESSAGE(JERR_NO_HUFF_TABLE, "Huffman table 0x%02x was not defined")
+JMESSAGE(JERR_NO_IMAGE, "JPEG datastream contains no image")
+JMESSAGE(JERR_NO_QUANT_TABLE, "Quantization table 0x%02x was not defined")
+JMESSAGE(JERR_NO_SOI, "Not a JPEG file: starts with 0x%02x 0x%02x")
+JMESSAGE(JERR_OUT_OF_MEMORY, "Insufficient memory (case %d)")
+JMESSAGE(JERR_QUANT_COMPONENTS,
+	 "Cannot quantize more than %d color components")
+JMESSAGE(JERR_QUANT_FEW_COLORS, "Cannot quantize to fewer than %d colors")
+JMESSAGE(JERR_QUANT_MANY_COLORS, "Cannot quantize to more than %d colors")
+JMESSAGE(JERR_SOF_DUPLICATE, "Invalid JPEG file structure: two SOF markers")
+JMESSAGE(JERR_SOF_NO_SOS, "Invalid JPEG file structure: missing SOS marker")
+JMESSAGE(JERR_SOF_UNSUPPORTED, "Unsupported JPEG process: SOF type 0x%02x")
+JMESSAGE(JERR_SOI_DUPLICATE, "Invalid JPEG file structure: two SOI markers")
+JMESSAGE(JERR_SOS_NO_SOF, "Invalid JPEG file structure: SOS before SOF")
+JMESSAGE(JERR_TFILE_CREATE, "Failed to create temporary file %s")
+JMESSAGE(JERR_TFILE_READ, "Read failed on temporary file")
+JMESSAGE(JERR_TFILE_SEEK, "Seek failed on temporary file")
+JMESSAGE(JERR_TFILE_WRITE,
+	 "Write failed on temporary file --- out of disk space?")
+JMESSAGE(JERR_TOO_LITTLE_DATA, "Application transferred too few scanlines")
+JMESSAGE(JERR_UNKNOWN_MARKER, "Unsupported marker type 0x%02x")
+JMESSAGE(JERR_VIRTUAL_BUG, "Virtual array controller messed up")
+JMESSAGE(JERR_WIDTH_OVERFLOW, "Image too wide for this implementation")
+JMESSAGE(JERR_XMS_READ, "Read from XMS failed")
+JMESSAGE(JERR_XMS_WRITE, "Write to XMS failed")
+JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT)
+JMESSAGE(JMSG_VERSION, JVERSION)
+JMESSAGE(JTRC_16BIT_TABLES,
+	 "Caution: quantization tables are too coarse for baseline JPEG")
+JMESSAGE(JTRC_ADOBE,
+	 "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
+JMESSAGE(JTRC_APP0, "Unknown APP0 marker (not JFIF), length %u")
+JMESSAGE(JTRC_APP14, "Unknown APP14 marker (not Adobe), length %u")
+JMESSAGE(JTRC_DAC, "Define Arithmetic Table 0x%02x: 0x%02x")
+JMESSAGE(JTRC_DHT, "Define Huffman Table 0x%02x")
+JMESSAGE(JTRC_DQT, "Define Quantization Table %d  precision %d")
+JMESSAGE(JTRC_DRI, "Define Restart Interval %u")
+JMESSAGE(JTRC_EMS_CLOSE, "Freed EMS handle %u")
+JMESSAGE(JTRC_EMS_OPEN, "Obtained EMS handle %u")
+JMESSAGE(JTRC_EOI, "End Of Image")
+JMESSAGE(JTRC_HUFFBITS, "        %3d %3d %3d %3d %3d %3d %3d %3d")
+JMESSAGE(JTRC_JFIF, "JFIF APP0 marker: version %d.%02d, density %dx%d  %d")
+JMESSAGE(JTRC_JFIF_BADTHUMBNAILSIZE,
+	 "Warning: thumbnail image size does not match data length %u")
+JMESSAGE(JTRC_JFIF_EXTENSION,
+	 "JFIF extension marker: type 0x%02x, length %u")
+JMESSAGE(JTRC_JFIF_THUMBNAIL, "    with %d x %d thumbnail image")
+JMESSAGE(JTRC_MISC_MARKER, "Miscellaneous marker 0x%02x, length %u")
+JMESSAGE(JTRC_PARMLESS_MARKER, "Unexpected marker 0x%02x")
+JMESSAGE(JTRC_QUANTVALS, "        %4u %4u %4u %4u %4u %4u %4u %4u")
+JMESSAGE(JTRC_QUANT_3_NCOLORS, "Quantizing to %d = %d*%d*%d colors")
+JMESSAGE(JTRC_QUANT_NCOLORS, "Quantizing to %d colors")
+JMESSAGE(JTRC_QUANT_SELECTED, "Selected %d colors for quantization")
+JMESSAGE(JTRC_RECOVERY_ACTION, "At marker 0x%02x, recovery action %d")
+JMESSAGE(JTRC_RST, "RST%d")
+JMESSAGE(JTRC_SMOOTH_NOTIMPL,
+	 "Smoothing not supported with nonstandard sampling ratios")
+JMESSAGE(JTRC_SOF, "Start Of Frame 0x%02x: width=%u, height=%u, components=%d")
+JMESSAGE(JTRC_SOF_COMPONENT, "    Component %d: %dhx%dv q=%d")
+JMESSAGE(JTRC_SOI, "Start of Image")
+JMESSAGE(JTRC_SOS, "Start Of Scan: %d components")
+JMESSAGE(JTRC_SOS_COMPONENT, "    Component %d: dc=%d ac=%d")
+JMESSAGE(JTRC_SOS_PARAMS, "  Ss=%d, Se=%d, Ah=%d, Al=%d")
+JMESSAGE(JTRC_TFILE_CLOSE, "Closed temporary file %s")
+JMESSAGE(JTRC_TFILE_OPEN, "Opened temporary file %s")
+JMESSAGE(JTRC_THUMB_JPEG,
+	 "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
+JMESSAGE(JTRC_THUMB_PALETTE,
+	 "JFIF extension marker: palette thumbnail image, length %u")
+JMESSAGE(JTRC_THUMB_RGB,
+	 "JFIF extension marker: RGB thumbnail image, length %u")
+JMESSAGE(JTRC_UNKNOWN_IDS,
+	 "Unrecognized component IDs %d %d %d, assuming YCbCr")
+JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u")
+JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u")
+JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d")
+JMESSAGE(JWRN_BOGUS_PROGRESSION,
+	 "Inconsistent progression sequence for component %d coefficient %d")
+JMESSAGE(JWRN_EXTRANEOUS_DATA,
+	 "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
+JMESSAGE(JWRN_HIT_MARKER, "Corrupt JPEG data: premature end of data segment")
+JMESSAGE(JWRN_HUFF_BAD_CODE, "Corrupt JPEG data: bad Huffman code")
+JMESSAGE(JWRN_JFIF_MAJOR, "Warning: unknown JFIF revision number %d.%02d")
+JMESSAGE(JWRN_JPEG_EOF, "Premature end of JPEG file")
+JMESSAGE(JWRN_MUST_RESYNC,
+	 "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
+JMESSAGE(JWRN_NOT_SEQUENTIAL, "Invalid SOS parameters for sequential JPEG")
+JMESSAGE(JWRN_TOO_MUCH_DATA, "Application transferred too many scanlines")
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTMSGCODE
+} J_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE
+
+
+#ifndef JERROR_H
+#define JERROR_H
+
+/* Macros to simplify using the error and trace message stuff */
+/* The first parameter is either type of cinfo pointer */
+
+/* Fatal errors (print message and exit) */
+#define ERREXIT(cinfo,code)  \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT1(cinfo,code,p1)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT2(cinfo,code,p1,p2)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT3(cinfo,code,p1,p2,p3)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (cinfo)->err->msg_parm.i[2] = (p3), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXIT4(cinfo,code,p1,p2,p3,p4)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (cinfo)->err->msg_parm.i[2] = (p3), \
+   (cinfo)->err->msg_parm.i[3] = (p4), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+#define ERREXITS(cinfo,code,str)  \
+  ((cinfo)->err->msg_code = (code), \
+   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
+   (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
+
+#define MAKESTMT(stuff)		do { stuff } while (0)
+
+/* Nonfatal errors (we can keep going, but the data is probably corrupt) */
+#define WARNMS(cinfo,code)  \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1))
+#define WARNMS1(cinfo,code,p1)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1))
+#define WARNMS2(cinfo,code,p1,p2)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), -1))
+
+/* Informational/debugging messages */
+#define TRACEMS(cinfo,lvl,code)  \
+  ((cinfo)->err->msg_code = (code), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+#define TRACEMS1(cinfo,lvl,code,p1)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+#define TRACEMS2(cinfo,lvl,code,p1,p2)  \
+  ((cinfo)->err->msg_code = (code), \
+   (cinfo)->err->msg_parm.i[0] = (p1), \
+   (cinfo)->err->msg_parm.i[1] = (p2), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+#define TRACEMS3(cinfo,lvl,code,p1,p2,p3)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMS4(cinfo,lvl,code,p1,p2,p3,p4)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMS5(cinfo,lvl,code,p1,p2,p3,p4,p5)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+	   _mp[4] = (p5); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMS8(cinfo,lvl,code,p1,p2,p3,p4,p5,p6,p7,p8)  \
+  MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
+	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+	   _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \
+	   (cinfo)->err->msg_code = (code); \
+	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+#define TRACEMSS(cinfo,lvl,code,str)  \
+  ((cinfo)->err->msg_code = (code), \
+   strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
+   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
+
+#endif /* JERROR_H */
diff --git a/jpeg/jfdctflt.c b/jpeg/jfdctflt.c
new file mode 100644
index 000000000..79d7a0078
--- /dev/null
+++ b/jpeg/jfdctflt.c
@@ -0,0 +1,168 @@
+/*
+ * jfdctflt.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a floating-point implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * This implementation should be more accurate than either of the integer
+ * DCT implementations.  However, it may not give the same results on all
+ * machines because of differences in roundoff behavior.  Speed will depend
+ * on the hardware's floating point capacity.
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with a fixed-point
+ * implementation, accuracy is lost due to imprecise representation of the
+ * scaled quantization values.  However, that problem does not arise if
+ * we use floating point arithmetic.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_float (FAST_FLOAT * data)
+{
+  FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
+  FAST_FLOAT z1, z2, z3, z4, z5, z11, z13;
+  FAST_FLOAT *dataptr;
+  int ctr;
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[4] = tmp10 - tmp11;
+    
+    z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
+    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = (tmp10 - tmp12) * ((FAST_FLOAT) 0.382683433); /* c6 */
+    z2 = ((FAST_FLOAT) 0.541196100) * tmp10 + z5; /* c2-c6 */
+    z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
+    z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[3] = z13 - z2;
+    dataptr[1] = z11 + z4;
+    dataptr[7] = z11 - z4;
+
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[DCTSIZE*4] = tmp10 - tmp11;
+    
+    z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
+    dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
+    dataptr[DCTSIZE*6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = (tmp10 - tmp12) * ((FAST_FLOAT) 0.382683433); /* c6 */
+    z2 = ((FAST_FLOAT) 0.541196100) * tmp10 + z5; /* c2-c6 */
+    z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
+    z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
+    dataptr[DCTSIZE*3] = z13 - z2;
+    dataptr[DCTSIZE*1] = z11 + z4;
+    dataptr[DCTSIZE*7] = z11 - z4;
+
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
diff --git a/jpeg/jfdctfst.c b/jpeg/jfdctfst.c
new file mode 100644
index 000000000..ccb378a3b
--- /dev/null
+++ b/jpeg/jfdctfst.c
@@ -0,0 +1,224 @@
+/*
+ * jfdctfst.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_IFAST_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jfdctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * Again to save a few shifts, the intermediate results between pass 1 and
+ * pass 2 are not upscaled, but are represented only to integral precision.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#define CONST_BITS  8
+
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_0_382683433  ((INT32)   98)		/* FIX(0.382683433) */
+#define FIX_0_541196100  ((INT32)  139)		/* FIX(0.541196100) */
+#define FIX_0_707106781  ((INT32)  181)		/* FIX(0.707106781) */
+#define FIX_1_306562965  ((INT32)  334)		/* FIX(1.306562965) */
+#else
+#define FIX_0_382683433  FIX(0.382683433)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_707106781  FIX(0.707106781)
+#define FIX_1_306562965  FIX(1.306562965)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x,n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var,const)  ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_ifast (DCTELEM * data)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z1, z2, z3, z4, z5, z11, z13;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[4] = tmp10 - tmp11;
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[3] = z13 - z2;
+    dataptr[1] = z11 + z4;
+    dataptr[7] = z11 - z4;
+
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part */
+    
+    tmp10 = tmp0 + tmp3;	/* phase 2 */
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
+    dataptr[DCTSIZE*4] = tmp10 - tmp11;
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
+    dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
+    dataptr[DCTSIZE*6] = tmp13 - z1;
+    
+    /* Odd part */
+
+    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp11 = tmp5 + tmp6;
+    tmp12 = tmp6 + tmp7;
+
+    /* The rotator is modified from fig 4-8 to avoid extra negations. */
+    z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
+    z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
+    z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
+    z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
+
+    z11 = tmp7 + z3;		/* phase 5 */
+    z13 = tmp7 - z3;
+
+    dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
+    dataptr[DCTSIZE*3] = z13 - z2;
+    dataptr[DCTSIZE*1] = z11 + z4;
+    dataptr[DCTSIZE*7] = z11 - z4;
+
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_IFAST_SUPPORTED */
diff --git a/jpeg/jfdctint.c b/jpeg/jfdctint.c
new file mode 100644
index 000000000..0a78b64ae
--- /dev/null
+++ b/jpeg/jfdctint.c
@@ -0,0 +1,283 @@
+/*
+ * jfdctint.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a slow-but-accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform).
+ *
+ * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
+ * on each column.  Direct algorithms are also available, but they are
+ * much more complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_ISLOW_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D DCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true DCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D DCT,
+ * because the y0 and y4 outputs need not be divided by sqrt(N).
+ * In the IJG code, this factor of 8 is removed by the quantization step
+ * (in jcdctmgr.c), NOT in this module.
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (For 12-bit sample data, the intermediate
+ * array is INT32 anyway.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
+#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
+#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
+#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
+#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
+#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
+#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
+#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
+#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
+#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
+#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
+#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#else
+#define FIX_0_298631336  FIX(0.298631336)
+#define FIX_0_390180644  FIX(0.390180644)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_175875602  FIX(1.175875602)
+#define FIX_1_501321110  FIX(1.501321110)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_1_961570560  FIX(1.961570560)
+#define FIX_2_053119869  FIX(2.053119869)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_072711026  FIX(3.072711026)
+#endif
+
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
+#else
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/*
+ * Perform the forward DCT on one block of samples.
+ */
+
+GLOBAL(void)
+jpeg_fdct_islow (DCTELEM * data)
+{
+  INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  INT32 tmp10, tmp11, tmp12, tmp13;
+  INT32 z1, z2, z3, z4, z5;
+  DCTELEM *dataptr;
+  int ctr;
+  SHIFT_TEMPS
+
+  /* Pass 1: process rows. */
+  /* Note results are scaled up by sqrt(8) compared to a true DCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[0] + dataptr[7];
+    tmp7 = dataptr[0] - dataptr[7];
+    tmp1 = dataptr[1] + dataptr[6];
+    tmp6 = dataptr[1] - dataptr[6];
+    tmp2 = dataptr[2] + dataptr[5];
+    tmp5 = dataptr[2] - dataptr[5];
+    tmp3 = dataptr[3] + dataptr[4];
+    tmp4 = dataptr[3] - dataptr[4];
+    
+    /* Even part per LL&M figure 1 --- note that published figure is faulty;
+     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
+     */
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
+    dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
+    dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
+				   CONST_BITS-PASS1_BITS);
+    dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
+				   CONST_BITS-PASS1_BITS);
+    
+    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
+     * cK represents cos(K*pi/16).
+     * i0..i3 in the paper are tmp4..tmp7 here.
+     */
+    
+    z1 = tmp4 + tmp7;
+    z2 = tmp5 + tmp6;
+    z3 = tmp4 + tmp6;
+    z4 = tmp5 + tmp7;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
+    dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
+    dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
+    dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
+    
+    dataptr += DCTSIZE;		/* advance pointer to next row */
+  }
+
+  /* Pass 2: process columns.
+   * We remove the PASS1_BITS scaling, but leave the results scaled up
+   * by an overall factor of 8.
+   */
+
+  dataptr = data;
+  for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
+    tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
+    tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
+    tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
+    tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
+    tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
+    tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
+    tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
+    tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
+    
+    /* Even part per LL&M figure 1 --- note that published figure is faulty;
+     * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
+     */
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
+    dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
+    
+    z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
+    dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
+					   CONST_BITS+PASS1_BITS);
+    
+    /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
+     * cK represents cos(K*pi/16).
+     * i0..i3 in the paper are tmp4..tmp7 here.
+     */
+    
+    z1 = tmp4 + tmp7;
+    z2 = tmp5 + tmp6;
+    z3 = tmp4 + tmp6;
+    z4 = tmp5 + tmp7;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
+					   CONST_BITS+PASS1_BITS);
+    dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
+					   CONST_BITS+PASS1_BITS);
+    
+    dataptr++;			/* advance pointer to next column */
+  }
+}
+
+#endif /* DCT_ISLOW_SUPPORTED */
diff --git a/jpeg/jidctflt.c b/jpeg/jidctflt.c
new file mode 100644
index 000000000..0188ce3df
--- /dev/null
+++ b/jpeg/jidctflt.c
@@ -0,0 +1,242 @@
+/*
+ * jidctflt.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a floating-point implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * This implementation should be more accurate than either of the integer
+ * IDCT implementations.  However, it may not give the same results on all
+ * machines because of differences in roundoff behavior.  Speed will depend
+ * on the hardware's floating point capacity.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with a fixed-point
+ * implementation, accuracy is lost due to imprecise representation of the
+ * scaled quantization values.  However, that problem does not arise if
+ * we use floating point arithmetic.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_FLOAT_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce a float result.
+ */
+
+#define DEQUANTIZE(coef,quantval)  (((FAST_FLOAT) (coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
+  FAST_FLOAT z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  FLOAT_MULT_TYPE * quantptr;
+  FAST_FLOAT * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  FAST_FLOAT workspace[DCTSIZE2]; /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (FLOAT_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+    
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+	inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero */
+      FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      wsptr[DCTSIZE*4] = dcval;
+      wsptr[DCTSIZE*5] = dcval;
+      wsptr[DCTSIZE*6] = dcval;
+      wsptr[DCTSIZE*7] = dcval;
+      
+      inptr++;			/* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+    
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp11 = tmp0 - tmp2;
+
+    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT) 1.414213562) - tmp13; /* 2*c4 */
+
+    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+    
+    /* Odd part */
+
+    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    z13 = tmp6 + tmp5;		/* phase 6 */
+    z10 = tmp6 - tmp5;
+    z11 = tmp4 + tmp7;
+    z12 = tmp4 - tmp7;
+
+    tmp7 = z11 + z13;		/* phase 5 */
+    tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); /* 2*c4 */
+
+    z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
+    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
+    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    wsptr[DCTSIZE*0] = tmp0 + tmp7;
+    wsptr[DCTSIZE*7] = tmp0 - tmp7;
+    wsptr[DCTSIZE*1] = tmp1 + tmp6;
+    wsptr[DCTSIZE*6] = tmp1 - tmp6;
+    wsptr[DCTSIZE*2] = tmp2 + tmp5;
+    wsptr[DCTSIZE*5] = tmp2 - tmp5;
+    wsptr[DCTSIZE*4] = tmp3 + tmp4;
+    wsptr[DCTSIZE*3] = tmp3 - tmp4;
+
+    inptr++;			/* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+  
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * And testing floats for zero is relatively expensive, so we don't bother.
+     */
+    
+    /* Even part */
+
+    tmp10 = wsptr[0] + wsptr[4];
+    tmp11 = wsptr[0] - wsptr[4];
+
+    tmp13 = wsptr[2] + wsptr[6];
+    tmp12 = (wsptr[2] - wsptr[6]) * ((FAST_FLOAT) 1.414213562) - tmp13;
+
+    tmp0 = tmp10 + tmp13;
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z13 = wsptr[5] + wsptr[3];
+    z10 = wsptr[5] - wsptr[3];
+    z11 = wsptr[1] + wsptr[7];
+    z12 = wsptr[1] - wsptr[7];
+
+    tmp7 = z11 + z13;
+    tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562);
+
+    z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
+    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
+    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    /* Final output stage: scale down by a factor of 8 and range-limit */
+
+    outptr[0] = range_limit[(int) DESCALE((INT32) (tmp0 + tmp7), 3)
+			    & RANGE_MASK];
+    outptr[7] = range_limit[(int) DESCALE((INT32) (tmp0 - tmp7), 3)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE((INT32) (tmp1 + tmp6), 3)
+			    & RANGE_MASK];
+    outptr[6] = range_limit[(int) DESCALE((INT32) (tmp1 - tmp6), 3)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[(int) DESCALE((INT32) (tmp2 + tmp5), 3)
+			    & RANGE_MASK];
+    outptr[5] = range_limit[(int) DESCALE((INT32) (tmp2 - tmp5), 3)
+			    & RANGE_MASK];
+    outptr[4] = range_limit[(int) DESCALE((INT32) (tmp3 + tmp4), 3)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[(int) DESCALE((INT32) (tmp3 - tmp4), 3)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_FLOAT_SUPPORTED */
diff --git a/jpeg/jidctfst.c b/jpeg/jidctfst.c
new file mode 100644
index 000000000..dba4216fb
--- /dev/null
+++ b/jpeg/jidctfst.c
@@ -0,0 +1,368 @@
+/*
+ * jidctfst.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a fast, not so accurate integer implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on Arai, Agui, and Nakajima's algorithm for
+ * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
+ * Japanese, but the algorithm is described in the Pennebaker & Mitchell
+ * JPEG textbook (see REFERENCES section in file README).  The following code
+ * is based directly on figure 4-8 in P&M.
+ * While an 8-point DCT cannot be done in less than 11 multiplies, it is
+ * possible to arrange the computation so that many of the multiplies are
+ * simple scalings of the final outputs.  These multiplies can then be
+ * folded into the multiplications or divisions by the JPEG quantization
+ * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
+ * to be done in the DCT itself.
+ * The primary disadvantage of this method is that with fixed-point math,
+ * accuracy is lost due to imprecise representation of the scaled
+ * quantization values.  The smaller the quantization table entry, the less
+ * precise the scaled value, so this implementation does worse with high-
+ * quality-setting files than with low-quality ones.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_IFAST_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling decisions are generally the same as in the LL&M algorithm;
+ * see jidctint.c for more details.  However, we choose to descale
+ * (right shift) multiplication products as soon as they are formed,
+ * rather than carrying additional fractional bits into subsequent additions.
+ * This compromises accuracy slightly, but it lets us save a few shifts.
+ * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
+ * everywhere except in the multiplications proper; this saves a good deal
+ * of work on 16-bit-int machines.
+ *
+ * The dequantized coefficients are not integers because the AA&N scaling
+ * factors have been incorporated.  We represent them scaled up by PASS1_BITS,
+ * so that the first and second IDCT rounds have the same input scaling.
+ * For 8-bit JSAMPLEs, we choose IFAST_SCALE_BITS = PASS1_BITS so as to
+ * avoid a descaling shift; this compromises accuracy rather drastically
+ * for small quantization table entries, but it saves a lot of shifts.
+ * For 12-bit JSAMPLEs, there's no hope of using 16x16 multiplies anyway,
+ * so we use a much larger scaling factor to preserve accuracy.
+ *
+ * A final compromise is to represent the multiplicative constants to only
+ * 8 fractional bits, rather than 13.  This saves some shifting work on some
+ * machines, and may also reduce the cost of multiplication (since there
+ * are fewer one-bits in the constants).
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  8
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  8
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 8
+#define FIX_1_082392200  ((INT32)  277)		/* FIX(1.082392200) */
+#define FIX_1_414213562  ((INT32)  362)		/* FIX(1.414213562) */
+#define FIX_1_847759065  ((INT32)  473)		/* FIX(1.847759065) */
+#define FIX_2_613125930  ((INT32)  669)		/* FIX(2.613125930) */
+#else
+#define FIX_1_082392200  FIX(1.082392200)
+#define FIX_1_414213562  FIX(1.414213562)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_2_613125930  FIX(2.613125930)
+#endif
+
+
+/* We can gain a little more speed, with a further compromise in accuracy,
+ * by omitting the addition in a descaling shift.  This yields an incorrectly
+ * rounded result half the time...
+ */
+
+#ifndef USE_ACCURATE_ROUNDING
+#undef DESCALE
+#define DESCALE(x,n)  RIGHT_SHIFT(x, n)
+#endif
+
+
+/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+ * descale to yield a DCTELEM result.
+ */
+
+#define MULTIPLY(var,const)  ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce a DCTELEM result.  For 8-bit data a 16x16->16
+ * multiplication will do.  For 12-bit data, the multiplier table is
+ * declared INT32, so a 32-bit multiply will be used.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define DEQUANTIZE(coef,quantval)  (((IFAST_MULT_TYPE) (coef)) * (quantval))
+#else
+#define DEQUANTIZE(coef,quantval)  \
+	DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS)
+#endif
+
+
+/* Like DESCALE, but applies to a DCTELEM and produces an int.
+ * We assume that int right shift is unsigned if INT32 right shift is.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define ISHIFT_TEMPS	DCTELEM ishift_temp;
+#if BITS_IN_JSAMPLE == 8
+#define DCTELEMBITS  16		/* DCTELEM may be 16 or 32 bits */
+#else
+#define DCTELEMBITS  32		/* DCTELEM must be 32 bits */
+#endif
+#define IRIGHT_SHIFT(x,shft)  \
+    ((ishift_temp = (x)) < 0 ? \
+     (ishift_temp >> (shft)) | ((~((DCTELEM) 0)) << (DCTELEMBITS-(shft))) : \
+     (ishift_temp >> (shft)))
+#else
+#define ISHIFT_TEMPS
+#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#endif
+
+#ifdef USE_ACCURATE_ROUNDING
+#define IDESCALE(x,n)  ((int) IRIGHT_SHIFT((x) + (1 << ((n)-1)), n))
+#else
+#define IDESCALE(x,n)  ((int) IRIGHT_SHIFT(x, n))
+#endif
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  DCTELEM tmp10, tmp11, tmp12, tmp13;
+  DCTELEM z5, z10, z11, z12, z13;
+  JCOEFPTR inptr;
+  IFAST_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE2];	/* buffers data between passes */
+  SHIFT_TEMPS			/* for DESCALE */
+  ISHIFT_TEMPS			/* for IDESCALE */
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (IFAST_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+    
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+	inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero */
+      int dcval = (int) DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      wsptr[DCTSIZE*4] = dcval;
+      wsptr[DCTSIZE*5] = dcval;
+      wsptr[DCTSIZE*6] = dcval;
+      wsptr[DCTSIZE*7] = dcval;
+      
+      inptr++;			/* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+    
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp11 = tmp0 - tmp2;
+
+    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp12 = MULTIPLY(tmp1 - tmp3, FIX_1_414213562) - tmp13; /* 2*c4 */
+
+    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+    
+    /* Odd part */
+
+    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    z13 = tmp6 + tmp5;		/* phase 6 */
+    z10 = tmp6 - tmp5;
+    z11 = tmp4 + tmp7;
+    z12 = tmp4 - tmp7;
+
+    tmp7 = z11 + z13;		/* phase 5 */
+    tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
+
+    z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
+    tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
+    tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    wsptr[DCTSIZE*0] = (int) (tmp0 + tmp7);
+    wsptr[DCTSIZE*7] = (int) (tmp0 - tmp7);
+    wsptr[DCTSIZE*1] = (int) (tmp1 + tmp6);
+    wsptr[DCTSIZE*6] = (int) (tmp1 - tmp6);
+    wsptr[DCTSIZE*2] = (int) (tmp2 + tmp5);
+    wsptr[DCTSIZE*5] = (int) (tmp2 - tmp5);
+    wsptr[DCTSIZE*4] = (int) (tmp3 + tmp4);
+    wsptr[DCTSIZE*3] = (int) (tmp3 - tmp4);
+
+    inptr++;			/* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+  
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+    
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
+	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[IDESCALE(wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      outptr[2] = dcval;
+      outptr[3] = dcval;
+      outptr[4] = dcval;
+      outptr[5] = dcval;
+      outptr[6] = dcval;
+      outptr[7] = dcval;
+
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part */
+
+    tmp10 = ((DCTELEM) wsptr[0] + (DCTELEM) wsptr[4]);
+    tmp11 = ((DCTELEM) wsptr[0] - (DCTELEM) wsptr[4]);
+
+    tmp13 = ((DCTELEM) wsptr[2] + (DCTELEM) wsptr[6]);
+    tmp12 = MULTIPLY((DCTELEM) wsptr[2] - (DCTELEM) wsptr[6], FIX_1_414213562)
+	    - tmp13;
+
+    tmp0 = tmp10 + tmp13;
+    tmp3 = tmp10 - tmp13;
+    tmp1 = tmp11 + tmp12;
+    tmp2 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z13 = (DCTELEM) wsptr[5] + (DCTELEM) wsptr[3];
+    z10 = (DCTELEM) wsptr[5] - (DCTELEM) wsptr[3];
+    z11 = (DCTELEM) wsptr[1] + (DCTELEM) wsptr[7];
+    z12 = (DCTELEM) wsptr[1] - (DCTELEM) wsptr[7];
+
+    tmp7 = z11 + z13;		/* phase 5 */
+    tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
+
+    z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
+    tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
+    tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
+
+    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp5 = tmp11 - tmp6;
+    tmp4 = tmp10 + tmp5;
+
+    /* Final output stage: scale down by a factor of 8 and range-limit */
+
+    outptr[0] = range_limit[IDESCALE(tmp0 + tmp7, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[7] = range_limit[IDESCALE(tmp0 - tmp7, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[IDESCALE(tmp1 + tmp6, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[6] = range_limit[IDESCALE(tmp1 - tmp6, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[IDESCALE(tmp2 + tmp5, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[5] = range_limit[IDESCALE(tmp2 - tmp5, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[4] = range_limit[IDESCALE(tmp3 + tmp4, PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[IDESCALE(tmp3 - tmp4, PASS1_BITS+3)
+			    & RANGE_MASK];
+
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_IFAST_SUPPORTED */
diff --git a/jpeg/jidctint.c b/jpeg/jidctint.c
new file mode 100644
index 000000000..a72b3207c
--- /dev/null
+++ b/jpeg/jidctint.c
@@ -0,0 +1,389 @@
+/*
+ * jidctint.c
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains a slow-but-accurate integer implementation of the
+ * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
+ * must also perform dequantization of the input coefficients.
+ *
+ * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
+ * on each row (or vice versa, but it's more convenient to emit a row at
+ * a time).  Direct algorithms are also available, but they are much more
+ * complex and seem not to be any faster when reduced to code.
+ *
+ * This implementation is based on an algorithm described in
+ *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
+ *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
+ *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
+ * The primary algorithm described there uses 11 multiplies and 29 adds.
+ * We use their alternate method with 12 multiplies and 32 adds.
+ * The advantage of this method is that no data path contains more than one
+ * multiplication; this allows a very simple and accurate implementation in
+ * scaled fixed-point arithmetic, with a minimal number of shifts.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef DCT_ISLOW_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * The poop on this scaling stuff is as follows:
+ *
+ * Each 1-D IDCT step produces outputs which are a factor of sqrt(N)
+ * larger than the true IDCT outputs.  The final outputs are therefore
+ * a factor of N larger than desired; since N=8 this can be cured by
+ * a simple right shift at the end of the algorithm.  The advantage of
+ * this arrangement is that we save two multiplications per 1-D IDCT,
+ * because the y0 and y4 inputs need not be divided by sqrt(N).
+ *
+ * We have to do addition and subtraction of the integer inputs, which
+ * is no problem, and multiplication by fractional constants, which is
+ * a problem to do in integer arithmetic.  We multiply all the constants
+ * by CONST_SCALE and convert them to integer constants (thus retaining
+ * CONST_BITS bits of precision in the constants).  After doing a
+ * multiplication we have to divide the product by CONST_SCALE, with proper
+ * rounding, to produce the correct output.  This division can be done
+ * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
+ * as long as possible so that partial sums can be added together with
+ * full fractional precision.
+ *
+ * The outputs of the first pass are scaled up by PASS1_BITS bits so that
+ * they are represented to better-than-integral precision.  These outputs
+ * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
+ * with the recommended scaling.  (To scale up 12-bit sample data further, an
+ * intermediate INT32 array would be needed.)
+ *
+ * To avoid overflow of the 32-bit intermediate results in pass 2, we must
+ * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
+ * shows that the values given below are the most effective.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
+#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
+#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
+#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
+#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
+#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
+#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
+#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
+#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
+#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
+#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
+#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#else
+#define FIX_0_298631336  FIX(0.298631336)
+#define FIX_0_390180644  FIX(0.390180644)
+#define FIX_0_541196100  FIX(0.541196100)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_175875602  FIX(1.175875602)
+#define FIX_1_501321110  FIX(1.501321110)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_1_961570560  FIX(1.961570560)
+#define FIX_2_053119869  FIX(2.053119869)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_072711026  FIX(3.072711026)
+#endif
+
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
+#else
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef,quantval)  (((ISLOW_MULT_TYPE) (coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ */
+
+GLOBAL(void)
+jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  INT32 tmp0, tmp1, tmp2, tmp3;
+  INT32 tmp10, tmp11, tmp12, tmp13;
+  INT32 z1, z2, z3, z4, z5;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE2];	/* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+  /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
+  /* furthermore, we scale the results by 2**PASS1_BITS. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; ctr--) {
+    /* Due to quantization, we will usually find that many of the input
+     * coefficients are zero, especially the AC terms.  We can exploit this
+     * by short-circuiting the IDCT calculation for any column in which all
+     * the AC terms are zero.  In that case each output is equal to the
+     * DC coefficient (with scale factor as needed).
+     * With typical images and quantization tables, half or more of the
+     * column DCT calculations can be simplified this way.
+     */
+    
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+	inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero */
+      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      wsptr[DCTSIZE*4] = dcval;
+      wsptr[DCTSIZE*5] = dcval;
+      wsptr[DCTSIZE*6] = dcval;
+      wsptr[DCTSIZE*7] = dcval;
+      
+      inptr++;			/* advance pointers to next column */
+      quantptr++;
+      wsptr++;
+      continue;
+    }
+    
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+    
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    
+    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
+    tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
+    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
+    
+    z2 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+
+    tmp0 = (z2 + z3) << CONST_BITS;
+    tmp1 = (z2 - z3) << CONST_BITS;
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+    
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    
+    z1 = tmp0 + tmp3;
+    z2 = tmp1 + tmp2;
+    z3 = tmp0 + tmp2;
+    z4 = tmp1 + tmp3;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    tmp0 += z1 + z3;
+    tmp1 += z2 + z4;
+    tmp2 += z2 + z3;
+    tmp3 += z1 + z4;
+    
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+    
+    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
+    
+    inptr++;			/* advance pointers to next column */
+    quantptr++;
+    wsptr++;
+  }
+  
+  /* Pass 2: process rows from work array, store into output array. */
+  /* Note that we must descale the results by a factor of 8 == 2**3, */
+  /* and also undo the PASS1_BITS scaling. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < DCTSIZE; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* Rows of zeroes can be exploited in the same way as we did with columns.
+     * However, the column calculation has created many nonzero AC terms, so
+     * the simplification applies less often (typically 5% to 10% of the time).
+     * On machines with very fast multiplication, it's possible that the
+     * test takes more time than it's worth.  In that case this section
+     * may be commented out.
+     */
+    
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
+	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      outptr[2] = dcval;
+      outptr[3] = dcval;
+      outptr[4] = dcval;
+      outptr[5] = dcval;
+      outptr[6] = dcval;
+      outptr[7] = dcval;
+
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part: reverse the even part of the forward DCT. */
+    /* The rotator is sqrt(2)*c(-6). */
+    
+    z2 = (INT32) wsptr[2];
+    z3 = (INT32) wsptr[6];
+    
+    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
+    tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
+    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
+    
+    tmp0 = ((INT32) wsptr[0] + (INT32) wsptr[4]) << CONST_BITS;
+    tmp1 = ((INT32) wsptr[0] - (INT32) wsptr[4]) << CONST_BITS;
+    
+    tmp10 = tmp0 + tmp3;
+    tmp13 = tmp0 - tmp3;
+    tmp11 = tmp1 + tmp2;
+    tmp12 = tmp1 - tmp2;
+    
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+    
+    tmp0 = (INT32) wsptr[7];
+    tmp1 = (INT32) wsptr[5];
+    tmp2 = (INT32) wsptr[3];
+    tmp3 = (INT32) wsptr[1];
+    
+    z1 = tmp0 + tmp3;
+    z2 = tmp1 + tmp2;
+    z3 = tmp0 + tmp2;
+    z4 = tmp1 + tmp3;
+    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
+    
+    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
+    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
+    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
+    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
+    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
+    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
+    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
+    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
+    
+    z3 += z5;
+    z4 += z5;
+    
+    tmp0 += z1 + z3;
+    tmp1 += z2 + z4;
+    tmp2 += z2 + z3;
+    tmp3 += z1 + z4;
+    
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+    
+    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp3,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[7] = range_limit[(int) DESCALE(tmp10 - tmp3,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE(tmp11 + tmp2,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[6] = range_limit[(int) DESCALE(tmp11 - tmp2,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[(int) DESCALE(tmp12 + tmp1,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[5] = range_limit[(int) DESCALE(tmp12 - tmp1,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[(int) DESCALE(tmp13 + tmp0,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    outptr[4] = range_limit[(int) DESCALE(tmp13 - tmp0,
+					  CONST_BITS+PASS1_BITS+3)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+#endif /* DCT_ISLOW_SUPPORTED */
diff --git a/jpeg/jidctred.c b/jpeg/jidctred.c
new file mode 100644
index 000000000..421f3c7ca
--- /dev/null
+++ b/jpeg/jidctred.c
@@ -0,0 +1,398 @@
+/*
+ * jidctred.c
+ *
+ * Copyright (C) 1994-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains inverse-DCT routines that produce reduced-size output:
+ * either 4x4, 2x2, or 1x1 pixels from an 8x8 DCT block.
+ *
+ * The implementation is based on the Loeffler, Ligtenberg and Moschytz (LL&M)
+ * algorithm used in jidctint.c.  We simply replace each 8-to-8 1-D IDCT step
+ * with an 8-to-4 step that produces the four averages of two adjacent outputs
+ * (or an 8-to-2 step producing two averages of four outputs, for 2x2 output).
+ * These steps were derived by computing the corresponding values at the end
+ * of the normal LL&M code, then simplifying as much as possible.
+ *
+ * 1x1 is trivial: just take the DC coefficient divided by 8.
+ *
+ * See jidctint.c for additional comments.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jdct.h"		/* Private declarations for DCT subsystem */
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+
+/*
+ * This module is specialized to the case DCTSIZE = 8.
+ */
+
+#if DCTSIZE != 8
+  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+#endif
+
+
+/* Scaling is the same as in jidctint.c. */
+
+#if BITS_IN_JSAMPLE == 8
+#define CONST_BITS  13
+#define PASS1_BITS  2
+#else
+#define CONST_BITS  13
+#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#endif
+
+/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
+ * causing a lot of useless floating-point operations at run time.
+ * To get around this we use the following pre-calculated constants.
+ * If you change CONST_BITS you may want to add appropriate values.
+ * (With a reasonable C compiler, you can just rely on the FIX() macro...)
+ */
+
+#if CONST_BITS == 13
+#define FIX_0_211164243  ((INT32)  1730)	/* FIX(0.211164243) */
+#define FIX_0_509795579  ((INT32)  4176)	/* FIX(0.509795579) */
+#define FIX_0_601344887  ((INT32)  4926)	/* FIX(0.601344887) */
+#define FIX_0_720959822  ((INT32)  5906)	/* FIX(0.720959822) */
+#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
+#define FIX_0_850430095  ((INT32)  6967)	/* FIX(0.850430095) */
+#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
+#define FIX_1_061594337  ((INT32)  8697)	/* FIX(1.061594337) */
+#define FIX_1_272758580  ((INT32)  10426)	/* FIX(1.272758580) */
+#define FIX_1_451774981  ((INT32)  11893)	/* FIX(1.451774981) */
+#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
+#define FIX_2_172734803  ((INT32)  17799)	/* FIX(2.172734803) */
+#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
+#define FIX_3_624509785  ((INT32)  29692)	/* FIX(3.624509785) */
+#else
+#define FIX_0_211164243  FIX(0.211164243)
+#define FIX_0_509795579  FIX(0.509795579)
+#define FIX_0_601344887  FIX(0.601344887)
+#define FIX_0_720959822  FIX(0.720959822)
+#define FIX_0_765366865  FIX(0.765366865)
+#define FIX_0_850430095  FIX(0.850430095)
+#define FIX_0_899976223  FIX(0.899976223)
+#define FIX_1_061594337  FIX(1.061594337)
+#define FIX_1_272758580  FIX(1.272758580)
+#define FIX_1_451774981  FIX(1.451774981)
+#define FIX_1_847759065  FIX(1.847759065)
+#define FIX_2_172734803  FIX(2.172734803)
+#define FIX_2_562915447  FIX(2.562915447)
+#define FIX_3_624509785  FIX(3.624509785)
+#endif
+
+
+/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+ * For 8-bit samples with the recommended scaling, all the variable
+ * and constant values involved are no more than 16 bits wide, so a
+ * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
+ * For 12-bit samples, a full 32-bit multiplication will be needed.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
+#else
+#define MULTIPLY(var,const)  ((var) * (const))
+#endif
+
+
+/* Dequantize a coefficient by multiplying it by the multiplier-table
+ * entry; produce an int result.  In this module, both inputs and result
+ * are 16 bits or less, so either int or short multiply will work.
+ */
+
+#define DEQUANTIZE(coef,quantval)  (((ISLOW_MULT_TYPE) (coef)) * (quantval))
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 4x4 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JCOEFPTR coef_block,
+	       JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  INT32 tmp0, tmp2, tmp10, tmp12;
+  INT32 z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE*4];	/* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) {
+    /* Don't bother to process column 4, because second pass won't use it */
+    if (ctr == DCTSIZE-4)
+      continue;
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
+	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 &&
+	inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero; we need not examine term 4 for 4x4 output */
+      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      wsptr[DCTSIZE*2] = dcval;
+      wsptr[DCTSIZE*3] = dcval;
+      
+      continue;
+    }
+    
+    /* Even part */
+    
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 <<= (CONST_BITS+1);
+    
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp2 = MULTIPLY(z2, FIX_1_847759065) + MULTIPLY(z3, - FIX_0_765366865);
+    
+    tmp10 = tmp0 + tmp2;
+    tmp12 = tmp0 - tmp2;
+    
+    /* Odd part */
+    
+    z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    
+    tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
+	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+    
+    tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
+	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+
+    /* Final output stage */
+    
+    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp2, CONST_BITS-PASS1_BITS+1);
+    wsptr[DCTSIZE*3] = (int) DESCALE(tmp10 - tmp2, CONST_BITS-PASS1_BITS+1);
+    wsptr[DCTSIZE*1] = (int) DESCALE(tmp12 + tmp0, CONST_BITS-PASS1_BITS+1);
+    wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 - tmp0, CONST_BITS-PASS1_BITS+1);
+  }
+  
+  /* Pass 2: process 4 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 4; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* It's not clear whether a zero row test is worthwhile here ... */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 &&
+	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      outptr[2] = dcval;
+      outptr[3] = dcval;
+      
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part */
+    
+    tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1);
+    
+    tmp2 = MULTIPLY((INT32) wsptr[2], FIX_1_847759065)
+	 + MULTIPLY((INT32) wsptr[6], - FIX_0_765366865);
+    
+    tmp10 = tmp0 + tmp2;
+    tmp12 = tmp0 - tmp2;
+    
+    /* Odd part */
+    
+    z1 = (INT32) wsptr[7];
+    z2 = (INT32) wsptr[5];
+    z3 = (INT32) wsptr[3];
+    z4 = (INT32) wsptr[1];
+    
+    tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
+	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+    
+    tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
+	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+
+    /* Final output stage */
+    
+    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp2,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    outptr[3] = range_limit[(int) DESCALE(tmp10 - tmp2,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE(tmp12 + tmp0,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    outptr[2] = range_limit[(int) DESCALE(tmp12 - tmp0,
+					  CONST_BITS+PASS1_BITS+3+1)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 2x2 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JCOEFPTR coef_block,
+	       JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  INT32 tmp0, tmp10, z1;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE * quantptr;
+  int * wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[DCTSIZE*2];	/* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = DCTSIZE; ctr > 0; inptr++, quantptr++, wsptr++, ctr--) {
+    /* Don't bother to process columns 2,4,6 */
+    if (ctr == DCTSIZE-2 || ctr == DCTSIZE-4 || ctr == DCTSIZE-6)
+      continue;
+    if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*3] == 0 &&
+	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) {
+      /* AC terms all zero; we need not examine terms 2,4,6 for 2x2 output */
+      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
+      
+      wsptr[DCTSIZE*0] = dcval;
+      wsptr[DCTSIZE*1] = dcval;
+      
+      continue;
+    }
+    
+    /* Even part */
+    
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp10 = z1 << (CONST_BITS+2);
+    
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp0 = MULTIPLY(z1, - FIX_0_720959822); /* sqrt(2) * (c7-c5+c3-c1) */
+    z1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp0 += MULTIPLY(z1, FIX_0_850430095); /* sqrt(2) * (-c1+c3+c5+c7) */
+    z1 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    tmp0 += MULTIPLY(z1, - FIX_1_272758580); /* sqrt(2) * (-c1+c3-c5-c7) */
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp0 += MULTIPLY(z1, FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
+
+    /* Final output stage */
+    
+    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp0, CONST_BITS-PASS1_BITS+2);
+    wsptr[DCTSIZE*1] = (int) DESCALE(tmp10 - tmp0, CONST_BITS-PASS1_BITS+2);
+  }
+  
+  /* Pass 2: process 2 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 2; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+    /* It's not clear whether a zero row test is worthwhile here ... */
+
+#ifndef NO_ZERO_ROW_TEST
+    if (wsptr[1] == 0 && wsptr[3] == 0 && wsptr[5] == 0 && wsptr[7] == 0) {
+      /* AC terms all zero */
+      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
+				  & RANGE_MASK];
+      
+      outptr[0] = dcval;
+      outptr[1] = dcval;
+      
+      wsptr += DCTSIZE;		/* advance pointer to next row */
+      continue;
+    }
+#endif
+    
+    /* Even part */
+    
+    tmp10 = ((INT32) wsptr[0]) << (CONST_BITS+2);
+    
+    /* Odd part */
+
+    tmp0 = MULTIPLY((INT32) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */
+	 + MULTIPLY((INT32) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */
+	 + MULTIPLY((INT32) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */
+	 + MULTIPLY((INT32) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
+
+    /* Final output stage */
+    
+    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp0,
+					  CONST_BITS+PASS1_BITS+3+2)
+			    & RANGE_MASK];
+    outptr[1] = range_limit[(int) DESCALE(tmp10 - tmp0,
+					  CONST_BITS+PASS1_BITS+3+2)
+			    & RANGE_MASK];
+    
+    wsptr += DCTSIZE;		/* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 1x1 output block.
+ */
+
+GLOBAL(void)
+jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+	       JCOEFPTR coef_block,
+	       JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  int dcval;
+  ISLOW_MULT_TYPE * quantptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  SHIFT_TEMPS
+
+  /* We hardly need an inverse DCT routine for this: just take the
+   * average pixel value, which is one-eighth of the DC coefficient.
+   */
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  dcval = DEQUANTIZE(coef_block[0], quantptr[0]);
+  dcval = (int) DESCALE((INT32) dcval, 3);
+
+  output_buf[0][output_col] = range_limit[dcval & RANGE_MASK];
+}
+
+#endif /* IDCT_SCALING_SUPPORTED */
diff --git a/jpeg/jinclude.h b/jpeg/jinclude.h
new file mode 100644
index 000000000..0a4f15146
--- /dev/null
+++ b/jpeg/jinclude.h
@@ -0,0 +1,91 @@
+/*
+ * jinclude.h
+ *
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file exists to provide a single place to fix any problems with
+ * including the wrong system include files.  (Common problems are taken
+ * care of by the standard jconfig symbols, but on really weird systems
+ * you may have to edit this file.)
+ *
+ * NOTE: this file is NOT intended to be included by applications using the
+ * JPEG library.  Most applications need only include jpeglib.h.
+ */
+
+
+/* Include auto-config file to find out which system include files we need. */
+
+#include "jconfig.h"		/* auto configuration options */
+#define JCONFIG_INCLUDED	/* so that jpeglib.h doesn't do it again */
+
+/*
+ * We need the NULL macro and size_t typedef.
+ * On an ANSI-conforming system it is sufficient to include <stddef.h>.
+ * Otherwise, we get them from <stdlib.h> or <stdio.h>; we may have to
+ * pull in <sys/types.h> as well.
+ * Note that the core JPEG library does not require <stdio.h>;
+ * only the default error handler and data source/destination modules do.
+ * But we must pull it in because of the references to FILE in jpeglib.h.
+ * You can remove those references if you want to compile without <stdio.h>.
+ */
+
+#ifdef HAVE_STDDEF_H
+#include <stddef.h>
+#endif
+
+#ifdef HAVE_STDLIB_H
+#include <stdlib.h>
+#endif
+
+#ifdef NEED_SYS_TYPES_H
+#include <sys/types.h>
+#endif
+
+#include <stdio.h>
+
+/*
+ * We need memory copying and zeroing functions, plus strncpy().
+ * ANSI and System V implementations declare these in <string.h>.
+ * BSD doesn't have the mem() functions, but it does have bcopy()/bzero().
+ * Some systems may declare memset and memcpy in <memory.h>.
+ *
+ * NOTE: we assume the size parameters to these functions are of type size_t.
+ * Change the casts in these macros if not!
+ */
+
+#ifdef NEED_BSD_STRINGS
+
+#include <strings.h>
+#define MEMZERO(target,size)	bzero((void *)(target), (size_t)(size))
+#define MEMCOPY(dest,src,size)	bcopy((const void *)(src), (void *)(dest), (size_t)(size))
+
+#else /* not BSD, assume ANSI/SysV string lib */
+
+#include <string.h>
+#define MEMZERO(target,size)	memset((void *)(target), 0, (size_t)(size))
+#define MEMCOPY(dest,src,size)	memcpy((void *)(dest), (const void *)(src), (size_t)(size))
+
+#endif
+
+/*
+ * In ANSI C, and indeed any rational implementation, size_t is also the
+ * type returned by sizeof().  However, it seems there are some irrational
+ * implementations out there, in which sizeof() returns an int even though
+ * size_t is defined as long or unsigned long.  To ensure consistent results
+ * we always use this SIZEOF() macro in place of using sizeof() directly.
+ */
+
+#define SIZEOF(object)	((size_t) sizeof(object))
+
+/*
+ * The modules that use fread() and fwrite() always invoke them through
+ * these macros.  On some systems you may need to twiddle the argument casts.
+ * CAUTION: argument order is different from underlying functions!
+ */
+
+#define JFREAD(file,buf,sizeofbuf)  \
+  ((size_t) fread((void *) (buf), (size_t) 1, (size_t) (sizeofbuf), (file)))
+#define JFWRITE(file,buf,sizeofbuf)  \
+  ((size_t) fwrite((const void *) (buf), (size_t) 1, (size_t) (sizeofbuf), (file)))
diff --git a/jpeg/jmemmgr.c b/jpeg/jmemmgr.c
new file mode 100644
index 000000000..d801b322d
--- /dev/null
+++ b/jpeg/jmemmgr.c
@@ -0,0 +1,1118 @@
+/*
+ * jmemmgr.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains the JPEG system-independent memory management
+ * routines.  This code is usable across a wide variety of machines; most
+ * of the system dependencies have been isolated in a separate file.
+ * The major functions provided here are:
+ *   * pool-based allocation and freeing of memory;
+ *   * policy decisions about how to divide available memory among the
+ *     virtual arrays;
+ *   * control logic for swapping virtual arrays between main memory and
+ *     backing storage.
+ * The separate system-dependent file provides the actual backing-storage
+ * access code, and it contains the policy decision about how much total
+ * main memory to use.
+ * This file is system-dependent in the sense that some of its functions
+ * are unnecessary in some systems.  For example, if there is enough virtual
+ * memory so that backing storage will never be used, much of the virtual
+ * array control logic could be removed.  (Of course, if you have that much
+ * memory then you shouldn't care about a little bit of unused code...)
+ */
+
+#define JPEG_INTERNALS
+#define AM_MEMORY_MANAGER	/* we define jvirt_Xarray_control structs */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef NO_GETENV
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare getenv() */
+extern char * getenv JPP((const char * name));
+#endif
+#endif
+
+
+/*
+ * Some important notes:
+ *   The allocation routines provided here must never return NULL.
+ *   They should exit to error_exit if unsuccessful.
+ *
+ *   It's not a good idea to try to merge the sarray and barray routines,
+ *   even though they are textually almost the same, because samples are
+ *   usually stored as bytes while coefficients are shorts or ints.  Thus,
+ *   in machines where byte pointers have a different representation from
+ *   word pointers, the resulting machine code could not be the same.
+ */
+
+
+/*
+ * Many machines require storage alignment: longs must start on 4-byte
+ * boundaries, doubles on 8-byte boundaries, etc.  On such machines, malloc()
+ * always returns pointers that are multiples of the worst-case alignment
+ * requirement, and we had better do so too.
+ * There isn't any really portable way to determine the worst-case alignment
+ * requirement.  This module assumes that the alignment requirement is
+ * multiples of sizeof(ALIGN_TYPE).
+ * By default, we define ALIGN_TYPE as double.  This is necessary on some
+ * workstations (where doubles really do need 8-byte alignment) and will work
+ * fine on nearly everything.  If your machine has lesser alignment needs,
+ * you can save a few bytes by making ALIGN_TYPE smaller.
+ * The only place I know of where this will NOT work is certain Macintosh
+ * 680x0 compilers that define double as a 10-byte IEEE extended float.
+ * Doing 10-byte alignment is counterproductive because longwords won't be
+ * aligned well.  Put "#define ALIGN_TYPE long" in jconfig.h if you have
+ * such a compiler.
+ */
+
+#ifndef ALIGN_TYPE		/* so can override from jconfig.h */
+#define ALIGN_TYPE  double
+#endif
+
+
+/*
+ * We allocate objects from "pools", where each pool is gotten with a single
+ * request to jpeg_get_small() or jpeg_get_large().  There is no per-object
+ * overhead within a pool, except for alignment padding.  Each pool has a
+ * header with a link to the next pool of the same class.
+ * Small and large pool headers are identical except that the latter's
+ * link pointer must be FAR on 80x86 machines.
+ * Notice that the "real" header fields are union'ed with a dummy ALIGN_TYPE
+ * field.  This forces the compiler to make SIZEOF(small_pool_hdr) a multiple
+ * of the alignment requirement of ALIGN_TYPE.
+ */
+
+typedef union small_pool_struct * small_pool_ptr;
+
+typedef union small_pool_struct {
+  struct {
+    small_pool_ptr next;	/* next in list of pools */
+    size_t bytes_used;		/* how many bytes already used within pool */
+    size_t bytes_left;		/* bytes still available in this pool */
+  } hdr;
+  ALIGN_TYPE dummy;		/* included in union to ensure alignment */
+} small_pool_hdr;
+
+typedef union large_pool_struct FAR * large_pool_ptr;
+
+typedef union large_pool_struct {
+  struct {
+    large_pool_ptr next;	/* next in list of pools */
+    size_t bytes_used;		/* how many bytes already used within pool */
+    size_t bytes_left;		/* bytes still available in this pool */
+  } hdr;
+  ALIGN_TYPE dummy;		/* included in union to ensure alignment */
+} large_pool_hdr;
+
+
+/*
+ * Here is the full definition of a memory manager object.
+ */
+
+typedef struct {
+  struct jpeg_memory_mgr pub;	/* public fields */
+
+  /* Each pool identifier (lifetime class) names a linked list of pools. */
+  small_pool_ptr small_list[JPOOL_NUMPOOLS];
+  large_pool_ptr large_list[JPOOL_NUMPOOLS];
+
+  /* Since we only have one lifetime class of virtual arrays, only one
+   * linked list is necessary (for each datatype).  Note that the virtual
+   * array control blocks being linked together are actually stored somewhere
+   * in the small-pool list.
+   */
+  jvirt_sarray_ptr virt_sarray_list;
+  jvirt_barray_ptr virt_barray_list;
+
+  /* This counts total space obtained from jpeg_get_small/large */
+  long total_space_allocated;
+
+  /* alloc_sarray and alloc_barray set this value for use by virtual
+   * array routines.
+   */
+  JDIMENSION last_rowsperchunk;	/* from most recent alloc_sarray/barray */
+} my_memory_mgr;
+
+typedef my_memory_mgr * my_mem_ptr;
+
+
+/*
+ * The control blocks for virtual arrays.
+ * Note that these blocks are allocated in the "small" pool area.
+ * System-dependent info for the associated backing store (if any) is hidden
+ * inside the backing_store_info struct.
+ */
+
+struct jvirt_sarray_control {
+  JSAMPARRAY mem_buffer;	/* => the in-memory buffer */
+  JDIMENSION rows_in_array;	/* total virtual array height */
+  JDIMENSION samplesperrow;	/* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;		/* max rows accessed by access_virt_sarray */
+  JDIMENSION rows_in_mem;	/* height of memory buffer */
+  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
+  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
+  boolean pre_zero;		/* pre-zero mode requested? */
+  boolean dirty;		/* do current buffer contents need written? */
+  boolean b_s_open;		/* is backing-store data valid? */
+  jvirt_sarray_ptr next;	/* link to next virtual sarray control block */
+  backing_store_info b_s_info;	/* System-dependent control info */
+};
+
+struct jvirt_barray_control {
+  JBLOCKARRAY mem_buffer;	/* => the in-memory buffer */
+  JDIMENSION rows_in_array;	/* total virtual array height */
+  JDIMENSION blocksperrow;	/* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;		/* max rows accessed by access_virt_barray */
+  JDIMENSION rows_in_mem;	/* height of memory buffer */
+  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
+  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
+  boolean pre_zero;		/* pre-zero mode requested? */
+  boolean dirty;		/* do current buffer contents need written? */
+  boolean b_s_open;		/* is backing-store data valid? */
+  jvirt_barray_ptr next;	/* link to next virtual barray control block */
+  backing_store_info b_s_info;	/* System-dependent control info */
+};
+
+
+#ifdef MEM_STATS		/* optional extra stuff for statistics */
+
+LOCAL(void)
+print_mem_stats (j_common_ptr cinfo, int pool_id)
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  small_pool_ptr shdr_ptr;
+  large_pool_ptr lhdr_ptr;
+
+  /* Since this is only a debugging stub, we can cheat a little by using
+   * fprintf directly rather than going through the trace message code.
+   * This is helpful because message parm array can't handle longs.
+   */
+  fprintf(stderr, "Freeing pool %d, total space = %ld\n",
+	  pool_id, mem->total_space_allocated);
+
+  for (lhdr_ptr = mem->large_list[pool_id]; lhdr_ptr != NULL;
+       lhdr_ptr = lhdr_ptr->hdr.next) {
+    fprintf(stderr, "  Large chunk used %ld\n",
+	    (long) lhdr_ptr->hdr.bytes_used);
+  }
+
+  for (shdr_ptr = mem->small_list[pool_id]; shdr_ptr != NULL;
+       shdr_ptr = shdr_ptr->hdr.next) {
+    fprintf(stderr, "  Small chunk used %ld free %ld\n",
+	    (long) shdr_ptr->hdr.bytes_used,
+	    (long) shdr_ptr->hdr.bytes_left);
+  }
+}
+
+#endif /* MEM_STATS */
+
+
+LOCAL(void)
+out_of_memory (j_common_ptr cinfo, int which)
+/* Report an out-of-memory error and stop execution */
+/* If we compiled MEM_STATS support, report alloc requests before dying */
+{
+#ifdef MEM_STATS
+  cinfo->err->trace_level = 2;	/* force self_destruct to report stats */
+#endif
+  ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, which);
+}
+
+
+/*
+ * Allocation of "small" objects.
+ *
+ * For these, we use pooled storage.  When a new pool must be created,
+ * we try to get enough space for the current request plus a "slop" factor,
+ * where the slop will be the amount of leftover space in the new pool.
+ * The speed vs. space tradeoff is largely determined by the slop values.
+ * A different slop value is provided for each pool class (lifetime),
+ * and we also distinguish the first pool of a class from later ones.
+ * NOTE: the values given work fairly well on both 16- and 32-bit-int
+ * machines, but may be too small if longs are 64 bits or more.
+ */
+
+static const size_t first_pool_slop[JPOOL_NUMPOOLS] = 
+{
+	1600,			/* first PERMANENT pool */
+	16000			/* first IMAGE pool */
+};
+
+static const size_t extra_pool_slop[JPOOL_NUMPOOLS] = 
+{
+	0,			/* additional PERMANENT pools */
+	5000			/* additional IMAGE pools */
+};
+
+#define MIN_SLOP  50		/* greater than 0 to avoid futile looping */
+
+
+METHODDEF(void *)
+alloc_small (j_common_ptr cinfo, int pool_id, size_t sizeofobject)
+/* Allocate a "small" object */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  small_pool_ptr hdr_ptr, prev_hdr_ptr;
+  char * data_ptr;
+  size_t odd_bytes, min_request, slop;
+
+  /* Check for unsatisfiable request (do now to ensure no overflow below) */
+  if (sizeofobject > (size_t) (MAX_ALLOC_CHUNK-SIZEOF(small_pool_hdr)))
+    out_of_memory(cinfo, 1);	/* request exceeds malloc's ability */
+
+  /* Round up the requested size to a multiple of SIZEOF(ALIGN_TYPE) */
+  odd_bytes = sizeofobject % SIZEOF(ALIGN_TYPE);
+  if (odd_bytes > 0)
+    sizeofobject += SIZEOF(ALIGN_TYPE) - odd_bytes;
+
+  /* See if space is available in any existing pool */
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+  prev_hdr_ptr = NULL;
+  hdr_ptr = mem->small_list[pool_id];
+  while (hdr_ptr != NULL) {
+    if (hdr_ptr->hdr.bytes_left >= sizeofobject)
+      break;			/* found pool with enough space */
+    prev_hdr_ptr = hdr_ptr;
+    hdr_ptr = hdr_ptr->hdr.next;
+  }
+
+  /* Time to make a new pool? */
+  if (hdr_ptr == NULL) {
+    /* min_request is what we need now, slop is what will be leftover */
+    min_request = sizeofobject + SIZEOF(small_pool_hdr);
+    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+      slop = first_pool_slop[pool_id];
+    else
+      slop = extra_pool_slop[pool_id];
+    /* Don't ask for more than MAX_ALLOC_CHUNK */
+    if (slop > (size_t) (MAX_ALLOC_CHUNK-min_request))
+      slop = (size_t) (MAX_ALLOC_CHUNK-min_request);
+    /* Try to get space, if fail reduce slop and try again */
+    for (;;) {
+      hdr_ptr = (small_pool_ptr) jpeg_get_small(cinfo, min_request + slop);
+      if (hdr_ptr != NULL)
+	break;
+      slop /= 2;
+      if (slop < MIN_SLOP)	/* give up when it gets real small */
+	out_of_memory(cinfo, 2); /* jpeg_get_small failed */
+    }
+    mem->total_space_allocated += min_request + slop;
+    /* Success, initialize the new pool header and add to end of list */
+    hdr_ptr->hdr.next = NULL;
+    hdr_ptr->hdr.bytes_used = 0;
+    hdr_ptr->hdr.bytes_left = sizeofobject + slop;
+    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+      mem->small_list[pool_id] = hdr_ptr;
+    else
+      prev_hdr_ptr->hdr.next = hdr_ptr;
+  }
+
+  /* OK, allocate the object from the current pool */
+  data_ptr = (char *) (hdr_ptr + 1); /* point to first data byte in pool */
+  data_ptr += hdr_ptr->hdr.bytes_used; /* point to place for object */
+  hdr_ptr->hdr.bytes_used += sizeofobject;
+  hdr_ptr->hdr.bytes_left -= sizeofobject;
+
+  return (void *) data_ptr;
+}
+
+
+/*
+ * Allocation of "large" objects.
+ *
+ * The external semantics of these are the same as "small" objects,
+ * except that FAR pointers are used on 80x86.  However the pool
+ * management heuristics are quite different.  We assume that each
+ * request is large enough that it may as well be passed directly to
+ * jpeg_get_large; the pool management just links everything together
+ * so that we can free it all on demand.
+ * Note: the major use of "large" objects is in JSAMPARRAY and JBLOCKARRAY
+ * structures.  The routines that create these structures (see below)
+ * deliberately bunch rows together to ensure a large request size.
+ */
+
+METHODDEF(void FAR *)
+alloc_large (j_common_ptr cinfo, int pool_id, size_t sizeofobject)
+/* Allocate a "large" object */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  large_pool_ptr hdr_ptr;
+  size_t odd_bytes;
+
+  /* Check for unsatisfiable request (do now to ensure no overflow below) */
+  if (sizeofobject > (size_t) (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)))
+    out_of_memory(cinfo, 3);	/* request exceeds malloc's ability */
+
+  /* Round up the requested size to a multiple of SIZEOF(ALIGN_TYPE) */
+  odd_bytes = sizeofobject % SIZEOF(ALIGN_TYPE);
+  if (odd_bytes > 0)
+    sizeofobject += SIZEOF(ALIGN_TYPE) - odd_bytes;
+
+  /* Always make a new pool */
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+  hdr_ptr = (large_pool_ptr) jpeg_get_large(cinfo, sizeofobject +
+					    SIZEOF(large_pool_hdr));
+  if (hdr_ptr == NULL)
+    out_of_memory(cinfo, 4);	/* jpeg_get_large failed */
+  mem->total_space_allocated += sizeofobject + SIZEOF(large_pool_hdr);
+
+  /* Success, initialize the new pool header and add to list */
+  hdr_ptr->hdr.next = mem->large_list[pool_id];
+  /* We maintain space counts in each pool header for statistical purposes,
+   * even though they are not needed for allocation.
+   */
+  hdr_ptr->hdr.bytes_used = sizeofobject;
+  hdr_ptr->hdr.bytes_left = 0;
+  mem->large_list[pool_id] = hdr_ptr;
+
+  return (void FAR *) (hdr_ptr + 1); /* point to first data byte in pool */
+}
+
+
+/*
+ * Creation of 2-D sample arrays.
+ * The pointers are in near heap, the samples themselves in FAR heap.
+ *
+ * To minimize allocation overhead and to allow I/O of large contiguous
+ * blocks, we allocate the sample rows in groups of as many rows as possible
+ * without exceeding MAX_ALLOC_CHUNK total bytes per allocation request.
+ * NB: the virtual array control routines, later in this file, know about
+ * this chunking of rows.  The rowsperchunk value is left in the mem manager
+ * object so that it can be saved away if this sarray is the workspace for
+ * a virtual array.
+ */
+
+METHODDEF(JSAMPARRAY)
+alloc_sarray (j_common_ptr cinfo, int pool_id,
+	      JDIMENSION samplesperrow, JDIMENSION numrows)
+/* Allocate a 2-D sample array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  JSAMPARRAY result;
+  JSAMPROW workspace;
+  JDIMENSION rowsperchunk, currow, i;
+  long ltemp;
+
+  /* Calculate max # of rows allowed in one allocation chunk */
+  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
+	  ((long) samplesperrow * SIZEOF(JSAMPLE));
+  if (ltemp <= 0)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  if (ltemp < (long) numrows)
+    rowsperchunk = (JDIMENSION) ltemp;
+  else
+    rowsperchunk = numrows;
+  mem->last_rowsperchunk = rowsperchunk;
+
+  /* Get space for row pointers (small object) */
+  result = (JSAMPARRAY) alloc_small(cinfo, pool_id,
+				    (size_t) (numrows * SIZEOF(JSAMPROW)));
+
+  /* Get the rows themselves (large objects) */
+  currow = 0;
+  while (currow < numrows) {
+    rowsperchunk = MIN(rowsperchunk, numrows - currow);
+    workspace = (JSAMPROW) alloc_large(cinfo, pool_id,
+	(size_t) ((size_t) rowsperchunk * (size_t) samplesperrow
+		  * SIZEOF(JSAMPLE)));
+    for (i = rowsperchunk; i > 0; i--) {
+      result[currow++] = workspace;
+      workspace += samplesperrow;
+    }
+  }
+
+  return result;
+}
+
+
+/*
+ * Creation of 2-D coefficient-block arrays.
+ * This is essentially the same as the code for sample arrays, above.
+ */
+
+METHODDEF(JBLOCKARRAY)
+alloc_barray (j_common_ptr cinfo, int pool_id,
+	      JDIMENSION blocksperrow, JDIMENSION numrows)
+/* Allocate a 2-D coefficient-block array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  JBLOCKARRAY result;
+  JBLOCKROW workspace;
+  JDIMENSION rowsperchunk, currow, i;
+  long ltemp;
+
+  /* Calculate max # of rows allowed in one allocation chunk */
+  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
+	  ((long) blocksperrow * SIZEOF(JBLOCK));
+  if (ltemp <= 0)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+  if (ltemp < (long) numrows)
+    rowsperchunk = (JDIMENSION) ltemp;
+  else
+    rowsperchunk = numrows;
+  mem->last_rowsperchunk = rowsperchunk;
+
+  /* Get space for row pointers (small object) */
+  result = (JBLOCKARRAY) alloc_small(cinfo, pool_id,
+				     (size_t) (numrows * SIZEOF(JBLOCKROW)));
+
+  /* Get the rows themselves (large objects) */
+  currow = 0;
+  while (currow < numrows) {
+    rowsperchunk = MIN(rowsperchunk, numrows - currow);
+    workspace = (JBLOCKROW) alloc_large(cinfo, pool_id,
+	(size_t) ((size_t) rowsperchunk * (size_t) blocksperrow
+		  * SIZEOF(JBLOCK)));
+    for (i = rowsperchunk; i > 0; i--) {
+      result[currow++] = workspace;
+      workspace += blocksperrow;
+    }
+  }
+
+  return result;
+}
+
+
+/*
+ * About virtual array management:
+ *
+ * The above "normal" array routines are only used to allocate strip buffers
+ * (as wide as the image, but just a few rows high).  Full-image-sized buffers
+ * are handled as "virtual" arrays.  The array is still accessed a strip at a
+ * time, but the memory manager must save the whole array for repeated
+ * accesses.  The intended implementation is that there is a strip buffer in
+ * memory (as high as is possible given the desired memory limit), plus a
+ * backing file that holds the rest of the array.
+ *
+ * The request_virt_array routines are told the total size of the image and
+ * the maximum number of rows that will be accessed at once.  The in-memory
+ * buffer must be at least as large as the maxaccess value.
+ *
+ * The request routines create control blocks but not the in-memory buffers.
+ * That is postponed until realize_virt_arrays is called.  At that time the
+ * total amount of space needed is known (approximately, anyway), so free
+ * memory can be divided up fairly.
+ *
+ * The access_virt_array routines are responsible for making a specific strip
+ * area accessible (after reading or writing the backing file, if necessary).
+ * Note that the access routines are told whether the caller intends to modify
+ * the accessed strip; during a read-only pass this saves having to rewrite
+ * data to disk.  The access routines are also responsible for pre-zeroing
+ * any newly accessed rows, if pre-zeroing was requested.
+ *
+ * In current usage, the access requests are usually for nonoverlapping
+ * strips; that is, successive access start_row numbers differ by exactly
+ * num_rows = maxaccess.  This means we can get good performance with simple
+ * buffer dump/reload logic, by making the in-memory buffer be a multiple
+ * of the access height; then there will never be accesses across bufferload
+ * boundaries.  The code will still work with overlapping access requests,
+ * but it doesn't handle bufferload overlaps very efficiently.
+ */
+
+
+METHODDEF(jvirt_sarray_ptr)
+request_virt_sarray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
+		     JDIMENSION samplesperrow, JDIMENSION numrows,
+		     JDIMENSION maxaccess)
+/* Request a virtual 2-D sample array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  jvirt_sarray_ptr result;
+
+  /* Only IMAGE-lifetime virtual arrays are currently supported */
+  if (pool_id != JPOOL_IMAGE)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+  /* get control block */
+  result = (jvirt_sarray_ptr) alloc_small(cinfo, pool_id,
+					  SIZEOF(struct jvirt_sarray_control));
+
+  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->rows_in_array = numrows;
+  result->samplesperrow = samplesperrow;
+  result->maxaccess = maxaccess;
+  result->pre_zero = pre_zero;
+  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->next = mem->virt_sarray_list; /* add to list of virtual arrays */
+  mem->virt_sarray_list = result;
+
+  return result;
+}
+
+
+METHODDEF(jvirt_barray_ptr)
+request_virt_barray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
+		     JDIMENSION blocksperrow, JDIMENSION numrows,
+		     JDIMENSION maxaccess)
+/* Request a virtual 2-D coefficient-block array */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  jvirt_barray_ptr result;
+
+  /* Only IMAGE-lifetime virtual arrays are currently supported */
+  if (pool_id != JPOOL_IMAGE)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+  /* get control block */
+  result = (jvirt_barray_ptr) alloc_small(cinfo, pool_id,
+					  SIZEOF(struct jvirt_barray_control));
+
+  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->rows_in_array = numrows;
+  result->blocksperrow = blocksperrow;
+  result->maxaccess = maxaccess;
+  result->pre_zero = pre_zero;
+  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->next = mem->virt_barray_list; /* add to list of virtual arrays */
+  mem->virt_barray_list = result;
+
+  return result;
+}
+
+
+METHODDEF(void)
+realize_virt_arrays (j_common_ptr cinfo)
+/* Allocate the in-memory buffers for any unrealized virtual arrays */
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  long space_per_minheight, maximum_space, avail_mem;
+  long minheights, max_minheights;
+  jvirt_sarray_ptr sptr;
+  jvirt_barray_ptr bptr;
+
+  /* Compute the minimum space needed (maxaccess rows in each buffer)
+   * and the maximum space needed (full image height in each buffer).
+   * These may be of use to the system-dependent jpeg_mem_available routine.
+   */
+  space_per_minheight = 0;
+  maximum_space = 0;
+  for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+    if (sptr->mem_buffer == NULL) { /* if not realized yet */
+      space_per_minheight += (long) sptr->maxaccess *
+			     (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+      maximum_space += (long) sptr->rows_in_array *
+		       (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+    }
+  }
+  for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+    if (bptr->mem_buffer == NULL) { /* if not realized yet */
+      space_per_minheight += (long) bptr->maxaccess *
+			     (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+      maximum_space += (long) bptr->rows_in_array *
+		       (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+    }
+  }
+
+  if (space_per_minheight <= 0)
+    return;			/* no unrealized arrays, no work */
+
+  /* Determine amount of memory to actually use; this is system-dependent. */
+  avail_mem = jpeg_mem_available(cinfo, space_per_minheight, maximum_space,
+				 mem->total_space_allocated);
+
+  /* If the maximum space needed is available, make all the buffers full
+   * height; otherwise parcel it out with the same number of minheights
+   * in each buffer.
+   */
+  if (avail_mem >= maximum_space)
+    max_minheights = 1000000000L;
+  else {
+    max_minheights = avail_mem / space_per_minheight;
+    /* If there doesn't seem to be enough space, try to get the minimum
+     * anyway.  This allows a "stub" implementation of jpeg_mem_available().
+     */
+    if (max_minheights <= 0)
+      max_minheights = 1;
+  }
+
+  /* Allocate the in-memory buffers and initialize backing store as needed. */
+
+  for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+    if (sptr->mem_buffer == NULL) { /* if not realized yet */
+      minheights = ((long) sptr->rows_in_array - 1L) / sptr->maxaccess + 1L;
+      if (minheights <= max_minheights) {
+	/* This buffer fits in memory */
+	sptr->rows_in_mem = sptr->rows_in_array;
+      } else {
+	/* It doesn't fit in memory, create backing store. */
+	sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess);
+	jpeg_open_backing_store(cinfo, & sptr->b_s_info,
+				(long) sptr->rows_in_array *
+				(long) sptr->samplesperrow *
+				(long) SIZEOF(JSAMPLE));
+	sptr->b_s_open = TRUE;
+      }
+      sptr->mem_buffer = alloc_sarray(cinfo, JPOOL_IMAGE,
+				      sptr->samplesperrow, sptr->rows_in_mem);
+      sptr->rowsperchunk = mem->last_rowsperchunk;
+      sptr->cur_start_row = 0;
+      sptr->first_undef_row = 0;
+      sptr->dirty = FALSE;
+    }
+  }
+
+  for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+    if (bptr->mem_buffer == NULL) { /* if not realized yet */
+      minheights = ((long) bptr->rows_in_array - 1L) / bptr->maxaccess + 1L;
+      if (minheights <= max_minheights) {
+	/* This buffer fits in memory */
+	bptr->rows_in_mem = bptr->rows_in_array;
+      } else {
+	/* It doesn't fit in memory, create backing store. */
+	bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess);
+	jpeg_open_backing_store(cinfo, & bptr->b_s_info,
+				(long) bptr->rows_in_array *
+				(long) bptr->blocksperrow *
+				(long) SIZEOF(JBLOCK));
+	bptr->b_s_open = TRUE;
+      }
+      bptr->mem_buffer = alloc_barray(cinfo, JPOOL_IMAGE,
+				      bptr->blocksperrow, bptr->rows_in_mem);
+      bptr->rowsperchunk = mem->last_rowsperchunk;
+      bptr->cur_start_row = 0;
+      bptr->first_undef_row = 0;
+      bptr->dirty = FALSE;
+    }
+  }
+}
+
+
+LOCAL(void)
+do_sarray_io (j_common_ptr cinfo, jvirt_sarray_ptr ptr, boolean writing)
+/* Do backing store read or write of a virtual sample array */
+{
+  long bytesperrow, file_offset, byte_count, rows, thisrow, i;
+
+  bytesperrow = (long) ptr->samplesperrow * SIZEOF(JSAMPLE);
+  file_offset = ptr->cur_start_row * bytesperrow;
+  /* Loop to read or write each allocation chunk in mem_buffer */
+  for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
+    /* One chunk, but check for short chunk at end of buffer */
+    rows = MIN((long) ptr->rowsperchunk, (long) ptr->rows_in_mem - i);
+    /* Transfer no more than is currently defined */
+    thisrow = (long) ptr->cur_start_row + i;
+    rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
+    /* Transfer no more than fits in file */
+    rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
+    if (rows <= 0)		/* this chunk might be past end of file! */
+      break;
+    byte_count = rows * bytesperrow;
+    if (writing)
+      (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
+					    (void FAR *) ptr->mem_buffer[i],
+					    file_offset, byte_count);
+    else
+      (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
+					   (void FAR *) ptr->mem_buffer[i],
+					   file_offset, byte_count);
+    file_offset += byte_count;
+  }
+}
+
+
+LOCAL(void)
+do_barray_io (j_common_ptr cinfo, jvirt_barray_ptr ptr, boolean writing)
+/* Do backing store read or write of a virtual coefficient-block array */
+{
+  long bytesperrow, file_offset, byte_count, rows, thisrow, i;
+
+  bytesperrow = (long) ptr->blocksperrow * SIZEOF(JBLOCK);
+  file_offset = ptr->cur_start_row * bytesperrow;
+  /* Loop to read or write each allocation chunk in mem_buffer */
+  for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
+    /* One chunk, but check for short chunk at end of buffer */
+    rows = MIN((long) ptr->rowsperchunk, (long) ptr->rows_in_mem - i);
+    /* Transfer no more than is currently defined */
+    thisrow = (long) ptr->cur_start_row + i;
+    rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
+    /* Transfer no more than fits in file */
+    rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
+    if (rows <= 0)		/* this chunk might be past end of file! */
+      break;
+    byte_count = rows * bytesperrow;
+    if (writing)
+      (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
+					    (void FAR *) ptr->mem_buffer[i],
+					    file_offset, byte_count);
+    else
+      (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
+					   (void FAR *) ptr->mem_buffer[i],
+					   file_offset, byte_count);
+    file_offset += byte_count;
+  }
+}
+
+
+METHODDEF(JSAMPARRAY)
+access_virt_sarray (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
+		    JDIMENSION start_row, JDIMENSION num_rows,
+		    boolean writable)
+/* Access the part of a virtual sample array starting at start_row */
+/* and extending for num_rows rows.  writable is true if  */
+/* caller intends to modify the accessed area. */
+{
+  JDIMENSION end_row = start_row + num_rows;
+  JDIMENSION undef_row;
+
+  /* debugging check */
+  if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess ||
+      ptr->mem_buffer == NULL)
+    ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+
+  /* Make the desired part of the virtual array accessible */
+  if (start_row < ptr->cur_start_row ||
+      end_row > ptr->cur_start_row+ptr->rows_in_mem) {
+    if (! ptr->b_s_open)
+      ERREXIT(cinfo, JERR_VIRTUAL_BUG);
+    /* Flush old buffer contents if necessary */
+    if (ptr->dirty) {
+      do_sarray_io(cinfo, ptr, TRUE);
+      ptr->dirty = FALSE;
+    }
+    /* Decide what part of virtual array to access.
+     * Algorithm: if target address > current window, assume forward scan,
+     * load starting at target address.  If target address < current window,
+     * assume backward scan, load so that target area is top of window.
+     * Note that when switching from forward write to forward read, will have
+     * start_row = 0, so the limiting case applies and we load from 0 anyway.
+     */
+    if (start_row > ptr->cur_start_row) {
+      ptr->cur_start_row = start_row;
+    } else {
+      /* use long arithmetic here to avoid overflow & unsigned problems */
+      long ltemp;
+
+      ltemp = (long) end_row - (long) ptr->rows_in_mem;
+      if (ltemp < 0)
+	ltemp = 0;		/* don't fall off front end of file */
+      ptr->cur_start_row = (JDIMENSION) ltemp;
+    }
+    /* Read in the selected part of the array.
+     * During the initial write pass, we will do no actual read
+     * because the selected part is all undefined.
+     */
+    do_sarray_io(cinfo, ptr, FALSE);
+  }
+  /* Ensure the accessed part of the array is defined; prezero if needed.
+   * To improve locality of access, we only prezero the part of the array
+   * that the caller is about to access, not the entire in-memory array.
+   */
+  if (ptr->first_undef_row < end_row) {
+    if (ptr->first_undef_row < start_row) {
+      if (writable)		/* writer skipped over a section of array */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;	/* but reader is allowed to read ahead */
+    } else {
+      undef_row = ptr->first_undef_row;
+    }
+    if (writable)
+      ptr->first_undef_row = end_row;
+    if (ptr->pre_zero) {
+      size_t bytesperrow = (size_t) ptr->samplesperrow * SIZEOF(JSAMPLE);
+      undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
+      end_row -= ptr->cur_start_row;
+      while (undef_row < end_row) {
+	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
+	undef_row++;
+      }
+    } else {
+      if (! writable)		/* reader looking at undefined data */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+    }
+  }
+  /* Flag the buffer dirty if caller will write in it */
+  if (writable)
+    ptr->dirty = TRUE;
+  /* Return address of proper part of the buffer */
+  return ptr->mem_buffer + (start_row - ptr->cur_start_row);
+}
+
+
+METHODDEF(JBLOCKARRAY)
+access_virt_barray (j_common_ptr cinfo, jvirt_barray_ptr ptr,
+		    JDIMENSION start_row, JDIMENSION num_rows,
+		    boolean writable)
+/* Access the part of a virtual block array starting at start_row */
+/* and extending for num_rows rows.  writable is true if  */
+/* caller intends to modify the accessed area. */
+{
+  JDIMENSION end_row = start_row + num_rows;
+  JDIMENSION undef_row;
+
+  /* debugging check */
+  if (end_row > ptr->rows_in_array || num_rows > ptr->maxaccess ||
+      ptr->mem_buffer == NULL)
+    ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+
+  /* Make the desired part of the virtual array accessible */
+  if (start_row < ptr->cur_start_row ||
+      end_row > ptr->cur_start_row+ptr->rows_in_mem) {
+    if (! ptr->b_s_open)
+      ERREXIT(cinfo, JERR_VIRTUAL_BUG);
+    /* Flush old buffer contents if necessary */
+    if (ptr->dirty) {
+      do_barray_io(cinfo, ptr, TRUE);
+      ptr->dirty = FALSE;
+    }
+    /* Decide what part of virtual array to access.
+     * Algorithm: if target address > current window, assume forward scan,
+     * load starting at target address.  If target address < current window,
+     * assume backward scan, load so that target area is top of window.
+     * Note that when switching from forward write to forward read, will have
+     * start_row = 0, so the limiting case applies and we load from 0 anyway.
+     */
+    if (start_row > ptr->cur_start_row) {
+      ptr->cur_start_row = start_row;
+    } else {
+      /* use long arithmetic here to avoid overflow & unsigned problems */
+      long ltemp;
+
+      ltemp = (long) end_row - (long) ptr->rows_in_mem;
+      if (ltemp < 0)
+	ltemp = 0;		/* don't fall off front end of file */
+      ptr->cur_start_row = (JDIMENSION) ltemp;
+    }
+    /* Read in the selected part of the array.
+     * During the initial write pass, we will do no actual read
+     * because the selected part is all undefined.
+     */
+    do_barray_io(cinfo, ptr, FALSE);
+  }
+  /* Ensure the accessed part of the array is defined; prezero if needed.
+   * To improve locality of access, we only prezero the part of the array
+   * that the caller is about to access, not the entire in-memory array.
+   */
+  if (ptr->first_undef_row < end_row) {
+    if (ptr->first_undef_row < start_row) {
+      if (writable)		/* writer skipped over a section of array */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;	/* but reader is allowed to read ahead */
+    } else {
+      undef_row = ptr->first_undef_row;
+    }
+    if (writable)
+      ptr->first_undef_row = end_row;
+    if (ptr->pre_zero) {
+      size_t bytesperrow = (size_t) ptr->blocksperrow * SIZEOF(JBLOCK);
+      undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
+      end_row -= ptr->cur_start_row;
+      while (undef_row < end_row) {
+	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
+	undef_row++;
+      }
+    } else {
+      if (! writable)		/* reader looking at undefined data */
+	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+    }
+  }
+  /* Flag the buffer dirty if caller will write in it */
+  if (writable)
+    ptr->dirty = TRUE;
+  /* Return address of proper part of the buffer */
+  return ptr->mem_buffer + (start_row - ptr->cur_start_row);
+}
+
+
+/*
+ * Release all objects belonging to a specified pool.
+ */
+
+METHODDEF(void)
+free_pool (j_common_ptr cinfo, int pool_id)
+{
+  my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
+  small_pool_ptr shdr_ptr;
+  large_pool_ptr lhdr_ptr;
+  size_t space_freed;
+
+  if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+
+#ifdef MEM_STATS
+  if (cinfo->err->trace_level > 1)
+    print_mem_stats(cinfo, pool_id); /* print pool's memory usage statistics */
+#endif
+
+  /* If freeing IMAGE pool, close any virtual arrays first */
+  if (pool_id == JPOOL_IMAGE) {
+    jvirt_sarray_ptr sptr;
+    jvirt_barray_ptr bptr;
+
+    for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
+      if (sptr->b_s_open) {	/* there may be no backing store */
+	sptr->b_s_open = FALSE;	/* prevent recursive close if error */
+	(*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info);
+      }
+    }
+    mem->virt_sarray_list = NULL;
+    for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
+      if (bptr->b_s_open) {	/* there may be no backing store */
+	bptr->b_s_open = FALSE;	/* prevent recursive close if error */
+	(*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info);
+      }
+    }
+    mem->virt_barray_list = NULL;
+  }
+
+  /* Release large objects */
+  lhdr_ptr = mem->large_list[pool_id];
+  mem->large_list[pool_id] = NULL;
+
+  while (lhdr_ptr != NULL) {
+    large_pool_ptr next_lhdr_ptr = lhdr_ptr->hdr.next;
+    space_freed = lhdr_ptr->hdr.bytes_used +
+		  lhdr_ptr->hdr.bytes_left +
+		  SIZEOF(large_pool_hdr);
+    jpeg_free_large(cinfo, (void FAR *) lhdr_ptr, space_freed);
+    mem->total_space_allocated -= space_freed;
+    lhdr_ptr = next_lhdr_ptr;
+  }
+
+  /* Release small objects */
+  shdr_ptr = mem->small_list[pool_id];
+  mem->small_list[pool_id] = NULL;
+
+  while (shdr_ptr != NULL) {
+    small_pool_ptr next_shdr_ptr = shdr_ptr->hdr.next;
+    space_freed = shdr_ptr->hdr.bytes_used +
+		  shdr_ptr->hdr.bytes_left +
+		  SIZEOF(small_pool_hdr);
+    jpeg_free_small(cinfo, (void *) shdr_ptr, space_freed);
+    mem->total_space_allocated -= space_freed;
+    shdr_ptr = next_shdr_ptr;
+  }
+}
+
+
+/*
+ * Close up shop entirely.
+ * Note that this cannot be called unless cinfo->mem is non-NULL.
+ */
+
+METHODDEF(void)
+self_destruct (j_common_ptr cinfo)
+{
+  int pool;
+
+  /* Close all backing store, release all memory.
+   * Releasing pools in reverse order might help avoid fragmentation
+   * with some (brain-damaged) malloc libraries.
+   */
+  for (pool = JPOOL_NUMPOOLS-1; pool >= JPOOL_PERMANENT; pool--) {
+    free_pool(cinfo, pool);
+  }
+
+  /* Release the memory manager control block too. */
+  jpeg_free_small(cinfo, (void *) cinfo->mem, SIZEOF(my_memory_mgr));
+  cinfo->mem = NULL;		/* ensures I will be called only once */
+
+  jpeg_mem_term(cinfo);		/* system-dependent cleanup */
+}
+
+
+/*
+ * Memory manager initialization.
+ * When this is called, only the error manager pointer is valid in cinfo!
+ */
+
+GLOBAL(void)
+jinit_memory_mgr (j_common_ptr cinfo)
+{
+  my_mem_ptr mem;
+  long max_to_use;
+  int pool;
+  size_t test_mac;
+
+  cinfo->mem = NULL;		/* for safety if init fails */
+
+  /* Check for configuration errors.
+   * SIZEOF(ALIGN_TYPE) should be a power of 2; otherwise, it probably
+   * doesn't reflect any real hardware alignment requirement.
+   * The test is a little tricky: for X>0, X and X-1 have no one-bits
+   * in common if and only if X is a power of 2, ie has only one one-bit.
+   * Some compilers may give an "unreachable code" warning here; ignore it.
+   */
+  if ((SIZEOF(ALIGN_TYPE) & (SIZEOF(ALIGN_TYPE)-1)) != 0)
+    ERREXIT(cinfo, JERR_BAD_ALIGN_TYPE);
+  /* MAX_ALLOC_CHUNK must be representable as type size_t, and must be
+   * a multiple of SIZEOF(ALIGN_TYPE).
+   * Again, an "unreachable code" warning may be ignored here.
+   * But a "constant too large" warning means you need to fix MAX_ALLOC_CHUNK.
+   */
+  test_mac = (size_t) MAX_ALLOC_CHUNK;
+  if ((long) test_mac != MAX_ALLOC_CHUNK ||
+      (MAX_ALLOC_CHUNK % SIZEOF(ALIGN_TYPE)) != 0)
+    ERREXIT(cinfo, JERR_BAD_ALLOC_CHUNK);
+
+  max_to_use = jpeg_mem_init(cinfo); /* system-dependent initialization */
+
+  /* Attempt to allocate memory manager's control block */
+  mem = (my_mem_ptr) jpeg_get_small(cinfo, SIZEOF(my_memory_mgr));
+
+  if (mem == NULL) {
+    jpeg_mem_term(cinfo);	/* system-dependent cleanup */
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 0);
+  }
+
+  /* OK, fill in the method pointers */
+  mem->pub.alloc_small = alloc_small;
+  mem->pub.alloc_large = alloc_large;
+  mem->pub.alloc_sarray = alloc_sarray;
+  mem->pub.alloc_barray = alloc_barray;
+  mem->pub.request_virt_sarray = request_virt_sarray;
+  mem->pub.request_virt_barray = request_virt_barray;
+  mem->pub.realize_virt_arrays = realize_virt_arrays;
+  mem->pub.access_virt_sarray = access_virt_sarray;
+  mem->pub.access_virt_barray = access_virt_barray;
+  mem->pub.free_pool = free_pool;
+  mem->pub.self_destruct = self_destruct;
+
+  /* Make MAX_ALLOC_CHUNK accessible to other modules */
+  mem->pub.max_alloc_chunk = MAX_ALLOC_CHUNK;
+
+  /* Initialize working state */
+  mem->pub.max_memory_to_use = max_to_use;
+
+  for (pool = JPOOL_NUMPOOLS-1; pool >= JPOOL_PERMANENT; pool--) {
+    mem->small_list[pool] = NULL;
+    mem->large_list[pool] = NULL;
+  }
+  mem->virt_sarray_list = NULL;
+  mem->virt_barray_list = NULL;
+
+  mem->total_space_allocated = SIZEOF(my_memory_mgr);
+
+  /* Declare ourselves open for business */
+  cinfo->mem = & mem->pub;
+
+  /* Check for an environment variable JPEGMEM; if found, override the
+   * default max_memory setting from jpeg_mem_init.  Note that the
+   * surrounding application may again override this value.
+   * If your system doesn't support getenv(), define NO_GETENV to disable
+   * this feature.
+   */
+#ifndef NO_GETENV
+  { char * memenv;
+
+    if ((memenv = getenv("JPEGMEM")) != NULL) {
+      char ch = 'x';
+
+      if (sscanf(memenv, "%ld%c", &max_to_use, &ch) > 0) {
+	if (ch == 'm' || ch == 'M')
+	  max_to_use *= 1000L;
+	mem->pub.max_memory_to_use = max_to_use * 1000L;
+      }
+    }
+  }
+#endif
+
+}
diff --git a/jpeg/jmemnobs.c b/jpeg/jmemnobs.c
new file mode 100644
index 000000000..eb8c33772
--- /dev/null
+++ b/jpeg/jmemnobs.c
@@ -0,0 +1,109 @@
+/*
+ * jmemnobs.c
+ *
+ * Copyright (C) 1992-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides a really simple implementation of the system-
+ * dependent portion of the JPEG memory manager.  This implementation
+ * assumes that no backing-store files are needed: all required space
+ * can be obtained from malloc().
+ * This is very portable in the sense that it'll compile on almost anything,
+ * but you'd better have lots of main memory (or virtual memory) if you want
+ * to process big images.
+ * Note that the max_memory_to_use option is ignored by this implementation.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jmemsys.h"		/* import the system-dependent declarations */
+
+#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
+extern void * malloc JPP((size_t size));
+extern void free JPP((void *ptr));
+#endif
+
+
+/*
+ * Memory allocation and freeing are controlled by the regular library
+ * routines malloc() and free().
+ */
+
+GLOBAL(void *)
+jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * "Large" objects are treated the same as "small" ones.
+ * NB: although we include FAR keywords in the routine declarations,
+ * this file won't actually work in 80x86 small/medium model; at least,
+ * you probably won't be able to process useful-size images in only 64KB.
+ */
+
+GLOBAL(void FAR *)
+jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
+{
+  return (void FAR *) malloc(sizeofobject);
+}
+
+GLOBAL(void)
+jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+{
+  free(object);
+}
+
+
+/*
+ * This routine computes the total memory space available for allocation.
+ * Here we always say, "we got all you want bud!"
+ */
+
+GLOBAL(long)
+jpeg_mem_available (j_common_ptr cinfo, long min_bytes_needed,
+		    long max_bytes_needed, long already_allocated)
+{
+  return max_bytes_needed;
+}
+
+
+/*
+ * Backing store (temporary file) management.
+ * Since jpeg_mem_available always promised the moon,
+ * this should never be called and we can just error out.
+ */
+
+GLOBAL(void)
+jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
+			 long total_bytes_needed)
+{
+  ERREXIT(cinfo, JERR_NO_BACKING_STORE);
+}
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.  Here, there isn't any.
+ */
+
+GLOBAL(long)
+jpeg_mem_init (j_common_ptr cinfo)
+{
+  return 0;			/* just set max_memory_to_use to 0 */
+}
+
+GLOBAL(void)
+jpeg_mem_term (j_common_ptr cinfo)
+{
+  /* no work */
+}
diff --git a/jpeg/jmemsys.h b/jpeg/jmemsys.h
new file mode 100644
index 000000000..6c3c6d348
--- /dev/null
+++ b/jpeg/jmemsys.h
@@ -0,0 +1,198 @@
+/*
+ * jmemsys.h
+ *
+ * Copyright (C) 1992-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This include file defines the interface between the system-independent
+ * and system-dependent portions of the JPEG memory manager.  No other
+ * modules need include it.  (The system-independent portion is jmemmgr.c;
+ * there are several different versions of the system-dependent portion.)
+ *
+ * This file works as-is for the system-dependent memory managers supplied
+ * in the IJG distribution.  You may need to modify it if you write a
+ * custom memory manager.  If system-dependent changes are needed in
+ * this file, the best method is to #ifdef them based on a configuration
+ * symbol supplied in jconfig.h, as we have done with USE_MSDOS_MEMMGR
+ * and USE_MAC_MEMMGR.
+ */
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_get_small		jGetSmall
+#define jpeg_free_small		jFreeSmall
+#define jpeg_get_large		jGetLarge
+#define jpeg_free_large		jFreeLarge
+#define jpeg_mem_available	jMemAvail
+#define jpeg_open_backing_store	jOpenBackStore
+#define jpeg_mem_init		jMemInit
+#define jpeg_mem_term		jMemTerm
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/*
+ * These two functions are used to allocate and release small chunks of
+ * memory.  (Typically the total amount requested through jpeg_get_small is
+ * no more than 20K or so; this will be requested in chunks of a few K each.)
+ * Behavior should be the same as for the standard library functions malloc
+ * and free; in particular, jpeg_get_small must return NULL on failure.
+ * On most systems, these ARE malloc and free.  jpeg_free_small is passed the
+ * size of the object being freed, just in case it's needed.
+ * On an 80x86 machine using small-data memory model, these manage near heap.
+ */
+
+EXTERN(void *) jpeg_get_small JPP((j_common_ptr cinfo, size_t sizeofobject));
+EXTERN(void) jpeg_free_small JPP((j_common_ptr cinfo, void * object,
+				  size_t sizeofobject));
+
+/*
+ * These two functions are used to allocate and release large chunks of
+ * memory (up to the total free space designated by jpeg_mem_available).
+ * The interface is the same as above, except that on an 80x86 machine,
+ * far pointers are used.  On most other machines these are identical to
+ * the jpeg_get/free_small routines; but we keep them separate anyway,
+ * in case a different allocation strategy is desirable for large chunks.
+ */
+
+EXTERN(void FAR *) jpeg_get_large JPP((j_common_ptr cinfo,
+				       size_t sizeofobject));
+EXTERN(void) jpeg_free_large JPP((j_common_ptr cinfo, void FAR * object,
+				  size_t sizeofobject));
+
+/*
+ * The macro MAX_ALLOC_CHUNK designates the maximum number of bytes that may
+ * be requested in a single call to jpeg_get_large (and jpeg_get_small for that
+ * matter, but that case should never come into play).  This macro is needed
+ * to model the 64Kb-segment-size limit of far addressing on 80x86 machines.
+ * On those machines, we expect that jconfig.h will provide a proper value.
+ * On machines with 32-bit flat address spaces, any large constant may be used.
+ *
+ * NB: jmemmgr.c expects that MAX_ALLOC_CHUNK will be representable as type
+ * size_t and will be a multiple of sizeof(align_type).
+ */
+
+#ifndef MAX_ALLOC_CHUNK		/* may be overridden in jconfig.h */
+#define MAX_ALLOC_CHUNK  1000000000L
+#endif
+
+/*
+ * This routine computes the total space still available for allocation by
+ * jpeg_get_large.  If more space than this is needed, backing store will be
+ * used.  NOTE: any memory already allocated must not be counted.
+ *
+ * There is a minimum space requirement, corresponding to the minimum
+ * feasible buffer sizes; jmemmgr.c will request that much space even if
+ * jpeg_mem_available returns zero.  The maximum space needed, enough to hold
+ * all working storage in memory, is also passed in case it is useful.
+ * Finally, the total space already allocated is passed.  If no better
+ * method is available, cinfo->mem->max_memory_to_use - already_allocated
+ * is often a suitable calculation.
+ *
+ * It is OK for jpeg_mem_available to underestimate the space available
+ * (that'll just lead to more backing-store access than is really necessary).
+ * However, an overestimate will lead to failure.  Hence it's wise to subtract
+ * a slop factor from the true available space.  5% should be enough.
+ *
+ * On machines with lots of virtual memory, any large constant may be returned.
+ * Conversely, zero may be returned to always use the minimum amount of memory.
+ */
+
+EXTERN(long) jpeg_mem_available JPP((j_common_ptr cinfo,
+				     long min_bytes_needed,
+				     long max_bytes_needed,
+				     long already_allocated));
+
+
+/*
+ * This structure holds whatever state is needed to access a single
+ * backing-store object.  The read/write/close method pointers are called
+ * by jmemmgr.c to manipulate the backing-store object; all other fields
+ * are private to the system-dependent backing store routines.
+ */
+
+#define TEMP_NAME_LENGTH   64	/* max length of a temporary file's name */
+
+
+#ifdef USE_MSDOS_MEMMGR		/* DOS-specific junk */
+
+typedef unsigned short XMSH;	/* type of extended-memory handles */
+typedef unsigned short EMSH;	/* type of expanded-memory handles */
+
+typedef union {
+  short file_handle;		/* DOS file handle if it's a temp file */
+  XMSH xms_handle;		/* handle if it's a chunk of XMS */
+  EMSH ems_handle;		/* handle if it's a chunk of EMS */
+} handle_union;
+
+#endif /* USE_MSDOS_MEMMGR */
+
+#ifdef USE_MAC_MEMMGR		/* Mac-specific junk */
+#include <Files.h>
+#endif /* USE_MAC_MEMMGR */
+
+
+typedef struct backing_store_struct * backing_store_ptr;
+
+typedef struct backing_store_struct {
+  /* Methods for reading/writing/closing this backing-store object */
+  JMETHOD(void, read_backing_store, (j_common_ptr cinfo,
+				     backing_store_ptr info,
+				     void FAR * buffer_address,
+				     long file_offset, long byte_count));
+  JMETHOD(void, write_backing_store, (j_common_ptr cinfo,
+				      backing_store_ptr info,
+				      void FAR * buffer_address,
+				      long file_offset, long byte_count));
+  JMETHOD(void, close_backing_store, (j_common_ptr cinfo,
+				      backing_store_ptr info));
+
+  /* Private fields for system-dependent backing-store management */
+#ifdef USE_MSDOS_MEMMGR
+  /* For the MS-DOS manager (jmemdos.c), we need: */
+  handle_union handle;		/* reference to backing-store storage object */
+  char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
+#else
+#ifdef USE_MAC_MEMMGR
+  /* For the Mac manager (jmemmac.c), we need: */
+  short temp_file;		/* file reference number to temp file */
+  FSSpec tempSpec;		/* the FSSpec for the temp file */
+  char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
+#else
+  /* For a typical implementation with temp files, we need: */
+  FILE * temp_file;		/* stdio reference to temp file */
+  char temp_name[TEMP_NAME_LENGTH]; /* name of temp file */
+#endif
+#endif
+} backing_store_info;
+
+
+/*
+ * Initial opening of a backing-store object.  This must fill in the
+ * read/write/close pointers in the object.  The read/write routines
+ * may take an error exit if the specified maximum file size is exceeded.
+ * (If jpeg_mem_available always returns a large value, this routine can
+ * just take an error exit.)
+ */
+
+EXTERN(void) jpeg_open_backing_store JPP((j_common_ptr cinfo,
+					  backing_store_ptr info,
+					  long total_bytes_needed));
+
+
+/*
+ * These routines take care of any system-dependent initialization and
+ * cleanup required.  jpeg_mem_init will be called before anything is
+ * allocated (and, therefore, nothing in cinfo is of use except the error
+ * manager pointer).  It should return a suitable default value for
+ * max_memory_to_use; this may subsequently be overridden by the surrounding
+ * application.  (Note that max_memory_to_use is only important if
+ * jpeg_mem_available chooses to consult it ... no one else will.)
+ * jpeg_mem_term may assume that all requested memory has been freed and that
+ * all opened backing-store objects have been closed.
+ */
+
+EXTERN(long) jpeg_mem_init JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_mem_term JPP((j_common_ptr cinfo));
diff --git a/jpeg/jmorecfg.h b/jpeg/jmorecfg.h
new file mode 100644
index 000000000..4f491fc91
--- /dev/null
+++ b/jpeg/jmorecfg.h
@@ -0,0 +1,316 @@
+/*
+ * jmorecfg.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains additional configuration options that customize the
+ * JPEG software for special applications or support machine-dependent
+ * optimizations.  Most users will not need to touch this file.
+ */
+
+
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  8	/* use 8 or 12 */
+
+
+/*
+ * Maximum number of components (color channels) allowed in JPEG image.
+ * To meet the letter of the JPEG spec, set this to 255.  However, darn
+ * few applications need more than 4 channels (maybe 5 for CMYK + alpha
+ * mask).  We recommend 10 as a reasonable compromise; use 4 if you are
+ * really short on memory.  (Each allowed component costs a hundred or so
+ * bytes of storage, whether actually used in an image or not.)
+ */
+
+#define MAX_COMPONENTS  10	/* maximum number of image components */
+
+
+/*
+ * Basic data types.
+ * You may need to change these if you have a machine with unusual data
+ * type sizes; for example, "char" not 8 bits, "short" not 16 bits,
+ * or "long" not 32 bits.  We don't care whether "int" is 16 or 32 bits,
+ * but it had better be at least 16.
+ */
+
+/* Representation of a single sample (pixel element value).
+ * We frequently allocate large arrays of these, so it's important to keep
+ * them small.  But if you have memory to burn and access to char or short
+ * arrays is very slow on your hardware, you might want to change these.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+/* JSAMPLE should be the smallest type that will hold the values 0..255.
+ * You can use a signed char by having GETJSAMPLE mask it with 0xFF.
+ */
+
+typedef unsigned char JSAMPLE;
+#define GETJSAMPLE(value)  ((int) (value))
+
+#define MAXJSAMPLE	255
+#define CENTERJSAMPLE	128
+
+#endif /* BITS_IN_JSAMPLE == 8 */
+
+
+#if BITS_IN_JSAMPLE == 12
+/* JSAMPLE should be the smallest type that will hold the values 0..4095.
+ * On nearly all machines "short" will do nicely.
+ */
+
+typedef short JSAMPLE;
+#define GETJSAMPLE(value)  ((int) (value))
+
+#define MAXJSAMPLE	4095
+#define CENTERJSAMPLE	2048
+
+#endif /* BITS_IN_JSAMPLE == 12 */
+
+
+/* Representation of a DCT frequency coefficient.
+ * This should be a signed value of at least 16 bits; "short" is usually OK.
+ * Again, we allocate large arrays of these, but you can change to int
+ * if you have memory to burn and "short" is really slow.
+ */
+
+typedef short JCOEF;
+
+
+/* Compressed datastreams are represented as arrays of JOCTET.
+ * These must be EXACTLY 8 bits wide, at least once they are written to
+ * external storage.  Note that when using the stdio data source/destination
+ * managers, this is also the data type passed to fread/fwrite.
+ */
+
+typedef unsigned char JOCTET;
+#define GETJOCTET(value)  (value)
+
+/* These typedefs are used for various table entries and so forth.
+ * They must be at least as wide as specified; but making them too big
+ * won't cost a huge amount of memory, so we don't provide special
+ * extraction code like we did for JSAMPLE.  (In other words, these
+ * typedefs live at a different point on the speed/space tradeoff curve.)
+ */
+
+/* UINT8 must hold at least the values 0..255. */
+
+typedef unsigned char UINT8;
+
+/* UINT16 must hold at least the values 0..65535. */
+
+typedef unsigned short UINT16;
+
+/* INT16 must hold at least the values -32768..32767. */
+
+#ifndef XMD_H			/* X11/xmd.h correctly defines INT16 */
+typedef short INT16;
+#endif
+
+/* INT32 must hold at least signed 32-bit values. */
+
+#ifndef XMD_H
+typedef int INT32;
+#endif
+
+/* Datatype used for image dimensions.  The JPEG standard only supports
+ * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
+ * "unsigned int" is sufficient on all machines.  However, if you need to
+ * handle larger images and you don't mind deviating from the spec, you
+ * can change this datatype.
+ */
+
+typedef unsigned int JDIMENSION;
+
+#define JPEG_MAX_DIMENSION  65500L  /* a tad under 64K to prevent overflows */
+
+
+/* These macros are used in all function definitions and extern declarations.
+ * You could modify them if you need to change function linkage conventions;
+ * in particular, you'll need to do that to make the library a Windows DLL.
+ * Another application is to make all functions global for use with debuggers
+ * or code profilers that require it.
+ */
+
+/* a function called through method pointers: */
+#define METHODDEF(type)		static type
+/* a function used only in its module: */
+#define LOCAL(type)		static type
+/* a function referenced thru EXTERNs: */
+#define GLOBAL(type)		type
+/* a reference to a GLOBAL function: */
+#define EXTERN(type)		extern type
+
+
+/* This macro is used to declare a "method", that is, a function pointer.
+ * We want to supply prototype parameters if the compiler can cope.
+ * Note that the arglist parameter must be parenthesized!
+ * Again, you can customize this if you need special linkage keywords.
+ */
+
+#define JMETHOD(type,methodname,arglist)  type (*methodname) arglist
+
+/* Here is the pseudo-keyword for declaring pointers that must be "far"
+ * on 80x86 machines.  Most of the specialized coding for 80x86 is handled
+ * by just saying "FAR *" where such a pointer is needed.  In a few places
+ * explicit coding is needed; see uses of the NEED_FAR_POINTERS symbol.
+ */
+
+#undef FAR
+#define FAR
+
+
+/*
+ * On a few systems, type boolean and/or its values FALSE, TRUE may appear
+ * in standard header files.  Or you may have conflicts with application-
+ * specific header files that you want to include together with these files.
+ * Defining HAVE_BOOLEAN before including jpeglib.h should make it work.
+ */
+
+#ifndef HAVE_BOOLEAN
+typedef char boolean;
+#endif
+#ifndef FALSE			/* in case these macros already exist */
+#define FALSE	0		/* values of boolean */
+#endif
+#ifndef TRUE
+#define TRUE	1
+#endif
+
+
+/*
+ * The remaining options affect code selection within the JPEG library,
+ * but they don't need to be visible to most applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS or JPEG_INTERNAL_OPTIONS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+#define JPEG_INTERNAL_OPTIONS
+#endif
+
+#ifdef JPEG_INTERNAL_OPTIONS
+
+
+/*
+ * These defines indicate whether to include various optional functions.
+ * Undefining some of these symbols will produce a smaller but less capable
+ * library.  Note that you can leave certain source files out of the
+ * compilation/linking process if you've #undef'd the corresponding symbols.
+ * (You may HAVE to do that if your compiler doesn't like null source files.)
+ */
+
+/* Arithmetic coding is unsupported for legal reasons.  Complaints to IBM. */
+
+/* Capability options common to encoder and decoder: */
+
+#define DCT_ISLOW_SUPPORTED	/* slow but accurate integer algorithm */
+#define DCT_IFAST_SUPPORTED	/* faster, less accurate integer method */
+#define DCT_FLOAT_SUPPORTED	/* floating-point: accurate, fast on fast HW */
+
+/* Encoder capability options: */
+
+#undef  C_ARITH_CODING_SUPPORTED    /* Arithmetic coding back end? */
+#define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
+#define C_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
+#define ENTROPY_OPT_SUPPORTED	    /* Optimization of entropy coding parms? */
+/* Note: if you selected 12-bit data precision, it is dangerous to turn off
+ * ENTROPY_OPT_SUPPORTED.  The standard Huffman tables are only good for 8-bit
+ * precision, so jchuff.c normally uses entropy optimization to compute
+ * usable tables for higher precision.  If you don't want to do optimization,
+ * you'll have to supply different default Huffman tables.
+ * The exact same statements apply for progressive JPEG: the default tables
+ * don't work for progressive mode.  (This may get fixed, however.)
+ */
+#define INPUT_SMOOTHING_SUPPORTED   /* Input image smoothing option? */
+
+/* Decoder capability options: */
+
+#undef  D_ARITH_CODING_SUPPORTED    /* Arithmetic coding back end? */
+#define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
+#define D_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
+#define SAVE_MARKERS_SUPPORTED	    /* jpeg_save_markers() needed? */
+#define BLOCK_SMOOTHING_SUPPORTED   /* Block smoothing? (Progressive only) */
+#define IDCT_SCALING_SUPPORTED	    /* Output rescaling via IDCT? */
+#undef  UPSAMPLE_SCALING_SUPPORTED  /* Output rescaling at upsample stage? */
+#define UPSAMPLE_MERGING_SUPPORTED  /* Fast path for sloppy upsampling? */
+#define QUANT_1PASS_SUPPORTED	    /* 1-pass color quantization? */
+#define QUANT_2PASS_SUPPORTED	    /* 2-pass color quantization? */
+
+/* more capability options later, no doubt */
+
+
+/*
+ * Ordering of RGB data in scanlines passed to or from the application.
+ * If your application wants to deal with data in the order B,G,R, just
+ * change these macros.  You can also deal with formats such as R,G,B,X
+ * (one extra byte per pixel) by changing RGB_PIXELSIZE.  Note that changing
+ * the offsets will also change the order in which colormap data is organized.
+ * RESTRICTIONS:
+ * 1. The sample applications cjpeg,djpeg do NOT support modified RGB formats.
+ * 2. These macros only affect RGB<=>YCbCr color conversion, so they are not
+ *    useful if you are using JPEG color spaces other than YCbCr or grayscale.
+ * 3. The color quantizer modules will not behave desirably if RGB_PIXELSIZE
+ *    is not 3 (they don't understand about dummy color components!).  So you
+ *    can't use color quantization if you change that value.
+ */
+
+#define RGB_RED		0	/* Offset of Red in an RGB scanline element */
+#define RGB_GREEN	1	/* Offset of Green */
+#define RGB_BLUE	2	/* Offset of Blue */
+#define RGB_PIXELSIZE	3	/* JSAMPLEs per RGB scanline element */
+
+
+/* Definitions for speed-related optimizations. */
+
+
+/* If your compiler supports inline functions, define INLINE
+ * as the inline keyword; otherwise define it as empty.
+ */
+
+#ifndef INLINE
+#ifdef __GNUC__			/* for instance, GNU C knows about inline */
+#define INLINE __inline__
+#endif
+#ifndef INLINE
+#define INLINE			/* default is to define it as empty */
+#endif
+#endif
+
+
+/* On some machines (notably 68000 series) "int" is 32 bits, but multiplying
+ * two 16-bit shorts is faster than multiplying two ints.  Define MULTIPLIER
+ * as short on such a machine.  MULTIPLIER must be at least 16 bits wide.
+ */
+
+#ifndef MULTIPLIER
+#define MULTIPLIER  int		/* type for fastest integer multiply */
+#endif
+
+
+/* FAST_FLOAT should be either float or double, whichever is done faster
+ * by your compiler.  (Note that this type is only used in the floating point
+ * DCT routines, so it only matters if you've defined DCT_FLOAT_SUPPORTED.)
+ * Typically, float is faster in ANSI C compilers, while double is faster in
+ * pre-ANSI compilers (because they insist on converting to double anyway).
+ * The code below therefore chooses float if we have ANSI-style prototypes.
+ */
+
+#ifndef FAST_FLOAT
+#ifdef HAVE_PROTOTYPES
+#define FAST_FLOAT  float
+#else
+#define FAST_FLOAT  double
+#endif
+#endif
+
+#endif /* JPEG_INTERNAL_OPTIONS */
diff --git a/jpeg/jpegint.h b/jpeg/jpegint.h
new file mode 100644
index 000000000..95b00d405
--- /dev/null
+++ b/jpeg/jpegint.h
@@ -0,0 +1,392 @@
+/*
+ * jpegint.h
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file provides common declarations for the various JPEG modules.
+ * These declarations are considered internal to the JPEG library; most
+ * applications using the library shouldn't need to include this file.
+ */
+
+
+/* Declarations for both compression & decompression */
+
+typedef enum {			/* Operating modes for buffer controllers */
+	JBUF_PASS_THRU,		/* Plain stripwise operation */
+	/* Remaining modes require a full-image buffer to have been created */
+	JBUF_SAVE_SOURCE,	/* Run source subobject only, save output */
+	JBUF_CRANK_DEST,	/* Run dest subobject only, using saved data */
+	JBUF_SAVE_AND_PASS	/* Run both subobjects, save output */
+} J_BUF_MODE;
+
+/* Values of global_state field (jdapi.c has some dependencies on ordering!) */
+#define CSTATE_START	100	/* after create_compress */
+#define CSTATE_SCANNING	101	/* start_compress done, write_scanlines OK */
+#define CSTATE_RAW_OK	102	/* start_compress done, write_raw_data OK */
+#define CSTATE_WRCOEFS	103	/* jpeg_write_coefficients done */
+#define DSTATE_START	200	/* after create_decompress */
+#define DSTATE_INHEADER	201	/* reading header markers, no SOS yet */
+#define DSTATE_READY	202	/* found SOS, ready for start_decompress */
+#define DSTATE_PRELOAD	203	/* reading multiscan file in start_decompress*/
+#define DSTATE_PRESCAN	204	/* performing dummy pass for 2-pass quant */
+#define DSTATE_SCANNING	205	/* start_decompress done, read_scanlines OK */
+#define DSTATE_RAW_OK	206	/* start_decompress done, read_raw_data OK */
+#define DSTATE_BUFIMAGE	207	/* expecting jpeg_start_output */
+#define DSTATE_BUFPOST	208	/* looking for SOS/EOI in jpeg_finish_output */
+#define DSTATE_RDCOEFS	209	/* reading file in jpeg_read_coefficients */
+#define DSTATE_STOPPING	210	/* looking for EOI in jpeg_finish_decompress */
+
+
+/* Declarations for compression modules */
+
+/* Master control module */
+struct jpeg_comp_master {
+  JMETHOD(void, prepare_for_pass, (j_compress_ptr cinfo));
+  JMETHOD(void, pass_startup, (j_compress_ptr cinfo));
+  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+
+  /* State variables made visible to other modules */
+  boolean call_pass_startup;	/* True if pass_startup must be called */
+  boolean is_last_pass;		/* True during last pass */
+};
+
+/* Main buffer control (downsampled-data buffer) */
+struct jpeg_c_main_controller {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, process_data, (j_compress_ptr cinfo,
+			       JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+			       JDIMENSION in_rows_avail));
+};
+
+/* Compression preprocessing (downsampling input buffer control) */
+struct jpeg_c_prep_controller {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, pre_process_data, (j_compress_ptr cinfo,
+				   JSAMPARRAY input_buf,
+				   JDIMENSION *in_row_ctr,
+				   JDIMENSION in_rows_avail,
+				   JSAMPIMAGE output_buf,
+				   JDIMENSION *out_row_group_ctr,
+				   JDIMENSION out_row_groups_avail));
+};
+
+/* Coefficient buffer control */
+struct jpeg_c_coef_controller {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(boolean, compress_data, (j_compress_ptr cinfo,
+				   JSAMPIMAGE input_buf));
+};
+
+/* Colorspace conversion */
+struct jpeg_color_converter {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  JMETHOD(void, color_convert, (j_compress_ptr cinfo,
+				JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+				JDIMENSION output_row, int num_rows));
+};
+
+/* Downsampling */
+struct jpeg_downsampler {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  JMETHOD(void, downsample, (j_compress_ptr cinfo,
+			     JSAMPIMAGE input_buf, JDIMENSION in_row_index,
+			     JSAMPIMAGE output_buf,
+			     JDIMENSION out_row_group_index));
+
+  boolean need_context_rows;	/* TRUE if need rows above & below */
+};
+
+/* Forward DCT (also controls coefficient quantization) */
+struct jpeg_forward_dct {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  /* perhaps this should be an array??? */
+  JMETHOD(void, forward_DCT, (j_compress_ptr cinfo,
+			      jpeg_component_info * compptr,
+			      JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+			      JDIMENSION start_row, JDIMENSION start_col,
+			      JDIMENSION num_blocks));
+};
+
+/* Entropy encoding */
+struct jpeg_entropy_encoder {
+  JMETHOD(void, start_pass, (j_compress_ptr cinfo, boolean gather_statistics));
+  JMETHOD(boolean, encode_mcu, (j_compress_ptr cinfo, JBLOCKROW *MCU_data));
+  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+};
+
+/* Marker writing */
+struct jpeg_marker_writer {
+  JMETHOD(void, write_file_header, (j_compress_ptr cinfo));
+  JMETHOD(void, write_frame_header, (j_compress_ptr cinfo));
+  JMETHOD(void, write_scan_header, (j_compress_ptr cinfo));
+  JMETHOD(void, write_file_trailer, (j_compress_ptr cinfo));
+  JMETHOD(void, write_tables_only, (j_compress_ptr cinfo));
+  /* These routines are exported to allow insertion of extra markers */
+  /* Probably only COM and APPn markers should be written this way */
+  JMETHOD(void, write_marker_header, (j_compress_ptr cinfo, int marker,
+				      unsigned int datalen));
+  JMETHOD(void, write_marker_byte, (j_compress_ptr cinfo, int val));
+};
+
+
+/* Declarations for decompression modules */
+
+/* Master control module */
+struct jpeg_decomp_master {
+  JMETHOD(void, prepare_for_output_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, finish_output_pass, (j_decompress_ptr cinfo));
+
+  /* State variables made visible to other modules */
+  boolean is_dummy_pass;	/* True during 1st pass for 2-pass quant */
+};
+
+/* Input control module */
+struct jpeg_input_controller {
+  JMETHOD(int, consume_input, (j_decompress_ptr cinfo));
+  JMETHOD(void, reset_input_controller, (j_decompress_ptr cinfo));
+  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, finish_input_pass, (j_decompress_ptr cinfo));
+
+  /* State variables made visible to other modules */
+  boolean has_multiple_scans;	/* True if file has multiple scans */
+  boolean eoi_reached;		/* True when EOI has been consumed */
+};
+
+/* Main buffer control (downsampled-data buffer) */
+struct jpeg_d_main_controller {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, process_data, (j_decompress_ptr cinfo,
+			       JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+			       JDIMENSION out_rows_avail));
+};
+
+/* Coefficient buffer control */
+struct jpeg_d_coef_controller {
+  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
+  JMETHOD(int, consume_data, (j_decompress_ptr cinfo));
+  JMETHOD(void, start_output_pass, (j_decompress_ptr cinfo));
+  JMETHOD(int, decompress_data, (j_decompress_ptr cinfo,
+				 JSAMPIMAGE output_buf));
+  /* Pointer to array of coefficient virtual arrays, or NULL if none */
+  jvirt_barray_ptr *coef_arrays;
+};
+
+/* Decompression postprocessing (color quantization buffer control) */
+struct jpeg_d_post_controller {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
+  JMETHOD(void, post_process_data, (j_decompress_ptr cinfo,
+				    JSAMPIMAGE input_buf,
+				    JDIMENSION *in_row_group_ctr,
+				    JDIMENSION in_row_groups_avail,
+				    JSAMPARRAY output_buf,
+				    JDIMENSION *out_row_ctr,
+				    JDIMENSION out_rows_avail));
+};
+
+/* Marker reading & parsing */
+struct jpeg_marker_reader {
+  JMETHOD(void, reset_marker_reader, (j_decompress_ptr cinfo));
+  /* Read markers until SOS or EOI.
+   * Returns same codes as are defined for jpeg_consume_input:
+   * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
+   */
+  JMETHOD(int, read_markers, (j_decompress_ptr cinfo));
+  /* Read a restart marker --- exported for use by entropy decoder only */
+  jpeg_marker_parser_method read_restart_marker;
+
+  /* State of marker reader --- nominally internal, but applications
+   * supplying COM or APPn handlers might like to know the state.
+   */
+  boolean saw_SOI;		/* found SOI? */
+  boolean saw_SOF;		/* found SOF? */
+  int next_restart_num;		/* next restart number expected (0-7) */
+  unsigned int discarded_bytes;	/* # of bytes skipped looking for a marker */
+};
+
+/* Entropy decoding */
+struct jpeg_entropy_decoder {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  JMETHOD(boolean, decode_mcu, (j_decompress_ptr cinfo,
+				JBLOCKROW *MCU_data));
+
+  /* This is here to share code between baseline and progressive decoders; */
+  /* other modules probably should not use it */
+  boolean insufficient_data;	/* set TRUE after emitting warning */
+};
+
+/* Inverse DCT (also performs dequantization) */
+typedef JMETHOD(void, inverse_DCT_method_ptr,
+		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
+		 JCOEFPTR coef_block,
+		 JSAMPARRAY output_buf, JDIMENSION output_col));
+
+struct jpeg_inverse_dct {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  /* It is useful to allow each component to have a separate IDCT method. */
+  inverse_DCT_method_ptr inverse_DCT[MAX_COMPONENTS];
+};
+
+/* Upsampling (note that upsampler must also call color converter) */
+struct jpeg_upsampler {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, upsample, (j_decompress_ptr cinfo,
+			   JSAMPIMAGE input_buf,
+			   JDIMENSION *in_row_group_ctr,
+			   JDIMENSION in_row_groups_avail,
+			   JSAMPARRAY output_buf,
+			   JDIMENSION *out_row_ctr,
+			   JDIMENSION out_rows_avail));
+
+  boolean need_context_rows;	/* TRUE if need rows above & below */
+};
+
+/* Colorspace conversion */
+struct jpeg_color_deconverter {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, color_convert, (j_decompress_ptr cinfo,
+				JSAMPIMAGE input_buf, JDIMENSION input_row,
+				JSAMPARRAY output_buf, int num_rows));
+};
+
+/* Color quantization or color precision reduction */
+struct jpeg_color_quantizer {
+  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, boolean is_pre_scan));
+  JMETHOD(void, color_quantize, (j_decompress_ptr cinfo,
+				 JSAMPARRAY input_buf, JSAMPARRAY output_buf,
+				 int num_rows));
+  JMETHOD(void, finish_pass, (j_decompress_ptr cinfo));
+  JMETHOD(void, new_color_map, (j_decompress_ptr cinfo));
+};
+
+
+/* Miscellaneous useful macros */
+
+#undef MAX
+#define MAX(a,b)	((a) > (b) ? (a) : (b))
+#undef MIN
+#define MIN(a,b)	((a) < (b) ? (a) : (b))
+
+
+/* We assume that right shift corresponds to signed division by 2 with
+ * rounding towards minus infinity.  This is correct for typical "arithmetic
+ * shift" instructions that shift in copies of the sign bit.  But some
+ * C compilers implement >> with an unsigned shift.  For these machines you
+ * must define RIGHT_SHIFT_IS_UNSIGNED.
+ * RIGHT_SHIFT provides a proper signed right shift of an INT32 quantity.
+ * It is only applied with constant shift counts.  SHIFT_TEMPS must be
+ * included in the variables of any routine using RIGHT_SHIFT.
+ */
+
+#ifdef RIGHT_SHIFT_IS_UNSIGNED
+#define SHIFT_TEMPS	INT32 shift_temp;
+#define RIGHT_SHIFT(x,shft)  \
+	((shift_temp = (x)) < 0 ? \
+	 (shift_temp >> (shft)) | ((~((INT32) 0)) << (32-(shft))) : \
+	 (shift_temp >> (shft)))
+#else
+#define SHIFT_TEMPS
+#define RIGHT_SHIFT(x,shft)	((x) >> (shft))
+#endif
+
+
+/* Short forms of external names for systems with brain-damaged linkers. */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jinit_compress_master	jICompress
+#define jinit_c_master_control	jICMaster
+#define jinit_c_main_controller	jICMainC
+#define jinit_c_prep_controller	jICPrepC
+#define jinit_c_coef_controller	jICCoefC
+#define jinit_color_converter	jICColor
+#define jinit_downsampler	jIDownsampler
+#define jinit_forward_dct	jIFDCT
+#define jinit_huff_encoder	jIHEncoder
+#define jinit_phuff_encoder	jIPHEncoder
+#define jinit_marker_writer	jIMWriter
+#define jinit_master_decompress	jIDMaster
+#define jinit_d_main_controller	jIDMainC
+#define jinit_d_coef_controller	jIDCoefC
+#define jinit_d_post_controller	jIDPostC
+#define jinit_input_controller	jIInCtlr
+#define jinit_marker_reader	jIMReader
+#define jinit_huff_decoder	jIHDecoder
+#define jinit_phuff_decoder	jIPHDecoder
+#define jinit_inverse_dct	jIIDCT
+#define jinit_upsampler		jIUpsampler
+#define jinit_color_deconverter	jIDColor
+#define jinit_1pass_quantizer	jI1Quant
+#define jinit_2pass_quantizer	jI2Quant
+#define jinit_merged_upsampler	jIMUpsampler
+#define jinit_memory_mgr	jIMemMgr
+#define jdiv_round_up		jDivRound
+#define jround_up		jRound
+#define jcopy_sample_rows	jCopySamples
+#define jcopy_block_row		jCopyBlocks
+#define jzero_far		jZeroFar
+#define jpeg_zigzag_order	jZIGTable
+#define jpeg_natural_order	jZAGTable
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/* Compression module initialization routines */
+EXTERN(void) jinit_compress_master JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_c_master_control JPP((j_compress_ptr cinfo,
+					 boolean transcode_only));
+EXTERN(void) jinit_c_main_controller JPP((j_compress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_c_prep_controller JPP((j_compress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_c_coef_controller JPP((j_compress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_color_converter JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_downsampler JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_forward_dct JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_huff_encoder JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_phuff_encoder JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_marker_writer JPP((j_compress_ptr cinfo));
+/* Decompression module initialization routines */
+EXTERN(void) jinit_master_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_d_main_controller JPP((j_decompress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_d_coef_controller JPP((j_decompress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_d_post_controller JPP((j_decompress_ptr cinfo,
+					  boolean need_full_buffer));
+EXTERN(void) jinit_input_controller JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_marker_reader JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_huff_decoder JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_phuff_decoder JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_inverse_dct JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_upsampler JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_color_deconverter JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_1pass_quantizer JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_2pass_quantizer JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_merged_upsampler JPP((j_decompress_ptr cinfo));
+/* Memory manager initialization */
+EXTERN(void) jinit_memory_mgr JPP((j_common_ptr cinfo));
+
+/* Utility routines in jutils.c */
+EXTERN(long) jdiv_round_up JPP((long a, long b));
+EXTERN(long) jround_up JPP((long a, long b));
+EXTERN(void) jcopy_sample_rows JPP((JSAMPARRAY input_array, int source_row,
+				    JSAMPARRAY output_array, int dest_row,
+				    int num_rows, JDIMENSION num_cols));
+EXTERN(void) jcopy_block_row JPP((JBLOCKROW input_row, JBLOCKROW output_row,
+				  JDIMENSION num_blocks));
+EXTERN(void) jzero_far JPP((void FAR * target, size_t bytestozero));
+/* Constant tables in jutils.c */
+#if 0				/* This table is not actually needed in v6a */
+extern const int jpeg_zigzag_order[]; /* natural coef order to zigzag order */
+#endif
+extern const int jpeg_natural_order[]; /* zigzag coef order to natural order */
+
+/* Suppress undefined-structure complaints if necessary. */
+
+#ifdef INCOMPLETE_TYPES_BROKEN
+#ifndef AM_MEMORY_MANAGER	/* only jmemmgr.c defines these */
+struct jvirt_sarray_control { long dummy; };
+struct jvirt_barray_control { long dummy; };
+#endif
+#endif /* INCOMPLETE_TYPES_BROKEN */
diff --git a/jpeg/jpeglib.h b/jpeg/jpeglib.h
new file mode 100644
index 000000000..d1be8ddef
--- /dev/null
+++ b/jpeg/jpeglib.h
@@ -0,0 +1,1096 @@
+/*
+ * jpeglib.h
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file defines the application interface for the JPEG library.
+ * Most applications using the library need only include this file,
+ * and perhaps jerror.h if they want to know the exact error codes.
+ */
+
+#ifndef JPEGLIB_H
+#define JPEGLIB_H
+
+/*
+ * First we include the configuration files that record how this
+ * installation of the JPEG library is set up.  jconfig.h can be
+ * generated automatically for many systems.  jmorecfg.h contains
+ * manual configuration options that most people need not worry about.
+ */
+
+#ifndef JCONFIG_INCLUDED	/* in case jinclude.h already did */
+#include "jconfig.h"		/* widely used configuration options */
+#endif
+#include "jmorecfg.h"		/* seldom changed options */
+
+
+/* Version ID for the JPEG library.
+ * Might be useful for tests like "#if JPEG_LIB_VERSION >= 60".
+ */
+
+#define JPEG_LIB_VERSION  62	/* Version 6b */
+
+
+/* Various constants determining the sizes of things.
+ * All of these are specified by the JPEG standard, so don't change them
+ * if you want to be compatible.
+ */
+
+#define DCTSIZE		    8	/* The basic DCT block is 8x8 samples */
+#define DCTSIZE2	    64	/* DCTSIZE squared; # of elements in a block */
+#define NUM_QUANT_TBLS      4	/* Quantization tables are numbered 0..3 */
+#define NUM_HUFF_TBLS       4	/* Huffman tables are numbered 0..3 */
+#define NUM_ARITH_TBLS      16	/* Arith-coding tables are numbered 0..15 */
+#define MAX_COMPS_IN_SCAN   4	/* JPEG limit on # of components in one scan */
+#define MAX_SAMP_FACTOR     4	/* JPEG limit on sampling factors */
+/* Unfortunately, some bozo at Adobe saw no reason to be bound by the standard;
+ * the PostScript DCT filter can emit files with many more than 10 blocks/MCU.
+ * If you happen to run across such a file, you can up D_MAX_BLOCKS_IN_MCU
+ * to handle it.  We even let you do this from the jconfig.h file.  However,
+ * we strongly discourage changing C_MAX_BLOCKS_IN_MCU; just because Adobe
+ * sometimes emits noncompliant files doesn't mean you should too.
+ */
+#define C_MAX_BLOCKS_IN_MCU   10 /* compressor's limit on blocks per MCU */
+#ifndef D_MAX_BLOCKS_IN_MCU
+#define D_MAX_BLOCKS_IN_MCU   10 /* decompressor's limit on blocks per MCU */
+#endif
+
+
+/* Data structures for images (arrays of samples and of DCT coefficients).
+ * On 80x86 machines, the image arrays are too big for near pointers,
+ * but the pointer arrays can fit in near memory.
+ */
+
+typedef JSAMPLE FAR *JSAMPROW;	/* ptr to one image row of pixel samples. */
+typedef JSAMPROW *JSAMPARRAY;	/* ptr to some rows (a 2-D sample array) */
+typedef JSAMPARRAY *JSAMPIMAGE;	/* a 3-D sample array: top index is color */
+
+typedef JCOEF JBLOCK[DCTSIZE2];	/* one block of coefficients */
+typedef JBLOCK FAR *JBLOCKROW;	/* pointer to one row of coefficient blocks */
+typedef JBLOCKROW *JBLOCKARRAY;		/* a 2-D array of coefficient blocks */
+typedef JBLOCKARRAY *JBLOCKIMAGE;	/* a 3-D array of coefficient blocks */
+
+typedef JCOEF FAR *JCOEFPTR;	/* useful in a couple of places */
+
+
+/* Types for JPEG compression parameters and working tables. */
+
+
+/* DCT coefficient quantization tables. */
+
+typedef struct {
+  /* This array gives the coefficient quantizers in natural array order
+   * (not the zigzag order in which they are stored in a JPEG DQT marker).
+   * CAUTION: IJG versions prior to v6a kept this array in zigzag order.
+   */
+  UINT16 quantval[DCTSIZE2];	/* quantization step for each coefficient */
+  /* This field is used only during compression.  It's initialized FALSE when
+   * the table is created, and set TRUE when it's been output to the file.
+   * You could suppress output of a table by setting this to TRUE.
+   * (See jpeg_suppress_tables for an example.)
+   */
+  boolean sent_table;		/* TRUE when table has been output */
+} JQUANT_TBL;
+
+
+/* Huffman coding tables. */
+
+typedef struct {
+  /* These two fields directly represent the contents of a JPEG DHT marker */
+  UINT8 bits[17];		/* bits[k] = # of symbols with codes of */
+				/* length k bits; bits[0] is unused */
+  UINT8 huffval[256];		/* The symbols, in order of incr code length */
+  /* This field is used only during compression.  It's initialized FALSE when
+   * the table is created, and set TRUE when it's been output to the file.
+   * You could suppress output of a table by setting this to TRUE.
+   * (See jpeg_suppress_tables for an example.)
+   */
+  boolean sent_table;		/* TRUE when table has been output */
+} JHUFF_TBL;
+
+
+/* Basic info about one component (color channel). */
+
+typedef struct {
+  /* These values are fixed over the whole image. */
+  /* For compression, they must be supplied by parameter setup; */
+  /* for decompression, they are read from the SOF marker. */
+  int component_id;		/* identifier for this component (0..255) */
+  int component_index;		/* its index in SOF or cinfo->comp_info[] */
+  int h_samp_factor;		/* horizontal sampling factor (1..4) */
+  int v_samp_factor;		/* vertical sampling factor (1..4) */
+  int quant_tbl_no;		/* quantization table selector (0..3) */
+  /* These values may vary between scans. */
+  /* For compression, they must be supplied by parameter setup; */
+  /* for decompression, they are read from the SOS marker. */
+  /* The decompressor output side may not use these variables. */
+  int dc_tbl_no;		/* DC entropy table selector (0..3) */
+  int ac_tbl_no;		/* AC entropy table selector (0..3) */
+  
+  /* Remaining fields should be treated as private by applications. */
+  
+  /* These values are computed during compression or decompression startup: */
+  /* Component's size in DCT blocks.
+   * Any dummy blocks added to complete an MCU are not counted; therefore
+   * these values do not depend on whether a scan is interleaved or not.
+   */
+  JDIMENSION width_in_blocks;
+  JDIMENSION height_in_blocks;
+  /* Size of a DCT block in samples.  Always DCTSIZE for compression.
+   * For decompression this is the size of the output from one DCT block,
+   * reflecting any scaling we choose to apply during the IDCT step.
+   * Values of 1,2,4,8 are likely to be supported.  Note that different
+   * components may receive different IDCT scalings.
+   */
+  int DCT_scaled_size;
+  /* The downsampled dimensions are the component's actual, unpadded number
+   * of samples at the main buffer (preprocessing/compression interface), thus
+   * downsampled_width = ceil(image_width * Hi/Hmax)
+   * and similarly for height.  For decompression, IDCT scaling is included, so
+   * downsampled_width = ceil(image_width * Hi/Hmax * DCT_scaled_size/DCTSIZE)
+   */
+  JDIMENSION downsampled_width;	 /* actual width in samples */
+  JDIMENSION downsampled_height; /* actual height in samples */
+  /* This flag is used only for decompression.  In cases where some of the
+   * components will be ignored (eg grayscale output from YCbCr image),
+   * we can skip most computations for the unused components.
+   */
+  boolean component_needed;	/* do we need the value of this component? */
+
+  /* These values are computed before starting a scan of the component. */
+  /* The decompressor output side may not use these variables. */
+  int MCU_width;		/* number of blocks per MCU, horizontally */
+  int MCU_height;		/* number of blocks per MCU, vertically */
+  int MCU_blocks;		/* MCU_width * MCU_height */
+  int MCU_sample_width;		/* MCU width in samples, MCU_width*DCT_scaled_size */
+  int last_col_width;		/* # of non-dummy blocks across in last MCU */
+  int last_row_height;		/* # of non-dummy blocks down in last MCU */
+
+  /* Saved quantization table for component; NULL if none yet saved.
+   * See jdinput.c comments about the need for this information.
+   * This field is currently used only for decompression.
+   */
+  JQUANT_TBL * quant_table;
+
+  /* Private per-component storage for DCT or IDCT subsystem. */
+  void * dct_table;
+} jpeg_component_info;
+
+
+/* The script for encoding a multiple-scan file is an array of these: */
+
+typedef struct {
+  int comps_in_scan;		/* number of components encoded in this scan */
+  int component_index[MAX_COMPS_IN_SCAN]; /* their SOF/comp_info[] indexes */
+  int Ss, Se;			/* progressive JPEG spectral selection parms */
+  int Ah, Al;			/* progressive JPEG successive approx. parms */
+} jpeg_scan_info;
+
+/* The decompressor can save APPn and COM markers in a list of these: */
+
+typedef struct jpeg_marker_struct FAR * jpeg_saved_marker_ptr;
+
+struct jpeg_marker_struct {
+  jpeg_saved_marker_ptr next;	/* next in list, or NULL */
+  UINT8 marker;			/* marker code: JPEG_COM, or JPEG_APP0+n */
+  unsigned int original_length;	/* # bytes of data in the file */
+  unsigned int data_length;	/* # bytes of data saved at data[] */
+  JOCTET FAR * data;		/* the data contained in the marker */
+  /* the marker length word is not counted in data_length or original_length */
+};
+
+/* Known color spaces. */
+
+typedef enum {
+	JCS_UNKNOWN,		/* error/unspecified */
+	JCS_GRAYSCALE,		/* monochrome */
+	JCS_RGB,		/* red/green/blue */
+	JCS_YCbCr,		/* Y/Cb/Cr (also known as YUV) */
+	JCS_CMYK,		/* C/M/Y/K */
+	JCS_YCCK		/* Y/Cb/Cr/K */
+} J_COLOR_SPACE;
+
+/* DCT/IDCT algorithm options. */
+
+typedef enum {
+	JDCT_ISLOW,		/* slow but accurate integer algorithm */
+	JDCT_IFAST,		/* faster, less accurate integer method */
+	JDCT_FLOAT		/* floating-point: accurate, fast on fast HW */
+} J_DCT_METHOD;
+
+#ifndef JDCT_DEFAULT		/* may be overridden in jconfig.h */
+#define JDCT_DEFAULT  JDCT_ISLOW
+#endif
+#ifndef JDCT_FASTEST		/* may be overridden in jconfig.h */
+#define JDCT_FASTEST  JDCT_IFAST
+#endif
+
+/* Dithering options for decompression. */
+
+typedef enum {
+	JDITHER_NONE,		/* no dithering */
+	JDITHER_ORDERED,	/* simple ordered dither */
+	JDITHER_FS		/* Floyd-Steinberg error diffusion dither */
+} J_DITHER_MODE;
+
+
+/* Common fields between JPEG compression and decompression master structs. */
+
+#define jpeg_common_fields \
+  struct jpeg_error_mgr * err;	/* Error handler module */\
+  struct jpeg_memory_mgr * mem;	/* Memory manager module */\
+  struct jpeg_progress_mgr * progress; /* Progress monitor, or NULL if none */\
+  void * client_data;		/* Available for use by application */\
+  boolean is_decompressor;	/* So common code can tell which is which */\
+  int global_state		/* For checking call sequence validity */
+
+/* Routines that are to be used by both halves of the library are declared
+ * to receive a pointer to this structure.  There are no actual instances of
+ * jpeg_common_struct, only of jpeg_compress_struct and jpeg_decompress_struct.
+ */
+struct jpeg_common_struct {
+  jpeg_common_fields;		/* Fields common to both master struct types */
+  /* Additional fields follow in an actual jpeg_compress_struct or
+   * jpeg_decompress_struct.  All three structs must agree on these
+   * initial fields!  (This would be a lot cleaner in C++.)
+   */
+};
+
+typedef struct jpeg_common_struct * j_common_ptr;
+typedef struct jpeg_compress_struct * j_compress_ptr;
+typedef struct jpeg_decompress_struct * j_decompress_ptr;
+
+
+/* Master record for a compression instance */
+
+struct jpeg_compress_struct {
+  jpeg_common_fields;		/* Fields shared with jpeg_decompress_struct */
+
+  /* Destination for compressed data */
+  struct jpeg_destination_mgr * dest;
+
+  /* Description of source image --- these fields must be filled in by
+   * outer application before starting compression.  in_color_space must
+   * be correct before you can even call jpeg_set_defaults().
+   */
+
+  JDIMENSION image_width;	/* input image width */
+  JDIMENSION image_height;	/* input image height */
+  int input_components;		/* # of color components in input image */
+  J_COLOR_SPACE in_color_space;	/* colorspace of input image */
+
+  double input_gamma;		/* image gamma of input image */
+
+  /* Compression parameters --- these fields must be set before calling
+   * jpeg_start_compress().  We recommend calling jpeg_set_defaults() to
+   * initialize everything to reasonable defaults, then changing anything
+   * the application specifically wants to change.  That way you won't get
+   * burnt when new parameters are added.  Also note that there are several
+   * helper routines to simplify changing parameters.
+   */
+
+  int data_precision;		/* bits of precision in image data */
+
+  int num_components;		/* # of color components in JPEG image */
+  J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
+
+  jpeg_component_info * comp_info;
+  /* comp_info[i] describes component that appears i'th in SOF */
+  
+  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  /* ptrs to coefficient quantization tables, or NULL if not defined */
+  
+  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  /* ptrs to Huffman coding tables, or NULL if not defined */
+  
+  UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
+  UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
+  UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */
+
+  int num_scans;		/* # of entries in scan_info array */
+  const jpeg_scan_info * scan_info; /* script for multi-scan file, or NULL */
+  /* The default value of scan_info is NULL, which causes a single-scan
+   * sequential JPEG file to be emitted.  To create a multi-scan file,
+   * set num_scans and scan_info to point to an array of scan definitions.
+   */
+
+  boolean raw_data_in;		/* TRUE=caller supplies downsampled data */
+  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean optimize_coding;	/* TRUE=optimize entropy encoding parms */
+  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+  int smoothing_factor;		/* 1..100, or 0 for no input smoothing */
+  J_DCT_METHOD dct_method;	/* DCT algorithm selector */
+
+  /* The restart interval can be specified in absolute MCUs by setting
+   * restart_interval, or in MCU rows by setting restart_in_rows
+   * (in which case the correct restart_interval will be figured
+   * for each scan).
+   */
+  unsigned int restart_interval; /* MCUs per restart, or 0 for no restart */
+  int restart_in_rows;		/* if > 0, MCU rows per restart interval */
+
+  /* Parameters controlling emission of special markers. */
+
+  boolean write_JFIF_header;	/* should a JFIF marker be written? */
+  UINT8 JFIF_major_version;	/* What to write for the JFIF version number */
+  UINT8 JFIF_minor_version;
+  /* These three values are not used by the JPEG code, merely copied */
+  /* into the JFIF APP0 marker.  density_unit can be 0 for unknown, */
+  /* 1 for dots/inch, or 2 for dots/cm.  Note that the pixel aspect */
+  /* ratio is defined by X_density/Y_density even when density_unit=0. */
+  UINT8 density_unit;		/* JFIF code for pixel size units */
+  UINT16 X_density;		/* Horizontal pixel density */
+  UINT16 Y_density;		/* Vertical pixel density */
+  boolean write_Adobe_marker;	/* should an Adobe marker be written? */
+  
+  /* State variable: index of next scanline to be written to
+   * jpeg_write_scanlines().  Application may use this to control its
+   * processing loop, e.g., "while (next_scanline < image_height)".
+   */
+
+  JDIMENSION next_scanline;	/* 0 .. image_height-1  */
+
+  /* Remaining fields are known throughout compressor, but generally
+   * should not be touched by a surrounding application.
+   */
+
+  /*
+   * These fields are computed during compression startup
+   */
+  boolean progressive_mode;	/* TRUE if scan script uses progressive mode */
+  int max_h_samp_factor;	/* largest h_samp_factor */
+  int max_v_samp_factor;	/* largest v_samp_factor */
+
+  JDIMENSION total_iMCU_rows;	/* # of iMCU rows to be input to coef ctlr */
+  /* The coefficient controller receives data in units of MCU rows as defined
+   * for fully interleaved scans (whether the JPEG file is interleaved or not).
+   * There are v_samp_factor * DCTSIZE sample rows of each component in an
+   * "iMCU" (interleaved MCU) row.
+   */
+  
+  /*
+   * These fields are valid during any one scan.
+   * They describe the components and MCUs actually appearing in the scan.
+   */
+  int comps_in_scan;		/* # of JPEG components in this scan */
+  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  /* *cur_comp_info[i] describes component that appears i'th in SOS */
+  
+  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
+  
+  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+  int MCU_membership[C_MAX_BLOCKS_IN_MCU];
+  /* MCU_membership[i] is index in cur_comp_info of component owning */
+  /* i'th block in an MCU */
+
+  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+
+  /*
+   * Links to compression subobjects (methods and private variables of modules)
+   */
+  struct jpeg_comp_master * master;
+  struct jpeg_c_main_controller * main;
+  struct jpeg_c_prep_controller * prep;
+  struct jpeg_c_coef_controller * coef;
+  struct jpeg_marker_writer * marker;
+  struct jpeg_color_converter * cconvert;
+  struct jpeg_downsampler * downsample;
+  struct jpeg_forward_dct * fdct;
+  struct jpeg_entropy_encoder * entropy;
+  jpeg_scan_info * script_space; /* workspace for jpeg_simple_progression */
+  int script_space_size;
+};
+
+
+/* Master record for a decompression instance */
+
+struct jpeg_decompress_struct {
+  jpeg_common_fields;		/* Fields shared with jpeg_compress_struct */
+
+  /* Source of compressed data */
+  struct jpeg_source_mgr * src;
+
+  /* Basic description of image --- filled in by jpeg_read_header(). */
+  /* Application may inspect these values to decide how to process image. */
+
+  JDIMENSION image_width;	/* nominal image width (from SOF marker) */
+  JDIMENSION image_height;	/* nominal image height */
+  int num_components;		/* # of color components in JPEG image */
+  J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
+
+  /* Decompression processing parameters --- these fields must be set before
+   * calling jpeg_start_decompress().  Note that jpeg_read_header() initializes
+   * them to default values.
+   */
+
+  J_COLOR_SPACE out_color_space; /* colorspace for output */
+
+  unsigned int scale_num, scale_denom; /* fraction by which to scale image */
+
+  double output_gamma;		/* image gamma wanted in output */
+
+  boolean buffered_image;	/* TRUE=multiple output passes */
+  boolean raw_data_out;		/* TRUE=downsampled data wanted */
+
+  J_DCT_METHOD dct_method;	/* IDCT algorithm selector */
+  boolean do_fancy_upsampling;	/* TRUE=apply fancy upsampling */
+  boolean do_block_smoothing;	/* TRUE=apply interblock smoothing */
+
+  boolean quantize_colors;	/* TRUE=colormapped output wanted */
+  /* the following are ignored if not quantize_colors: */
+  J_DITHER_MODE dither_mode;	/* type of color dithering to use */
+  boolean two_pass_quantize;	/* TRUE=use two-pass color quantization */
+  int desired_number_of_colors;	/* max # colors to use in created colormap */
+  /* these are significant only in buffered-image mode: */
+  boolean enable_1pass_quant;	/* enable future use of 1-pass quantizer */
+  boolean enable_external_quant;/* enable future use of external colormap */
+  boolean enable_2pass_quant;	/* enable future use of 2-pass quantizer */
+
+  /* Description of actual output image that will be returned to application.
+   * These fields are computed by jpeg_start_decompress().
+   * You can also use jpeg_calc_output_dimensions() to determine these values
+   * in advance of calling jpeg_start_decompress().
+   */
+
+  JDIMENSION output_width;	/* scaled image width */
+  JDIMENSION output_height;	/* scaled image height */
+  int out_color_components;	/* # of color components in out_color_space */
+  int output_components;	/* # of color components returned */
+  /* output_components is 1 (a colormap index) when quantizing colors;
+   * otherwise it equals out_color_components.
+   */
+  int rec_outbuf_height;	/* min recommended height of scanline buffer */
+  /* If the buffer passed to jpeg_read_scanlines() is less than this many rows
+   * high, space and time will be wasted due to unnecessary data copying.
+   * Usually rec_outbuf_height will be 1 or 2, at most 4.
+   */
+
+  /* When quantizing colors, the output colormap is described by these fields.
+   * The application can supply a colormap by setting colormap non-NULL before
+   * calling jpeg_start_decompress; otherwise a colormap is created during
+   * jpeg_start_decompress or jpeg_start_output.
+   * The map has out_color_components rows and actual_number_of_colors columns.
+   */
+  int actual_number_of_colors;	/* number of entries in use */
+  JSAMPARRAY colormap;		/* The color map as a 2-D pixel array */
+
+  /* State variables: these variables indicate the progress of decompression.
+   * The application may examine these but must not modify them.
+   */
+
+  /* Row index of next scanline to be read from jpeg_read_scanlines().
+   * Application may use this to control its processing loop, e.g.,
+   * "while (output_scanline < output_height)".
+   */
+  JDIMENSION output_scanline;	/* 0 .. output_height-1  */
+
+  /* Current input scan number and number of iMCU rows completed in scan.
+   * These indicate the progress of the decompressor input side.
+   */
+  int input_scan_number;	/* Number of SOS markers seen so far */
+  JDIMENSION input_iMCU_row;	/* Number of iMCU rows completed */
+
+  /* The "output scan number" is the notional scan being displayed by the
+   * output side.  The decompressor will not allow output scan/row number
+   * to get ahead of input scan/row, but it can fall arbitrarily far behind.
+   */
+  int output_scan_number;	/* Nominal scan number being displayed */
+  JDIMENSION output_iMCU_row;	/* Number of iMCU rows read */
+
+  /* Current progression status.  coef_bits[c][i] indicates the precision
+   * with which component c's DCT coefficient i (in zigzag order) is known.
+   * It is -1 when no data has yet been received, otherwise it is the point
+   * transform (shift) value for the most recent scan of the coefficient
+   * (thus, 0 at completion of the progression).
+   * This pointer is NULL when reading a non-progressive file.
+   */
+  int (*coef_bits)[DCTSIZE2];	/* -1 or current Al value for each coef */
+
+  /* Internal JPEG parameters --- the application usually need not look at
+   * these fields.  Note that the decompressor output side may not use
+   * any parameters that can change between scans.
+   */
+
+  /* Quantization and Huffman tables are carried forward across input
+   * datastreams when processing abbreviated JPEG datastreams.
+   */
+
+  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  /* ptrs to coefficient quantization tables, or NULL if not defined */
+
+  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  /* ptrs to Huffman coding tables, or NULL if not defined */
+
+  /* These parameters are never carried across datastreams, since they
+   * are given in SOF/SOS markers or defined to be reset by SOI.
+   */
+
+  int data_precision;		/* bits of precision in image data */
+
+  jpeg_component_info * comp_info;
+  /* comp_info[i] describes component that appears i'th in SOF */
+
+  boolean progressive_mode;	/* TRUE if SOFn specifies progressive mode */
+  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
+
+  UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
+  UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
+  UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */
+
+  unsigned int restart_interval; /* MCUs per restart interval, or 0 for no restart */
+
+  /* These fields record data obtained from optional markers recognized by
+   * the JPEG library.
+   */
+  boolean saw_JFIF_marker;	/* TRUE iff a JFIF APP0 marker was found */
+  /* Data copied from JFIF marker; only valid if saw_JFIF_marker is TRUE: */
+  UINT8 JFIF_major_version;	/* JFIF version number */
+  UINT8 JFIF_minor_version;
+  UINT8 density_unit;		/* JFIF code for pixel size units */
+  UINT16 X_density;		/* Horizontal pixel density */
+  UINT16 Y_density;		/* Vertical pixel density */
+  boolean saw_Adobe_marker;	/* TRUE iff an Adobe APP14 marker was found */
+  UINT8 Adobe_transform;	/* Color transform code from Adobe marker */
+
+  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+
+  /* Aside from the specific data retained from APPn markers known to the
+   * library, the uninterpreted contents of any or all APPn and COM markers
+   * can be saved in a list for examination by the application.
+   */
+  jpeg_saved_marker_ptr marker_list; /* Head of list of saved markers */
+
+  /* Remaining fields are known throughout decompressor, but generally
+   * should not be touched by a surrounding application.
+   */
+
+  /*
+   * These fields are computed during decompression startup
+   */
+  int max_h_samp_factor;	/* largest h_samp_factor */
+  int max_v_samp_factor;	/* largest v_samp_factor */
+
+  int min_DCT_scaled_size;	/* smallest DCT_scaled_size of any component */
+
+  JDIMENSION total_iMCU_rows;	/* # of iMCU rows in image */
+  /* The coefficient controller's input and output progress is measured in
+   * units of "iMCU" (interleaved MCU) rows.  These are the same as MCU rows
+   * in fully interleaved JPEG scans, but are used whether the scan is
+   * interleaved or not.  We define an iMCU row as v_samp_factor DCT block
+   * rows of each component.  Therefore, the IDCT output contains
+   * v_samp_factor*DCT_scaled_size sample rows of a component per iMCU row.
+   */
+
+  JSAMPLE * sample_range_limit; /* table for fast range-limiting */
+
+  /*
+   * These fields are valid during any one scan.
+   * They describe the components and MCUs actually appearing in the scan.
+   * Note that the decompressor output side must not use these fields.
+   */
+  int comps_in_scan;		/* # of JPEG components in this scan */
+  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  /* *cur_comp_info[i] describes component that appears i'th in SOS */
+
+  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
+
+  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+  int MCU_membership[D_MAX_BLOCKS_IN_MCU];
+  /* MCU_membership[i] is index in cur_comp_info of component owning */
+  /* i'th block in an MCU */
+
+  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+
+  /* This field is shared between entropy decoder and marker parser.
+   * It is either zero or the code of a JPEG marker that has been
+   * read from the data source, but has not yet been processed.
+   */
+  int unread_marker;
+
+  /*
+   * Links to decompression subobjects (methods, private variables of modules)
+   */
+  struct jpeg_decomp_master * master;
+  struct jpeg_d_main_controller * main;
+  struct jpeg_d_coef_controller * coef;
+  struct jpeg_d_post_controller * post;
+  struct jpeg_input_controller * inputctl;
+  struct jpeg_marker_reader * marker;
+  struct jpeg_entropy_decoder * entropy;
+  struct jpeg_inverse_dct * idct;
+  struct jpeg_upsampler * upsample;
+  struct jpeg_color_deconverter * cconvert;
+  struct jpeg_color_quantizer * cquantize;
+};
+
+
+/* "Object" declarations for JPEG modules that may be supplied or called
+ * directly by the surrounding application.
+ * As with all objects in the JPEG library, these structs only define the
+ * publicly visible methods and state variables of a module.  Additional
+ * private fields may exist after the public ones.
+ */
+
+
+/* Error handler object */
+
+struct jpeg_error_mgr {
+  /* Error exit handler: does not return to caller */
+  JMETHOD(void, error_exit, (j_common_ptr cinfo));
+  /* Conditionally emit a trace or warning message */
+  JMETHOD(void, emit_message, (j_common_ptr cinfo, int msg_level));
+  /* Routine that actually outputs a trace or error message */
+  JMETHOD(void, output_message, (j_common_ptr cinfo));
+  /* Format a message string for the most recent JPEG error or message */
+  JMETHOD(void, format_message, (j_common_ptr cinfo, char * buffer));
+#define JMSG_LENGTH_MAX  200	/* recommended size of format_message buffer */
+  /* Reset error state variables at start of a new image */
+  JMETHOD(void, reset_error_mgr, (j_common_ptr cinfo));
+  
+  /* The message ID code and any parameters are saved here.
+   * A message can have one string parameter or up to 8 int parameters.
+   */
+  int msg_code;
+#define JMSG_STR_PARM_MAX  80
+  union {
+    int i[8];
+    char s[JMSG_STR_PARM_MAX];
+  } msg_parm;
+  
+  /* Standard state variables for error facility */
+  
+  int trace_level;		/* max msg_level that will be displayed */
+  
+  /* For recoverable corrupt-data errors, we emit a warning message,
+   * but keep going unless emit_message chooses to abort.  emit_message
+   * should count warnings in num_warnings.  The surrounding application
+   * can check for bad data by seeing if num_warnings is nonzero at the
+   * end of processing.
+   */
+  long num_warnings;		/* number of corrupt-data warnings */
+
+  /* These fields point to the table(s) of error message strings.
+   * An application can change the table pointer to switch to a different
+   * message list (typically, to change the language in which errors are
+   * reported).  Some applications may wish to add additional error codes
+   * that will be handled by the JPEG library error mechanism; the second
+   * table pointer is used for this purpose.
+   *
+   * First table includes all errors generated by JPEG library itself.
+   * Error code 0 is reserved for a "no such error string" message.
+   */
+  const char * const * jpeg_message_table; /* Library errors */
+  int last_jpeg_message;    /* Table contains strings 0..last_jpeg_message */
+  /* Second table can be added by application (see cjpeg/djpeg for example).
+   * It contains strings numbered first_addon_message..last_addon_message.
+   */
+  const char * const * addon_message_table; /* Non-library errors */
+  int first_addon_message;	/* code for first string in addon table */
+  int last_addon_message;	/* code for last string in addon table */
+};
+
+
+/* Progress monitor object */
+
+struct jpeg_progress_mgr {
+  JMETHOD(void, progress_monitor, (j_common_ptr cinfo));
+
+  long pass_counter;		/* work units completed in this pass */
+  long pass_limit;		/* total number of work units in this pass */
+  int completed_passes;		/* passes completed so far */
+  int total_passes;		/* total number of passes expected */
+};
+
+
+/* Data destination object for compression */
+
+struct jpeg_destination_mgr {
+  JOCTET * next_output_byte;	/* => next byte to write in buffer */
+  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+
+  JMETHOD(void, init_destination, (j_compress_ptr cinfo));
+  JMETHOD(boolean, empty_output_buffer, (j_compress_ptr cinfo));
+  JMETHOD(void, term_destination, (j_compress_ptr cinfo));
+};
+
+
+/* Data source object for decompression */
+
+struct jpeg_source_mgr {
+  const JOCTET * next_input_byte; /* => next byte to read from buffer */
+  size_t bytes_in_buffer;	/* # of bytes remaining in buffer */
+
+  JMETHOD(void, init_source, (j_decompress_ptr cinfo));
+  JMETHOD(boolean, fill_input_buffer, (j_decompress_ptr cinfo));
+  JMETHOD(void, skip_input_data, (j_decompress_ptr cinfo, long num_bytes));
+  JMETHOD(boolean, resync_to_restart, (j_decompress_ptr cinfo, int desired));
+  JMETHOD(void, term_source, (j_decompress_ptr cinfo));
+};
+
+
+/* Memory manager object.
+ * Allocates "small" objects (a few K total), "large" objects (tens of K),
+ * and "really big" objects (virtual arrays with backing store if needed).
+ * The memory manager does not allow individual objects to be freed; rather,
+ * each created object is assigned to a pool, and whole pools can be freed
+ * at once.  This is faster and more convenient than remembering exactly what
+ * to free, especially where malloc()/free() are not too speedy.
+ * NB: alloc routines never return NULL.  They exit to error_exit if not
+ * successful.
+ */
+
+#define JPOOL_PERMANENT	0	/* lasts until master record is destroyed */
+#define JPOOL_IMAGE	1	/* lasts until done with image/datastream */
+#define JPOOL_NUMPOOLS	2
+
+typedef struct jvirt_sarray_control * jvirt_sarray_ptr;
+typedef struct jvirt_barray_control * jvirt_barray_ptr;
+
+
+struct jpeg_memory_mgr {
+  /* Method pointers */
+  JMETHOD(void *, alloc_small, (j_common_ptr cinfo, int pool_id,
+				size_t sizeofobject));
+  JMETHOD(void FAR *, alloc_large, (j_common_ptr cinfo, int pool_id,
+				     size_t sizeofobject));
+  JMETHOD(JSAMPARRAY, alloc_sarray, (j_common_ptr cinfo, int pool_id,
+				     JDIMENSION samplesperrow,
+				     JDIMENSION numrows));
+  JMETHOD(JBLOCKARRAY, alloc_barray, (j_common_ptr cinfo, int pool_id,
+				      JDIMENSION blocksperrow,
+				      JDIMENSION numrows));
+  JMETHOD(jvirt_sarray_ptr, request_virt_sarray, (j_common_ptr cinfo,
+						  int pool_id,
+						  boolean pre_zero,
+						  JDIMENSION samplesperrow,
+						  JDIMENSION numrows,
+						  JDIMENSION maxaccess));
+  JMETHOD(jvirt_barray_ptr, request_virt_barray, (j_common_ptr cinfo,
+						  int pool_id,
+						  boolean pre_zero,
+						  JDIMENSION blocksperrow,
+						  JDIMENSION numrows,
+						  JDIMENSION maxaccess));
+  JMETHOD(void, realize_virt_arrays, (j_common_ptr cinfo));
+  JMETHOD(JSAMPARRAY, access_virt_sarray, (j_common_ptr cinfo,
+					   jvirt_sarray_ptr ptr,
+					   JDIMENSION start_row,
+					   JDIMENSION num_rows,
+					   boolean writable));
+  JMETHOD(JBLOCKARRAY, access_virt_barray, (j_common_ptr cinfo,
+					    jvirt_barray_ptr ptr,
+					    JDIMENSION start_row,
+					    JDIMENSION num_rows,
+					    boolean writable));
+  JMETHOD(void, free_pool, (j_common_ptr cinfo, int pool_id));
+  JMETHOD(void, self_destruct, (j_common_ptr cinfo));
+
+  /* Limit on memory allocation for this JPEG object.  (Note that this is
+   * merely advisory, not a guaranteed maximum; it only affects the space
+   * used for virtual-array buffers.)  May be changed by outer application
+   * after creating the JPEG object.
+   */
+  long max_memory_to_use;
+
+  /* Maximum allocation request accepted by alloc_large. */
+  long max_alloc_chunk;
+};
+
+
+/* Routine signature for application-supplied marker processing methods.
+ * Need not pass marker code since it is stored in cinfo->unread_marker.
+ */
+typedef JMETHOD(boolean, jpeg_marker_parser_method, (j_decompress_ptr cinfo));
+
+
+/* Declarations for routines called by application.
+ * The JPP macro hides prototype parameters from compilers that can't cope.
+ * Note JPP requires double parentheses.
+ */
+
+#ifdef HAVE_PROTOTYPES
+#define JPP(arglist)	arglist
+#else
+#define JPP(arglist)	()
+#endif
+
+
+/* Short forms of external names for systems with brain-damaged linkers.
+ * We shorten external names to be unique in the first six letters, which
+ * is good enough for all known systems.
+ * (If your compiler itself needs names to be unique in less than 15 
+ * characters, you are out of luck.  Get a better compiler.)
+ */
+
+#ifdef NEED_SHORT_EXTERNAL_NAMES
+#define jpeg_std_error		jStdError
+#define jpeg_CreateCompress	jCreaCompress
+#define jpeg_CreateDecompress	jCreaDecompress
+#define jpeg_destroy_compress	jDestCompress
+#define jpeg_destroy_decompress	jDestDecompress
+#define jpeg_stdio_dest		jStdDest
+#define jpeg_stdio_src		jStdSrc
+#define jpeg_set_defaults	jSetDefaults
+#define jpeg_set_colorspace	jSetColorspace
+#define jpeg_default_colorspace	jDefColorspace
+#define jpeg_set_quality	jSetQuality
+#define jpeg_set_linear_quality	jSetLQuality
+#define jpeg_add_quant_table	jAddQuantTable
+#define jpeg_quality_scaling	jQualityScaling
+#define jpeg_simple_progression	jSimProgress
+#define jpeg_suppress_tables	jSuppressTables
+#define jpeg_alloc_quant_table	jAlcQTable
+#define jpeg_alloc_huff_table	jAlcHTable
+#define jpeg_start_compress	jStrtCompress
+#define jpeg_write_scanlines	jWrtScanlines
+#define jpeg_finish_compress	jFinCompress
+#define jpeg_write_raw_data	jWrtRawData
+#define jpeg_write_marker	jWrtMarker
+#define jpeg_write_m_header	jWrtMHeader
+#define jpeg_write_m_byte	jWrtMByte
+#define jpeg_write_tables	jWrtTables
+#define jpeg_read_header	jReadHeader
+#define jpeg_start_decompress	jStrtDecompress
+#define jpeg_read_scanlines	jReadScanlines
+#define jpeg_finish_decompress	jFinDecompress
+#define jpeg_read_raw_data	jReadRawData
+#define jpeg_has_multiple_scans	jHasMultScn
+#define jpeg_start_output	jStrtOutput
+#define jpeg_finish_output	jFinOutput
+#define jpeg_input_complete	jInComplete
+#define jpeg_new_colormap	jNewCMap
+#define jpeg_consume_input	jConsumeInput
+#define jpeg_calc_output_dimensions	jCalcDimensions
+#define jpeg_save_markers	jSaveMarkers
+#define jpeg_set_marker_processor	jSetMarker
+#define jpeg_read_coefficients	jReadCoefs
+#define jpeg_write_coefficients	jWrtCoefs
+#define jpeg_copy_critical_parameters	jCopyCrit
+#define jpeg_abort_compress	jAbrtCompress
+#define jpeg_abort_decompress	jAbrtDecompress
+#define jpeg_abort		jAbort
+#define jpeg_destroy		jDestroy
+#define jpeg_resync_to_restart	jResyncRestart
+#endif /* NEED_SHORT_EXTERNAL_NAMES */
+
+
+/* Default error-management setup */
+EXTERN(struct jpeg_error_mgr *) jpeg_std_error
+	JPP((struct jpeg_error_mgr * err));
+
+/* Initialization of JPEG compression objects.
+ * jpeg_create_compress() and jpeg_create_decompress() are the exported
+ * names that applications should call.  These expand to calls on
+ * jpeg_CreateCompress and jpeg_CreateDecompress with additional information
+ * passed for version mismatch checking.
+ * NB: you must set up the error-manager BEFORE calling jpeg_create_xxx.
+ */
+#define jpeg_create_compress(cinfo) \
+    jpeg_CreateCompress((cinfo), JPEG_LIB_VERSION, \
+			(size_t) sizeof(struct jpeg_compress_struct))
+#define jpeg_create_decompress(cinfo) \
+    jpeg_CreateDecompress((cinfo), JPEG_LIB_VERSION, \
+			  (size_t) sizeof(struct jpeg_decompress_struct))
+EXTERN(void) jpeg_CreateCompress JPP((j_compress_ptr cinfo,
+				      int version, size_t structsize));
+EXTERN(void) jpeg_CreateDecompress JPP((j_decompress_ptr cinfo,
+					int version, size_t structsize));
+/* Destruction of JPEG compression objects */
+EXTERN(void) jpeg_destroy_compress JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_destroy_decompress JPP((j_decompress_ptr cinfo));
+
+/* Standard data source and destination managers: stdio streams. */
+/* Caller is responsible for opening the file before and closing after. */
+EXTERN(void) jpeg_stdio_dest JPP((j_compress_ptr cinfo, FILE * outfile));
+EXTERN(void) jpeg_stdio_src JPP((j_decompress_ptr cinfo, FILE * infile));
+
+/* Default parameter setup for compression */
+EXTERN(void) jpeg_set_defaults JPP((j_compress_ptr cinfo));
+/* Compression parameter setup aids */
+EXTERN(void) jpeg_set_colorspace JPP((j_compress_ptr cinfo,
+				      J_COLOR_SPACE colorspace));
+EXTERN(void) jpeg_default_colorspace JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_set_quality JPP((j_compress_ptr cinfo, int quality,
+				   boolean force_baseline));
+EXTERN(void) jpeg_set_linear_quality JPP((j_compress_ptr cinfo,
+					  int scale_factor,
+					  boolean force_baseline));
+EXTERN(void) jpeg_add_quant_table JPP((j_compress_ptr cinfo, int which_tbl,
+				       const unsigned int *basic_table,
+				       int scale_factor,
+				       boolean force_baseline));
+EXTERN(int) jpeg_quality_scaling JPP((int quality));
+EXTERN(void) jpeg_simple_progression JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_suppress_tables JPP((j_compress_ptr cinfo,
+				       boolean suppress));
+EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table JPP((j_common_ptr cinfo));
+EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table JPP((j_common_ptr cinfo));
+
+/* Main entry points for compression */
+EXTERN(void) jpeg_start_compress JPP((j_compress_ptr cinfo,
+				      boolean write_all_tables));
+EXTERN(JDIMENSION) jpeg_write_scanlines JPP((j_compress_ptr cinfo,
+					     JSAMPARRAY scanlines,
+					     JDIMENSION num_lines));
+EXTERN(void) jpeg_finish_compress JPP((j_compress_ptr cinfo));
+
+/* Replaces jpeg_write_scanlines when writing raw downsampled data. */
+EXTERN(JDIMENSION) jpeg_write_raw_data JPP((j_compress_ptr cinfo,
+					    JSAMPIMAGE data,
+					    JDIMENSION num_lines));
+
+/* Write a special marker.  See libjpeg.doc concerning safe usage. */
+EXTERN(void) jpeg_write_marker
+	JPP((j_compress_ptr cinfo, int marker,
+	     const JOCTET * dataptr, unsigned int datalen));
+/* Same, but piecemeal. */
+EXTERN(void) jpeg_write_m_header
+	JPP((j_compress_ptr cinfo, int marker, unsigned int datalen));
+EXTERN(void) jpeg_write_m_byte
+	JPP((j_compress_ptr cinfo, int val));
+
+/* Alternate compression function: just write an abbreviated table file */
+EXTERN(void) jpeg_write_tables JPP((j_compress_ptr cinfo));
+
+/* Decompression startup: read start of JPEG datastream to see what's there */
+EXTERN(int) jpeg_read_header JPP((j_decompress_ptr cinfo,
+				  boolean require_image));
+/* Return value is one of: */
+#define JPEG_SUSPENDED		0 /* Suspended due to lack of input data */
+#define JPEG_HEADER_OK		1 /* Found valid image datastream */
+#define JPEG_HEADER_TABLES_ONLY	2 /* Found valid table-specs-only datastream */
+/* If you pass require_image = TRUE (normal case), you need not check for
+ * a TABLES_ONLY return code; an abbreviated file will cause an error exit.
+ * JPEG_SUSPENDED is only possible if you use a data source module that can
+ * give a suspension return (the stdio source module doesn't).
+ */
+
+/* Main entry points for decompression */
+EXTERN(boolean) jpeg_start_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(JDIMENSION) jpeg_read_scanlines JPP((j_decompress_ptr cinfo,
+					    JSAMPARRAY scanlines,
+					    JDIMENSION max_lines));
+EXTERN(boolean) jpeg_finish_decompress JPP((j_decompress_ptr cinfo));
+
+/* Replaces jpeg_read_scanlines when reading raw downsampled data. */
+EXTERN(JDIMENSION) jpeg_read_raw_data JPP((j_decompress_ptr cinfo,
+					   JSAMPIMAGE data,
+					   JDIMENSION max_lines));
+
+/* Additional entry points for buffered-image mode. */
+EXTERN(boolean) jpeg_has_multiple_scans JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_start_output JPP((j_decompress_ptr cinfo,
+				       int scan_number));
+EXTERN(boolean) jpeg_finish_output JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_input_complete JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_new_colormap JPP((j_decompress_ptr cinfo));
+EXTERN(int) jpeg_consume_input JPP((j_decompress_ptr cinfo));
+/* Return value is one of: */
+/* #define JPEG_SUSPENDED	0    Suspended due to lack of input data */
+#define JPEG_REACHED_SOS	1 /* Reached start of new scan */
+#define JPEG_REACHED_EOI	2 /* Reached end of image */
+#define JPEG_ROW_COMPLETED	3 /* Completed one iMCU row */
+#define JPEG_SCAN_COMPLETED	4 /* Completed last iMCU row of a scan */
+
+/* Precalculate output dimensions for current decompression parameters. */
+EXTERN(void) jpeg_calc_output_dimensions JPP((j_decompress_ptr cinfo));
+
+/* Control saving of COM and APPn markers into marker_list. */
+EXTERN(void) jpeg_save_markers
+	JPP((j_decompress_ptr cinfo, int marker_code,
+	     unsigned int length_limit));
+
+/* Install a special processing method for COM or APPn markers. */
+EXTERN(void) jpeg_set_marker_processor
+	JPP((j_decompress_ptr cinfo, int marker_code,
+	     jpeg_marker_parser_method routine));
+
+/* Read or write raw DCT coefficients --- useful for lossless transcoding. */
+EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_write_coefficients JPP((j_compress_ptr cinfo,
+					  jvirt_barray_ptr * coef_arrays));
+EXTERN(void) jpeg_copy_critical_parameters JPP((j_decompress_ptr srcinfo,
+						j_compress_ptr dstinfo));
+
+/* If you choose to abort compression or decompression before completing
+ * jpeg_finish_(de)compress, then you need to clean up to release memory,
+ * temporary files, etc.  You can just call jpeg_destroy_(de)compress
+ * if you're done with the JPEG object, but if you want to clean it up and
+ * reuse it, call this:
+ */
+EXTERN(void) jpeg_abort_compress JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_abort_decompress JPP((j_decompress_ptr cinfo));
+
+/* Generic versions of jpeg_abort and jpeg_destroy that work on either
+ * flavor of JPEG object.  These may be more convenient in some places.
+ */
+EXTERN(void) jpeg_abort JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_destroy JPP((j_common_ptr cinfo));
+
+/* Default restart-marker-resync procedure for use by data source modules */
+EXTERN(boolean) jpeg_resync_to_restart JPP((j_decompress_ptr cinfo,
+					    int desired));
+
+
+/* These marker codes are exported since applications and data source modules
+ * are likely to want to use them.
+ */
+
+#define JPEG_RST0	0xD0	/* RST0 marker code */
+#define JPEG_EOI	0xD9	/* EOI marker code */
+#define JPEG_APP0	0xE0	/* APP0 marker code */
+#define JPEG_COM	0xFE	/* COM marker code */
+
+
+/* If we have a brain-damaged compiler that emits warnings (or worse, errors)
+ * for structure definitions that are never filled in, keep it quiet by
+ * supplying dummy definitions for the various substructures.
+ */
+
+#ifdef INCOMPLETE_TYPES_BROKEN
+#ifndef JPEG_INTERNALS		/* will be defined in jpegint.h */
+struct jvirt_sarray_control { long dummy; };
+struct jvirt_barray_control { long dummy; };
+struct jpeg_comp_master { long dummy; };
+struct jpeg_c_main_controller { long dummy; };
+struct jpeg_c_prep_controller { long dummy; };
+struct jpeg_c_coef_controller { long dummy; };
+struct jpeg_marker_writer { long dummy; };
+struct jpeg_color_converter { long dummy; };
+struct jpeg_downsampler { long dummy; };
+struct jpeg_forward_dct { long dummy; };
+struct jpeg_entropy_encoder { long dummy; };
+struct jpeg_decomp_master { long dummy; };
+struct jpeg_d_main_controller { long dummy; };
+struct jpeg_d_coef_controller { long dummy; };
+struct jpeg_d_post_controller { long dummy; };
+struct jpeg_input_controller { long dummy; };
+struct jpeg_marker_reader { long dummy; };
+struct jpeg_entropy_decoder { long dummy; };
+struct jpeg_inverse_dct { long dummy; };
+struct jpeg_upsampler { long dummy; };
+struct jpeg_color_deconverter { long dummy; };
+struct jpeg_color_quantizer { long dummy; };
+#endif /* JPEG_INTERNALS */
+#endif /* INCOMPLETE_TYPES_BROKEN */
+
+
+/*
+ * The JPEG library modules define JPEG_INTERNALS before including this file.
+ * The internal structure declarations are read only when that is true.
+ * Applications using the library should not include jpegint.h, but may wish
+ * to include jerror.h.
+ */
+
+#ifdef JPEG_INTERNALS
+#include "jpegint.h"		/* fetch private declarations */
+#include "jerror.h"		/* fetch error codes too */
+#endif
+
+#endif /* JPEGLIB_H */
diff --git a/jpeg/jquant1.c b/jpeg/jquant1.c
new file mode 100644
index 000000000..b2f96aa15
--- /dev/null
+++ b/jpeg/jquant1.c
@@ -0,0 +1,856 @@
+/*
+ * jquant1.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains 1-pass color quantization (color mapping) routines.
+ * These routines provide mapping to a fixed color map using equally spaced
+ * color values.  Optional Floyd-Steinberg or ordered dithering is available.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef QUANT_1PASS_SUPPORTED
+
+
+/*
+ * The main purpose of 1-pass quantization is to provide a fast, if not very
+ * high quality, colormapped output capability.  A 2-pass quantizer usually
+ * gives better visual quality; however, for quantized grayscale output this
+ * quantizer is perfectly adequate.  Dithering is highly recommended with this
+ * quantizer, though you can turn it off if you really want to.
+ *
+ * In 1-pass quantization the colormap must be chosen in advance of seeing the
+ * image.  We use a map consisting of all combinations of Ncolors[i] color
+ * values for the i'th component.  The Ncolors[] values are chosen so that
+ * their product, the total number of colors, is no more than that requested.
+ * (In most cases, the product will be somewhat less.)
+ *
+ * Since the colormap is orthogonal, the representative value for each color
+ * component can be determined without considering the other components;
+ * then these indexes can be combined into a colormap index by a standard
+ * N-dimensional-array-subscript calculation.  Most of the arithmetic involved
+ * can be precalculated and stored in the lookup table colorindex[].
+ * colorindex[i][j] maps pixel value j in component i to the nearest
+ * representative value (grid plane) for that component; this index is
+ * multiplied by the array stride for component i, so that the
+ * index of the colormap entry closest to a given pixel value is just
+ *    sum( colorindex[component-number][pixel-component-value] )
+ * Aside from being fast, this scheme allows for variable spacing between
+ * representative values with no additional lookup cost.
+ *
+ * If gamma correction has been applied in color conversion, it might be wise
+ * to adjust the color grid spacing so that the representative colors are
+ * equidistant in linear space.  At this writing, gamma correction is not
+ * implemented by jdcolor, so nothing is done here.
+ */
+
+
+/* Declarations for ordered dithering.
+ *
+ * We use a standard 16x16 ordered dither array.  The basic concept of ordered
+ * dithering is described in many references, for instance Dale Schumacher's
+ * chapter II.2 of Graphics Gems II (James Arvo, ed. Academic Press, 1991).
+ * In place of Schumacher's comparisons against a "threshold" value, we add a
+ * "dither" value to the input pixel and then round the result to the nearest
+ * output value.  The dither value is equivalent to (0.5 - threshold) times
+ * the distance between output values.  For ordered dithering, we assume that
+ * the output colors are equally spaced; if not, results will probably be
+ * worse, since the dither may be too much or too little at a given point.
+ *
+ * The normal calculation would be to form pixel value + dither, range-limit
+ * this to 0..MAXJSAMPLE, and then index into the colorindex table as usual.
+ * We can skip the separate range-limiting step by extending the colorindex
+ * table in both directions.
+ */
+
+#define ODITHER_SIZE  16	/* dimension of dither matrix */
+/* NB: if ODITHER_SIZE is not a power of 2, ODITHER_MASK uses will break */
+#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE)	/* # cells in matrix */
+#define ODITHER_MASK  (ODITHER_SIZE-1) /* mask for wrapping around counters */
+
+typedef int ODITHER_MATRIX[ODITHER_SIZE][ODITHER_SIZE];
+typedef int (*ODITHER_MATRIX_PTR)[ODITHER_SIZE];
+
+static const UINT8 base_dither_matrix[ODITHER_SIZE][ODITHER_SIZE] = {
+  /* Bayer's order-4 dither array.  Generated by the code given in
+   * Stephen Hawley's article "Ordered Dithering" in Graphics Gems I.
+   * The values in this array must range from 0 to ODITHER_CELLS-1.
+   */
+  {   0,192, 48,240, 12,204, 60,252,  3,195, 51,243, 15,207, 63,255 },
+  { 128, 64,176,112,140, 76,188,124,131, 67,179,115,143, 79,191,127 },
+  {  32,224, 16,208, 44,236, 28,220, 35,227, 19,211, 47,239, 31,223 },
+  { 160, 96,144, 80,172,108,156, 92,163, 99,147, 83,175,111,159, 95 },
+  {   8,200, 56,248,  4,196, 52,244, 11,203, 59,251,  7,199, 55,247 },
+  { 136, 72,184,120,132, 68,180,116,139, 75,187,123,135, 71,183,119 },
+  {  40,232, 24,216, 36,228, 20,212, 43,235, 27,219, 39,231, 23,215 },
+  { 168,104,152, 88,164,100,148, 84,171,107,155, 91,167,103,151, 87 },
+  {   2,194, 50,242, 14,206, 62,254,  1,193, 49,241, 13,205, 61,253 },
+  { 130, 66,178,114,142, 78,190,126,129, 65,177,113,141, 77,189,125 },
+  {  34,226, 18,210, 46,238, 30,222, 33,225, 17,209, 45,237, 29,221 },
+  { 162, 98,146, 82,174,110,158, 94,161, 97,145, 81,173,109,157, 93 },
+  {  10,202, 58,250,  6,198, 54,246,  9,201, 57,249,  5,197, 53,245 },
+  { 138, 74,186,122,134, 70,182,118,137, 73,185,121,133, 69,181,117 },
+  {  42,234, 26,218, 38,230, 22,214, 41,233, 25,217, 37,229, 21,213 },
+  { 170,106,154, 90,166,102,150, 86,169,105,153, 89,165,101,149, 85 }
+};
+
+
+/* Declarations for Floyd-Steinberg dithering.
+ *
+ * Errors are accumulated into the array fserrors[], at a resolution of
+ * 1/16th of a pixel count.  The error at a given pixel is propagated
+ * to its not-yet-processed neighbors using the standard F-S fractions,
+ *		...	(here)	7/16
+ *		3/16	5/16	1/16
+ * We work left-to-right on even rows, right-to-left on odd rows.
+ *
+ * We can get away with a single array (holding one row's worth of errors)
+ * by using it to store the current row's errors at pixel columns not yet
+ * processed, but the next row's errors at columns already processed.  We
+ * need only a few extra variables to hold the errors immediately around the
+ * current column.  (If we are lucky, those variables are in registers, but
+ * even if not, they're probably cheaper to access than array elements are.)
+ *
+ * The fserrors[] array is indexed [component#][position].
+ * We provide (#columns + 2) entries per component; the extra entry at each
+ * end saves us from special-casing the first and last pixels.
+ *
+ * Note: on a wide image, we might not have enough room in a PC's near data
+ * segment to hold the error array; so it is allocated with alloc_large.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef INT16 FSERROR;		/* 16 bits should be enough */
+typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+#else
+typedef INT32 FSERROR;		/* may need more than 16 bits */
+typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+#endif
+
+typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+
+
+/* Private subobject */
+
+#define MAX_Q_COMPS 4		/* max components I can handle */
+
+typedef struct {
+  struct jpeg_color_quantizer pub; /* public fields */
+
+  /* Initially allocated colormap is saved here */
+  JSAMPARRAY sv_colormap;	/* The color map as a 2-D pixel array */
+  int sv_actual;		/* number of entries in use */
+
+  JSAMPARRAY colorindex;	/* Precomputed mapping for speed */
+  /* colorindex[i][j] = index of color closest to pixel value j in component i,
+   * premultiplied as described above.  Since colormap indexes must fit into
+   * JSAMPLEs, the entries of this array will too.
+   */
+  boolean is_padded;		/* is the colorindex padded for odither? */
+
+  int Ncolors[MAX_Q_COMPS];	/* # of values alloced to each component */
+
+  /* Variables for ordered dithering */
+  int row_index;		/* cur row's vertical index in dither matrix */
+  ODITHER_MATRIX_PTR odither[MAX_Q_COMPS]; /* one dither array per component */
+
+  /* Variables for Floyd-Steinberg dithering */
+  FSERRPTR fserrors[MAX_Q_COMPS]; /* accumulated errors */
+  boolean on_odd_row;		/* flag to remember which row we are on */
+} my_cquantizer;
+
+typedef my_cquantizer * my_cquantize_ptr;
+
+
+/*
+ * Policy-making subroutines for create_colormap and create_colorindex.
+ * These routines determine the colormap to be used.  The rest of the module
+ * only assumes that the colormap is orthogonal.
+ *
+ *  * select_ncolors decides how to divvy up the available colors
+ *    among the components.
+ *  * output_value defines the set of representative values for a component.
+ *  * largest_input_value defines the mapping from input values to
+ *    representative values for a component.
+ * Note that the latter two routines may impose different policies for
+ * different components, though this is not currently done.
+ */
+
+
+LOCAL(int)
+select_ncolors (j_decompress_ptr cinfo, int Ncolors[])
+/* Determine allocation of desired colors to components, */
+/* and fill in Ncolors[] array to indicate choice. */
+/* Return value is total number of colors (product of Ncolors[] values). */
+{
+  int nc = cinfo->out_color_components; /* number of color components */
+  int max_colors = cinfo->desired_number_of_colors;
+  int total_colors, iroot, i, j;
+  boolean changed;
+  long temp;
+  static const int RGB_order[3] = { RGB_GREEN, RGB_RED, RGB_BLUE };
+
+  /* We can allocate at least the nc'th root of max_colors per component. */
+  /* Compute floor(nc'th root of max_colors). */
+  iroot = 1;
+  do {
+    iroot++;
+    temp = iroot;		/* set temp = iroot ** nc */
+    for (i = 1; i < nc; i++)
+      temp *= iroot;
+  } while (temp <= (long) max_colors); /* repeat till iroot exceeds root */
+  iroot--;			/* now iroot = floor(root) */
+
+  /* Must have at least 2 color values per component */
+  if (iroot < 2)
+    ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, (int) temp);
+
+  /* Initialize to iroot color values for each component */
+  total_colors = 1;
+  for (i = 0; i < nc; i++) {
+    Ncolors[i] = iroot;
+    total_colors *= iroot;
+  }
+  /* We may be able to increment the count for one or more components without
+   * exceeding max_colors, though we know not all can be incremented.
+   * Sometimes, the first component can be incremented more than once!
+   * (Example: for 16 colors, we start at 2*2*2, go to 3*2*2, then 4*2*2.)
+   * In RGB colorspace, try to increment G first, then R, then B.
+   */
+  do {
+    changed = FALSE;
+    for (i = 0; i < nc; i++) {
+      j = (cinfo->out_color_space == JCS_RGB ? RGB_order[i] : i);
+      /* calculate new total_colors if Ncolors[j] is incremented */
+      temp = total_colors / Ncolors[j];
+      temp *= Ncolors[j]+1;	/* done in long arith to avoid oflo */
+      if (temp > (long) max_colors)
+	break;			/* won't fit, done with this pass */
+      Ncolors[j]++;		/* OK, apply the increment */
+      total_colors = (int) temp;
+      changed = TRUE;
+    }
+  } while (changed);
+
+  return total_colors;
+}
+
+
+LOCAL(int)
+output_value (j_decompress_ptr cinfo, int ci, int j, int maxj)
+/* Return j'th output value, where j will range from 0 to maxj */
+/* The output values must fall in 0..MAXJSAMPLE in increasing order */
+{
+  /* We always provide values 0 and MAXJSAMPLE for each component;
+   * any additional values are equally spaced between these limits.
+   * (Forcing the upper and lower values to the limits ensures that
+   * dithering can't produce a color outside the selected gamut.)
+   */
+  return (int) (((INT32) j * MAXJSAMPLE + maxj/2) / maxj);
+}
+
+
+LOCAL(int)
+largest_input_value (j_decompress_ptr cinfo, int ci, int j, int maxj)
+/* Return largest input value that should map to j'th output value */
+/* Must have largest(j=0) >= 0, and largest(j=maxj) >= MAXJSAMPLE */
+{
+  /* Breakpoints are halfway between values returned by output_value */
+  return (int) (((INT32) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj));
+}
+
+
+/*
+ * Create the colormap.
+ */
+
+LOCAL(void)
+create_colormap (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  JSAMPARRAY colormap;		/* Created colormap */
+  int total_colors;		/* Number of distinct output colors */
+  int i,j,k, nci, blksize, blkdist, ptr, val;
+
+  /* Select number of colors for each component */
+  total_colors = select_ncolors(cinfo, cquantize->Ncolors);
+
+  /* Report selected color counts */
+  if (cinfo->out_color_components == 3)
+    TRACEMS4(cinfo, 1, JTRC_QUANT_3_NCOLORS,
+	     total_colors, cquantize->Ncolors[0],
+	     cquantize->Ncolors[1], cquantize->Ncolors[2]);
+  else
+    TRACEMS1(cinfo, 1, JTRC_QUANT_NCOLORS, total_colors);
+
+  /* Allocate and fill in the colormap. */
+  /* The colors are ordered in the map in standard row-major order, */
+  /* i.e. rightmost (highest-indexed) color changes most rapidly. */
+
+  colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) total_colors, (JDIMENSION) cinfo->out_color_components);
+
+  /* blksize is number of adjacent repeated entries for a component */
+  /* blkdist is distance between groups of identical entries for a component */
+  blkdist = total_colors;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    /* fill in colormap entries for i'th color component */
+    nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
+    blksize = blkdist / nci;
+    for (j = 0; j < nci; j++) {
+      /* Compute j'th output value (out of nci) for component */
+      val = output_value(cinfo, i, j, nci-1);
+      /* Fill in all colormap entries that have this value of this component */
+      for (ptr = j * blksize; ptr < total_colors; ptr += blkdist) {
+	/* fill in blksize entries beginning at ptr */
+	for (k = 0; k < blksize; k++)
+	  colormap[i][ptr+k] = (JSAMPLE) val;
+      }
+    }
+    blkdist = blksize;		/* blksize of this color is blkdist of next */
+  }
+
+  /* Save the colormap in private storage,
+   * where it will survive color quantization mode changes.
+   */
+  cquantize->sv_colormap = colormap;
+  cquantize->sv_actual = total_colors;
+}
+
+
+/*
+ * Create the color index table.
+ */
+
+LOCAL(void)
+create_colorindex (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  JSAMPROW indexptr;
+  int i,j,k, nci, blksize, val, pad;
+
+  /* For ordered dither, we pad the color index tables by MAXJSAMPLE in
+   * each direction (input index values can be -MAXJSAMPLE .. 2*MAXJSAMPLE).
+   * This is not necessary in the other dithering modes.  However, we
+   * flag whether it was done in case user changes dithering mode.
+   */
+  if (cinfo->dither_mode == JDITHER_ORDERED) {
+    pad = MAXJSAMPLE*2;
+    cquantize->is_padded = TRUE;
+  } else {
+    pad = 0;
+    cquantize->is_padded = FALSE;
+  }
+
+  cquantize->colorindex = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (MAXJSAMPLE+1 + pad),
+     (JDIMENSION) cinfo->out_color_components);
+
+  /* blksize is number of adjacent repeated entries for a component */
+  blksize = cquantize->sv_actual;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    /* fill in colorindex entries for i'th color component */
+    nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
+    blksize = blksize / nci;
+
+    /* adjust colorindex pointers to provide padding at negative indexes. */
+    if (pad)
+      cquantize->colorindex[i] += MAXJSAMPLE;
+
+    /* in loop, val = index of current output value, */
+    /* and k = largest j that maps to current val */
+    indexptr = cquantize->colorindex[i];
+    val = 0;
+    k = largest_input_value(cinfo, i, 0, nci-1);
+    for (j = 0; j <= MAXJSAMPLE; j++) {
+      while (j > k)		/* advance val if past boundary */
+	k = largest_input_value(cinfo, i, ++val, nci-1);
+      /* premultiply so that no multiplication needed in main processing */
+      indexptr[j] = (JSAMPLE) (val * blksize);
+    }
+    /* Pad at both ends if necessary */
+    if (pad)
+      for (j = 1; j <= MAXJSAMPLE; j++) {
+	indexptr[-j] = indexptr[0];
+	indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE];
+      }
+  }
+}
+
+
+/*
+ * Create an ordered-dither array for a component having ncolors
+ * distinct output values.
+ */
+
+LOCAL(ODITHER_MATRIX_PTR)
+make_odither_array (j_decompress_ptr cinfo, int ncolors)
+{
+  ODITHER_MATRIX_PTR odither;
+  int j,k;
+  INT32 num,den;
+
+  odither = (ODITHER_MATRIX_PTR)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(ODITHER_MATRIX));
+  /* The inter-value distance for this color is MAXJSAMPLE/(ncolors-1).
+   * Hence the dither value for the matrix cell with fill order f
+   * (f=0..N-1) should be (N-1-2*f)/(2*N) * MAXJSAMPLE/(ncolors-1).
+   * On 16-bit-int machine, be careful to avoid overflow.
+   */
+  den = 2 * ODITHER_CELLS * ((INT32) (ncolors - 1));
+  for (j = 0; j < ODITHER_SIZE; j++) {
+    for (k = 0; k < ODITHER_SIZE; k++) {
+      num = ((INT32) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k])))
+	    * MAXJSAMPLE;
+      /* Ensure round towards zero despite C's lack of consistency
+       * about rounding negative values in integer division...
+       */
+      odither[j][k] = (int) (num<0 ? -((-num)/den) : num/den);
+    }
+  }
+  return odither;
+}
+
+
+/*
+ * Create the ordered-dither tables.
+ * Components having the same number of representative colors may 
+ * share a dither table.
+ */
+
+LOCAL(void)
+create_odither_tables (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  ODITHER_MATRIX_PTR odither;
+  int i, j, nci;
+
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
+    odither = NULL;		/* search for matching prior component */
+    for (j = 0; j < i; j++) {
+      if (nci == cquantize->Ncolors[j]) {
+	odither = cquantize->odither[j];
+	break;
+      }
+    }
+    if (odither == NULL)	/* need a new table? */
+      odither = make_odither_array(cinfo, nci);
+    cquantize->odither[i] = odither;
+  }
+}
+
+
+/*
+ * Map some rows of pixels to the output colormapped representation.
+ */
+
+METHODDEF(void)
+color_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		JSAMPARRAY output_buf, int num_rows)
+/* General case, no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  JSAMPARRAY colorindex = cquantize->colorindex;
+  register int pixcode, ci;
+  register JSAMPROW ptrin, ptrout;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  register int nc = cinfo->out_color_components;
+
+  for (row = 0; row < num_rows; row++) {
+    ptrin = input_buf[row];
+    ptrout = output_buf[row];
+    for (col = width; col > 0; col--) {
+      pixcode = 0;
+      for (ci = 0; ci < nc; ci++) {
+	pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
+      }
+      *ptrout++ = (JSAMPLE) pixcode;
+    }
+  }
+}
+
+
+METHODDEF(void)
+color_quantize3 (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		 JSAMPARRAY output_buf, int num_rows)
+/* Fast path for out_color_components==3, no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register int pixcode;
+  register JSAMPROW ptrin, ptrout;
+  JSAMPROW colorindex0 = cquantize->colorindex[0];
+  JSAMPROW colorindex1 = cquantize->colorindex[1];
+  JSAMPROW colorindex2 = cquantize->colorindex[2];
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    ptrin = input_buf[row];
+    ptrout = output_buf[row];
+    for (col = width; col > 0; col--) {
+      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*ptrin++)]);
+      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*ptrin++)]);
+      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*ptrin++)]);
+      *ptrout++ = (JSAMPLE) pixcode;
+    }
+  }
+}
+
+
+METHODDEF(void)
+quantize_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		     JSAMPARRAY output_buf, int num_rows)
+/* General case, with ordered dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex_ci;
+  int * dither;			/* points to active row of dither matrix */
+  int row_index, col_index;	/* current indexes into dither matrix */
+  int nc = cinfo->out_color_components;
+  int ci;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    /* Initialize output values to 0 so can process components separately */
+    jzero_far((void FAR *) output_buf[row],
+	      (size_t) (width * SIZEOF(JSAMPLE)));
+    row_index = cquantize->row_index;
+    for (ci = 0; ci < nc; ci++) {
+      input_ptr = input_buf[row] + ci;
+      output_ptr = output_buf[row];
+      colorindex_ci = cquantize->colorindex[ci];
+      dither = cquantize->odither[ci][row_index];
+      col_index = 0;
+
+      for (col = width; col > 0; col--) {
+	/* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
+	 * select output value, accumulate into output code for this pixel.
+	 * Range-limiting need not be done explicitly, as we have extended
+	 * the colorindex table to produce the right answers for out-of-range
+	 * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
+	 * required amount of padding.
+	 */
+	*output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]];
+	input_ptr += nc;
+	output_ptr++;
+	col_index = (col_index + 1) & ODITHER_MASK;
+      }
+    }
+    /* Advance row index for next row */
+    row_index = (row_index + 1) & ODITHER_MASK;
+    cquantize->row_index = row_index;
+  }
+}
+
+
+METHODDEF(void)
+quantize3_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		      JSAMPARRAY output_buf, int num_rows)
+/* Fast path for out_color_components==3, with ordered dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register int pixcode;
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex0 = cquantize->colorindex[0];
+  JSAMPROW colorindex1 = cquantize->colorindex[1];
+  JSAMPROW colorindex2 = cquantize->colorindex[2];
+  int * dither0;		/* points to active row of dither matrix */
+  int * dither1;
+  int * dither2;
+  int row_index, col_index;	/* current indexes into dither matrix */
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    row_index = cquantize->row_index;
+    input_ptr = input_buf[row];
+    output_ptr = output_buf[row];
+    dither0 = cquantize->odither[0][row_index];
+    dither1 = cquantize->odither[1][row_index];
+    dither2 = cquantize->odither[2][row_index];
+    col_index = 0;
+
+    for (col = width; col > 0; col--) {
+      pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) +
+					dither0[col_index]]);
+      pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) +
+					dither1[col_index]]);
+      pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) +
+					dither2[col_index]]);
+      *output_ptr++ = (JSAMPLE) pixcode;
+      col_index = (col_index + 1) & ODITHER_MASK;
+    }
+    row_index = (row_index + 1) & ODITHER_MASK;
+    cquantize->row_index = row_index;
+  }
+}
+
+
+METHODDEF(void)
+quantize_fs_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		    JSAMPARRAY output_buf, int num_rows)
+/* General case, with Floyd-Steinberg dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register LOCFSERROR cur;	/* current error or pixel value */
+  LOCFSERROR belowerr;		/* error for pixel below cur */
+  LOCFSERROR bpreverr;		/* error for below/prev col */
+  LOCFSERROR bnexterr;		/* error for below/next col */
+  LOCFSERROR delta;
+  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
+  register JSAMPROW input_ptr;
+  register JSAMPROW output_ptr;
+  JSAMPROW colorindex_ci;
+  JSAMPROW colormap_ci;
+  int pixcode;
+  int nc = cinfo->out_color_components;
+  int dir;			/* 1 for left-to-right, -1 for right-to-left */
+  int dirnc;			/* dir * nc */
+  int ci;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  SHIFT_TEMPS
+
+  for (row = 0; row < num_rows; row++) {
+    /* Initialize output values to 0 so can process components separately */
+    jzero_far((void FAR *) output_buf[row],
+	      (size_t) (width * SIZEOF(JSAMPLE)));
+    for (ci = 0; ci < nc; ci++) {
+      input_ptr = input_buf[row] + ci;
+      output_ptr = output_buf[row];
+      if (cquantize->on_odd_row) {
+	/* work right to left in this row */
+	input_ptr += (width-1) * nc; /* so point to rightmost pixel */
+	output_ptr += width-1;
+	dir = -1;
+	dirnc = -nc;
+	errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
+      } else {
+	/* work left to right in this row */
+	dir = 1;
+	dirnc = nc;
+	errorptr = cquantize->fserrors[ci]; /* => entry before first column */
+      }
+      colorindex_ci = cquantize->colorindex[ci];
+      colormap_ci = cquantize->sv_colormap[ci];
+      /* Preset error values: no error propagated to first pixel from left */
+      cur = 0;
+      /* and no error propagated to row below yet */
+      belowerr = bpreverr = 0;
+
+      for (col = width; col > 0; col--) {
+	/* cur holds the error propagated from the previous pixel on the
+	 * current line.  Add the error propagated from the previous line
+	 * to form the complete error correction term for this pixel, and
+	 * round the error term (which is expressed * 16) to an integer.
+	 * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+	 * for either sign of the error value.
+	 * Note: errorptr points to *previous* column's array entry.
+	 */
+	cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
+	/* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+	 * The maximum error is +- MAXJSAMPLE; this sets the required size
+	 * of the range_limit array.
+	 */
+	cur += GETJSAMPLE(*input_ptr);
+	cur = GETJSAMPLE(range_limit[cur]);
+	/* Select output value, accumulate into output code for this pixel */
+	pixcode = GETJSAMPLE(colorindex_ci[cur]);
+	*output_ptr += (JSAMPLE) pixcode;
+	/* Compute actual representation error at this pixel */
+	/* Note: we can do this even though we don't have the final */
+	/* pixel code, because the colormap is orthogonal. */
+	cur -= GETJSAMPLE(colormap_ci[pixcode]);
+	/* Compute error fractions to be propagated to adjacent pixels.
+	 * Add these into the running sums, and simultaneously shift the
+	 * next-line error sums left by 1 column.
+	 */
+	bnexterr = cur;
+	delta = cur * 2;
+	cur += delta;		/* form error * 3 */
+	errorptr[0] = (FSERROR) (bpreverr + cur);
+	cur += delta;		/* form error * 5 */
+	bpreverr = belowerr + cur;
+	belowerr = bnexterr;
+	cur += delta;		/* form error * 7 */
+	/* At this point cur contains the 7/16 error value to be propagated
+	 * to the next pixel on the current line, and all the errors for the
+	 * next line have been shifted over. We are therefore ready to move on.
+	 */
+	input_ptr += dirnc;	/* advance input ptr to next column */
+	output_ptr += dir;	/* advance output ptr to next column */
+	errorptr += dir;	/* advance errorptr to current column */
+      }
+      /* Post-loop cleanup: we must unload the final error value into the
+       * final fserrors[] entry.  Note we need not unload belowerr because
+       * it is for the dummy column before or after the actual array.
+       */
+      errorptr[0] = (FSERROR) bpreverr; /* unload prev err into array */
+    }
+    cquantize->on_odd_row = (cquantize->on_odd_row ? FALSE : TRUE);
+  }
+}
+
+
+/*
+ * Allocate workspace for Floyd-Steinberg errors.
+ */
+
+LOCAL(void)
+alloc_fs_workspace (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  size_t arraysize;
+  int i;
+
+  arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+  for (i = 0; i < cinfo->out_color_components; i++) {
+    cquantize->fserrors[i] = (FSERRPTR)
+      (*cinfo->mem->alloc_large)((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
+  }
+}
+
+
+/*
+ * Initialize for one-pass color quantization.
+ */
+
+METHODDEF(void)
+start_pass_1_quant (j_decompress_ptr cinfo, boolean is_pre_scan)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  size_t arraysize;
+  int i;
+
+  /* Install my colormap. */
+  cinfo->colormap = cquantize->sv_colormap;
+  cinfo->actual_number_of_colors = cquantize->sv_actual;
+
+  /* Initialize for desired dithering mode. */
+  switch (cinfo->dither_mode) {
+  case JDITHER_NONE:
+    if (cinfo->out_color_components == 3)
+      cquantize->pub.color_quantize = color_quantize3;
+    else
+      cquantize->pub.color_quantize = color_quantize;
+    break;
+  case JDITHER_ORDERED:
+    if (cinfo->out_color_components == 3)
+      cquantize->pub.color_quantize = quantize3_ord_dither;
+    else
+      cquantize->pub.color_quantize = quantize_ord_dither;
+    cquantize->row_index = 0;	/* initialize state for ordered dither */
+    /* If user changed to ordered dither from another mode,
+     * we must recreate the color index table with padding.
+     * This will cost extra space, but probably isn't very likely.
+     */
+    if (! cquantize->is_padded)
+      create_colorindex(cinfo);
+    /* Create ordered-dither tables if we didn't already. */
+    if (cquantize->odither[0] == NULL)
+      create_odither_tables(cinfo);
+    break;
+  case JDITHER_FS:
+    cquantize->pub.color_quantize = quantize_fs_dither;
+    cquantize->on_odd_row = FALSE; /* initialize state for F-S dither */
+    /* Allocate Floyd-Steinberg workspace if didn't already. */
+    if (cquantize->fserrors[0] == NULL)
+      alloc_fs_workspace(cinfo);
+    /* Initialize the propagated errors to zero. */
+    arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+    for (i = 0; i < cinfo->out_color_components; i++)
+      jzero_far((void FAR *) cquantize->fserrors[i], arraysize);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_NOT_COMPILED);
+    break;
+  }
+}
+
+
+/*
+ * Finish up at the end of the pass.
+ */
+
+METHODDEF(void)
+finish_pass_1_quant (j_decompress_ptr cinfo)
+{
+  /* no work in 1-pass case */
+}
+
+
+/*
+ * Switch to a new external colormap between output passes.
+ * Shouldn't get to this module!
+ */
+
+METHODDEF(void)
+new_color_map_1_quant (j_decompress_ptr cinfo)
+{
+  ERREXIT(cinfo, JERR_MODE_CHANGE);
+}
+
+
+/*
+ * Module initialization routine for 1-pass color quantization.
+ */
+
+GLOBAL(void)
+jinit_1pass_quantizer (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize;
+
+  cquantize = (my_cquantize_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_cquantizer));
+  cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
+  cquantize->pub.start_pass = start_pass_1_quant;
+  cquantize->pub.finish_pass = finish_pass_1_quant;
+  cquantize->pub.new_color_map = new_color_map_1_quant;
+  cquantize->fserrors[0] = NULL; /* Flag FS workspace not allocated */
+  cquantize->odither[0] = NULL;	/* Also flag odither arrays not allocated */
+
+  /* Make sure my internal arrays won't overflow */
+  if (cinfo->out_color_components > MAX_Q_COMPS)
+    ERREXIT1(cinfo, JERR_QUANT_COMPONENTS, MAX_Q_COMPS);
+  /* Make sure colormap indexes can be represented by JSAMPLEs */
+  if (cinfo->desired_number_of_colors > (MAXJSAMPLE+1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXJSAMPLE+1);
+
+  /* Create the colormap and color index table. */
+  create_colormap(cinfo);
+  create_colorindex(cinfo);
+
+  /* Allocate Floyd-Steinberg workspace now if requested.
+   * We do this now since it is FAR storage and may affect the memory
+   * manager's space calculations.  If the user changes to FS dither
+   * mode in a later pass, we will allocate the space then, and will
+   * possibly overrun the max_memory_to_use setting.
+   */
+  if (cinfo->dither_mode == JDITHER_FS)
+    alloc_fs_workspace(cinfo);
+}
+
+#endif /* QUANT_1PASS_SUPPORTED */
diff --git a/jpeg/jquant2.c b/jpeg/jquant2.c
new file mode 100644
index 000000000..af601e334
--- /dev/null
+++ b/jpeg/jquant2.c
@@ -0,0 +1,1310 @@
+/*
+ * jquant2.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains 2-pass color quantization (color mapping) routines.
+ * These routines provide selection of a custom color map for an image,
+ * followed by mapping of the image to that color map, with optional
+ * Floyd-Steinberg dithering.
+ * It is also possible to use just the second pass to map to an arbitrary
+ * externally-given color map.
+ *
+ * Note: ordered dithering is not supported, since there isn't any fast
+ * way to compute intercolor distances; it's unclear that ordered dither's
+ * fundamental assumptions even hold with an irregularly spaced color map.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+#ifdef QUANT_2PASS_SUPPORTED
+
+
+/*
+ * This module implements the well-known Heckbert paradigm for color
+ * quantization.  Most of the ideas used here can be traced back to
+ * Heckbert's seminal paper
+ *   Heckbert, Paul.  "Color Image Quantization for Frame Buffer Display",
+ *   Proc. SIGGRAPH '82, Computer Graphics v.16 #3 (July 1982), pp 297-304.
+ *
+ * In the first pass over the image, we accumulate a histogram showing the
+ * usage count of each possible color.  To keep the histogram to a reasonable
+ * size, we reduce the precision of the input; typical practice is to retain
+ * 5 or 6 bits per color, so that 8 or 4 different input values are counted
+ * in the same histogram cell.
+ *
+ * Next, the color-selection step begins with a box representing the whole
+ * color space, and repeatedly splits the "largest" remaining box until we
+ * have as many boxes as desired colors.  Then the mean color in each
+ * remaining box becomes one of the possible output colors.
+ * 
+ * The second pass over the image maps each input pixel to the closest output
+ * color (optionally after applying a Floyd-Steinberg dithering correction).
+ * This mapping is logically trivial, but making it go fast enough requires
+ * considerable care.
+ *
+ * Heckbert-style quantizers vary a good deal in their policies for choosing
+ * the "largest" box and deciding where to cut it.  The particular policies
+ * used here have proved out well in experimental comparisons, but better ones
+ * may yet be found.
+ *
+ * In earlier versions of the IJG code, this module quantized in YCbCr color
+ * space, processing the raw upsampled data without a color conversion step.
+ * This allowed the color conversion math to be done only once per colormap
+ * entry, not once per pixel.  However, that optimization precluded other
+ * useful optimizations (such as merging color conversion with upsampling)
+ * and it also interfered with desired capabilities such as quantizing to an
+ * externally-supplied colormap.  We have therefore abandoned that approach.
+ * The present code works in the post-conversion color space, typically RGB.
+ *
+ * To improve the visual quality of the results, we actually work in scaled
+ * RGB space, giving G distances more weight than R, and R in turn more than
+ * B.  To do everything in integer math, we must use integer scale factors.
+ * The 2/3/1 scale factors used here correspond loosely to the relative
+ * weights of the colors in the NTSC grayscale equation.
+ * If you want to use this code to quantize a non-RGB color space, you'll
+ * probably need to change these scale factors.
+ */
+
+#define R_SCALE 2		/* scale R distances by this much */
+#define G_SCALE 3		/* scale G distances by this much */
+#define B_SCALE 1		/* and B by this much */
+
+/* Relabel R/G/B as components 0/1/2, respecting the RGB ordering defined
+ * in jmorecfg.h.  As the code stands, it will do the right thing for R,G,B
+ * and B,G,R orders.  If you define some other weird order in jmorecfg.h,
+ * you'll get compile errors until you extend this logic.  In that case
+ * you'll probably want to tweak the histogram sizes too.
+ */
+
+#if RGB_RED == 0
+#define C0_SCALE R_SCALE
+#endif
+#if RGB_BLUE == 0
+#define C0_SCALE B_SCALE
+#endif
+#if RGB_GREEN == 1
+#define C1_SCALE G_SCALE
+#endif
+#if RGB_RED == 2
+#define C2_SCALE R_SCALE
+#endif
+#if RGB_BLUE == 2
+#define C2_SCALE B_SCALE
+#endif
+
+
+/*
+ * First we have the histogram data structure and routines for creating it.
+ *
+ * The number of bits of precision can be adjusted by changing these symbols.
+ * We recommend keeping 6 bits for G and 5 each for R and B.
+ * If you have plenty of memory and cycles, 6 bits all around gives marginally
+ * better results; if you are short of memory, 5 bits all around will save
+ * some space but degrade the results.
+ * To maintain a fully accurate histogram, we'd need to allocate a "long"
+ * (preferably unsigned long) for each cell.  In practice this is overkill;
+ * we can get by with 16 bits per cell.  Few of the cell counts will overflow,
+ * and clamping those that do overflow to the maximum value will give close-
+ * enough results.  This reduces the recommended histogram size from 256Kb
+ * to 128Kb, which is a useful savings on PC-class machines.
+ * (In the second pass the histogram space is re-used for pixel mapping data;
+ * in that capacity, each cell must be able to store zero to the number of
+ * desired colors.  16 bits/cell is plenty for that too.)
+ * Since the JPEG code is intended to run in small memory model on 80x86
+ * machines, we can't just allocate the histogram in one chunk.  Instead
+ * of a true 3-D array, we use a row of pointers to 2-D arrays.  Each
+ * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and
+ * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.  Note that
+ * on 80x86 machines, the pointer row is in near memory but the actual
+ * arrays are in far memory (same arrangement as we use for image arrays).
+ */
+
+#define MAXNUMCOLORS  (MAXJSAMPLE+1) /* maximum size of colormap */
+
+/* These will do the right thing for either R,G,B or B,G,R color order,
+ * but you may not like the results for other color orders.
+ */
+#define HIST_C0_BITS  5		/* bits of precision in R/B histogram */
+#define HIST_C1_BITS  6		/* bits of precision in G histogram */
+#define HIST_C2_BITS  5		/* bits of precision in B/R histogram */
+
+/* Number of elements along histogram axes. */
+#define HIST_C0_ELEMS  (1<<HIST_C0_BITS)
+#define HIST_C1_ELEMS  (1<<HIST_C1_BITS)
+#define HIST_C2_ELEMS  (1<<HIST_C2_BITS)
+
+/* These are the amounts to shift an input value to get a histogram index. */
+#define C0_SHIFT  (BITS_IN_JSAMPLE-HIST_C0_BITS)
+#define C1_SHIFT  (BITS_IN_JSAMPLE-HIST_C1_BITS)
+#define C2_SHIFT  (BITS_IN_JSAMPLE-HIST_C2_BITS)
+
+
+typedef UINT16 histcell;	/* histogram cell; prefer an unsigned type */
+
+typedef histcell FAR * histptr;	/* for pointers to histogram cells */
+
+typedef histcell hist1d[HIST_C2_ELEMS]; /* typedefs for the array */
+typedef hist1d FAR * hist2d;	/* type for the 2nd-level pointers */
+typedef hist2d * hist3d;	/* type for top-level pointer */
+
+
+/* Declarations for Floyd-Steinberg dithering.
+ *
+ * Errors are accumulated into the array fserrors[], at a resolution of
+ * 1/16th of a pixel count.  The error at a given pixel is propagated
+ * to its not-yet-processed neighbors using the standard F-S fractions,
+ *		...	(here)	7/16
+ *		3/16	5/16	1/16
+ * We work left-to-right on even rows, right-to-left on odd rows.
+ *
+ * We can get away with a single array (holding one row's worth of errors)
+ * by using it to store the current row's errors at pixel columns not yet
+ * processed, but the next row's errors at columns already processed.  We
+ * need only a few extra variables to hold the errors immediately around the
+ * current column.  (If we are lucky, those variables are in registers, but
+ * even if not, they're probably cheaper to access than array elements are.)
+ *
+ * The fserrors[] array has (#columns + 2) entries; the extra entry at
+ * each end saves us from special-casing the first and last pixels.
+ * Each entry is three values long, one value for each color component.
+ *
+ * Note: on a wide image, we might not have enough room in a PC's near data
+ * segment to hold the error array; so it is allocated with alloc_large.
+ */
+
+#if BITS_IN_JSAMPLE == 8
+typedef INT16 FSERROR;		/* 16 bits should be enough */
+typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+#else
+typedef INT32 FSERROR;		/* may need more than 16 bits */
+typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+#endif
+
+typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_color_quantizer pub; /* public fields */
+
+  /* Space for the eventually created colormap is stashed here */
+  JSAMPARRAY sv_colormap;	/* colormap allocated at init time */
+  int desired;			/* desired # of colors = size of colormap */
+
+  /* Variables for accumulating image statistics */
+  hist3d histogram;		/* pointer to the histogram */
+
+  boolean needs_zeroed;		/* TRUE if next pass must zero histogram */
+
+  /* Variables for Floyd-Steinberg dithering */
+  FSERRPTR fserrors;		/* accumulated errors */
+  boolean on_odd_row;		/* flag to remember which row we are on */
+  int * error_limiter;		/* table for clamping the applied error */
+} my_cquantizer;
+
+typedef my_cquantizer * my_cquantize_ptr;
+
+
+/*
+ * Prescan some rows of pixels.
+ * In this module the prescan simply updates the histogram, which has been
+ * initialized to zeroes by start_pass.
+ * An output_buf parameter is required by the method signature, but no data
+ * is actually output (in fact the buffer controller is probably passing a
+ * NULL pointer).
+ */
+
+METHODDEF(void)
+prescan_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+		  JSAMPARRAY output_buf, int num_rows)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  register JSAMPROW ptr;
+  register histptr histp;
+  register hist3d histogram = cquantize->histogram;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    ptr = input_buf[row];
+    for (col = width; col > 0; col--) {
+      /* get pixel value and index into the histogram */
+      histp = & histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
+			 [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
+			 [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
+      /* increment, check for overflow and undo increment if so. */
+      if (++(*histp) <= 0)
+	(*histp)--;
+      ptr += 3;
+    }
+  }
+}
+
+
+/*
+ * Next we have the really interesting routines: selection of a colormap
+ * given the completed histogram.
+ * These routines work with a list of "boxes", each representing a rectangular
+ * subset of the input color space (to histogram precision).
+ */
+
+typedef struct {
+  /* The bounds of the box (inclusive); expressed as histogram indexes */
+  int c0min, c0max;
+  int c1min, c1max;
+  int c2min, c2max;
+  /* The volume (actually 2-norm) of the box */
+  INT32 volume;
+  /* The number of nonzero histogram cells within this box */
+  long colorcount;
+} box;
+
+typedef box * boxptr;
+
+
+LOCAL(boxptr)
+find_biggest_color_pop (boxptr boxlist, int numboxes)
+/* Find the splittable box with the largest color population */
+/* Returns NULL if no splittable boxes remain */
+{
+  register boxptr boxp;
+  register int i;
+  register long maxc = 0;
+  boxptr which = NULL;
+  
+  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
+    if (boxp->colorcount > maxc && boxp->volume > 0) {
+      which = boxp;
+      maxc = boxp->colorcount;
+    }
+  }
+  return which;
+}
+
+
+LOCAL(boxptr)
+find_biggest_volume (boxptr boxlist, int numboxes)
+/* Find the splittable box with the largest (scaled) volume */
+/* Returns NULL if no splittable boxes remain */
+{
+  register boxptr boxp;
+  register int i;
+  register INT32 maxv = 0;
+  boxptr which = NULL;
+  
+  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
+    if (boxp->volume > maxv) {
+      which = boxp;
+      maxv = boxp->volume;
+    }
+  }
+  return which;
+}
+
+
+LOCAL(void)
+update_box (j_decompress_ptr cinfo, boxptr boxp)
+/* Shrink the min/max bounds of a box to enclose only nonzero elements, */
+/* and recompute its volume and population */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  histptr histp;
+  int c0,c1,c2;
+  int c0min,c0max,c1min,c1max,c2min,c2max;
+  INT32 dist0,dist1,dist2;
+  long ccount;
+  
+  c0min = boxp->c0min;  c0max = boxp->c0max;
+  c1min = boxp->c1min;  c1max = boxp->c1max;
+  c2min = boxp->c2min;  c2max = boxp->c2max;
+  
+  if (c0max > c0min)
+    for (c0 = c0min; c0 <= c0max; c0++)
+      for (c1 = c1min; c1 <= c1max; c1++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c0min = c0min = c0;
+	    goto have_c0min;
+	  }
+      }
+ have_c0min:
+  if (c0max > c0min)
+    for (c0 = c0max; c0 >= c0min; c0--)
+      for (c1 = c1min; c1 <= c1max; c1++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c0max = c0max = c0;
+	    goto have_c0max;
+	  }
+      }
+ have_c0max:
+  if (c1max > c1min)
+    for (c1 = c1min; c1 <= c1max; c1++)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c1min = c1min = c1;
+	    goto have_c1min;
+	  }
+      }
+ have_c1min:
+  if (c1max > c1min)
+    for (c1 = c1max; c1 >= c1min; c1--)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1][c2min];
+	for (c2 = c2min; c2 <= c2max; c2++)
+	  if (*histp++ != 0) {
+	    boxp->c1max = c1max = c1;
+	    goto have_c1max;
+	  }
+      }
+ have_c1max:
+  if (c2max > c2min)
+    for (c2 = c2min; c2 <= c2max; c2++)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1min][c2];
+	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+	  if (*histp != 0) {
+	    boxp->c2min = c2min = c2;
+	    goto have_c2min;
+	  }
+      }
+ have_c2min:
+  if (c2max > c2min)
+    for (c2 = c2max; c2 >= c2min; c2--)
+      for (c0 = c0min; c0 <= c0max; c0++) {
+	histp = & histogram[c0][c1min][c2];
+	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+	  if (*histp != 0) {
+	    boxp->c2max = c2max = c2;
+	    goto have_c2max;
+	  }
+      }
+ have_c2max:
+
+  /* Update box volume.
+   * We use 2-norm rather than real volume here; this biases the method
+   * against making long narrow boxes, and it has the side benefit that
+   * a box is splittable iff norm > 0.
+   * Since the differences are expressed in histogram-cell units,
+   * we have to shift back to JSAMPLE units to get consistent distances;
+   * after which, we scale according to the selected distance scale factors.
+   */
+  dist0 = ((c0max - c0min) << C0_SHIFT) * C0_SCALE;
+  dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE;
+  dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE;
+  boxp->volume = dist0*dist0 + dist1*dist1 + dist2*dist2;
+  
+  /* Now scan remaining volume of box and compute population */
+  ccount = 0;
+  for (c0 = c0min; c0 <= c0max; c0++)
+    for (c1 = c1min; c1 <= c1max; c1++) {
+      histp = & histogram[c0][c1][c2min];
+      for (c2 = c2min; c2 <= c2max; c2++, histp++)
+	if (*histp != 0) {
+	  ccount++;
+	}
+    }
+  boxp->colorcount = ccount;
+}
+
+
+LOCAL(int)
+median_cut (j_decompress_ptr cinfo, boxptr boxlist, int numboxes,
+	    int desired_colors)
+/* Repeatedly select and split the largest box until we have enough boxes */
+{
+  int n,lb;
+  int c0,c1,c2,cmax;
+  register boxptr b1,b2;
+
+  while (numboxes < desired_colors) {
+    /* Select box to split.
+     * Current algorithm: by population for first half, then by volume.
+     */
+    if (numboxes*2 <= desired_colors) {
+      b1 = find_biggest_color_pop(boxlist, numboxes);
+    } else {
+      b1 = find_biggest_volume(boxlist, numboxes);
+    }
+    if (b1 == NULL)		/* no splittable boxes left! */
+      break;
+    b2 = &boxlist[numboxes];	/* where new box will go */
+    /* Copy the color bounds to the new box. */
+    b2->c0max = b1->c0max; b2->c1max = b1->c1max; b2->c2max = b1->c2max;
+    b2->c0min = b1->c0min; b2->c1min = b1->c1min; b2->c2min = b1->c2min;
+    /* Choose which axis to split the box on.
+     * Current algorithm: longest scaled axis.
+     * See notes in update_box about scaling distances.
+     */
+    c0 = ((b1->c0max - b1->c0min) << C0_SHIFT) * C0_SCALE;
+    c1 = ((b1->c1max - b1->c1min) << C1_SHIFT) * C1_SCALE;
+    c2 = ((b1->c2max - b1->c2min) << C2_SHIFT) * C2_SCALE;
+    /* We want to break any ties in favor of green, then red, blue last.
+     * This code does the right thing for R,G,B or B,G,R color orders only.
+     */
+#if RGB_RED == 0
+    cmax = c1; n = 1;
+    if (c0 > cmax) { cmax = c0; n = 0; }
+    if (c2 > cmax) { n = 2; }
+#else
+    cmax = c1; n = 1;
+    if (c2 > cmax) { cmax = c2; n = 2; }
+    if (c0 > cmax) { n = 0; }
+#endif
+    /* Choose split point along selected axis, and update box bounds.
+     * Current algorithm: split at halfway point.
+     * (Since the box has been shrunk to minimum volume,
+     * any split will produce two nonempty subboxes.)
+     * Note that lb value is max for lower box, so must be < old max.
+     */
+    switch (n) {
+    case 0:
+      lb = (b1->c0max + b1->c0min) / 2;
+      b1->c0max = lb;
+      b2->c0min = lb+1;
+      break;
+    case 1:
+      lb = (b1->c1max + b1->c1min) / 2;
+      b1->c1max = lb;
+      b2->c1min = lb+1;
+      break;
+    case 2:
+      lb = (b1->c2max + b1->c2min) / 2;
+      b1->c2max = lb;
+      b2->c2min = lb+1;
+      break;
+    }
+    /* Update stats for boxes */
+    update_box(cinfo, b1);
+    update_box(cinfo, b2);
+    numboxes++;
+  }
+  return numboxes;
+}
+
+
+LOCAL(void)
+compute_color (j_decompress_ptr cinfo, boxptr boxp, int icolor)
+/* Compute representative color for a box, put it in colormap[icolor] */
+{
+  /* Current algorithm: mean weighted by pixels (not colors) */
+  /* Note it is important to get the rounding correct! */
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  histptr histp;
+  int c0,c1,c2;
+  int c0min,c0max,c1min,c1max,c2min,c2max;
+  long count;
+  long total = 0;
+  long c0total = 0;
+  long c1total = 0;
+  long c2total = 0;
+  
+  c0min = boxp->c0min;  c0max = boxp->c0max;
+  c1min = boxp->c1min;  c1max = boxp->c1max;
+  c2min = boxp->c2min;  c2max = boxp->c2max;
+  
+  for (c0 = c0min; c0 <= c0max; c0++)
+    for (c1 = c1min; c1 <= c1max; c1++) {
+      histp = & histogram[c0][c1][c2min];
+      for (c2 = c2min; c2 <= c2max; c2++) {
+	if ((count = *histp++) != 0) {
+	  total += count;
+	  c0total += ((c0 << C0_SHIFT) + ((1<<C0_SHIFT)>>1)) * count;
+	  c1total += ((c1 << C1_SHIFT) + ((1<<C1_SHIFT)>>1)) * count;
+	  c2total += ((c2 << C2_SHIFT) + ((1<<C2_SHIFT)>>1)) * count;
+	}
+      }
+    }
+  
+  cinfo->colormap[0][icolor] = (JSAMPLE) ((c0total + (total>>1)) / total);
+  cinfo->colormap[1][icolor] = (JSAMPLE) ((c1total + (total>>1)) / total);
+  cinfo->colormap[2][icolor] = (JSAMPLE) ((c2total + (total>>1)) / total);
+}
+
+
+LOCAL(void)
+select_colors (j_decompress_ptr cinfo, int desired_colors)
+/* Master routine for color selection */
+{
+  boxptr boxlist;
+  int numboxes;
+  int i;
+
+  /* Allocate workspace for box list */
+  boxlist = (boxptr) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * SIZEOF(box));
+  /* Initialize one box containing whole space */
+  numboxes = 1;
+  boxlist[0].c0min = 0;
+  boxlist[0].c0max = MAXJSAMPLE >> C0_SHIFT;
+  boxlist[0].c1min = 0;
+  boxlist[0].c1max = MAXJSAMPLE >> C1_SHIFT;
+  boxlist[0].c2min = 0;
+  boxlist[0].c2max = MAXJSAMPLE >> C2_SHIFT;
+  /* Shrink it to actually-used volume and set its statistics */
+  update_box(cinfo, & boxlist[0]);
+  /* Perform median-cut to produce final box list */
+  numboxes = median_cut(cinfo, boxlist, numboxes, desired_colors);
+  /* Compute the representative color for each box, fill colormap */
+  for (i = 0; i < numboxes; i++)
+    compute_color(cinfo, & boxlist[i], i);
+  cinfo->actual_number_of_colors = numboxes;
+  TRACEMS1(cinfo, 1, JTRC_QUANT_SELECTED, numboxes);
+}
+
+
+/*
+ * These routines are concerned with the time-critical task of mapping input
+ * colors to the nearest color in the selected colormap.
+ *
+ * We re-use the histogram space as an "inverse color map", essentially a
+ * cache for the results of nearest-color searches.  All colors within a
+ * histogram cell will be mapped to the same colormap entry, namely the one
+ * closest to the cell's center.  This may not be quite the closest entry to
+ * the actual input color, but it's almost as good.  A zero in the cache
+ * indicates we haven't found the nearest color for that cell yet; the array
+ * is cleared to zeroes before starting the mapping pass.  When we find the
+ * nearest color for a cell, its colormap index plus one is recorded in the
+ * cache for future use.  The pass2 scanning routines call fill_inverse_cmap
+ * when they need to use an unfilled entry in the cache.
+ *
+ * Our method of efficiently finding nearest colors is based on the "locally
+ * sorted search" idea described by Heckbert and on the incremental distance
+ * calculation described by Spencer W. Thomas in chapter III.1 of Graphics
+ * Gems II (James Arvo, ed.  Academic Press, 1991).  Thomas points out that
+ * the distances from a given colormap entry to each cell of the histogram can
+ * be computed quickly using an incremental method: the differences between
+ * distances to adjacent cells themselves differ by a constant.  This allows a
+ * fairly fast implementation of the "brute force" approach of computing the
+ * distance from every colormap entry to every histogram cell.  Unfortunately,
+ * it needs a work array to hold the best-distance-so-far for each histogram
+ * cell (because the inner loop has to be over cells, not colormap entries).
+ * The work array elements have to be INT32s, so the work array would need
+ * 256Kb at our recommended precision.  This is not feasible in DOS machines.
+ *
+ * To get around these problems, we apply Thomas' method to compute the
+ * nearest colors for only the cells within a small subbox of the histogram.
+ * The work array need be only as big as the subbox, so the memory usage
+ * problem is solved.  Furthermore, we need not fill subboxes that are never
+ * referenced in pass2; many images use only part of the color gamut, so a
+ * fair amount of work is saved.  An additional advantage of this
+ * approach is that we can apply Heckbert's locality criterion to quickly
+ * eliminate colormap entries that are far away from the subbox; typically
+ * three-fourths of the colormap entries are rejected by Heckbert's criterion,
+ * and we need not compute their distances to individual cells in the subbox.
+ * The speed of this approach is heavily influenced by the subbox size: too
+ * small means too much overhead, too big loses because Heckbert's criterion
+ * can't eliminate as many colormap entries.  Empirically the best subbox
+ * size seems to be about 1/512th of the histogram (1/8th in each direction).
+ *
+ * Thomas' article also describes a refined method which is asymptotically
+ * faster than the brute-force method, but it is also far more complex and
+ * cannot efficiently be applied to small subboxes.  It is therefore not
+ * useful for programs intended to be portable to DOS machines.  On machines
+ * with plenty of memory, filling the whole histogram in one shot with Thomas'
+ * refined method might be faster than the present code --- but then again,
+ * it might not be any faster, and it's certainly more complicated.
+ */
+
+
+/* log2(histogram cells in update box) for each axis; this can be adjusted */
+#define BOX_C0_LOG  (HIST_C0_BITS-3)
+#define BOX_C1_LOG  (HIST_C1_BITS-3)
+#define BOX_C2_LOG  (HIST_C2_BITS-3)
+
+#define BOX_C0_ELEMS  (1<<BOX_C0_LOG) /* # of hist cells in update box */
+#define BOX_C1_ELEMS  (1<<BOX_C1_LOG)
+#define BOX_C2_ELEMS  (1<<BOX_C2_LOG)
+
+#define BOX_C0_SHIFT  (C0_SHIFT + BOX_C0_LOG)
+#define BOX_C1_SHIFT  (C1_SHIFT + BOX_C1_LOG)
+#define BOX_C2_SHIFT  (C2_SHIFT + BOX_C2_LOG)
+
+
+/*
+ * The next three routines implement inverse colormap filling.  They could
+ * all be folded into one big routine, but splitting them up this way saves
+ * some stack space (the mindist[] and bestdist[] arrays need not coexist)
+ * and may allow some compilers to produce better code by registerizing more
+ * inner-loop variables.
+ */
+
+LOCAL(int)
+find_nearby_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
+		    JSAMPLE colorlist[])
+/* Locate the colormap entries close enough to an update box to be candidates
+ * for the nearest entry to some cell(s) in the update box.  The update box
+ * is specified by the center coordinates of its first cell.  The number of
+ * candidate colormap entries is returned, and their colormap indexes are
+ * placed in colorlist[].
+ * This routine uses Heckbert's "locally sorted search" criterion to select
+ * the colors that need further consideration.
+ */
+{
+  int numcolors = cinfo->actual_number_of_colors;
+  int maxc0, maxc1, maxc2;
+  int centerc0, centerc1, centerc2;
+  int i, x, ncolors;
+  INT32 minmaxdist, min_dist, max_dist, tdist;
+  INT32 mindist[MAXNUMCOLORS];	/* min distance to colormap entry i */
+
+  /* Compute true coordinates of update box's upper corner and center.
+   * Actually we compute the coordinates of the center of the upper-corner
+   * histogram cell, which are the upper bounds of the volume we care about.
+   * Note that since ">>" rounds down, the "center" values may be closer to
+   * min than to max; hence comparisons to them must be "<=", not "<".
+   */
+  maxc0 = minc0 + ((1 << BOX_C0_SHIFT) - (1 << C0_SHIFT));
+  centerc0 = (minc0 + maxc0) >> 1;
+  maxc1 = minc1 + ((1 << BOX_C1_SHIFT) - (1 << C1_SHIFT));
+  centerc1 = (minc1 + maxc1) >> 1;
+  maxc2 = minc2 + ((1 << BOX_C2_SHIFT) - (1 << C2_SHIFT));
+  centerc2 = (minc2 + maxc2) >> 1;
+
+  /* For each color in colormap, find:
+   *  1. its minimum squared-distance to any point in the update box
+   *     (zero if color is within update box);
+   *  2. its maximum squared-distance to any point in the update box.
+   * Both of these can be found by considering only the corners of the box.
+   * We save the minimum distance for each color in mindist[];
+   * only the smallest maximum distance is of interest.
+   */
+  minmaxdist = 0x7FFFFFFFL;
+
+  for (i = 0; i < numcolors; i++) {
+    /* We compute the squared-c0-distance term, then add in the other two. */
+    x = GETJSAMPLE(cinfo->colormap[0][i]);
+    if (x < minc0) {
+      tdist = (x - minc0) * C0_SCALE;
+      min_dist = tdist*tdist;
+      tdist = (x - maxc0) * C0_SCALE;
+      max_dist = tdist*tdist;
+    } else if (x > maxc0) {
+      tdist = (x - maxc0) * C0_SCALE;
+      min_dist = tdist*tdist;
+      tdist = (x - minc0) * C0_SCALE;
+      max_dist = tdist*tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      min_dist = 0;
+      if (x <= centerc0) {
+	tdist = (x - maxc0) * C0_SCALE;
+	max_dist = tdist*tdist;
+      } else {
+	tdist = (x - minc0) * C0_SCALE;
+	max_dist = tdist*tdist;
+      }
+    }
+
+    x = GETJSAMPLE(cinfo->colormap[1][i]);
+    if (x < minc1) {
+      tdist = (x - minc1) * C1_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - maxc1) * C1_SCALE;
+      max_dist += tdist*tdist;
+    } else if (x > maxc1) {
+      tdist = (x - maxc1) * C1_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - minc1) * C1_SCALE;
+      max_dist += tdist*tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      if (x <= centerc1) {
+	tdist = (x - maxc1) * C1_SCALE;
+	max_dist += tdist*tdist;
+      } else {
+	tdist = (x - minc1) * C1_SCALE;
+	max_dist += tdist*tdist;
+      }
+    }
+
+    x = GETJSAMPLE(cinfo->colormap[2][i]);
+    if (x < minc2) {
+      tdist = (x - minc2) * C2_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - maxc2) * C2_SCALE;
+      max_dist += tdist*tdist;
+    } else if (x > maxc2) {
+      tdist = (x - maxc2) * C2_SCALE;
+      min_dist += tdist*tdist;
+      tdist = (x - minc2) * C2_SCALE;
+      max_dist += tdist*tdist;
+    } else {
+      /* within cell range so no contribution to min_dist */
+      if (x <= centerc2) {
+	tdist = (x - maxc2) * C2_SCALE;
+	max_dist += tdist*tdist;
+      } else {
+	tdist = (x - minc2) * C2_SCALE;
+	max_dist += tdist*tdist;
+      }
+    }
+
+    mindist[i] = min_dist;	/* save away the results */
+    if (max_dist < minmaxdist)
+      minmaxdist = max_dist;
+  }
+
+  /* Now we know that no cell in the update box is more than minmaxdist
+   * away from some colormap entry.  Therefore, only colors that are
+   * within minmaxdist of some part of the box need be considered.
+   */
+  ncolors = 0;
+  for (i = 0; i < numcolors; i++) {
+    if (mindist[i] <= minmaxdist)
+      colorlist[ncolors++] = (JSAMPLE) i;
+  }
+  return ncolors;
+}
+
+
+LOCAL(void)
+find_best_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
+		  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
+/* Find the closest colormap entry for each cell in the update box,
+ * given the list of candidate colors prepared by find_nearby_colors.
+ * Return the indexes of the closest entries in the bestcolor[] array.
+ * This routine uses Thomas' incremental distance calculation method to
+ * find the distance from a colormap entry to successive cells in the box.
+ */
+{
+  int ic0, ic1, ic2;
+  int i, icolor;
+  register INT32 * bptr;	/* pointer into bestdist[] array */
+  JSAMPLE * cptr;		/* pointer into bestcolor[] array */
+  INT32 dist0, dist1;		/* initial distance values */
+  register INT32 dist2;		/* current distance in inner loop */
+  INT32 xx0, xx1;		/* distance increments */
+  register INT32 xx2;
+  INT32 inc0, inc1, inc2;	/* initial values for increments */
+  /* This array holds the distance to the nearest-so-far color for each cell */
+  INT32 bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
+
+  /* Initialize best-distance for each cell of the update box */
+  bptr = bestdist;
+  for (i = BOX_C0_ELEMS*BOX_C1_ELEMS*BOX_C2_ELEMS-1; i >= 0; i--)
+    *bptr++ = 0x7FFFFFFFL;
+  
+  /* For each color selected by find_nearby_colors,
+   * compute its distance to the center of each cell in the box.
+   * If that's less than best-so-far, update best distance and color number.
+   */
+  
+  /* Nominal steps between cell centers ("x" in Thomas article) */
+#define STEP_C0  ((1 << C0_SHIFT) * C0_SCALE)
+#define STEP_C1  ((1 << C1_SHIFT) * C1_SCALE)
+#define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)
+  
+  for (i = 0; i < numcolors; i++) {
+    icolor = GETJSAMPLE(colorlist[i]);
+    /* Compute (square of) distance from minc0/c1/c2 to this color */
+    inc0 = (minc0 - GETJSAMPLE(cinfo->colormap[0][icolor])) * C0_SCALE;
+    dist0 = inc0*inc0;
+    inc1 = (minc1 - GETJSAMPLE(cinfo->colormap[1][icolor])) * C1_SCALE;
+    dist0 += inc1*inc1;
+    inc2 = (minc2 - GETJSAMPLE(cinfo->colormap[2][icolor])) * C2_SCALE;
+    dist0 += inc2*inc2;
+    /* Form the initial difference increments */
+    inc0 = inc0 * (2 * STEP_C0) + STEP_C0 * STEP_C0;
+    inc1 = inc1 * (2 * STEP_C1) + STEP_C1 * STEP_C1;
+    inc2 = inc2 * (2 * STEP_C2) + STEP_C2 * STEP_C2;
+    /* Now loop over all cells in box, updating distance per Thomas method */
+    bptr = bestdist;
+    cptr = bestcolor;
+    xx0 = inc0;
+    for (ic0 = BOX_C0_ELEMS-1; ic0 >= 0; ic0--) {
+      dist1 = dist0;
+      xx1 = inc1;
+      for (ic1 = BOX_C1_ELEMS-1; ic1 >= 0; ic1--) {
+	dist2 = dist1;
+	xx2 = inc2;
+	for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) {
+	  if (dist2 < *bptr) {
+	    *bptr = dist2;
+	    *cptr = (JSAMPLE) icolor;
+	  }
+	  dist2 += xx2;
+	  xx2 += 2 * STEP_C2 * STEP_C2;
+	  bptr++;
+	  cptr++;
+	}
+	dist1 += xx1;
+	xx1 += 2 * STEP_C1 * STEP_C1;
+      }
+      dist0 += xx0;
+      xx0 += 2 * STEP_C0 * STEP_C0;
+    }
+  }
+}
+
+
+LOCAL(void)
+fill_inverse_cmap (j_decompress_ptr cinfo, int c0, int c1, int c2)
+/* Fill the inverse-colormap entries in the update box that contains */
+/* histogram cell c0/c1/c2.  (Only that one cell MUST be filled, but */
+/* we can fill as many others as we wish.) */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  int minc0, minc1, minc2;	/* lower left corner of update box */
+  int ic0, ic1, ic2;
+  register JSAMPLE * cptr;	/* pointer into bestcolor[] array */
+  register histptr cachep;	/* pointer into main cache array */
+  /* This array lists the candidate colormap indexes. */
+  JSAMPLE colorlist[MAXNUMCOLORS];
+  int numcolors;		/* number of candidate colors */
+  /* This array holds the actually closest colormap index for each cell. */
+  JSAMPLE bestcolor[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
+
+  /* Convert cell coordinates to update box ID */
+  c0 >>= BOX_C0_LOG;
+  c1 >>= BOX_C1_LOG;
+  c2 >>= BOX_C2_LOG;
+
+  /* Compute true coordinates of update box's origin corner.
+   * Actually we compute the coordinates of the center of the corner
+   * histogram cell, which are the lower bounds of the volume we care about.
+   */
+  minc0 = (c0 << BOX_C0_SHIFT) + ((1 << C0_SHIFT) >> 1);
+  minc1 = (c1 << BOX_C1_SHIFT) + ((1 << C1_SHIFT) >> 1);
+  minc2 = (c2 << BOX_C2_SHIFT) + ((1 << C2_SHIFT) >> 1);
+  
+  /* Determine which colormap entries are close enough to be candidates
+   * for the nearest entry to some cell in the update box.
+   */
+  numcolors = find_nearby_colors(cinfo, minc0, minc1, minc2, colorlist);
+
+  /* Determine the actually nearest colors. */
+  find_best_colors(cinfo, minc0, minc1, minc2, numcolors, colorlist,
+		   bestcolor);
+
+  /* Save the best color numbers (plus 1) in the main cache array */
+  c0 <<= BOX_C0_LOG;		/* convert ID back to base cell indexes */
+  c1 <<= BOX_C1_LOG;
+  c2 <<= BOX_C2_LOG;
+  cptr = bestcolor;
+  for (ic0 = 0; ic0 < BOX_C0_ELEMS; ic0++) {
+    for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
+      cachep = & histogram[c0+ic0][c1+ic1][c2];
+      for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
+	*cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1);
+      }
+    }
+  }
+}
+
+
+/*
+ * Map some rows of pixels to the output colormapped representation.
+ */
+
+METHODDEF(void)
+pass2_no_dither (j_decompress_ptr cinfo,
+		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+/* This version performs no dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  register JSAMPROW inptr, outptr;
+  register histptr cachep;
+  register int c0, c1, c2;
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+
+  for (row = 0; row < num_rows; row++) {
+    inptr = input_buf[row];
+    outptr = output_buf[row];
+    for (col = width; col > 0; col--) {
+      /* get pixel value and index into the cache */
+      c0 = GETJSAMPLE(*inptr++) >> C0_SHIFT;
+      c1 = GETJSAMPLE(*inptr++) >> C1_SHIFT;
+      c2 = GETJSAMPLE(*inptr++) >> C2_SHIFT;
+      cachep = & histogram[c0][c1][c2];
+      /* If we have not seen this color before, find nearest colormap entry */
+      /* and update the cache */
+      if (*cachep == 0)
+	fill_inverse_cmap(cinfo, c0,c1,c2);
+      /* Now emit the colormap index for this cell */
+      *outptr++ = (JSAMPLE) (*cachep - 1);
+    }
+  }
+}
+
+
+METHODDEF(void)
+pass2_fs_dither (j_decompress_ptr cinfo,
+		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+/* This version performs Floyd-Steinberg dithering */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  register LOCFSERROR cur0, cur1, cur2;	/* current error or pixel value */
+  LOCFSERROR belowerr0, belowerr1, belowerr2; /* error for pixel below cur */
+  LOCFSERROR bpreverr0, bpreverr1, bpreverr2; /* error for below/prev col */
+  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
+  JSAMPROW inptr;		/* => current input pixel */
+  JSAMPROW outptr;		/* => current output pixel */
+  histptr cachep;
+  int dir;			/* +1 or -1 depending on direction */
+  int dir3;			/* 3*dir, for advancing inptr & errorptr */
+  int row;
+  JDIMENSION col;
+  JDIMENSION width = cinfo->output_width;
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
+  int *error_limit = cquantize->error_limiter;
+  JSAMPROW colormap0 = cinfo->colormap[0];
+  JSAMPROW colormap1 = cinfo->colormap[1];
+  JSAMPROW colormap2 = cinfo->colormap[2];
+  SHIFT_TEMPS
+
+  for (row = 0; row < num_rows; row++) {
+    inptr = input_buf[row];
+    outptr = output_buf[row];
+    if (cquantize->on_odd_row) {
+      /* work right to left in this row */
+      inptr += (width-1) * 3;	/* so point to rightmost pixel */
+      outptr += width-1;
+      dir = -1;
+      dir3 = -3;
+      errorptr = cquantize->fserrors + (width+1)*3; /* => entry after last column */
+      cquantize->on_odd_row = FALSE; /* flip for next time */
+    } else {
+      /* work left to right in this row */
+      dir = 1;
+      dir3 = 3;
+      errorptr = cquantize->fserrors; /* => entry before first real column */
+      cquantize->on_odd_row = TRUE; /* flip for next time */
+    }
+    /* Preset error values: no error propagated to first pixel from left */
+    cur0 = cur1 = cur2 = 0;
+    /* and no error propagated to row below yet */
+    belowerr0 = belowerr1 = belowerr2 = 0;
+    bpreverr0 = bpreverr1 = bpreverr2 = 0;
+
+    for (col = width; col > 0; col--) {
+      /* curN holds the error propagated from the previous pixel on the
+       * current line.  Add the error propagated from the previous line
+       * to form the complete error correction term for this pixel, and
+       * round the error term (which is expressed * 16) to an integer.
+       * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+       * for either sign of the error value.
+       * Note: errorptr points to *previous* column's array entry.
+       */
+      cur0 = RIGHT_SHIFT(cur0 + errorptr[dir3+0] + 8, 4);
+      cur1 = RIGHT_SHIFT(cur1 + errorptr[dir3+1] + 8, 4);
+      cur2 = RIGHT_SHIFT(cur2 + errorptr[dir3+2] + 8, 4);
+      /* Limit the error using transfer function set by init_error_limit.
+       * See comments with init_error_limit for rationale.
+       */
+      cur0 = error_limit[cur0];
+      cur1 = error_limit[cur1];
+      cur2 = error_limit[cur2];
+      /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+       * The maximum error is +- MAXJSAMPLE (or less with error limiting);
+       * this sets the required size of the range_limit array.
+       */
+      cur0 += GETJSAMPLE(inptr[0]);
+      cur1 += GETJSAMPLE(inptr[1]);
+      cur2 += GETJSAMPLE(inptr[2]);
+      cur0 = GETJSAMPLE(range_limit[cur0]);
+      cur1 = GETJSAMPLE(range_limit[cur1]);
+      cur2 = GETJSAMPLE(range_limit[cur2]);
+      /* Index into the cache with adjusted pixel value */
+      cachep = & histogram[cur0>>C0_SHIFT][cur1>>C1_SHIFT][cur2>>C2_SHIFT];
+      /* If we have not seen this color before, find nearest colormap */
+      /* entry and update the cache */
+      if (*cachep == 0)
+	fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT);
+      /* Now emit the colormap index for this cell */
+      { register int pixcode = *cachep - 1;
+	*outptr = (JSAMPLE) pixcode;
+	/* Compute representation error for this pixel */
+	cur0 -= GETJSAMPLE(colormap0[pixcode]);
+	cur1 -= GETJSAMPLE(colormap1[pixcode]);
+	cur2 -= GETJSAMPLE(colormap2[pixcode]);
+      }
+      /* Compute error fractions to be propagated to adjacent pixels.
+       * Add these into the running sums, and simultaneously shift the
+       * next-line error sums left by 1 column.
+       */
+      { register LOCFSERROR bnexterr, delta;
+
+	bnexterr = cur0;	/* Process component 0 */
+	delta = cur0 * 2;
+	cur0 += delta;		/* form error * 3 */
+	errorptr[0] = (FSERROR) (bpreverr0 + cur0);
+	cur0 += delta;		/* form error * 5 */
+	bpreverr0 = belowerr0 + cur0;
+	belowerr0 = bnexterr;
+	cur0 += delta;		/* form error * 7 */
+	bnexterr = cur1;	/* Process component 1 */
+	delta = cur1 * 2;
+	cur1 += delta;		/* form error * 3 */
+	errorptr[1] = (FSERROR) (bpreverr1 + cur1);
+	cur1 += delta;		/* form error * 5 */
+	bpreverr1 = belowerr1 + cur1;
+	belowerr1 = bnexterr;
+	cur1 += delta;		/* form error * 7 */
+	bnexterr = cur2;	/* Process component 2 */
+	delta = cur2 * 2;
+	cur2 += delta;		/* form error * 3 */
+	errorptr[2] = (FSERROR) (bpreverr2 + cur2);
+	cur2 += delta;		/* form error * 5 */
+	bpreverr2 = belowerr2 + cur2;
+	belowerr2 = bnexterr;
+	cur2 += delta;		/* form error * 7 */
+      }
+      /* At this point curN contains the 7/16 error value to be propagated
+       * to the next pixel on the current line, and all the errors for the
+       * next line have been shifted over.  We are therefore ready to move on.
+       */
+      inptr += dir3;		/* Advance pixel pointers to next column */
+      outptr += dir;
+      errorptr += dir3;		/* advance errorptr to current column */
+    }
+    /* Post-loop cleanup: we must unload the final error values into the
+     * final fserrors[] entry.  Note we need not unload belowerrN because
+     * it is for the dummy column before or after the actual array.
+     */
+    errorptr[0] = (FSERROR) bpreverr0; /* unload prev errs into array */
+    errorptr[1] = (FSERROR) bpreverr1;
+    errorptr[2] = (FSERROR) bpreverr2;
+  }
+}
+
+
+/*
+ * Initialize the error-limiting transfer function (lookup table).
+ * The raw F-S error computation can potentially compute error values of up to
+ * +- MAXJSAMPLE.  But we want the maximum correction applied to a pixel to be
+ * much less, otherwise obviously wrong pixels will be created.  (Typical
+ * effects include weird fringes at color-area boundaries, isolated bright
+ * pixels in a dark area, etc.)  The standard advice for avoiding this problem
+ * is to ensure that the "corners" of the color cube are allocated as output
+ * colors; then repeated errors in the same direction cannot cause cascading
+ * error buildup.  However, that only prevents the error from getting
+ * completely out of hand; Aaron Giles reports that error limiting improves
+ * the results even with corner colors allocated.
+ * A simple clamping of the error values to about +- MAXJSAMPLE/8 works pretty
+ * well, but the smoother transfer function used below is even better.  Thanks
+ * to Aaron Giles for this idea.
+ */
+
+LOCAL(void)
+init_error_limit (j_decompress_ptr cinfo)
+/* Allocate and fill in the error_limiter table */
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  int * table;
+  int in, out;
+
+  table = (int *) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * SIZEOF(int));
+  table += MAXJSAMPLE;		/* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
+  cquantize->error_limiter = table;
+
+#define STEPSIZE ((MAXJSAMPLE+1)/16)
+  /* Map errors 1:1 up to +- MAXJSAMPLE/16 */
+  out = 0;
+  for (in = 0; in < STEPSIZE; in++, out++) {
+    table[in] = out; table[-in] = -out;
+  }
+  /* Map errors 1:2 up to +- 3*MAXJSAMPLE/16 */
+  for (; in < STEPSIZE*3; in++, out += (in&1) ? 0 : 1) {
+    table[in] = out; table[-in] = -out;
+  }
+  /* Clamp the rest to final out value (which is (MAXJSAMPLE+1)/8) */
+  for (; in <= MAXJSAMPLE; in++) {
+    table[in] = out; table[-in] = -out;
+  }
+#undef STEPSIZE
+}
+
+
+/*
+ * Finish up at the end of each pass.
+ */
+
+METHODDEF(void)
+finish_pass1 (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+
+  /* Select the representative colors and fill in cinfo->colormap */
+  cinfo->colormap = cquantize->sv_colormap;
+  select_colors(cinfo, cquantize->desired);
+  /* Force next pass to zero the color index table */
+  cquantize->needs_zeroed = TRUE;
+}
+
+
+METHODDEF(void)
+finish_pass2 (j_decompress_ptr cinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * Initialize for each processing pass.
+ */
+
+METHODDEF(void)
+start_pass_2_quant (j_decompress_ptr cinfo, boolean is_pre_scan)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+  hist3d histogram = cquantize->histogram;
+  int i;
+
+  /* Only F-S dithering or no dithering is supported. */
+  /* If user asks for ordered dither, give him F-S. */
+  if (cinfo->dither_mode != JDITHER_NONE)
+    cinfo->dither_mode = JDITHER_FS;
+
+  if (is_pre_scan) {
+    /* Set up method pointers */
+    cquantize->pub.color_quantize = prescan_quantize;
+    cquantize->pub.finish_pass = finish_pass1;
+    cquantize->needs_zeroed = TRUE; /* Always zero histogram */
+  } else {
+    /* Set up method pointers */
+    if (cinfo->dither_mode == JDITHER_FS)
+      cquantize->pub.color_quantize = pass2_fs_dither;
+    else
+      cquantize->pub.color_quantize = pass2_no_dither;
+    cquantize->pub.finish_pass = finish_pass2;
+
+    /* Make sure color count is acceptable */
+    i = cinfo->actual_number_of_colors;
+    if (i < 1)
+      ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 1);
+    if (i > MAXNUMCOLORS)
+      ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS);
+
+    if (cinfo->dither_mode == JDITHER_FS) {
+      size_t arraysize = (size_t) ((cinfo->output_width + 2) *
+				   (3 * SIZEOF(FSERROR)));
+      /* Allocate Floyd-Steinberg workspace if we didn't already. */
+      if (cquantize->fserrors == NULL)
+	cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
+	  ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
+      /* Initialize the propagated errors to zero. */
+      jzero_far((void FAR *) cquantize->fserrors, arraysize);
+      /* Make the error-limit table if we didn't already. */
+      if (cquantize->error_limiter == NULL)
+	init_error_limit(cinfo);
+      cquantize->on_odd_row = FALSE;
+    }
+
+  }
+  /* Zero the histogram or inverse color map, if necessary */
+  if (cquantize->needs_zeroed) {
+    for (i = 0; i < HIST_C0_ELEMS; i++) {
+      jzero_far((void FAR *) histogram[i],
+		HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+    }
+    cquantize->needs_zeroed = FALSE;
+  }
+}
+
+
+/*
+ * Switch to a new external colormap between output passes.
+ */
+
+METHODDEF(void)
+new_color_map_2_quant (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
+
+  /* Reset the inverse color map */
+  cquantize->needs_zeroed = TRUE;
+}
+
+
+/*
+ * Module initialization routine for 2-pass color quantization.
+ */
+
+GLOBAL(void)
+jinit_2pass_quantizer (j_decompress_ptr cinfo)
+{
+  my_cquantize_ptr cquantize;
+  int i;
+
+  cquantize = (my_cquantize_ptr)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+				SIZEOF(my_cquantizer));
+  cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
+  cquantize->pub.start_pass = start_pass_2_quant;
+  cquantize->pub.new_color_map = new_color_map_2_quant;
+  cquantize->fserrors = NULL;	/* flag optional arrays not allocated */
+  cquantize->error_limiter = NULL;
+
+  /* Make sure jdmaster didn't give me a case I can't handle */
+  if (cinfo->out_color_components != 3)
+    ERREXIT(cinfo, JERR_NOTIMPL);
+
+  /* Allocate the histogram/inverse colormap storage */
+  cquantize->histogram = (hist3d) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * SIZEOF(hist2d));
+  for (i = 0; i < HIST_C0_ELEMS; i++) {
+    cquantize->histogram[i] = (hist2d) (*cinfo->mem->alloc_large)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+  }
+  cquantize->needs_zeroed = TRUE; /* histogram is garbage now */
+
+  /* Allocate storage for the completed colormap, if required.
+   * We do this now since it is FAR storage and may affect
+   * the memory manager's space calculations.
+   */
+  if (cinfo->enable_2pass_quant) {
+    /* Make sure color count is acceptable */
+    int desired = cinfo->desired_number_of_colors;
+    /* Lower bound on # of colors ... somewhat arbitrary as long as > 0 */
+    if (desired < 8)
+      ERREXIT1(cinfo, JERR_QUANT_FEW_COLORS, 8);
+    /* Make sure colormap indexes can be represented by JSAMPLEs */
+    if (desired > MAXNUMCOLORS)
+      ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, MAXNUMCOLORS);
+    cquantize->sv_colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo,JPOOL_IMAGE, (JDIMENSION) desired, (JDIMENSION) 3);
+    cquantize->desired = desired;
+  } else
+    cquantize->sv_colormap = NULL;
+
+  /* Only F-S dithering or no dithering is supported. */
+  /* If user asks for ordered dither, give him F-S. */
+  if (cinfo->dither_mode != JDITHER_NONE)
+    cinfo->dither_mode = JDITHER_FS;
+
+  /* Allocate Floyd-Steinberg workspace if necessary.
+   * This isn't really needed until pass 2, but again it is FAR storage.
+   * Although we will cope with a later change in dither_mode,
+   * we do not promise to honor max_memory_to_use if dither_mode changes.
+   */
+  if (cinfo->dither_mode == JDITHER_FS) {
+    cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (size_t) ((cinfo->output_width + 2) * (3 * SIZEOF(FSERROR))));
+    /* Might as well create the error-limiting table too. */
+    init_error_limit(cinfo);
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
diff --git a/jpeg/jutils.c b/jpeg/jutils.c
new file mode 100644
index 000000000..d18a95556
--- /dev/null
+++ b/jpeg/jutils.c
@@ -0,0 +1,179 @@
+/*
+ * jutils.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains tables and miscellaneous utility routines needed
+ * for both compression and decompression.
+ * Note we prefix all global names with "j" to minimize conflicts with
+ * a surrounding application.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+
+
+/*
+ * jpeg_zigzag_order[i] is the zigzag-order position of the i'th element
+ * of a DCT block read in natural order (left to right, top to bottom).
+ */
+
+#if 0				/* This table is not actually needed in v6a */
+
+const int jpeg_zigzag_order[DCTSIZE2] = {
+   0,  1,  5,  6, 14, 15, 27, 28,
+   2,  4,  7, 13, 16, 26, 29, 42,
+   3,  8, 12, 17, 25, 30, 41, 43,
+   9, 11, 18, 24, 31, 40, 44, 53,
+  10, 19, 23, 32, 39, 45, 52, 54,
+  20, 22, 33, 38, 46, 51, 55, 60,
+  21, 34, 37, 47, 50, 56, 59, 61,
+  35, 36, 48, 49, 57, 58, 62, 63
+};
+
+#endif
+
+/*
+ * jpeg_natural_order[i] is the natural-order position of the i'th element
+ * of zigzag order.
+ *
+ * When reading corrupted data, the Huffman decoders could attempt
+ * to reference an entry beyond the end of this array (if the decoded
+ * zero run length reaches past the end of the block).  To prevent
+ * wild stores without adding an inner-loop test, we put some extra
+ * "63"s after the real entries.  This will cause the extra coefficient
+ * to be stored in location 63 of the block, not somewhere random.
+ * The worst case would be a run-length of 15, which means we need 16
+ * fake entries.
+ */
+
+const int jpeg_natural_order[DCTSIZE2+16] = {
+  0,  1,  8, 16,  9,  2,  3, 10,
+ 17, 24, 32, 25, 18, 11,  4,  5,
+ 12, 19, 26, 33, 40, 48, 41, 34,
+ 27, 20, 13,  6,  7, 14, 21, 28,
+ 35, 42, 49, 56, 57, 50, 43, 36,
+ 29, 22, 15, 23, 30, 37, 44, 51,
+ 58, 59, 52, 45, 38, 31, 39, 46,
+ 53, 60, 61, 54, 47, 55, 62, 63,
+ 63, 63, 63, 63, 63, 63, 63, 63, /* extra entries for safety in decoder */
+ 63, 63, 63, 63, 63, 63, 63, 63
+};
+
+
+/*
+ * Arithmetic utilities
+ */
+
+GLOBAL(long)
+jdiv_round_up (long a, long b)
+/* Compute a/b rounded up to next integer, ie, ceil(a/b) */
+/* Assumes a >= 0, b > 0 */
+{
+  return (a + b - 1L) / b;
+}
+
+
+GLOBAL(long)
+jround_up (long a, long b)
+/* Compute a rounded up to next multiple of b, ie, ceil(a/b)*b */
+/* Assumes a >= 0, b > 0 */
+{
+  a += b - 1L;
+  return a - (a % b);
+}
+
+
+/* On normal machines we can apply MEMCOPY() and MEMZERO() to sample arrays
+ * and coefficient-block arrays.  This won't work on 80x86 because the arrays
+ * are FAR and we're assuming a small-pointer memory model.  However, some
+ * DOS compilers provide far-pointer versions of memcpy() and memset() even
+ * in the small-model libraries.  These will be used if USE_FMEM is defined.
+ * Otherwise, the routines below do it the hard way.  (The performance cost
+ * is not all that great, because these routines aren't very heavily used.)
+ */
+
+#ifndef NEED_FAR_POINTERS	/* normal case, same as regular macros */
+#define FMEMCOPY(dest,src,size)	MEMCOPY(dest,src,size)
+#define FMEMZERO(target,size)	MEMZERO(target,size)
+#else				/* 80x86 case, define if we can */
+#ifdef USE_FMEM
+#define FMEMCOPY(dest,src,size)	_fmemcpy((void FAR *)(dest), (const void FAR *)(src), (size_t)(size))
+#define FMEMZERO(target,size)	_fmemset((void FAR *)(target), 0, (size_t)(size))
+#endif
+#endif
+
+
+GLOBAL(void)
+jcopy_sample_rows (JSAMPARRAY input_array, int source_row,
+		   JSAMPARRAY output_array, int dest_row,
+		   int num_rows, JDIMENSION num_cols)
+/* Copy some rows of samples from one place to another.
+ * num_rows rows are copied from input_array[source_row++]
+ * to output_array[dest_row++]; these areas may overlap for duplication.
+ * The source and destination arrays must be at least as wide as num_cols.
+ */
+{
+  register JSAMPROW inptr, outptr;
+#ifdef FMEMCOPY
+  register size_t count = (size_t) (num_cols * SIZEOF(JSAMPLE));
+#else
+  register JDIMENSION count;
+#endif
+  register int row;
+
+  input_array += source_row;
+  output_array += dest_row;
+
+  for (row = num_rows; row > 0; row--) {
+    inptr = *input_array++;
+    outptr = *output_array++;
+#ifdef FMEMCOPY
+    FMEMCOPY(outptr, inptr, count);
+#else
+    for (count = num_cols; count > 0; count--)
+      *outptr++ = *inptr++;	/* needn't bother with GETJSAMPLE() here */
+#endif
+  }
+}
+
+
+GLOBAL(void)
+jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row,
+		 JDIMENSION num_blocks)
+/* Copy a row of coefficient blocks from one place to another. */
+{
+#ifdef FMEMCOPY
+  FMEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * SIZEOF(JCOEF)));
+#else
+  register JCOEFPTR inptr, outptr;
+  register long count;
+
+  inptr = (JCOEFPTR) input_row;
+  outptr = (JCOEFPTR) output_row;
+  for (count = (long) num_blocks * DCTSIZE2; count > 0; count--) {
+    *outptr++ = *inptr++;
+  }
+#endif
+}
+
+
+GLOBAL(void)
+jzero_far (void FAR * target, size_t bytestozero)
+/* Zero out a chunk of FAR memory. */
+/* This might be sample-array data, block-array data, or alloc_large data. */
+{
+#ifdef FMEMZERO
+  FMEMZERO(target, bytestozero);
+#else
+  register char FAR * ptr = (char FAR *) target;
+  register size_t count;
+
+  for (count = bytestozero; count > 0; count--) {
+    *ptr++ = 0;
+  }
+#endif
+}
diff --git a/jpeg/jversion.h b/jpeg/jversion.h
new file mode 100644
index 000000000..6472c58d3
--- /dev/null
+++ b/jpeg/jversion.h
@@ -0,0 +1,14 @@
+/*
+ * jversion.h
+ *
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains software version identification.
+ */
+
+
+#define JVERSION	"6b  27-Mar-1998"
+
+#define JCOPYRIGHT	"Copyright (C) 1998, Thomas G. Lane"
diff --git a/jpeg/libjpeg.doc b/jpeg/libjpeg.doc
new file mode 100644
index 000000000..689b206c0
--- /dev/null
+++ b/jpeg/libjpeg.doc
@@ -0,0 +1,3006 @@
+USING THE IJG JPEG LIBRARY
+
+Copyright (C) 1994-1998, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+This file describes how to use the IJG JPEG library within an application
+program.  Read it if you want to write a program that uses the library.
+
+The file example.c provides heavily commented skeleton code for calling the
+JPEG library.  Also see jpeglib.h (the include file to be used by application
+programs) for full details about data structures and function parameter lists.
+The library source code, of course, is the ultimate reference.
+
+Note that there have been *major* changes from the application interface
+presented by IJG version 4 and earlier versions.  The old design had several
+inherent limitations, and it had accumulated a lot of cruft as we added
+features while trying to minimize application-interface changes.  We have
+sacrificed backward compatibility in the version 5 rewrite, but we think the
+improvements justify this.
+
+
+TABLE OF CONTENTS
+-----------------
+
+Overview:
+	Functions provided by the library
+	Outline of typical usage
+Basic library usage:
+	Data formats
+	Compression details
+	Decompression details
+	Mechanics of usage: include files, linking, etc
+Advanced features:
+	Compression parameter selection
+	Decompression parameter selection
+	Special color spaces
+	Error handling
+	Compressed data handling (source and destination managers)
+	I/O suspension
+	Progressive JPEG support
+	Buffered-image mode
+	Abbreviated datastreams and multiple images
+	Special markers
+	Raw (downsampled) image data
+	Really raw data: DCT coefficients
+	Progress monitoring
+	Memory management
+	Memory usage
+	Library compile-time options
+	Portability considerations
+	Notes for MS-DOS implementors
+
+You should read at least the overview and basic usage sections before trying
+to program with the library.  The sections on advanced features can be read
+if and when you need them.
+
+
+OVERVIEW
+========
+
+Functions provided by the library
+---------------------------------
+
+The IJG JPEG library provides C code to read and write JPEG-compressed image
+files.  The surrounding application program receives or supplies image data a
+scanline at a time, using a straightforward uncompressed image format.  All
+details of color conversion and other preprocessing/postprocessing can be
+handled by the library.
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  The application indirectly selects use of this code
+by specifying the format in which it wishes to supply or receive image data.
+For example, if colormapped output is requested, then the decompression
+library automatically invokes color quantization.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+A word about functions *not* provided by the library.  We handle a subset of
+the ISO JPEG standard; most baseline, extended-sequential, and progressive
+JPEG processes are supported.  (Our subset includes all features now in common
+use.)  Unsupported ISO options include:
+	* Hierarchical storage
+	* Lossless JPEG
+	* Arithmetic entropy coding (unsupported for legal reasons)
+	* DNL marker
+	* Nonintegral subsampling ratios
+We support both 8- and 12-bit data precision, but this is a compile-time
+choice rather than a run-time choice; hence it is difficult to use both
+precisions in a single application.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, this library is
+used by the free LIBTIFF library to support JPEG compression in TIFF.)
+
+
+Outline of typical usage
+------------------------
+
+The rough outline of a JPEG compression operation is:
+
+	Allocate and initialize a JPEG compression object
+	Specify the destination for the compressed data (eg, a file)
+	Set parameters for compression, including image size & colorspace
+	jpeg_start_compress(...);
+	while (scan lines remain to be written)
+		jpeg_write_scanlines(...);
+	jpeg_finish_compress(...);
+	Release the JPEG compression object
+
+A JPEG compression object holds parameters and working state for the JPEG
+library.  We make creation/destruction of the object separate from starting
+or finishing compression of an image; the same object can be re-used for a
+series of image compression operations.  This makes it easy to re-use the
+same parameter settings for a sequence of images.  Re-use of a JPEG object
+also has important implications for processing abbreviated JPEG datastreams,
+as discussed later.
+
+The image data to be compressed is supplied to jpeg_write_scanlines() from
+in-memory buffers.  If the application is doing file-to-file compression,
+reading image data from the source file is the application's responsibility.
+The library emits compressed data by calling a "data destination manager",
+which typically will write the data into a file; but the application can
+provide its own destination manager to do something else.
+
+Similarly, the rough outline of a JPEG decompression operation is:
+
+	Allocate and initialize a JPEG decompression object
+	Specify the source of the compressed data (eg, a file)
+	Call jpeg_read_header() to obtain image info
+	Set parameters for decompression
+	jpeg_start_decompress(...);
+	while (scan lines remain to be read)
+		jpeg_read_scanlines(...);
+	jpeg_finish_decompress(...);
+	Release the JPEG decompression object
+
+This is comparable to the compression outline except that reading the
+datastream header is a separate step.  This is helpful because information
+about the image's size, colorspace, etc is available when the application
+selects decompression parameters.  For example, the application can choose an
+output scaling ratio that will fit the image into the available screen size.
+
+The decompression library obtains compressed data by calling a data source
+manager, which typically will read the data from a file; but other behaviors
+can be obtained with a custom source manager.  Decompressed data is delivered
+into in-memory buffers passed to jpeg_read_scanlines().
+
+It is possible to abort an incomplete compression or decompression operation
+by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
+simply release it by calling jpeg_destroy().
+
+JPEG compression and decompression objects are two separate struct types.
+However, they share some common fields, and certain routines such as
+jpeg_destroy() can work on either type of object.
+
+The JPEG library has no static variables: all state is in the compression
+or decompression object.  Therefore it is possible to process multiple
+compression and decompression operations concurrently, using multiple JPEG
+objects.
+
+Both compression and decompression can be done in an incremental memory-to-
+memory fashion, if suitable source/destination managers are used.  See the
+section on "I/O suspension" for more details.
+
+
+BASIC LIBRARY USAGE
+===================
+
+Data formats
+------------
+
+Before diving into procedural details, it is helpful to understand the
+image data format that the JPEG library expects or returns.
+
+The standard input image format is a rectangular array of pixels, with each
+pixel having the same number of "component" or "sample" values (color
+channels).  You must specify how many components there are and the colorspace
+interpretation of the components.  Most applications will use RGB data
+(three components per pixel) or grayscale data (one component per pixel).
+PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
+A remarkable number of people manage to miss this, only to find that their
+programs don't work with grayscale JPEG files.
+
+There is no provision for colormapped input.  JPEG files are always full-color
+or full grayscale (or sometimes another colorspace such as CMYK).  You can
+feed in a colormapped image by expanding it to full-color format.  However
+JPEG often doesn't work very well with source data that has been colormapped,
+because of dithering noise.  This is discussed in more detail in the JPEG FAQ
+and the other references mentioned in the README file.
+
+Pixels are stored by scanlines, with each scanline running from left to
+right.  The component values for each pixel are adjacent in the row; for
+example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color.  Each scanline is an
+array of data type JSAMPLE --- which is typically "unsigned char", unless
+you've changed jmorecfg.h.  (You can also change the RGB pixel layout, say
+to B,G,R order, by modifying jmorecfg.h.  But see the restrictions listed in
+that file before doing so.)
+
+A 2-D array of pixels is formed by making a list of pointers to the starts of
+scanlines; so the scanlines need not be physically adjacent in memory.  Even
+if you process just one scanline at a time, you must make a one-element
+pointer array to conform to this structure.  Pointers to JSAMPLE rows are of
+type JSAMPROW, and the pointer to the pointer array is of type JSAMPARRAY.
+
+The library accepts or supplies one or more complete scanlines per call.
+It is not possible to process part of a row at a time.  Scanlines are always
+processed top-to-bottom.  You can process an entire image in one call if you
+have it all in memory, but usually it's simplest to process one scanline at
+a time.
+
+For best results, source data values should have the precision specified by
+BITS_IN_JSAMPLE (normally 8 bits).  For instance, if you choose to compress
+data that's only 6 bits/channel, you should left-justify each value in a
+byte before passing it to the compressor.  If you need to compress data
+that has more than 8 bits/channel, compile with BITS_IN_JSAMPLE = 12.
+(See "Library compile-time options", later.)
+
+
+The data format returned by the decompressor is the same in all details,
+except that colormapped output is supported.  (Again, a JPEG file is never
+colormapped.  But you can ask the decompressor to perform on-the-fly color
+quantization to deliver colormapped output.)  If you request colormapped
+output then the returned data array contains a single JSAMPLE per pixel;
+its value is an index into a color map.  The color map is represented as
+a 2-D JSAMPARRAY in which each row holds the values of one color component,
+that is, colormap[i][j] is the value of the i'th color component for pixel
+value (map index) j.  Note that since the colormap indexes are stored in
+JSAMPLEs, the maximum number of colors is limited by the size of JSAMPLE
+(ie, at most 256 colors for an 8-bit JPEG library).
+
+
+Compression details
+-------------------
+
+Here we revisit the JPEG compression outline given in the overview.
+
+1. Allocate and initialize a JPEG compression object.
+
+A JPEG compression object is a "struct jpeg_compress_struct".  (It also has
+a bunch of subsidiary structures which are allocated via malloc(), but the
+application doesn't control those directly.)  This struct can be just a local
+variable in the calling routine, if a single routine is going to execute the
+whole JPEG compression sequence.  Otherwise it can be static or allocated
+from malloc().
+
+You will also need a structure representing a JPEG error handler.  The part
+of this that the library cares about is a "struct jpeg_error_mgr".  If you
+are providing your own error handler, you'll typically want to embed the
+jpeg_error_mgr struct in a larger structure; this is discussed later under
+"Error handling".  For now we'll assume you are just using the default error
+handler.  The default error handler will print JPEG error/warning messages
+on stderr, and it will call exit() if a fatal error occurs.
+
+You must initialize the error handler structure, store a pointer to it into
+the JPEG object's "err" field, and then call jpeg_create_compress() to
+initialize the rest of the JPEG object.
+
+Typical code for this step, if you are using the default error handler, is
+
+	struct jpeg_compress_struct cinfo;
+	struct jpeg_error_mgr jerr;
+	...
+	cinfo.err = jpeg_std_error(&jerr);
+	jpeg_create_compress(&cinfo);
+
+jpeg_create_compress allocates a small amount of memory, so it could fail
+if you are out of memory.  In that case it will exit via the error handler;
+that's why the error handler must be initialized first.
+
+
+2. Specify the destination for the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library delivers compressed data to a
+"data destination" module.  The library includes one data destination
+module which knows how to write to a stdio stream.  You can use your own
+destination module if you want to do something else, as discussed later.
+
+If you use the standard destination module, you must open the target stdio
+stream beforehand.  Typical code for this step looks like:
+
+	FILE * outfile;
+	...
+	if ((outfile = fopen(filename, "wb")) == NULL) {
+	    fprintf(stderr, "can't open %s\n", filename);
+	    exit(1);
+	}
+	jpeg_stdio_dest(&cinfo, outfile);
+
+where the last line invokes the standard destination module.
+
+WARNING: it is critical that the binary compressed data be delivered to the
+output file unchanged.  On non-Unix systems the stdio library may perform
+newline translation or otherwise corrupt binary data.  To suppress this
+behavior, you may need to use a "b" option to fopen (as shown above), or use
+setmode() or another routine to put the stdio stream in binary mode.  See
+cjpeg.c and djpeg.c for code that has been found to work on many systems.
+
+You can select the data destination after setting other parameters (step 3),
+if that's more convenient.  You may not change the destination between
+calling jpeg_start_compress() and jpeg_finish_compress().
+
+
+3. Set parameters for compression, including image size & colorspace.
+
+You must supply information about the source image by setting the following
+fields in the JPEG object (cinfo structure):
+
+	image_width		Width of image, in pixels
+	image_height		Height of image, in pixels
+	input_components	Number of color channels (samples per pixel)
+	in_color_space		Color space of source image
+
+The image dimensions are, hopefully, obvious.  JPEG supports image dimensions
+of 1 to 64K pixels in either direction.  The input color space is typically
+RGB or grayscale, and input_components is 3 or 1 accordingly.  (See "Special
+color spaces", later, for more info.)  The in_color_space field must be
+assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
+JCS_GRAYSCALE.
+
+JPEG has a large number of compression parameters that determine how the
+image is encoded.  Most applications don't need or want to know about all
+these parameters.  You can set all the parameters to reasonable defaults by
+calling jpeg_set_defaults(); then, if there are particular values you want
+to change, you can do so after that.  The "Compression parameter selection"
+section tells about all the parameters.
+
+You must set in_color_space correctly before calling jpeg_set_defaults(),
+because the defaults depend on the source image colorspace.  However the
+other three source image parameters need not be valid until you call
+jpeg_start_compress().  There's no harm in calling jpeg_set_defaults() more
+than once, if that happens to be convenient.
+
+Typical code for a 24-bit RGB source image is
+
+	cinfo.image_width = Width; 	/* image width and height, in pixels */
+	cinfo.image_height = Height;
+	cinfo.input_components = 3;	/* # of color components per pixel */
+	cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
+
+	jpeg_set_defaults(&cinfo);
+	/* Make optional parameter settings here */
+
+
+4. jpeg_start_compress(...);
+
+After you have established the data destination and set all the necessary
+source image info and other parameters, call jpeg_start_compress() to begin
+a compression cycle.  This will initialize internal state, allocate working
+storage, and emit the first few bytes of the JPEG datastream header.
+
+Typical code:
+
+	jpeg_start_compress(&cinfo, TRUE);
+
+The "TRUE" parameter ensures that a complete JPEG interchange datastream
+will be written.  This is appropriate in most cases.  If you think you might
+want to use an abbreviated datastream, read the section on abbreviated
+datastreams, below.
+
+Once you have called jpeg_start_compress(), you may not alter any JPEG
+parameters or other fields of the JPEG object until you have completed
+the compression cycle.
+
+
+5. while (scan lines remain to be written)
+	jpeg_write_scanlines(...);
+
+Now write all the required image data by calling jpeg_write_scanlines()
+one or more times.  You can pass one or more scanlines in each call, up
+to the total image height.  In most applications it is convenient to pass
+just one or a few scanlines at a time.  The expected format for the passed
+data is discussed under "Data formats", above.
+
+Image data should be written in top-to-bottom scanline order.  The JPEG spec
+contains some weasel wording about how top and bottom are application-defined
+terms (a curious interpretation of the English language...) but if you want
+your files to be compatible with everyone else's, you WILL use top-to-bottom
+order.  If the source data must be read in bottom-to-top order, you can use
+the JPEG library's virtual array mechanism to invert the data efficiently.
+Examples of this can be found in the sample application cjpeg.
+
+The library maintains a count of the number of scanlines written so far
+in the next_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.next_scanline < cinfo.image_height)".
+
+Code for this step depends heavily on the way that you store the source data.
+example.c shows the following code for the case of a full-size 2-D source
+array containing 3-byte RGB pixels:
+
+	JSAMPROW row_pointer[1];	/* pointer to a single row */
+	int row_stride;			/* physical row width in buffer */
+
+	row_stride = image_width * 3;	/* JSAMPLEs per row in image_buffer */
+
+	while (cinfo.next_scanline < cinfo.image_height) {
+	    row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+	    jpeg_write_scanlines(&cinfo, row_pointer, 1);
+	}
+
+jpeg_write_scanlines() returns the number of scanlines actually written.
+This will normally be equal to the number passed in, so you can usually
+ignore the return value.  It is different in just two cases:
+  * If you try to write more scanlines than the declared image height,
+    the additional scanlines are ignored.
+  * If you use a suspending data destination manager, output buffer overrun
+    will cause the compressor to return before accepting all the passed lines.
+    This feature is discussed under "I/O suspension", below.  The normal
+    stdio destination manager will NOT cause this to happen.
+In any case, the return value is the same as the change in the value of
+next_scanline.
+
+
+6. jpeg_finish_compress(...);
+
+After all the image data has been written, call jpeg_finish_compress() to
+complete the compression cycle.  This step is ESSENTIAL to ensure that the
+last bufferload of data is written to the data destination.
+jpeg_finish_compress() also releases working memory associated with the JPEG
+object.
+
+Typical code:
+
+	jpeg_finish_compress(&cinfo);
+
+If using the stdio destination manager, don't forget to close the output
+stdio stream (if necessary) afterwards.
+
+If you have requested a multi-pass operating mode, such as Huffman code
+optimization, jpeg_finish_compress() will perform the additional passes using
+data buffered by the first pass.  In this case jpeg_finish_compress() may take
+quite a while to complete.  With the default compression parameters, this will
+not happen.
+
+It is an error to call jpeg_finish_compress() before writing the necessary
+total number of scanlines.  If you wish to abort compression, call
+jpeg_abort() as discussed below.
+
+After completing a compression cycle, you may dispose of the JPEG object
+as discussed next, or you may use it to compress another image.  In that case
+return to step 2, 3, or 4 as appropriate.  If you do not change the
+destination manager, the new datastream will be written to the same target.
+If you do not change any JPEG parameters, the new datastream will be written
+with the same parameters as before.  Note that you can change the input image
+dimensions freely between cycles, but if you change the input colorspace, you
+should call jpeg_set_defaults() to adjust for the new colorspace; and then
+you'll need to repeat all of step 3.
+
+
+7. Release the JPEG compression object.
+
+When you are done with a JPEG compression object, destroy it by calling
+jpeg_destroy_compress().  This will free all subsidiary memory (regardless of
+the previous state of the object).  Or you can call jpeg_destroy(), which
+works for either compression or decompression objects --- this may be more
+convenient if you are sharing code between compression and decompression
+cases.  (Actually, these routines are equivalent except for the declared type
+of the passed pointer.  To avoid gripes from ANSI C compilers, jpeg_destroy()
+should be passed a j_common_ptr.)
+
+If you allocated the jpeg_compress_struct structure from malloc(), freeing
+it is your responsibility --- jpeg_destroy() won't.  Ditto for the error
+handler structure.
+
+Typical code:
+
+	jpeg_destroy_compress(&cinfo);
+
+
+8. Aborting.
+
+If you decide to abort a compression cycle before finishing, you can clean up
+in either of two ways:
+
+* If you don't need the JPEG object any more, just call
+  jpeg_destroy_compress() or jpeg_destroy() to release memory.  This is
+  legitimate at any point after calling jpeg_create_compress() --- in fact,
+  it's safe even if jpeg_create_compress() fails.
+
+* If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
+  jpeg_abort() which works on both compression and decompression objects.
+  This will return the object to an idle state, releasing any working memory.
+  jpeg_abort() is allowed at any time after successful object creation.
+
+Note that cleaning up the data destination, if required, is your
+responsibility; neither of these routines will call term_destination().
+(See "Compressed data handling", below, for more about that.)
+
+jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
+object that has reported an error by calling error_exit (see "Error handling"
+for more info).  The internal state of such an object is likely to be out of
+whack.  Either of these two routines will return the object to a known state.
+
+
+Decompression details
+---------------------
+
+Here we revisit the JPEG decompression outline given in the overview.
+
+1. Allocate and initialize a JPEG decompression object.
+
+This is just like initialization for compression, as discussed above,
+except that the object is a "struct jpeg_decompress_struct" and you
+call jpeg_create_decompress().  Error handling is exactly the same.
+
+Typical code:
+
+	struct jpeg_decompress_struct cinfo;
+	struct jpeg_error_mgr jerr;
+	...
+	cinfo.err = jpeg_std_error(&jerr);
+	jpeg_create_decompress(&cinfo);
+
+(Both here and in the IJG code, we usually use variable name "cinfo" for
+both compression and decompression objects.)
+
+
+2. Specify the source of the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library reads compressed data from a "data
+source" module.  The library includes one data source module which knows how
+to read from a stdio stream.  You can use your own source module if you want
+to do something else, as discussed later.
+
+If you use the standard source module, you must open the source stdio stream
+beforehand.  Typical code for this step looks like:
+
+	FILE * infile;
+	...
+	if ((infile = fopen(filename, "rb")) == NULL) {
+	    fprintf(stderr, "can't open %s\n", filename);
+	    exit(1);
+	}
+	jpeg_stdio_src(&cinfo, infile);
+
+where the last line invokes the standard source module.
+
+WARNING: it is critical that the binary compressed data be read unchanged.
+On non-Unix systems the stdio library may perform newline translation or
+otherwise corrupt binary data.  To suppress this behavior, you may need to use
+a "b" option to fopen (as shown above), or use setmode() or another routine to
+put the stdio stream in binary mode.  See cjpeg.c and djpeg.c for code that
+has been found to work on many systems.
+
+You may not change the data source between calling jpeg_read_header() and
+jpeg_finish_decompress().  If you wish to read a series of JPEG images from
+a single source file, you should repeat the jpeg_read_header() to
+jpeg_finish_decompress() sequence without reinitializing either the JPEG
+object or the data source module; this prevents buffered input data from
+being discarded.
+
+
+3. Call jpeg_read_header() to obtain image info.
+
+Typical code for this step is just
+
+	jpeg_read_header(&cinfo, TRUE);
+
+This will read the source datastream header markers, up to the beginning
+of the compressed data proper.  On return, the image dimensions and other
+info have been stored in the JPEG object.  The application may wish to
+consult this information before selecting decompression parameters.
+
+More complex code is necessary if
+  * A suspending data source is used --- in that case jpeg_read_header()
+    may return before it has read all the header data.  See "I/O suspension",
+    below.  The normal stdio source manager will NOT cause this to happen.
+  * Abbreviated JPEG files are to be processed --- see the section on
+    abbreviated datastreams.  Standard applications that deal only in
+    interchange JPEG files need not be concerned with this case either.
+
+It is permissible to stop at this point if you just wanted to find out the
+image dimensions and other header info for a JPEG file.  In that case,
+call jpeg_destroy() when you are done with the JPEG object, or call
+jpeg_abort() to return it to an idle state before selecting a new data
+source and reading another header.
+
+
+4. Set parameters for decompression.
+
+jpeg_read_header() sets appropriate default decompression parameters based on
+the properties of the image (in particular, its colorspace).  However, you
+may well want to alter these defaults before beginning the decompression.
+For example, the default is to produce full color output from a color file.
+If you want colormapped output you must ask for it.  Other options allow the
+returned image to be scaled and allow various speed/quality tradeoffs to be
+selected.  "Decompression parameter selection", below, gives details.
+
+If the defaults are appropriate, nothing need be done at this step.
+
+Note that all default values are set by each call to jpeg_read_header().
+If you reuse a decompression object, you cannot expect your parameter
+settings to be preserved across cycles, as you can for compression.
+You must set desired parameter values each time.
+
+
+5. jpeg_start_decompress(...);
+
+Once the parameter values are satisfactory, call jpeg_start_decompress() to
+begin decompression.  This will initialize internal state, allocate working
+memory, and prepare for returning data.
+
+Typical code is just
+
+	jpeg_start_decompress(&cinfo);
+
+If you have requested a multi-pass operating mode, such as 2-pass color
+quantization, jpeg_start_decompress() will do everything needed before data
+output can begin.  In this case jpeg_start_decompress() may take quite a while
+to complete.  With a single-scan (non progressive) JPEG file and default
+decompression parameters, this will not happen; jpeg_start_decompress() will
+return quickly.
+
+After this call, the final output image dimensions, including any requested
+scaling, are available in the JPEG object; so is the selected colormap, if
+colormapped output has been requested.  Useful fields include
+
+	output_width		image width and height, as scaled
+	output_height
+	out_color_components	# of color components in out_color_space
+	output_components	# of color components returned per pixel
+	colormap		the selected colormap, if any
+	actual_number_of_colors		number of entries in colormap
+
+output_components is 1 (a colormap index) when quantizing colors; otherwise it
+equals out_color_components.  It is the number of JSAMPLE values that will be
+emitted per pixel in the output arrays.
+
+Typically you will need to allocate data buffers to hold the incoming image.
+You will need output_width * output_components JSAMPLEs per scanline in your
+output buffer, and a total of output_height scanlines will be returned.
+
+Note: if you are using the JPEG library's internal memory manager to allocate
+data buffers (as djpeg does), then the manager's protocol requires that you
+request large buffers *before* calling jpeg_start_decompress().  This is a
+little tricky since the output_XXX fields are not normally valid then.  You
+can make them valid by calling jpeg_calc_output_dimensions() after setting the
+relevant parameters (scaling, output color space, and quantization flag).
+
+
+6. while (scan lines remain to be read)
+	jpeg_read_scanlines(...);
+
+Now you can read the decompressed image data by calling jpeg_read_scanlines()
+one or more times.  At each call, you pass in the maximum number of scanlines
+to be read (ie, the height of your working buffer); jpeg_read_scanlines()
+will return up to that many lines.  The return value is the number of lines
+actually read.  The format of the returned data is discussed under "Data
+formats", above.  Don't forget that grayscale and color JPEGs will return
+different data formats!
+
+Image data is returned in top-to-bottom scanline order.  If you must write
+out the image in bottom-to-top order, you can use the JPEG library's virtual
+array mechanism to invert the data efficiently.  Examples of this can be
+found in the sample application djpeg.
+
+The library maintains a count of the number of scanlines returned so far
+in the output_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.output_scanline < cinfo.output_height)".  (Note that the test
+should NOT be against image_height, unless you never use scaling.  The
+image_height field is the height of the original unscaled image.)
+The return value always equals the change in the value of output_scanline.
+
+If you don't use a suspending data source, it is safe to assume that
+jpeg_read_scanlines() reads at least one scanline per call, until the
+bottom of the image has been reached.
+
+If you use a buffer larger than one scanline, it is NOT safe to assume that
+jpeg_read_scanlines() fills it.  (The current implementation returns only a
+few scanlines per call, no matter how large a buffer you pass.)  So you must
+always provide a loop that calls jpeg_read_scanlines() repeatedly until the
+whole image has been read.
+
+
+7. jpeg_finish_decompress(...);
+
+After all the image data has been read, call jpeg_finish_decompress() to
+complete the decompression cycle.  This causes working memory associated
+with the JPEG object to be released.
+
+Typical code:
+
+	jpeg_finish_decompress(&cinfo);
+
+If using the stdio source manager, don't forget to close the source stdio
+stream if necessary.
+
+It is an error to call jpeg_finish_decompress() before reading the correct
+total number of scanlines.  If you wish to abort decompression, call
+jpeg_abort() as discussed below.
+
+After completing a decompression cycle, you may dispose of the JPEG object as
+discussed next, or you may use it to decompress another image.  In that case
+return to step 2 or 3 as appropriate.  If you do not change the source
+manager, the next image will be read from the same source.
+
+
+8. Release the JPEG decompression object.
+
+When you are done with a JPEG decompression object, destroy it by calling
+jpeg_destroy_decompress() or jpeg_destroy().  The previous discussion of
+destroying compression objects applies here too.
+
+Typical code:
+
+	jpeg_destroy_decompress(&cinfo);
+
+
+9. Aborting.
+
+You can abort a decompression cycle by calling jpeg_destroy_decompress() or
+jpeg_destroy() if you don't need the JPEG object any more, or
+jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
+The previous discussion of aborting compression cycles applies here too.
+
+
+Mechanics of usage: include files, linking, etc
+-----------------------------------------------
+
+Applications using the JPEG library should include the header file jpeglib.h
+to obtain declarations of data types and routines.  Before including
+jpeglib.h, include system headers that define at least the typedefs FILE and
+size_t.  On ANSI-conforming systems, including <stdio.h> is sufficient; on
+older Unix systems, you may need <sys/types.h> to define size_t.
+
+If the application needs to refer to individual JPEG library error codes, also
+include jerror.h to define those symbols.
+
+jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h.  If you are
+installing the JPEG header files in a system directory, you will want to
+install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
+
+The most convenient way to include the JPEG code into your executable program
+is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
+machines) and reference it at your link step.  If you use only half of the
+library (only compression or only decompression), only that much code will be
+included from the library, unless your linker is hopelessly brain-damaged.
+The supplied makefiles build libjpeg.a automatically (see install.doc).
+
+While you can build the JPEG library as a shared library if the whim strikes
+you, we don't really recommend it.  The trouble with shared libraries is that
+at some point you'll probably try to substitute a new version of the library
+without recompiling the calling applications.  That generally doesn't work
+because the parameter struct declarations usually change with each new
+version.  In other words, the library's API is *not* guaranteed binary
+compatible across versions; we only try to ensure source-code compatibility.
+(In hindsight, it might have been smarter to hide the parameter structs from
+applications and introduce a ton of access functions instead.  Too late now,
+however.)
+
+On some systems your application may need to set up a signal handler to ensure
+that temporary files are deleted if the program is interrupted.  This is most
+critical if you are on MS-DOS and use the jmemdos.c memory manager back end;
+it will try to grab extended memory for temp files, and that space will NOT be
+freed automatically.  See cjpeg.c or djpeg.c for an example signal handler.
+
+It may be worth pointing out that the core JPEG library does not actually
+require the stdio library: only the default source/destination managers and
+error handler need it.  You can use the library in a stdio-less environment
+if you replace those modules and use jmemnobs.c (or another memory manager of
+your own devising).  More info about the minimum system library requirements
+may be found in jinclude.h.
+
+
+ADVANCED FEATURES
+=================
+
+Compression parameter selection
+-------------------------------
+
+This section describes all the optional parameters you can set for JPEG
+compression, as well as the "helper" routines provided to assist in this
+task.  Proper setting of some parameters requires detailed understanding
+of the JPEG standard; if you don't know what a parameter is for, it's best
+not to mess with it!  See REFERENCES in the README file for pointers to
+more info about JPEG.
+
+It's a good idea to call jpeg_set_defaults() first, even if you plan to set
+all the parameters; that way your code is more likely to work with future JPEG
+libraries that have additional parameters.  For the same reason, we recommend
+you use a helper routine where one is provided, in preference to twiddling
+cinfo fields directly.
+
+The helper routines are:
+
+jpeg_set_defaults (j_compress_ptr cinfo)
+	This routine sets all JPEG parameters to reasonable defaults, using
+	only the input image's color space (field in_color_space, which must
+	already be set in cinfo).  Many applications will only need to use
+	this routine and perhaps jpeg_set_quality().
+
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+	Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
+	and sets other color-space-dependent parameters appropriately.  See
+	"Special color spaces", below, before using this.  A large number of
+	parameters, including all per-component parameters, are set by this
+	routine; if you want to twiddle individual parameters you should call
+	jpeg_set_colorspace() before rather than after.
+
+jpeg_default_colorspace (j_compress_ptr cinfo)
+	Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
+	and calls jpeg_set_colorspace().  This is actually a subroutine of
+	jpeg_set_defaults().  It's broken out in case you want to change
+	just the colorspace-dependent JPEG parameters.
+
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+	Constructs JPEG quantization tables appropriate for the indicated
+	quality setting.  The quality value is expressed on the 0..100 scale
+	recommended by IJG (cjpeg's "-quality" switch uses this routine).
+	Note that the exact mapping from quality values to tables may change
+	in future IJG releases as more is learned about DCT quantization.
+	If the force_baseline parameter is TRUE, then the quantization table
+	entries are constrained to the range 1..255 for full JPEG baseline
+	compatibility.  In the current implementation, this only makes a
+	difference for quality settings below 25, and it effectively prevents
+	very small/low quality files from being generated.  The IJG decoder
+	is capable of reading the non-baseline files generated at low quality
+	settings when force_baseline is FALSE, but other decoders may not be.
+
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+			 boolean force_baseline)
+	Same as jpeg_set_quality() except that the generated tables are the
+	sample tables given in the JPEC spec section K.1, multiplied by the
+	specified scale factor (which is expressed as a percentage; thus
+	scale_factor = 100 reproduces the spec's tables).  Note that larger
+	scale factors give lower quality.  This entry point is useful for
+	conforming to the Adobe PostScript DCT conventions, but we do not
+	recommend linear scaling as a user-visible quality scale otherwise.
+	force_baseline again constrains the computed table entries to 1..255.
+
+int jpeg_quality_scaling (int quality)
+	Converts a value on the IJG-recommended quality scale to a linear
+	scaling percentage.  Note that this routine may change or go away
+	in future releases --- IJG may choose to adopt a scaling method that
+	can't be expressed as a simple scalar multiplier, in which case the
+	premise of this routine collapses.  Caveat user.
+
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+		      const unsigned int *basic_table,
+		      int scale_factor, boolean force_baseline)
+	Allows an arbitrary quantization table to be created.  which_tbl
+	indicates which table slot to fill.  basic_table points to an array
+	of 64 unsigned ints given in normal array order.  These values are
+	multiplied by scale_factor/100 and then clamped to the range 1..65535
+	(or to 1..255 if force_baseline is TRUE).
+	CAUTION: prior to library version 6a, jpeg_add_quant_table expected
+	the basic table to be given in JPEG zigzag order.  If you need to
+	write code that works with either older or newer versions of this
+	routine, you must check the library version number.  Something like
+	"#if JPEG_LIB_VERSION >= 61" is the right test.
+
+jpeg_simple_progression (j_compress_ptr cinfo)
+	Generates a default scan script for writing a progressive-JPEG file.
+	This is the recommended method of creating a progressive file,
+	unless you want to make a custom scan sequence.  You must ensure that
+	the JPEG color space is set correctly before calling this routine.
+
+
+Compression parameters (cinfo fields) include:
+
+J_DCT_METHOD dct_method
+	Selects the algorithm used for the DCT step.  Choices are:
+		JDCT_ISLOW: slow but accurate integer algorithm
+		JDCT_IFAST: faster, less accurate integer method
+		JDCT_FLOAT: floating-point method
+		JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+		JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+	The FLOAT method is very slightly more accurate than the ISLOW method,
+	but may give different results on different machines due to varying
+	roundoff behavior.  The integer methods should give the same results
+	on all machines.  On machines with sufficiently fast FP hardware, the
+	floating-point method may also be the fastest.  The IFAST method is
+	considerably less accurate than the other two; its use is not
+	recommended if high quality is a concern.  JDCT_DEFAULT and
+	JDCT_FASTEST are macros configurable by each installation.
+
+J_COLOR_SPACE jpeg_color_space
+int num_components
+	The JPEG color space and corresponding number of components; see
+	"Special color spaces", below, for more info.  We recommend using
+	jpeg_set_color_space() if you want to change these.
+
+boolean optimize_coding
+	TRUE causes the compressor to compute optimal Huffman coding tables
+	for the image.  This requires an extra pass over the data and
+	therefore costs a good deal of space and time.  The default is
+	FALSE, which tells the compressor to use the supplied or default
+	Huffman tables.  In most cases optimal tables save only a few percent
+	of file size compared to the default tables.  Note that when this is
+	TRUE, you need not supply Huffman tables at all, and any you do
+	supply will be overwritten.
+
+unsigned int restart_interval
+int restart_in_rows
+	To emit restart markers in the JPEG file, set one of these nonzero.
+	Set restart_interval to specify the exact interval in MCU blocks.
+	Set restart_in_rows to specify the interval in MCU rows.  (If
+	restart_in_rows is not 0, then restart_interval is set after the
+	image width in MCUs is computed.)  Defaults are zero (no restarts).
+	One restart marker per MCU row is often a good choice.
+	NOTE: the overhead of restart markers is higher in grayscale JPEG
+	files than in color files, and MUCH higher in progressive JPEGs.
+	If you use restarts, you may want to use larger intervals in those
+	cases.
+
+const jpeg_scan_info * scan_info
+int num_scans
+	By default, scan_info is NULL; this causes the compressor to write a
+	single-scan sequential JPEG file.  If not NULL, scan_info points to
+	an array of scan definition records of length num_scans.  The
+	compressor will then write a JPEG file having one scan for each scan
+	definition record.  This is used to generate noninterleaved or
+	progressive JPEG files.  The library checks that the scan array
+	defines a valid JPEG scan sequence.  (jpeg_simple_progression creates
+	a suitable scan definition array for progressive JPEG.)  This is
+	discussed further under "Progressive JPEG support".
+
+int smoothing_factor
+	If non-zero, the input image is smoothed; the value should be 1 for
+	minimal smoothing to 100 for maximum smoothing.  Consult jcsample.c
+	for details of the smoothing algorithm.  The default is zero.
+
+boolean write_JFIF_header
+	If TRUE, a JFIF APP0 marker is emitted.  jpeg_set_defaults() and
+	jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
+	(ie, YCbCr or grayscale) is selected, otherwise FALSE.
+
+UINT8 JFIF_major_version
+UINT8 JFIF_minor_version
+	The version number to be written into the JFIF marker.
+	jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
+	You should set it to 1.02 (major=1, minor=2) if you plan to write
+	any JFIF 1.02 extension markers.
+
+UINT8 density_unit
+UINT16 X_density
+UINT16 Y_density
+	The resolution information to be written into the JFIF marker;
+	not used otherwise.  density_unit may be 0 for unknown,
+	1 for dots/inch, or 2 for dots/cm.  The default values are 0,1,1
+	indicating square pixels of unknown size.
+
+boolean write_Adobe_marker
+	If TRUE, an Adobe APP14 marker is emitted.  jpeg_set_defaults() and
+	jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
+	or YCCK is selected, otherwise FALSE.  It is generally a bad idea
+	to set both write_JFIF_header and write_Adobe_marker.  In fact,
+	you probably shouldn't change the default settings at all --- the
+	default behavior ensures that the JPEG file's color space can be
+	recognized by the decoder.
+
+JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS]
+	Pointers to coefficient quantization tables, one per table slot,
+	or NULL if no table is defined for a slot.  Usually these should
+	be set via one of the above helper routines; jpeg_add_quant_table()
+	is general enough to define any quantization table.  The other
+	routines will set up table slot 0 for luminance quality and table
+	slot 1 for chrominance.
+
+JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS]
+JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS]
+	Pointers to Huffman coding tables, one per table slot, or NULL if
+	no table is defined for a slot.  Slots 0 and 1 are filled with the
+	JPEG sample tables by jpeg_set_defaults().  If you need to allocate
+	more table structures, jpeg_alloc_huff_table() may be used.
+	Note that optimal Huffman tables can be computed for an image
+	by setting optimize_coding, as discussed above; there's seldom
+	any need to mess with providing your own Huffman tables.
+
+There are some additional cinfo fields which are not documented here
+because you currently can't change them; for example, you can't set
+arith_code TRUE because arithmetic coding is unsupported.
+
+
+Per-component parameters are stored in the struct cinfo.comp_info[i] for
+component number i.  Note that components here refer to components of the
+JPEG color space, *not* the source image color space.  A suitably large
+comp_info[] array is allocated by jpeg_set_defaults(); if you choose not
+to use that routine, it's up to you to allocate the array.
+
+int component_id
+	The one-byte identifier code to be recorded in the JPEG file for
+	this component.  For the standard color spaces, we recommend you
+	leave the default values alone.
+
+int h_samp_factor
+int v_samp_factor
+	Horizontal and vertical sampling factors for the component; must
+	be 1..4 according to the JPEG standard.  Note that larger sampling
+	factors indicate a higher-resolution component; many people find
+	this behavior quite unintuitive.  The default values are 2,2 for
+	luminance components and 1,1 for chrominance components, except
+	for grayscale where 1,1 is used.
+
+int quant_tbl_no
+	Quantization table number for component.  The default value is
+	0 for luminance components and 1 for chrominance components.
+
+int dc_tbl_no
+int ac_tbl_no
+	DC and AC entropy coding table numbers.  The default values are
+	0 for luminance components and 1 for chrominance components.
+
+int component_index
+	Must equal the component's index in comp_info[].  (Beginning in
+	release v6, the compressor library will fill this in automatically;
+	you don't have to.)
+
+
+Decompression parameter selection
+---------------------------------
+
+Decompression parameter selection is somewhat simpler than compression
+parameter selection, since all of the JPEG internal parameters are
+recorded in the source file and need not be supplied by the application.
+(Unless you are working with abbreviated files, in which case see
+"Abbreviated datastreams", below.)  Decompression parameters control
+the postprocessing done on the image to deliver it in a format suitable
+for the application's use.  Many of the parameters control speed/quality
+tradeoffs, in which faster decompression may be obtained at the price of
+a poorer-quality image.  The defaults select the highest quality (slowest)
+processing.
+
+The following fields in the JPEG object are set by jpeg_read_header() and
+may be useful to the application in choosing decompression parameters:
+
+JDIMENSION image_width			Width and height of image
+JDIMENSION image_height
+int num_components			Number of color components
+J_COLOR_SPACE jpeg_color_space		Colorspace of image
+boolean saw_JFIF_marker			TRUE if a JFIF APP0 marker was seen
+  UINT8 JFIF_major_version		Version information from JFIF marker
+  UINT8 JFIF_minor_version
+  UINT8 density_unit			Resolution data from JFIF marker
+  UINT16 X_density
+  UINT16 Y_density
+boolean saw_Adobe_marker		TRUE if an Adobe APP14 marker was seen
+  UINT8 Adobe_transform			Color transform code from Adobe marker
+
+The JPEG color space, unfortunately, is something of a guess since the JPEG
+standard proper does not provide a way to record it.  In practice most files
+adhere to the JFIF or Adobe conventions, and the decoder will recognize these
+correctly.  See "Special color spaces", below, for more info.
+
+
+The decompression parameters that determine the basic properties of the
+returned image are:
+
+J_COLOR_SPACE out_color_space
+	Output color space.  jpeg_read_header() sets an appropriate default
+	based on jpeg_color_space; typically it will be RGB or grayscale.
+	The application can change this field to request output in a different
+	colorspace.  For example, set it to JCS_GRAYSCALE to get grayscale
+	output from a color file.  (This is useful for previewing: grayscale
+	output is faster than full color since the color components need not
+	be processed.)  Note that not all possible color space transforms are
+	currently implemented; you may need to extend jdcolor.c if you want an
+	unusual conversion.
+
+unsigned int scale_num, scale_denom
+	Scale the image by the fraction scale_num/scale_denom.  Default is
+	1/1, or no scaling.  Currently, the only supported scaling ratios
+	are 1/1, 1/2, 1/4, and 1/8.  (The library design allows for arbitrary
+	scaling ratios but this is not likely to be implemented any time soon.)
+	Smaller scaling ratios permit significantly faster decoding since
+	fewer pixels need be processed and a simpler IDCT method can be used.
+
+boolean quantize_colors
+	If set TRUE, colormapped output will be delivered.  Default is FALSE,
+	meaning that full-color output will be delivered.
+
+The next three parameters are relevant only if quantize_colors is TRUE.
+
+int desired_number_of_colors
+	Maximum number of colors to use in generating a library-supplied color
+	map (the actual number of colors is returned in a different field).
+	Default 256.  Ignored when the application supplies its own color map.
+
+boolean two_pass_quantize
+	If TRUE, an extra pass over the image is made to select a custom color
+	map for the image.  This usually looks a lot better than the one-size-
+	fits-all colormap that is used otherwise.  Default is TRUE.  Ignored
+	when the application supplies its own color map.
+
+J_DITHER_MODE dither_mode
+	Selects color dithering method.  Supported values are:
+		JDITHER_NONE	no dithering: fast, very low quality
+		JDITHER_ORDERED	ordered dither: moderate speed and quality
+		JDITHER_FS	Floyd-Steinberg dither: slow, high quality
+	Default is JDITHER_FS.  (At present, ordered dither is implemented
+	only in the single-pass, standard-colormap case.  If you ask for
+	ordered dither when two_pass_quantize is TRUE or when you supply
+	an external color map, you'll get F-S dithering.)
+
+When quantize_colors is TRUE, the target color map is described by the next
+two fields.  colormap is set to NULL by jpeg_read_header().  The application
+can supply a color map by setting colormap non-NULL and setting
+actual_number_of_colors to the map size.  Otherwise, jpeg_start_decompress()
+selects a suitable color map and sets these two fields itself.
+[Implementation restriction: at present, an externally supplied colormap is
+only accepted for 3-component output color spaces.]
+
+JSAMPARRAY colormap
+	The color map, represented as a 2-D pixel array of out_color_components
+	rows and actual_number_of_colors columns.  Ignored if not quantizing.
+	CAUTION: if the JPEG library creates its own colormap, the storage
+	pointed to by this field is released by jpeg_finish_decompress().
+	Copy the colormap somewhere else first, if you want to save it.
+
+int actual_number_of_colors
+	The number of colors in the color map.
+
+Additional decompression parameters that the application may set include:
+
+J_DCT_METHOD dct_method
+	Selects the algorithm used for the DCT step.  Choices are the same
+	as described above for compression.
+
+boolean do_fancy_upsampling
+	If TRUE, do careful upsampling of chroma components.  If FALSE,
+	a faster but sloppier method is used.  Default is TRUE.  The visual
+	impact of the sloppier method is often very small.
+
+boolean do_block_smoothing
+	If TRUE, interblock smoothing is applied in early stages of decoding
+	progressive JPEG files; if FALSE, not.  Default is TRUE.  Early
+	progression stages look "fuzzy" with smoothing, "blocky" without.
+	In any case, block smoothing ceases to be applied after the first few
+	AC coefficients are known to full accuracy, so it is relevant only
+	when using buffered-image mode for progressive images.
+
+boolean enable_1pass_quant
+boolean enable_external_quant
+boolean enable_2pass_quant
+	These are significant only in buffered-image mode, which is
+	described in its own section below.
+
+
+The output image dimensions are given by the following fields.  These are
+computed from the source image dimensions and the decompression parameters
+by jpeg_start_decompress().  You can also call jpeg_calc_output_dimensions()
+to obtain the values that will result from the current parameter settings.
+This can be useful if you are trying to pick a scaling ratio that will get
+close to a desired target size.  It's also important if you are using the
+JPEG library's memory manager to allocate output buffer space, because you
+are supposed to request such buffers *before* jpeg_start_decompress().
+
+JDIMENSION output_width		Actual dimensions of output image.
+JDIMENSION output_height
+int out_color_components	Number of color components in out_color_space.
+int output_components		Number of color components returned.
+int rec_outbuf_height		Recommended height of scanline buffer.
+
+When quantizing colors, output_components is 1, indicating a single color map
+index per pixel.  Otherwise it equals out_color_components.  The output arrays
+are required to be output_width * output_components JSAMPLEs wide.
+
+rec_outbuf_height is the recommended minimum height (in scanlines) of the
+buffer passed to jpeg_read_scanlines().  If the buffer is smaller, the
+library will still work, but time will be wasted due to unnecessary data
+copying.  In high-quality modes, rec_outbuf_height is always 1, but some
+faster, lower-quality modes set it to larger values (typically 2 to 4).
+If you are going to ask for a high-speed processing mode, you may as well
+go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
+(An output buffer larger than rec_outbuf_height lines is OK, but won't
+provide any material speed improvement over that height.)
+
+
+Special color spaces
+--------------------
+
+The JPEG standard itself is "color blind" and doesn't specify any particular
+color space.  It is customary to convert color data to a luminance/chrominance
+color space before compressing, since this permits greater compression.  The
+existing de-facto JPEG file format standards specify YCbCr or grayscale data
+(JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe).  For special
+applications such as multispectral images, other color spaces can be used,
+but it must be understood that such files will be unportable.
+
+The JPEG library can handle the most common colorspace conversions (namely
+RGB <=> YCbCr and CMYK <=> YCCK).  It can also deal with data of an unknown
+color space, passing it through without conversion.  If you deal extensively
+with an unusual color space, you can easily extend the library to understand
+additional color spaces and perform appropriate conversions.
+
+For compression, the source data's color space is specified by field
+in_color_space.  This is transformed to the JPEG file's color space given
+by jpeg_color_space.  jpeg_set_defaults() chooses a reasonable JPEG color
+space depending on in_color_space, but you can override this by calling
+jpeg_set_colorspace().  Of course you must select a supported transformation.
+jccolor.c currently supports the following transformations:
+	RGB => YCbCr
+	RGB => GRAYSCALE
+	YCbCr => GRAYSCALE
+	CMYK => YCCK
+plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
+YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
+
+The de-facto file format standards (JFIF and Adobe) specify APPn markers that
+indicate the color space of the JPEG file.  It is important to ensure that
+these are written correctly, or omitted if the JPEG file's color space is not
+one of the ones supported by the de-facto standards.  jpeg_set_colorspace()
+will set the compression parameters to include or omit the APPn markers
+properly, so long as it is told the truth about the JPEG color space.
+For example, if you are writing some random 3-component color space without
+conversion, don't try to fake out the library by setting in_color_space and
+jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN.  You may want to write an
+APPn marker of your own devising to identify the colorspace --- see "Special
+markers", below.
+
+When told that the color space is UNKNOWN, the library will default to using
+luminance-quality compression parameters for all color components.  You may
+well want to change these parameters.  See the source code for
+jpeg_set_colorspace(), in jcparam.c, for details.
+
+For decompression, the JPEG file's color space is given in jpeg_color_space,
+and this is transformed to the output color space out_color_space.
+jpeg_read_header's setting of jpeg_color_space can be relied on if the file
+conforms to JFIF or Adobe conventions, but otherwise it is no better than a
+guess.  If you know the JPEG file's color space for certain, you can override
+jpeg_read_header's guess by setting jpeg_color_space.  jpeg_read_header also
+selects a default output color space based on (its guess of) jpeg_color_space;
+set out_color_space to override this.  Again, you must select a supported
+transformation.  jdcolor.c currently supports
+	YCbCr => GRAYSCALE
+	YCbCr => RGB
+	GRAYSCALE => RGB
+	YCCK => CMYK
+as well as the null transforms.  (Since GRAYSCALE=>RGB is provided, an
+application can force grayscale JPEGs to look like color JPEGs if it only
+wants to handle one case.)
+
+The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
+(it weights distances appropriately for RGB colors).  You'll need to modify
+the code if you want to use it for non-RGB output color spaces.  Note that
+jquant2.c is used to map to an application-supplied colormap as well as for
+the normal two-pass colormap selection process.
+
+CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
+files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
+This is arguably a bug in Photoshop, but if you need to work with Photoshop
+CMYK files, you will have to deal with it in your application.  We cannot
+"fix" this in the library by inverting the data during the CMYK<=>YCCK
+transform, because that would break other applications, notably Ghostscript.
+Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
+data in the same inverted-YCCK representation used in bare JPEG files, but
+the surrounding PostScript code performs an inversion using the PS image
+operator.  I am told that Photoshop 3.0 will write uninverted YCCK in
+EPS/JPEG files, and will omit the PS-level inversion.  (But the data
+polarity used in bare JPEG files will not change in 3.0.)  In either case,
+the JPEG library must not invert the data itself, or else Ghostscript would
+read these EPS files incorrectly.
+
+
+Error handling
+--------------
+
+When the default error handler is used, any error detected inside the JPEG
+routines will cause a message to be printed on stderr, followed by exit().
+You can supply your own error handling routines to override this behavior
+and to control the treatment of nonfatal warnings and trace/debug messages.
+The file example.c illustrates the most common case, which is to have the
+application regain control after an error rather than exiting.
+
+The JPEG library never writes any message directly; it always goes through
+the error handling routines.  Three classes of messages are recognized:
+  * Fatal errors: the library cannot continue.
+  * Warnings: the library can continue, but the data is corrupt, and a
+    damaged output image is likely to result.
+  * Trace/informational messages.  These come with a trace level indicating
+    the importance of the message; you can control the verbosity of the
+    program by adjusting the maximum trace level that will be displayed.
+
+You may, if you wish, simply replace the entire JPEG error handling module
+(jerror.c) with your own code.  However, you can avoid code duplication by
+only replacing some of the routines depending on the behavior you need.
+This is accomplished by calling jpeg_std_error() as usual, but then overriding
+some of the method pointers in the jpeg_error_mgr struct, as illustrated by
+example.c.
+
+All of the error handling routines will receive a pointer to the JPEG object
+(a j_common_ptr which points to either a jpeg_compress_struct or a
+jpeg_decompress_struct; if you need to tell which, test the is_decompressor
+field).  This struct includes a pointer to the error manager struct in its
+"err" field.  Frequently, custom error handler routines will need to access
+additional data which is not known to the JPEG library or the standard error
+handler.  The most convenient way to do this is to embed either the JPEG
+object or the jpeg_error_mgr struct in a larger structure that contains
+additional fields; then casting the passed pointer provides access to the
+additional fields.  Again, see example.c for one way to do it.  (Beginning
+with IJG version 6b, there is also a void pointer "client_data" in each
+JPEG object, which the application can also use to find related data.
+The library does not touch client_data at all.)
+
+The individual methods that you might wish to override are:
+
+error_exit (j_common_ptr cinfo)
+	Receives control for a fatal error.  Information sufficient to
+	generate the error message has been stored in cinfo->err; call
+	output_message to display it.  Control must NOT return to the caller;
+	generally this routine will exit() or longjmp() somewhere.
+	Typically you would override this routine to get rid of the exit()
+	default behavior.  Note that if you continue processing, you should
+	clean up the JPEG object with jpeg_abort() or jpeg_destroy().
+
+output_message (j_common_ptr cinfo)
+	Actual output of any JPEG message.  Override this to send messages
+	somewhere other than stderr.  Note that this method does not know
+	how to generate a message, only where to send it.
+
+format_message (j_common_ptr cinfo, char * buffer)
+	Constructs a readable error message string based on the error info
+	stored in cinfo->err.  This method is called by output_message.  Few
+	applications should need to override this method.  One possible
+	reason for doing so is to implement dynamic switching of error message
+	language.
+
+emit_message (j_common_ptr cinfo, int msg_level)
+	Decide whether or not to emit a warning or trace message; if so,
+	calls output_message.  The main reason for overriding this method
+	would be to abort on warnings.  msg_level is -1 for warnings,
+	0 and up for trace messages.
+
+Only error_exit() and emit_message() are called from the rest of the JPEG
+library; the other two are internal to the error handler.
+
+The actual message texts are stored in an array of strings which is pointed to
+by the field err->jpeg_message_table.  The messages are numbered from 0 to
+err->last_jpeg_message, and it is these code numbers that are used in the
+JPEG library code.  You could replace the message texts (for instance, with
+messages in French or German) by changing the message table pointer.  See
+jerror.h for the default texts.  CAUTION: this table will almost certainly
+change or grow from one library version to the next.
+
+It may be useful for an application to add its own message texts that are
+handled by the same mechanism.  The error handler supports a second "add-on"
+message table for this purpose.  To define an addon table, set the pointer
+err->addon_message_table and the message numbers err->first_addon_message and
+err->last_addon_message.  If you number the addon messages beginning at 1000
+or so, you won't have to worry about conflicts with the library's built-in
+messages.  See the sample applications cjpeg/djpeg for an example of using
+addon messages (the addon messages are defined in cderror.h).
+
+Actual invocation of the error handler is done via macros defined in jerror.h:
+	ERREXITn(...)	for fatal errors
+	WARNMSn(...)	for corrupt-data warnings
+	TRACEMSn(...)	for trace and informational messages.
+These macros store the message code and any additional parameters into the
+error handler struct, then invoke the error_exit() or emit_message() method.
+The variants of each macro are for varying numbers of additional parameters.
+The additional parameters are inserted into the generated message using
+standard printf() format codes.
+
+See jerror.h and jerror.c for further details.
+
+
+Compressed data handling (source and destination managers)
+----------------------------------------------------------
+
+The JPEG compression library sends its compressed data to a "destination
+manager" module.  The default destination manager just writes the data to a
+stdio stream, but you can provide your own manager to do something else.
+Similarly, the decompression library calls a "source manager" to obtain the
+compressed data; you can provide your own source manager if you want the data
+to come from somewhere other than a stdio stream.
+
+In both cases, compressed data is processed a bufferload at a time: the
+destination or source manager provides a work buffer, and the library invokes
+the manager only when the buffer is filled or emptied.  (You could define a
+one-character buffer to force the manager to be invoked for each byte, but
+that would be rather inefficient.)  The buffer's size and location are
+controlled by the manager, not by the library.  For example, if you desired to
+decompress a JPEG datastream that was all in memory, you could just make the
+buffer pointer and length point to the original data in memory.  Then the
+buffer-reload procedure would be invoked only if the decompressor ran off the
+end of the datastream, which would indicate an erroneous datastream.
+
+The work buffer is defined as an array of datatype JOCTET, which is generally
+"char" or "unsigned char".  On a machine where char is not exactly 8 bits
+wide, you must define JOCTET as a wider data type and then modify the data
+source and destination modules to transcribe the work arrays into 8-bit units
+on external storage.
+
+A data destination manager struct contains a pointer and count defining the
+next byte to write in the work buffer and the remaining free space:
+
+	JOCTET * next_output_byte;  /* => next byte to write in buffer */
+	size_t free_in_buffer;      /* # of byte spaces remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is filled.  The manager's empty_output_buffer method must reset the pointer
+and count.  The manager is expected to remember the buffer's starting address
+and total size in private fields not visible to the library.
+
+A data destination manager provides three methods:
+
+init_destination (j_compress_ptr cinfo)
+	Initialize destination.  This is called by jpeg_start_compress()
+	before any data is actually written.  It must initialize
+	next_output_byte and free_in_buffer.  free_in_buffer must be
+	initialized to a positive value.
+
+empty_output_buffer (j_compress_ptr cinfo)
+	This is called whenever the buffer has filled (free_in_buffer
+	reaches zero).  In typical applications, it should write out the
+	*entire* buffer (use the saved start address and buffer length;
+	ignore the current state of next_output_byte and free_in_buffer).
+	Then reset the pointer & count to the start of the buffer, and
+	return TRUE indicating that the buffer has been dumped.
+	free_in_buffer must be set to a positive value when TRUE is
+	returned.  A FALSE return should only be used when I/O suspension is
+	desired (this operating mode is discussed in the next section).
+
+term_destination (j_compress_ptr cinfo)
+	Terminate destination --- called by jpeg_finish_compress() after all
+	data has been written.  In most applications, this must flush any
+	data remaining in the buffer.  Use either next_output_byte or
+	free_in_buffer to determine how much data is in the buffer.
+
+term_destination() is NOT called by jpeg_abort() or jpeg_destroy().  If you
+want the destination manager to be cleaned up during an abort, you must do it
+yourself.
+
+You will also need code to create a jpeg_destination_mgr struct, fill in its
+method pointers, and insert a pointer to the struct into the "dest" field of
+the JPEG compression object.  This can be done in-line in your setup code if
+you like, but it's probably cleaner to provide a separate routine similar to
+the jpeg_stdio_dest() routine of the supplied destination manager.
+
+Decompression source managers follow a parallel design, but with some
+additional frammishes.  The source manager struct contains a pointer and count
+defining the next byte to read from the work buffer and the number of bytes
+remaining:
+
+	const JOCTET * next_input_byte; /* => next byte to read from buffer */
+	size_t bytes_in_buffer;         /* # of bytes remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is emptied.  The manager's fill_input_buffer method must reset the pointer and
+count.  In most applications, the manager must remember the buffer's starting
+address and total size in private fields not visible to the library.
+
+A data source manager provides five methods:
+
+init_source (j_decompress_ptr cinfo)
+	Initialize source.  This is called by jpeg_read_header() before any
+	data is actually read.  Unlike init_destination(), it may leave
+	bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
+	will occur immediately).
+
+fill_input_buffer (j_decompress_ptr cinfo)
+	This is called whenever bytes_in_buffer has reached zero and more
+	data is wanted.  In typical applications, it should read fresh data
+	into the buffer (ignoring the current state of next_input_byte and
+	bytes_in_buffer), reset the pointer & count to the start of the
+	buffer, and return TRUE indicating that the buffer has been reloaded.
+	It is not necessary to fill the buffer entirely, only to obtain at
+	least one more byte.  bytes_in_buffer MUST be set to a positive value
+	if TRUE is returned.  A FALSE return should only be used when I/O
+	suspension is desired (this mode is discussed in the next section).
+
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+	Skip num_bytes worth of data.  The buffer pointer and count should
+	be advanced over num_bytes input bytes, refilling the buffer as
+	needed.  This is used to skip over a potentially large amount of
+	uninteresting data (such as an APPn marker).  In some applications
+	it may be possible to optimize away the reading of the skipped data,
+	but it's not clear that being smart is worth much trouble; large
+	skips are uncommon.  bytes_in_buffer may be zero on return.
+	A zero or negative skip count should be treated as a no-op.
+
+resync_to_restart (j_decompress_ptr cinfo, int desired)
+	This routine is called only when the decompressor has failed to find
+	a restart (RSTn) marker where one is expected.  Its mission is to
+	find a suitable point for resuming decompression.  For most
+	applications, we recommend that you just use the default resync
+	procedure, jpeg_resync_to_restart().  However, if you are able to back
+	up in the input data stream, or if you have a-priori knowledge about
+	the likely location of restart markers, you may be able to do better.
+	Read the read_restart_marker() and jpeg_resync_to_restart() routines
+	in jdmarker.c if you think you'd like to implement your own resync
+	procedure.
+
+term_source (j_decompress_ptr cinfo)
+	Terminate source --- called by jpeg_finish_decompress() after all
+	data has been read.  Often a no-op.
+
+For both fill_input_buffer() and skip_input_data(), there is no such thing
+as an EOF return.  If the end of the file has been reached, the routine has
+a choice of exiting via ERREXIT() or inserting fake data into the buffer.
+In most cases, generating a warning message and inserting a fake EOI marker
+is the best course of action --- this will allow the decompressor to output
+however much of the image is there.  In pathological cases, the decompressor
+may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
+jdatasrc.c illustrates the recommended error recovery behavior.
+
+term_source() is NOT called by jpeg_abort() or jpeg_destroy().  If you want
+the source manager to be cleaned up during an abort, you must do it yourself.
+
+You will also need code to create a jpeg_source_mgr struct, fill in its method
+pointers, and insert a pointer to the struct into the "src" field of the JPEG
+decompression object.  This can be done in-line in your setup code if you
+like, but it's probably cleaner to provide a separate routine similar to the
+jpeg_stdio_src() routine of the supplied source manager.
+
+For more information, consult the stdio source and destination managers
+in jdatasrc.c and jdatadst.c.
+
+
+I/O suspension
+--------------
+
+Some applications need to use the JPEG library as an incremental memory-to-
+memory filter: when the compressed data buffer is filled or emptied, they want
+control to return to the outer loop, rather than expecting that the buffer can
+be emptied or reloaded within the data source/destination manager subroutine.
+The library supports this need by providing an "I/O suspension" mode, which we
+describe in this section.
+
+The I/O suspension mode is not a panacea: nothing is guaranteed about the
+maximum amount of time spent in any one call to the library, so it will not
+eliminate response-time problems in single-threaded applications.  If you
+need guaranteed response time, we suggest you "bite the bullet" and implement
+a real multi-tasking capability.
+
+To use I/O suspension, cooperation is needed between the calling application
+and the data source or destination manager; you will always need a custom
+source/destination manager.  (Please read the previous section if you haven't
+already.)  The basic idea is that the empty_output_buffer() or
+fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
+that it has done nothing.  Upon seeing this, the JPEG library suspends
+operation and returns to its caller.  The surrounding application is
+responsible for emptying or refilling the work buffer before calling the
+JPEG library again.
+
+Compression suspension:
+
+For compression suspension, use an empty_output_buffer() routine that returns
+FALSE; typically it will not do anything else.  This will cause the
+compressor to return to the caller of jpeg_write_scanlines(), with the return
+value indicating that not all the supplied scanlines have been accepted.
+The application must make more room in the output buffer, adjust the output
+buffer pointer/count appropriately, and then call jpeg_write_scanlines()
+again, pointing to the first unconsumed scanline.
+
+When forced to suspend, the compressor will backtrack to a convenient stopping
+point (usually the start of the current MCU); it will regenerate some output
+data when restarted.  Therefore, although empty_output_buffer() is only
+called when the buffer is filled, you should NOT write out the entire buffer
+after a suspension.  Write only the data up to the current position of
+next_output_byte/free_in_buffer.  The data beyond that point will be
+regenerated after resumption.
+
+Because of the backtracking behavior, a good-size output buffer is essential
+for efficiency; you don't want the compressor to suspend often.  (In fact, an
+overly small buffer could lead to infinite looping, if a single MCU required
+more data than would fit in the buffer.)  We recommend a buffer of at least
+several Kbytes.  You may want to insert explicit code to ensure that you don't
+call jpeg_write_scanlines() unless there is a reasonable amount of space in
+the output buffer; in other words, flush the buffer before trying to compress
+more data.
+
+The compressor does not allow suspension while it is trying to write JPEG
+markers at the beginning and end of the file.  This means that:
+  * At the beginning of a compression operation, there must be enough free
+    space in the output buffer to hold the header markers (typically 600 or
+    so bytes).  The recommended buffer size is bigger than this anyway, so
+    this is not a problem as long as you start with an empty buffer.  However,
+    this restriction might catch you if you insert large special markers, such
+    as a JFIF thumbnail image, without flushing the buffer afterwards.
+  * When you call jpeg_finish_compress(), there must be enough space in the
+    output buffer to emit any buffered data and the final EOI marker.  In the
+    current implementation, half a dozen bytes should suffice for this, but
+    for safety's sake we recommend ensuring that at least 100 bytes are free
+    before calling jpeg_finish_compress().
+
+A more significant restriction is that jpeg_finish_compress() cannot suspend.
+This means you cannot use suspension with multi-pass operating modes, namely
+Huffman code optimization and multiple-scan output.  Those modes write the
+whole file during jpeg_finish_compress(), which will certainly result in
+buffer overrun.  (Note that this restriction applies only to compression,
+not decompression.  The decompressor supports input suspension in all of its
+operating modes.)
+
+Decompression suspension:
+
+For decompression suspension, use a fill_input_buffer() routine that simply
+returns FALSE (except perhaps during error recovery, as discussed below).
+This will cause the decompressor to return to its caller with an indication
+that suspension has occurred.  This can happen at four places:
+  * jpeg_read_header(): will return JPEG_SUSPENDED.
+  * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
+  * jpeg_read_scanlines(): will return the number of scanlines already
+	completed (possibly 0).
+  * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
+The surrounding application must recognize these cases, load more data into
+the input buffer, and repeat the call.  In the case of jpeg_read_scanlines(),
+increment the passed pointers past any scanlines successfully read.
+
+Just as with compression, the decompressor will typically backtrack to a
+convenient restart point before suspending.  When fill_input_buffer() is
+called, next_input_byte/bytes_in_buffer point to the current restart point,
+which is where the decompressor will backtrack to if FALSE is returned.
+The data beyond that position must NOT be discarded if you suspend; it needs
+to be re-read upon resumption.  In most implementations, you'll need to shift
+this data down to the start of your work buffer and then load more data after
+it.  Again, this behavior means that a several-Kbyte work buffer is essential
+for decent performance; furthermore, you should load a reasonable amount of
+new data before resuming decompression.  (If you loaded, say, only one new
+byte each time around, you could waste a LOT of cycles.)
+
+The skip_input_data() source manager routine requires special care in a
+suspension scenario.  This routine is NOT granted the ability to suspend the
+decompressor; it can decrement bytes_in_buffer to zero, but no more.  If the
+requested skip distance exceeds the amount of data currently in the input
+buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
+additional skip distance somewhere else.  The decompressor will immediately
+call fill_input_buffer(), which should return FALSE, which will cause a
+suspension return.  The surrounding application must then arrange to discard
+the recorded number of bytes before it resumes loading the input buffer.
+(Yes, this design is rather baroque, but it avoids complexity in the far more
+common case where a non-suspending source manager is used.)
+
+If the input data has been exhausted, we recommend that you emit a warning
+and insert dummy EOI markers just as a non-suspending data source manager
+would do.  This can be handled either in the surrounding application logic or
+within fill_input_buffer(); the latter is probably more efficient.  If
+fill_input_buffer() knows that no more data is available, it can set the
+pointer/count to point to a dummy EOI marker and then return TRUE just as
+though it had read more data in a non-suspending situation.
+
+The decompressor does not attempt to suspend within standard JPEG markers;
+instead it will backtrack to the start of the marker and reprocess the whole
+marker next time.  Hence the input buffer must be large enough to hold the
+longest standard marker in the file.  Standard JPEG markers should normally
+not exceed a few hundred bytes each (DHT tables are typically the longest).
+We recommend at least a 2K buffer for performance reasons, which is much
+larger than any correct marker is likely to be.  For robustness against
+damaged marker length counts, you may wish to insert a test in your
+application for the case that the input buffer is completely full and yet
+the decoder has suspended without consuming any data --- otherwise, if this
+situation did occur, it would lead to an endless loop.  (The library can't
+provide this test since it has no idea whether "the buffer is full", or
+even whether there is a fixed-size input buffer.)
+
+The input buffer would need to be 64K to allow for arbitrary COM or APPn
+markers, but these are handled specially: they are either saved into allocated
+memory, or skipped over by calling skip_input_data().  In the former case,
+suspension is handled correctly, and in the latter case, the problem of
+buffer overrun is placed on skip_input_data's shoulders, as explained above.
+Note that if you provide your own marker handling routine for large markers,
+you should consider how to deal with buffer overflow.
+
+Multiple-buffer management:
+
+In some applications it is desirable to store the compressed data in a linked
+list of buffer areas, so as to avoid data copying.  This can be handled by
+having empty_output_buffer() or fill_input_buffer() set the pointer and count
+to reference the next available buffer; FALSE is returned only if no more
+buffers are available.  Although seemingly straightforward, there is a
+pitfall in this approach: the backtrack that occurs when FALSE is returned
+could back up into an earlier buffer.  For example, when fill_input_buffer()
+is called, the current pointer & count indicate the backtrack restart point.
+Since fill_input_buffer() will set the pointer and count to refer to a new
+buffer, the restart position must be saved somewhere else.  Suppose a second
+call to fill_input_buffer() occurs in the same library call, and no
+additional input data is available, so fill_input_buffer must return FALSE.
+If the JPEG library has not moved the pointer/count forward in the current
+buffer, then *the correct restart point is the saved position in the prior
+buffer*.  Prior buffers may be discarded only after the library establishes
+a restart point within a later buffer.  Similar remarks apply for output into
+a chain of buffers.
+
+The library will never attempt to backtrack over a skip_input_data() call,
+so any skipped data can be permanently discarded.  You still have to deal
+with the case of skipping not-yet-received data, however.
+
+It's much simpler to use only a single buffer; when fill_input_buffer() is
+called, move any unconsumed data (beyond the current pointer/count) down to
+the beginning of this buffer and then load new data into the remaining buffer
+space.  This approach requires a little more data copying but is far easier
+to get right.
+
+
+Progressive JPEG support
+------------------------
+
+Progressive JPEG rearranges the stored data into a series of scans of
+increasing quality.  In situations where a JPEG file is transmitted across a
+slow communications link, a decoder can generate a low-quality image very
+quickly from the first scan, then gradually improve the displayed quality as
+more scans are received.  The final image after all scans are complete is
+identical to that of a regular (sequential) JPEG file of the same quality
+setting.  Progressive JPEG files are often slightly smaller than equivalent
+sequential JPEG files, but the possibility of incremental display is the main
+reason for using progressive JPEG.
+
+The IJG encoder library generates progressive JPEG files when given a
+suitable "scan script" defining how to divide the data into scans.
+Creation of progressive JPEG files is otherwise transparent to the encoder.
+Progressive JPEG files can also be read transparently by the decoder library.
+If the decoding application simply uses the library as defined above, it
+will receive a final decoded image without any indication that the file was
+progressive.  Of course, this approach does not allow incremental display.
+To perform incremental display, an application needs to use the decoder
+library's "buffered-image" mode, in which it receives a decoded image
+multiple times.
+
+Each displayed scan requires about as much work to decode as a full JPEG
+image of the same size, so the decoder must be fairly fast in relation to the
+data transmission rate in order to make incremental display useful.  However,
+it is possible to skip displaying the image and simply add the incoming bits
+to the decoder's coefficient buffer.  This is fast because only Huffman
+decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
+The IJG decoder library allows the application to switch dynamically between
+displaying the image and simply absorbing the incoming bits.  A properly
+coded application can automatically adapt the number of display passes to
+suit the time available as the image is received.  Also, a final
+higher-quality display cycle can be performed from the buffered data after
+the end of the file is reached.
+
+Progressive compression:
+
+To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
+set the scan_info cinfo field to point to an array of scan descriptors, and
+perform compression as usual.  Instead of constructing your own scan list,
+you can call the jpeg_simple_progression() helper routine to create a
+recommended progression sequence; this method should be used by all
+applications that don't want to get involved in the nitty-gritty of
+progressive scan sequence design.  (If you want to provide user control of
+scan sequences, you may wish to borrow the scan script reading code found
+in rdswitch.c, so that you can read scan script files just like cjpeg's.)
+When scan_info is not NULL, the compression library will store DCT'd data
+into a buffer array as jpeg_write_scanlines() is called, and will emit all
+the requested scans during jpeg_finish_compress().  This implies that
+multiple-scan output cannot be created with a suspending data destination
+manager, since jpeg_finish_compress() does not support suspension.  We
+should also note that the compressor currently forces Huffman optimization
+mode when creating a progressive JPEG file, because the default Huffman
+tables are unsuitable for progressive files.
+
+Progressive decompression:
+
+When buffered-image mode is not used, the decoder library will read all of
+a multi-scan file during jpeg_start_decompress(), so that it can provide a
+final decoded image.  (Here "multi-scan" means either progressive or
+multi-scan sequential.)  This makes multi-scan files transparent to the
+decoding application.  However, existing applications that used suspending
+input with version 5 of the IJG library will need to be modified to check
+for a suspension return from jpeg_start_decompress().
+
+To perform incremental display, an application must use the library's
+buffered-image mode.  This is described in the next section.
+
+
+Buffered-image mode
+-------------------
+
+In buffered-image mode, the library stores the partially decoded image in a
+coefficient buffer, from which it can be read out as many times as desired.
+This mode is typically used for incremental display of progressive JPEG files,
+but it can be used with any JPEG file.  Each scan of a progressive JPEG file
+adds more data (more detail) to the buffered image.  The application can
+display in lockstep with the source file (one display pass per input scan),
+or it can allow input processing to outrun display processing.  By making
+input and display processing run independently, it is possible for the
+application to adapt progressive display to a wide range of data transmission
+rates.
+
+The basic control flow for buffered-image decoding is
+
+	jpeg_create_decompress()
+	set data source
+	jpeg_read_header()
+	set overall decompression parameters
+	cinfo.buffered_image = TRUE;	/* select buffered-image mode */
+	jpeg_start_decompress()
+	for (each output pass) {
+	    adjust output decompression parameters if required
+	    jpeg_start_output()		/* start a new output pass */
+	    for (all scanlines in image) {
+	        jpeg_read_scanlines()
+	        display scanlines
+	    }
+	    jpeg_finish_output()	/* terminate output pass */
+	}
+	jpeg_finish_decompress()
+	jpeg_destroy_decompress()
+
+This differs from ordinary unbuffered decoding in that there is an additional
+level of looping.  The application can choose how many output passes to make
+and how to display each pass.
+
+The simplest approach to displaying progressive images is to do one display
+pass for each scan appearing in the input file.  In this case the outer loop
+condition is typically
+	while (! jpeg_input_complete(&cinfo))
+and the start-output call should read
+	jpeg_start_output(&cinfo, cinfo.input_scan_number);
+The second parameter to jpeg_start_output() indicates which scan of the input
+file is to be displayed; the scans are numbered starting at 1 for this
+purpose.  (You can use a loop counter starting at 1 if you like, but using
+the library's input scan counter is easier.)  The library automatically reads
+data as necessary to complete each requested scan, and jpeg_finish_output()
+advances to the next scan or end-of-image marker (hence input_scan_number
+will be incremented by the time control arrives back at jpeg_start_output()).
+With this technique, data is read from the input file only as needed, and
+input and output processing run in lockstep.
+
+After reading the final scan and reaching the end of the input file, the
+buffered image remains available; it can be read additional times by
+repeating the jpeg_start_output()/jpeg_read_scanlines()/jpeg_finish_output()
+sequence.  For example, a useful technique is to use fast one-pass color
+quantization for display passes made while the image is arriving, followed by
+a final display pass using two-pass quantization for highest quality.  This
+is done by changing the library parameters before the final output pass.
+Changing parameters between passes is discussed in detail below.
+
+In general the last scan of a progressive file cannot be recognized as such
+until after it is read, so a post-input display pass is the best approach if
+you want special processing in the final pass.
+
+When done with the image, be sure to call jpeg_finish_decompress() to release
+the buffered image (or just use jpeg_destroy_decompress()).
+
+If input data arrives faster than it can be displayed, the application can
+cause the library to decode input data in advance of what's needed to produce
+output.  This is done by calling the routine jpeg_consume_input().
+The return value is one of the following:
+	JPEG_REACHED_SOS:    reached an SOS marker (the start of a new scan)
+	JPEG_REACHED_EOI:    reached the EOI marker (end of image)
+	JPEG_ROW_COMPLETED:  completed reading one MCU row of compressed data
+	JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
+	JPEG_SUSPENDED:      suspended before completing any of the above
+(JPEG_SUSPENDED can occur only if a suspending data source is used.)  This
+routine can be called at any time after initializing the JPEG object.  It
+reads some additional data and returns when one of the indicated significant
+events occurs.  (If called after the EOI marker is reached, it will
+immediately return JPEG_REACHED_EOI without attempting to read more data.)
+
+The library's output processing will automatically call jpeg_consume_input()
+whenever the output processing overtakes the input; thus, simple lockstep
+display requires no direct calls to jpeg_consume_input().  But by adding
+calls to jpeg_consume_input(), you can absorb data in advance of what is
+being displayed.  This has two benefits:
+  * You can limit buildup of unprocessed data in your input buffer.
+  * You can eliminate extra display passes by paying attention to the
+    state of the library's input processing.
+
+The first of these benefits only requires interspersing calls to
+jpeg_consume_input() with your display operations and any other processing
+you may be doing.  To avoid wasting cycles due to backtracking, it's best to
+call jpeg_consume_input() only after a hundred or so new bytes have arrived.
+This is discussed further under "I/O suspension", above.  (Note: the JPEG
+library currently is not thread-safe.  You must not call jpeg_consume_input()
+from one thread of control if a different library routine is working on the
+same JPEG object in another thread.)
+
+When input arrives fast enough that more than one new scan is available
+before you start a new output pass, you may as well skip the output pass
+corresponding to the completed scan.  This occurs for free if you pass
+cinfo.input_scan_number as the target scan number to jpeg_start_output().
+The input_scan_number field is simply the index of the scan currently being
+consumed by the input processor.  You can ensure that this is up-to-date by
+emptying the input buffer just before calling jpeg_start_output(): call
+jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
+JPEG_REACHED_EOI.
+
+The target scan number passed to jpeg_start_output() is saved in the
+cinfo.output_scan_number field.  The library's output processing calls
+jpeg_consume_input() whenever the current input scan number and row within
+that scan is less than or equal to the current output scan number and row.
+Thus, input processing can "get ahead" of the output processing but is not
+allowed to "fall behind".  You can achieve several different effects by
+manipulating this interlock rule.  For example, if you pass a target scan
+number greater than the current input scan number, the output processor will
+wait until that scan starts to arrive before producing any output.  (To avoid
+an infinite loop, the target scan number is automatically reset to the last
+scan number when the end of image is reached.  Thus, if you specify a large
+target scan number, the library will just absorb the entire input file and
+then perform an output pass.  This is effectively the same as what
+jpeg_start_decompress() does when you don't select buffered-image mode.)
+When you pass a target scan number equal to the current input scan number,
+the image is displayed no faster than the current input scan arrives.  The
+final possibility is to pass a target scan number less than the current input
+scan number; this disables the input/output interlock and causes the output
+processor to simply display whatever it finds in the image buffer, without
+waiting for input.  (However, the library will not accept a target scan
+number less than one, so you can't avoid waiting for the first scan.)
+
+When data is arriving faster than the output display processing can advance
+through the image, jpeg_consume_input() will store data into the buffered
+image beyond the point at which the output processing is reading data out
+again.  If the input arrives fast enough, it may "wrap around" the buffer to
+the point where the input is more than one whole scan ahead of the output.
+If the output processing simply proceeds through its display pass without
+paying attention to the input, the effect seen on-screen is that the lower
+part of the image is one or more scans better in quality than the upper part.
+Then, when the next output scan is started, you have a choice of what target
+scan number to use.  The recommended choice is to use the current input scan
+number at that time, which implies that you've skipped the output scans
+corresponding to the input scans that were completed while you processed the
+previous output scan.  In this way, the decoder automatically adapts its
+speed to the arriving data, by skipping output scans as necessary to keep up
+with the arriving data.
+
+When using this strategy, you'll want to be sure that you perform a final
+output pass after receiving all the data; otherwise your last display may not
+be full quality across the whole screen.  So the right outer loop logic is
+something like this:
+	do {
+	    absorb any waiting input by calling jpeg_consume_input()
+	    final_pass = jpeg_input_complete(&cinfo);
+	    adjust output decompression parameters if required
+	    jpeg_start_output(&cinfo, cinfo.input_scan_number);
+	    ...
+	    jpeg_finish_output()
+	} while (! final_pass);
+rather than quitting as soon as jpeg_input_complete() returns TRUE.  This
+arrangement makes it simple to use higher-quality decoding parameters
+for the final pass.  But if you don't want to use special parameters for
+the final pass, the right loop logic is like this:
+	for (;;) {
+	    absorb any waiting input by calling jpeg_consume_input()
+	    jpeg_start_output(&cinfo, cinfo.input_scan_number);
+	    ...
+	    jpeg_finish_output()
+	    if (jpeg_input_complete(&cinfo) &&
+	        cinfo.input_scan_number == cinfo.output_scan_number)
+	      break;
+	}
+In this case you don't need to know in advance whether an output pass is to
+be the last one, so it's not necessary to have reached EOF before starting
+the final output pass; rather, what you want to test is whether the output
+pass was performed in sync with the final input scan.  This form of the loop
+will avoid an extra output pass whenever the decoder is able (or nearly able)
+to keep up with the incoming data.
+
+When the data transmission speed is high, you might begin a display pass,
+then find that much or all of the file has arrived before you can complete
+the pass.  (You can detect this by noting the JPEG_REACHED_EOI return code
+from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
+In this situation you may wish to abort the current display pass and start a
+new one using the newly arrived information.  To do so, just call
+jpeg_finish_output() and then start a new pass with jpeg_start_output().
+
+A variant strategy is to abort and restart display if more than one complete
+scan arrives during an output pass; this can be detected by noting
+JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number.  This
+idea should be employed with caution, however, since the display process
+might never get to the bottom of the image before being aborted, resulting
+in the lower part of the screen being several passes worse than the upper.
+In most cases it's probably best to abort an output pass only if the whole
+file has arrived and you want to begin the final output pass immediately.
+
+When receiving data across a communication link, we recommend always using
+the current input scan number for the output target scan number; if a
+higher-quality final pass is to be done, it should be started (aborting any
+incomplete output pass) as soon as the end of file is received.  However,
+many other strategies are possible.  For example, the application can examine
+the parameters of the current input scan and decide whether to display it or
+not.  If the scan contains only chroma data, one might choose not to use it
+as the target scan, expecting that the scan will be small and will arrive
+quickly.  To skip to the next scan, call jpeg_consume_input() until it
+returns JPEG_REACHED_SOS or JPEG_REACHED_EOI.  Or just use the next higher
+number as the target scan for jpeg_start_output(); but that method doesn't
+let you inspect the next scan's parameters before deciding to display it.
+
+
+In buffered-image mode, jpeg_start_decompress() never performs input and
+thus never suspends.  An application that uses input suspension with
+buffered-image mode must be prepared for suspension returns from these
+routines:
+* jpeg_start_output() performs input only if you request 2-pass quantization
+  and the target scan isn't fully read yet.  (This is discussed below.)
+* jpeg_read_scanlines(), as always, returns the number of scanlines that it
+  was able to produce before suspending.
+* jpeg_finish_output() will read any markers following the target scan,
+  up to the end of the file or the SOS marker that begins another scan.
+  (But it reads no input if jpeg_consume_input() has already reached the
+  end of the file or a SOS marker beyond the target output scan.)
+* jpeg_finish_decompress() will read until the end of file, and thus can
+  suspend if the end hasn't already been reached (as can be tested by
+  calling jpeg_input_complete()).
+jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
+all return TRUE if they completed their tasks, FALSE if they had to suspend.
+In the event of a FALSE return, the application must load more input data
+and repeat the call.  Applications that use non-suspending data sources need
+not check the return values of these three routines.
+
+
+It is possible to change decoding parameters between output passes in the
+buffered-image mode.  The decoder library currently supports only very
+limited changes of parameters.  ONLY THE FOLLOWING parameter changes are
+allowed after jpeg_start_decompress() is called:
+* dct_method can be changed before each call to jpeg_start_output().
+  For example, one could use a fast DCT method for early scans, changing
+  to a higher quality method for the final scan.
+* dither_mode can be changed before each call to jpeg_start_output();
+  of course this has no impact if not using color quantization.  Typically
+  one would use ordered dither for initial passes, then switch to
+  Floyd-Steinberg dither for the final pass.  Caution: changing dither mode
+  can cause more memory to be allocated by the library.  Although the amount
+  of memory involved is not large (a scanline or so), it may cause the
+  initial max_memory_to_use specification to be exceeded, which in the worst
+  case would result in an out-of-memory failure.
+* do_block_smoothing can be changed before each call to jpeg_start_output().
+  This setting is relevant only when decoding a progressive JPEG image.
+  During the first DC-only scan, block smoothing provides a very "fuzzy" look
+  instead of the very "blocky" look seen without it; which is better seems a
+  matter of personal taste.  But block smoothing is nearly always a win
+  during later stages, especially when decoding a successive-approximation
+  image: smoothing helps to hide the slight blockiness that otherwise shows
+  up on smooth gradients until the lowest coefficient bits are sent.
+* Color quantization mode can be changed under the rules described below.
+  You *cannot* change between full-color and quantized output (because that
+  would alter the required I/O buffer sizes), but you can change which
+  quantization method is used.
+
+When generating color-quantized output, changing quantization method is a
+very useful way of switching between high-speed and high-quality display.
+The library allows you to change among its three quantization methods:
+1. Single-pass quantization to a fixed color cube.
+   Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
+2. Single-pass quantization to an application-supplied colormap.
+   Selected by setting cinfo.colormap to point to the colormap (the value of
+   two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
+3. Two-pass quantization to a colormap chosen specifically for the image.
+   Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
+   (This is the default setting selected by jpeg_read_header, but it is
+   probably NOT what you want for the first pass of progressive display!)
+These methods offer successively better quality and lesser speed.  However,
+only the first method is available for quantizing in non-RGB color spaces.
+
+IMPORTANT: because the different quantizer methods have very different
+working-storage requirements, the library requires you to indicate which
+one(s) you intend to use before you call jpeg_start_decompress().  (If we did
+not require this, the max_memory_to_use setting would be a complete fiction.)
+You do this by setting one or more of these three cinfo fields to TRUE:
+	enable_1pass_quant		Fixed color cube colormap
+	enable_external_quant		Externally-supplied colormap
+	enable_2pass_quant		Two-pass custom colormap
+All three are initialized FALSE by jpeg_read_header().  But
+jpeg_start_decompress() automatically sets TRUE the one selected by the
+current two_pass_quantize and colormap settings, so you only need to set the
+enable flags for any other quantization methods you plan to change to later.
+
+After setting the enable flags correctly at jpeg_start_decompress() time, you
+can change to any enabled quantization method by setting two_pass_quantize
+and colormap properly just before calling jpeg_start_output().  The following
+special rules apply:
+1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
+   or 2-pass mode from a different mode, or when you want the 2-pass
+   quantizer to be re-run to generate a new colormap.
+2. To switch to an external colormap, or to change to a different external
+   colormap than was used on the prior pass, you must call
+   jpeg_new_colormap() after setting cinfo.colormap.
+NOTE: if you want to use the same colormap as was used in the prior pass,
+you should not do either of these things.  This will save some nontrivial
+switchover costs.
+(These requirements exist because cinfo.colormap will always be non-NULL
+after completing a prior output pass, since both the 1-pass and 2-pass
+quantizers set it to point to their output colormaps.  Thus you have to
+do one of these two things to notify the library that something has changed.
+Yup, it's a bit klugy, but it's necessary to do it this way for backwards
+compatibility.)
+
+Note that in buffered-image mode, the library generates any requested colormap
+during jpeg_start_output(), not during jpeg_start_decompress().
+
+When using two-pass quantization, jpeg_start_output() makes a pass over the
+buffered image to determine the optimum color map; it therefore may take a
+significant amount of time, whereas ordinarily it does little work.  The
+progress monitor hook is called during this pass, if defined.  It is also
+important to realize that if the specified target scan number is greater than
+or equal to the current input scan number, jpeg_start_output() will attempt
+to consume input as it makes this pass.  If you use a suspending data source,
+you need to check for a FALSE return from jpeg_start_output() under these
+conditions.  The combination of 2-pass quantization and a not-yet-fully-read
+target scan is the only case in which jpeg_start_output() will consume input.
+
+
+Application authors who support buffered-image mode may be tempted to use it
+for all JPEG images, even single-scan ones.  This will work, but it is
+inefficient: there is no need to create an image-sized coefficient buffer for
+single-scan images.  Requesting buffered-image mode for such an image wastes
+memory.  Worse, it can cost time on large images, since the buffered data has
+to be swapped out or written to a temporary file.  If you are concerned about
+maximum performance on baseline JPEG files, you should use buffered-image
+mode only when the incoming file actually has multiple scans.  This can be
+tested by calling jpeg_has_multiple_scans(), which will return a correct
+result at any time after jpeg_read_header() completes.
+
+It is also worth noting that when you use jpeg_consume_input() to let input
+processing get ahead of output processing, the resulting pattern of access to
+the coefficient buffer is quite nonsequential.  It's best to use the memory
+manager jmemnobs.c if you can (ie, if you have enough real or virtual main
+memory).  If not, at least make sure that max_memory_to_use is set as high as
+possible.  If the JPEG memory manager has to use a temporary file, you will
+probably see a lot of disk traffic and poor performance.  (This could be
+improved with additional work on the memory manager, but we haven't gotten
+around to it yet.)
+
+In some applications it may be convenient to use jpeg_consume_input() for all
+input processing, including reading the initial markers; that is, you may
+wish to call jpeg_consume_input() instead of jpeg_read_header() during
+startup.  This works, but note that you must check for JPEG_REACHED_SOS and
+JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
+Once the first SOS marker has been reached, you must call
+jpeg_start_decompress() before jpeg_consume_input() will consume more input;
+it'll just keep returning JPEG_REACHED_SOS until you do.  If you read a
+tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
+without ever returning JPEG_REACHED_SOS; be sure to check for this case.
+If this happens, the decompressor will not read any more input until you call
+jpeg_abort() to reset it.  It is OK to call jpeg_consume_input() even when not
+using buffered-image mode, but in that case it's basically a no-op after the
+initial markers have been read: it will just return JPEG_SUSPENDED.
+
+
+Abbreviated datastreams and multiple images
+-------------------------------------------
+
+A JPEG compression or decompression object can be reused to process multiple
+images.  This saves a small amount of time per image by eliminating the
+"create" and "destroy" operations, but that isn't the real purpose of the
+feature.  Rather, reuse of an object provides support for abbreviated JPEG
+datastreams.  Object reuse can also simplify processing a series of images in
+a single input or output file.  This section explains these features.
+
+A JPEG file normally contains several hundred bytes worth of quantization
+and Huffman tables.  In a situation where many images will be stored or
+transmitted with identical tables, this may represent an annoying overhead.
+The JPEG standard therefore permits tables to be omitted.  The standard
+defines three classes of JPEG datastreams:
+  * "Interchange" datastreams contain an image and all tables needed to decode
+     the image.  These are the usual kind of JPEG file.
+  * "Abbreviated image" datastreams contain an image, but are missing some or
+    all of the tables needed to decode that image.
+  * "Abbreviated table specification" (henceforth "tables-only") datastreams
+    contain only table specifications.
+To decode an abbreviated image, it is necessary to load the missing table(s)
+into the decoder beforehand.  This can be accomplished by reading a separate
+tables-only file.  A variant scheme uses a series of images in which the first
+image is an interchange (complete) datastream, while subsequent ones are
+abbreviated and rely on the tables loaded by the first image.  It is assumed
+that once the decoder has read a table, it will remember that table until a
+new definition for the same table number is encountered.
+
+It is the application designer's responsibility to figure out how to associate
+the correct tables with an abbreviated image.  While abbreviated datastreams
+can be useful in a closed environment, their use is strongly discouraged in
+any situation where data exchange with other applications might be needed.
+Caveat designer.
+
+The JPEG library provides support for reading and writing any combination of
+tables-only datastreams and abbreviated images.  In both compression and
+decompression objects, a quantization or Huffman table will be retained for
+the lifetime of the object, unless it is overwritten by a new table definition.
+
+
+To create abbreviated image datastreams, it is only necessary to tell the
+compressor not to emit some or all of the tables it is using.  Each
+quantization and Huffman table struct contains a boolean field "sent_table",
+which normally is initialized to FALSE.  For each table used by the image, the
+header-writing process emits the table and sets sent_table = TRUE unless it is
+already TRUE.  (In normal usage, this prevents outputting the same table
+definition multiple times, as would otherwise occur because the chroma
+components typically share tables.)  Thus, setting this field to TRUE before
+calling jpeg_start_compress() will prevent the table from being written at
+all.
+
+If you want to create a "pure" abbreviated image file containing no tables,
+just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
+tables.  If you want to emit some but not all tables, you'll need to set the
+individual sent_table fields directly.
+
+To create an abbreviated image, you must also call jpeg_start_compress()
+with a second parameter of FALSE, not TRUE.  Otherwise jpeg_start_compress()
+will force all the sent_table fields to FALSE.  (This is a safety feature to
+prevent abbreviated images from being created accidentally.)
+
+To create a tables-only file, perform the same parameter setup that you
+normally would, but instead of calling jpeg_start_compress() and so on, call
+jpeg_write_tables(&cinfo).  This will write an abbreviated datastream
+containing only SOI, DQT and/or DHT markers, and EOI.  All the quantization
+and Huffman tables that are currently defined in the compression object will
+be emitted unless their sent_tables flag is already TRUE, and then all the
+sent_tables flags will be set TRUE.
+
+A sure-fire way to create matching tables-only and abbreviated image files
+is to proceed as follows:
+
+	create JPEG compression object
+	set JPEG parameters
+	set destination to tables-only file
+	jpeg_write_tables(&cinfo);
+	set destination to image file
+	jpeg_start_compress(&cinfo, FALSE);
+	write data...
+	jpeg_finish_compress(&cinfo);
+
+Since the JPEG parameters are not altered between writing the table file and
+the abbreviated image file, the same tables are sure to be used.  Of course,
+you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
+many times to produce many abbreviated image files matching the table file.
+
+You cannot suppress output of the computed Huffman tables when Huffman
+optimization is selected.  (If you could, there'd be no way to decode the
+image...)  Generally, you don't want to set optimize_coding = TRUE when
+you are trying to produce abbreviated files.
+
+In some cases you might want to compress an image using tables which are
+not stored in the application, but are defined in an interchange or
+tables-only file readable by the application.  This can be done by setting up
+a JPEG decompression object to read the specification file, then copying the
+tables into your compression object.  See jpeg_copy_critical_parameters()
+for an example of copying quantization tables.
+
+
+To read abbreviated image files, you simply need to load the proper tables
+into the decompression object before trying to read the abbreviated image.
+If the proper tables are stored in the application program, you can just
+allocate the table structs and fill in their contents directly.  For example,
+to load a fixed quantization table into table slot "n":
+
+    if (cinfo.quant_tbl_ptrs[n] == NULL)
+      cinfo.quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
+    quant_ptr = cinfo.quant_tbl_ptrs[n];	/* quant_ptr is JQUANT_TBL* */
+    for (i = 0; i < 64; i++) {
+      /* Qtable[] is desired quantization table, in natural array order */
+      quant_ptr->quantval[i] = Qtable[i];
+    }
+
+Code to load a fixed Huffman table is typically (for AC table "n"):
+
+    if (cinfo.ac_huff_tbl_ptrs[n] == NULL)
+      cinfo.ac_huff_tbl_ptrs[n] = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
+    huff_ptr = cinfo.ac_huff_tbl_ptrs[n];	/* huff_ptr is JHUFF_TBL* */
+    for (i = 1; i <= 16; i++) {
+      /* counts[i] is number of Huffman codes of length i bits, i=1..16 */
+      huff_ptr->bits[i] = counts[i];
+    }
+    for (i = 0; i < 256; i++) {
+      /* symbols[] is the list of Huffman symbols, in code-length order */
+      huff_ptr->huffval[i] = symbols[i];
+    }
+
+(Note that trying to set cinfo.quant_tbl_ptrs[n] to point directly at a
+constant JQUANT_TBL object is not safe.  If the incoming file happened to
+contain a quantization table definition, your master table would get
+overwritten!  Instead allocate a working table copy and copy the master table
+into it, as illustrated above.  Ditto for Huffman tables, of course.)
+
+You might want to read the tables from a tables-only file, rather than
+hard-wiring them into your application.  The jpeg_read_header() call is
+sufficient to read a tables-only file.  You must pass a second parameter of
+FALSE to indicate that you do not require an image to be present.  Thus, the
+typical scenario is
+
+	create JPEG decompression object
+	set source to tables-only file
+	jpeg_read_header(&cinfo, FALSE);
+	set source to abbreviated image file
+	jpeg_read_header(&cinfo, TRUE);
+	set decompression parameters
+	jpeg_start_decompress(&cinfo);
+	read data...
+	jpeg_finish_decompress(&cinfo);
+
+In some cases, you may want to read a file without knowing whether it contains
+an image or just tables.  In that case, pass FALSE and check the return value
+from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
+JPEG_HEADER_TABLES_ONLY if only tables were found.  (A third return value,
+JPEG_SUSPENDED, is possible when using a suspending data source manager.)
+Note that jpeg_read_header() will not complain if you read an abbreviated
+image for which you haven't loaded the missing tables; the missing-table check
+occurs later, in jpeg_start_decompress().
+
+
+It is possible to read a series of images from a single source file by
+repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
+without releasing/recreating the JPEG object or the data source module.
+(If you did reinitialize, any partial bufferload left in the data source
+buffer at the end of one image would be discarded, causing you to lose the
+start of the next image.)  When you use this method, stored tables are
+automatically carried forward, so some of the images can be abbreviated images
+that depend on tables from earlier images.
+
+If you intend to write a series of images into a single destination file,
+you might want to make a specialized data destination module that doesn't
+flush the output buffer at term_destination() time.  This would speed things
+up by some trifling amount.  Of course, you'd need to remember to flush the
+buffer after the last image.  You can make the later images be abbreviated
+ones by passing FALSE to jpeg_start_compress().
+
+
+Special markers
+---------------
+
+Some applications may need to insert or extract special data in the JPEG
+datastream.  The JPEG standard provides marker types "COM" (comment) and
+"APP0" through "APP15" (application) to hold application-specific data.
+Unfortunately, the use of these markers is not specified by the standard.
+COM markers are fairly widely used to hold user-supplied text.  The JFIF file
+format spec uses APP0 markers with specified initial strings to hold certain
+data.  Adobe applications use APP14 markers beginning with the string "Adobe"
+for miscellaneous data.  Other APPn markers are rarely seen, but might
+contain almost anything.
+
+If you wish to store user-supplied text, we recommend you use COM markers
+and place readable 7-bit ASCII text in them.  Newline conventions are not
+standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
+(Mac style).  A robust COM reader should be able to cope with random binary
+garbage, including nulls, since some applications generate COM markers
+containing non-ASCII junk.  (But yours should not be one of them.)
+
+For program-supplied data, use an APPn marker, and be sure to begin it with an
+identifying string so that you can tell whether the marker is actually yours.
+It's probably best to avoid using APP0 or APP14 for any private markers.
+(NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
+not use APP8 markers for any private purposes, either.)
+
+Keep in mind that at most 65533 bytes can be put into one marker, but you
+can have as many markers as you like.
+
+By default, the IJG compression library will write a JFIF APP0 marker if the
+selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
+the selected colorspace is RGB, CMYK, or YCCK.  You can disable this, but
+we don't recommend it.  The decompression library will recognize JFIF and
+Adobe markers and will set the JPEG colorspace properly when one is found.
+
+
+You can write special markers immediately following the datastream header by
+calling jpeg_write_marker() after jpeg_start_compress() and before the first
+call to jpeg_write_scanlines().  When you do this, the markers appear after
+the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
+all else.  Specify the marker type parameter as "JPEG_COM" for COM or
+"JPEG_APP0 + n" for APPn.  (Actually, jpeg_write_marker will let you write
+any marker type, but we don't recommend writing any other kinds of marker.)
+For example, to write a user comment string pointed to by comment_text:
+	jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
+
+If it's not convenient to store all the marker data in memory at once,
+you can instead call jpeg_write_m_header() followed by multiple calls to
+jpeg_write_m_byte().  If you do it this way, it's your responsibility to
+call jpeg_write_m_byte() exactly the number of times given in the length
+parameter to jpeg_write_m_header().  (This method lets you empty the
+output buffer partway through a marker, which might be important when
+using a suspending data destination module.  In any case, if you are using
+a suspending destination, you should flush its buffer after inserting
+any special markers.  See "I/O suspension".)
+
+Or, if you prefer to synthesize the marker byte sequence yourself,
+you can just cram it straight into the data destination module.
+
+If you are writing JFIF 1.02 extension markers (thumbnail images), don't
+forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
+correct JFIF version number in the JFIF header marker.  The library's default
+is to write version 1.01, but that's wrong if you insert any 1.02 extension
+markers.  (We could probably get away with just defaulting to 1.02, but there
+used to be broken decoders that would complain about unknown minor version
+numbers.  To reduce compatibility risks it's safest not to write 1.02 unless
+you are actually using 1.02 extensions.)
+
+
+When reading, two methods of handling special markers are available:
+1. You can ask the library to save the contents of COM and/or APPn markers
+into memory, and then examine them at your leisure afterwards.
+2. You can supply your own routine to process COM and/or APPn markers
+on-the-fly as they are read.
+The first method is simpler to use, especially if you are using a suspending
+data source; writing a marker processor that copes with input suspension is
+not easy (consider what happens if the marker is longer than your available
+input buffer).  However, the second method conserves memory since the marker
+data need not be kept around after it's been processed.
+
+For either method, you'd normally set up marker handling after creating a
+decompression object and before calling jpeg_read_header(), because the
+markers of interest will typically be near the head of the file and so will
+be scanned by jpeg_read_header.  Once you've established a marker handling
+method, it will be used for the life of that decompression object
+(potentially many datastreams), unless you change it.  Marker handling is
+determined separately for COM markers and for each APPn marker code.
+
+
+To save the contents of special markers in memory, call
+	jpeg_save_markers(cinfo, marker_code, length_limit)
+where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
+(To arrange to save all the special marker types, you need to call this
+routine 17 times, for COM and APP0-APP15.)  If the incoming marker is longer
+than length_limit data bytes, only length_limit bytes will be saved; this
+parameter allows you to avoid chewing up memory when you only need to see the
+first few bytes of a potentially large marker.  If you want to save all the
+data, set length_limit to 0xFFFF; that is enough since marker lengths are only
+16 bits.  As a special case, setting length_limit to 0 prevents that marker
+type from being saved at all.  (That is the default behavior, in fact.)
+
+After jpeg_read_header() completes, you can examine the special markers by
+following the cinfo->marker_list pointer chain.  All the special markers in
+the file appear in this list, in order of their occurrence in the file (but
+omitting any markers of types you didn't ask for).  Both the original data
+length and the saved data length are recorded for each list entry; the latter
+will not exceed length_limit for the particular marker type.  Note that these
+lengths exclude the marker length word, whereas the stored representation
+within the JPEG file includes it.  (Hence the maximum data length is really
+only 65533.)
+
+It is possible that additional special markers appear in the file beyond the
+SOS marker at which jpeg_read_header stops; if so, the marker list will be
+extended during reading of the rest of the file.  This is not expected to be
+common, however.  If you are short on memory you may want to reset the length
+limit to zero for all marker types after finishing jpeg_read_header, to
+ensure that the max_memory_to_use setting cannot be exceeded due to addition
+of later markers.
+
+The marker list remains stored until you call jpeg_finish_decompress or
+jpeg_abort, at which point the memory is freed and the list is set to empty.
+(jpeg_destroy also releases the storage, of course.)
+
+Note that the library is internally interested in APP0 and APP14 markers;
+if you try to set a small nonzero length limit on these types, the library
+will silently force the length up to the minimum it wants.  (But you can set
+a zero length limit to prevent them from being saved at all.)  Also, in a
+16-bit environment, the maximum length limit may be constrained to less than
+65533 by malloc() limitations.  It is therefore best not to assume that the
+effective length limit is exactly what you set it to be.
+
+
+If you want to supply your own marker-reading routine, you do it by calling
+jpeg_set_marker_processor().  A marker processor routine must have the
+signature
+	boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
+Although the marker code is not explicitly passed, the routine can find it
+in cinfo->unread_marker.  At the time of call, the marker proper has been
+read from the data source module.  The processor routine is responsible for
+reading the marker length word and the remaining parameter bytes, if any.
+Return TRUE to indicate success.  (FALSE should be returned only if you are
+using a suspending data source and it tells you to suspend.  See the standard
+marker processors in jdmarker.c for appropriate coding methods if you need to
+use a suspending data source.)
+
+If you override the default APP0 or APP14 processors, it is up to you to
+recognize JFIF and Adobe markers if you want colorspace recognition to occur
+properly.  We recommend copying and extending the default processors if you
+want to do that.  (A better idea is to save these marker types for later
+examination by calling jpeg_save_markers(); that method doesn't interfere
+with the library's own processing of these markers.)
+
+jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
+--- if you call one it overrides any previous call to the other, for the
+particular marker type specified.
+
+A simple example of an external COM processor can be found in djpeg.c.
+Also, see jpegtran.c for an example of using jpeg_save_markers.
+
+
+Raw (downsampled) image data
+----------------------------
+
+Some applications need to supply already-downsampled image data to the JPEG
+compressor, or to receive raw downsampled data from the decompressor.  The
+library supports this requirement by allowing the application to write or
+read raw data, bypassing the normal preprocessing or postprocessing steps.
+The interface is different from the standard one and is somewhat harder to
+use.  If your interest is merely in bypassing color conversion, we recommend
+that you use the standard interface and simply set jpeg_color_space =
+in_color_space (or jpeg_color_space = out_color_space for decompression).
+The mechanism described in this section is necessary only to supply or
+receive downsampled image data, in which not all components have the same
+dimensions.
+
+
+To compress raw data, you must supply the data in the colorspace to be used
+in the JPEG file (please read the earlier section on Special color spaces)
+and downsampled to the sampling factors specified in the JPEG parameters.
+You must supply the data in the format used internally by the JPEG library,
+namely a JSAMPIMAGE array.  This is an array of pointers to two-dimensional
+arrays, each of type JSAMPARRAY.  Each 2-D array holds the values for one
+color component.  This structure is necessary since the components are of
+different sizes.  If the image dimensions are not a multiple of the MCU size,
+you must also pad the data correctly (usually, this is done by replicating
+the last column and/or row).  The data must be padded to a multiple of a DCT
+block in each component: that is, each downsampled row must contain a
+multiple of 8 valid samples, and there must be a multiple of 8 sample rows
+for each component.  (For applications such as conversion of digital TV
+images, the standard image size is usually a multiple of the DCT block size,
+so that no padding need actually be done.)
+
+The procedure for compression of raw data is basically the same as normal
+compression, except that you call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  Before calling jpeg_start_compress(), you must do
+the following:
+  * Set cinfo->raw_data_in to TRUE.  (It is set FALSE by jpeg_set_defaults().)
+    This notifies the library that you will be supplying raw data.
+  * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
+    call is a good idea.  Note that since color conversion is bypassed,
+    in_color_space is ignored, except that jpeg_set_defaults() uses it to
+    choose the default jpeg_color_space setting.
+  * Ensure the sampling factors, cinfo->comp_info[i].h_samp_factor and
+    cinfo->comp_info[i].v_samp_factor, are correct.  Since these indicate the
+    dimensions of the data you are supplying, it's wise to set them
+    explicitly, rather than assuming the library's defaults are what you want.
+
+To pass raw data to the library, call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  The two routines work similarly except that
+jpeg_write_raw_data takes a JSAMPIMAGE data array rather than JSAMPARRAY.
+The scanlines count passed to and returned from jpeg_write_raw_data is
+measured in terms of the component with the largest v_samp_factor.
+
+jpeg_write_raw_data() processes one MCU row per call, which is to say
+v_samp_factor*DCTSIZE sample rows of each component.  The passed num_lines
+value must be at least max_v_samp_factor*DCTSIZE, and the return value will
+be exactly that amount (or possibly some multiple of that amount, in future
+library versions).  This is true even on the last call at the bottom of the
+image; don't forget to pad your data as necessary.
+
+The required dimensions of the supplied data can be computed for each
+component as
+	cinfo->comp_info[i].width_in_blocks*DCTSIZE  samples per row
+	cinfo->comp_info[i].height_in_blocks*DCTSIZE rows in image
+after jpeg_start_compress() has initialized those fields.  If the valid data
+is smaller than this, it must be padded appropriately.  For some sampling
+factors and image sizes, additional dummy DCT blocks are inserted to make
+the image a multiple of the MCU dimensions.  The library creates such dummy
+blocks itself; it does not read them from your supplied data.  Therefore you
+need never pad by more than DCTSIZE samples.  An example may help here.
+Assume 2h2v downsampling of YCbCr data, that is
+	cinfo->comp_info[0].h_samp_factor = 2		for Y
+	cinfo->comp_info[0].v_samp_factor = 2
+	cinfo->comp_info[1].h_samp_factor = 1		for Cb
+	cinfo->comp_info[1].v_samp_factor = 1
+	cinfo->comp_info[2].h_samp_factor = 1		for Cr
+	cinfo->comp_info[2].v_samp_factor = 1
+and suppose that the nominal image dimensions (cinfo->image_width and
+cinfo->image_height) are 101x101 pixels.  Then jpeg_start_compress() will
+compute downsampled_width = 101 and width_in_blocks = 13 for Y,
+downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
+for the height fields).  You must pad the Y data to at least 13*8 = 104
+columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows.  The
+MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
+scanlines on each call to jpeg_write_raw_data(), which is to say 16 actual
+sample rows of Y and 8 each of Cb and Cr.  A total of 7 MCU rows are needed,
+so you must pass a total of 7*16 = 112 "scanlines".  The last DCT block row
+of Y data is dummy, so it doesn't matter what you pass for it in the data
+arrays, but the scanlines count must total up to 112 so that all of the Cb
+and Cr data gets passed.
+
+Output suspension is supported with raw-data compression: if the data
+destination module suspends, jpeg_write_raw_data() will return 0.
+In this case the same data rows must be passed again on the next call.
+
+
+Decompression with raw data output implies bypassing all postprocessing:
+you cannot ask for rescaling or color quantization, for instance.  More
+seriously, you must deal with the color space and sampling factors present in
+the incoming file.  If your application only handles, say, 2h1v YCbCr data,
+you must check for and fail on other color spaces or other sampling factors.
+The library will not convert to a different color space for you.
+
+To obtain raw data output, set cinfo->raw_data_out = TRUE before
+jpeg_start_decompress() (it is set FALSE by jpeg_read_header()).  Be sure to
+verify that the color space and sampling factors are ones you can handle.
+Then call jpeg_read_raw_data() in place of jpeg_read_scanlines().  The
+decompression process is otherwise the same as usual.
+
+jpeg_read_raw_data() returns one MCU row per call, and thus you must pass a
+buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
+the same as for raw-data compression).  The buffer you pass must be large
+enough to hold the actual data plus padding to DCT-block boundaries.  As with
+compression, any entirely dummy DCT blocks are not processed so you need not
+allocate space for them, but the total scanline count includes them.  The
+above example of computing buffer dimensions for raw-data compression is
+equally valid for decompression.
+
+Input suspension is supported with raw-data decompression: if the data source
+module suspends, jpeg_read_raw_data() will return 0.  You can also use
+buffered-image mode to read raw data in multiple passes.
+
+
+Really raw data: DCT coefficients
+---------------------------------
+
+It is possible to read or write the contents of a JPEG file as raw DCT
+coefficients.  This facility is mainly intended for use in lossless
+transcoding between different JPEG file formats.  Other possible applications
+include lossless cropping of a JPEG image, lossless reassembly of a
+multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
+
+To read the contents of a JPEG file as DCT coefficients, open the file and do
+jpeg_read_header() as usual.  But instead of calling jpeg_start_decompress()
+and jpeg_read_scanlines(), call jpeg_read_coefficients().  This will read the
+entire image into a set of virtual coefficient-block arrays, one array per
+component.  The return value is a pointer to an array of virtual-array
+descriptors.  Each virtual array can be accessed directly using the JPEG
+memory manager's access_virt_barray method (see Memory management, below,
+and also read structure.doc's discussion of virtual array handling).  Or,
+for simple transcoding to a different JPEG file format, the array list can
+just be handed directly to jpeg_write_coefficients().
+
+Each block in the block arrays contains quantized coefficient values in
+normal array order (not JPEG zigzag order).  The block arrays contain only
+DCT blocks containing real data; any entirely-dummy blocks added to fill out
+interleaved MCUs at the right or bottom edges of the image are discarded
+during reading and are not stored in the block arrays.  (The size of each
+block array can be determined from the width_in_blocks and height_in_blocks
+fields of the component's comp_info entry.)  This is also the data format
+expected by jpeg_write_coefficients().
+
+When you are done using the virtual arrays, call jpeg_finish_decompress()
+to release the array storage and return the decompression object to an idle
+state; or just call jpeg_destroy() if you don't need to reuse the object.
+
+If you use a suspending data source, jpeg_read_coefficients() will return
+NULL if it is forced to suspend; a non-NULL return value indicates successful
+completion.  You need not test for a NULL return value when using a
+non-suspending data source.
+
+It is also possible to call jpeg_read_coefficients() to obtain access to the
+decoder's coefficient arrays during a normal decode cycle in buffered-image
+mode.  This frammish might be useful for progressively displaying an incoming
+image and then re-encoding it without loss.  To do this, decode in buffered-
+image mode as discussed previously, then call jpeg_read_coefficients() after
+the last jpeg_finish_output() call.  The arrays will be available for your use
+until you call jpeg_finish_decompress().
+
+
+To write the contents of a JPEG file as DCT coefficients, you must provide
+the DCT coefficients stored in virtual block arrays.  You can either pass
+block arrays read from an input JPEG file by jpeg_read_coefficients(), or
+allocate virtual arrays from the JPEG compression object and fill them
+yourself.  In either case, jpeg_write_coefficients() is substituted for
+jpeg_start_compress() and jpeg_write_scanlines().  Thus the sequence is
+  * Create compression object
+  * Set all compression parameters as necessary
+  * Request virtual arrays if needed
+  * jpeg_write_coefficients()
+  * jpeg_finish_compress()
+  * Destroy or re-use compression object
+jpeg_write_coefficients() is passed a pointer to an array of virtual block
+array descriptors; the number of arrays is equal to cinfo.num_components.
+
+The virtual arrays need only have been requested, not realized, before
+jpeg_write_coefficients() is called.  A side-effect of
+jpeg_write_coefficients() is to realize any virtual arrays that have been
+requested from the compression object's memory manager.  Thus, when obtaining
+the virtual arrays from the compression object, you should fill the arrays
+after calling jpeg_write_coefficients().  The data is actually written out
+when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
+the file header.
+
+When writing raw DCT coefficients, it is crucial that the JPEG quantization
+tables and sampling factors match the way the data was encoded, or the
+resulting file will be invalid.  For transcoding from an existing JPEG file,
+we recommend using jpeg_copy_critical_parameters().  This routine initializes
+all the compression parameters to default values (like jpeg_set_defaults()),
+then copies the critical information from a source decompression object.
+The decompression object should have just been used to read the entire
+JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
+
+jpeg_write_coefficients() marks all tables stored in the compression object
+as needing to be written to the output file (thus, it acts like
+jpeg_start_compress(cinfo, TRUE)).  This is for safety's sake, to avoid
+emitting abbreviated JPEG files by accident.  If you really want to emit an
+abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
+individual sent_table flags, between calling jpeg_write_coefficients() and
+jpeg_finish_compress().
+
+
+Progress monitoring
+-------------------
+
+Some applications may need to regain control from the JPEG library every so
+often.  The typical use of this feature is to produce a percent-done bar or
+other progress display.  (For a simple example, see cjpeg.c or djpeg.c.)
+Although you do get control back frequently during the data-transferring pass
+(the jpeg_read_scanlines or jpeg_write_scanlines loop), any additional passes
+will occur inside jpeg_finish_compress or jpeg_start_decompress; those
+routines may take a long time to execute, and you don't get control back
+until they are done.
+
+You can define a progress-monitor routine which will be called periodically
+by the library.  No guarantees are made about how often this call will occur,
+so we don't recommend you use it for mouse tracking or anything like that.
+At present, a call will occur once per MCU row, scanline, or sample row
+group, whichever unit is convenient for the current processing mode; so the
+wider the image, the longer the time between calls.  During the data
+transferring pass, only one call occurs per call of jpeg_read_scanlines or
+jpeg_write_scanlines, so don't pass a large number of scanlines at once if
+you want fine resolution in the progress count.  (If you really need to use
+the callback mechanism for time-critical tasks like mouse tracking, you could
+insert additional calls inside some of the library's inner loops.)
+
+To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
+fill in its progress_monitor field with a pointer to your callback routine,
+and set cinfo->progress to point to the struct.  The callback will be called
+whenever cinfo->progress is non-NULL.  (This pointer is set to NULL by
+jpeg_create_compress or jpeg_create_decompress; the library will not change
+it thereafter.  So if you allocate dynamic storage for the progress struct,
+make sure it will live as long as the JPEG object does.  Allocating from the
+JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.)  You
+can use the same callback routine for both compression and decompression.
+
+The jpeg_progress_mgr struct contains four fields which are set by the library:
+	long pass_counter;	/* work units completed in this pass */
+	long pass_limit;	/* total number of work units in this pass */
+	int completed_passes;	/* passes completed so far */
+	int total_passes;	/* total number of passes expected */
+During any one pass, pass_counter increases from 0 up to (not including)
+pass_limit; the step size is usually but not necessarily 1.  The pass_limit
+value may change from one pass to another.  The expected total number of
+passes is in total_passes, and the number of passes already completed is in
+completed_passes.  Thus the fraction of work completed may be estimated as
+		completed_passes + (pass_counter/pass_limit)
+		--------------------------------------------
+				total_passes
+ignoring the fact that the passes may not be equal amounts of work.
+
+When decompressing, pass_limit can even change within a pass, because it
+depends on the number of scans in the JPEG file, which isn't always known in
+advance.  The computed fraction-of-work-done may jump suddenly (if the library
+discovers it has overestimated the number of scans) or even decrease (in the
+opposite case).  It is not wise to put great faith in the work estimate.
+
+When using the decompressor's buffered-image mode, the progress monitor work
+estimate is likely to be completely unhelpful, because the library has no way
+to know how many output passes will be demanded of it.  Currently, the library
+sets total_passes based on the assumption that there will be one more output
+pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
+TRUE), but no more output passes if the file end has been reached when the
+output pass is started.  This means that total_passes will rise as additional
+output passes are requested.  If you have a way of determining the input file
+size, estimating progress based on the fraction of the file that's been read
+will probably be more useful than using the library's value.
+
+
+Memory management
+-----------------
+
+This section covers some key facts about the JPEG library's built-in memory
+manager.  For more info, please read structure.doc's section about the memory
+manager, and consult the source code if necessary.
+
+All memory and temporary file allocation within the library is done via the
+memory manager.  If necessary, you can replace the "back end" of the memory
+manager to control allocation yourself (for example, if you don't want the
+library to use malloc() and free() for some reason).
+
+Some data is allocated "permanently" and will not be freed until the JPEG
+object is destroyed.  Most data is allocated "per image" and is freed by
+jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort.  You can call the
+memory manager yourself to allocate structures that will automatically be
+freed at these times.  Typical code for this is
+  ptr = (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, size);
+Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
+Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
+There are also alloc_sarray and alloc_barray routines that automatically
+build 2-D sample or block arrays.
+
+The library's minimum space requirements to process an image depend on the
+image's width, but not on its height, because the library ordinarily works
+with "strip" buffers that are as wide as the image but just a few rows high.
+Some operating modes (eg, two-pass color quantization) require full-image
+buffers.  Such buffers are treated as "virtual arrays": only the current strip
+need be in memory, and the rest can be swapped out to a temporary file.
+
+If you use the simplest memory manager back end (jmemnobs.c), then no
+temporary files are used; virtual arrays are simply malloc()'d.  Images bigger
+than memory can be processed only if your system supports virtual memory.
+The other memory manager back ends support temporary files of various flavors
+and thus work in machines without virtual memory.  They may also be useful on
+Unix machines if you need to process images that exceed available swap space.
+
+When using temporary files, the library will make the in-memory buffers for
+its virtual arrays just big enough to stay within a "maximum memory" setting.
+Your application can set this limit by setting cinfo->mem->max_memory_to_use
+after creating the JPEG object.  (Of course, there is still a minimum size for
+the buffers, so the max-memory setting is effective only if it is bigger than
+the minimum space needed.)  If you allocate any large structures yourself, you
+must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
+order to have them counted against the max memory limit.  Also keep in mind
+that space allocated with alloc_small() is ignored, on the assumption that
+it's too small to be worth worrying about; so a reasonable safety margin
+should be left when setting max_memory_to_use.
+
+If you use the jmemname.c or jmemdos.c memory manager back end, it is
+important to clean up the JPEG object properly to ensure that the temporary
+files get deleted.  (This is especially crucial with jmemdos.c, where the
+"temporary files" may be extended-memory segments; if they are not freed,
+DOS will require a reboot to recover the memory.)  Thus, with these memory
+managers, it's a good idea to provide a signal handler that will trap any
+early exit from your program.  The handler should call either jpeg_abort()
+or jpeg_destroy() for any active JPEG objects.  A handler is not needed with
+jmemnobs.c, and shouldn't be necessary with jmemansi.c or jmemmac.c either,
+since the C library is supposed to take care of deleting files made with
+tmpfile().
+
+
+Memory usage
+------------
+
+Working memory requirements while performing compression or decompression
+depend on image dimensions, image characteristics (such as colorspace and
+JPEG process), and operating mode (application-selected options).
+
+As of v6b, the decompressor requires:
+ 1. About 24K in more-or-less-fixed-size data.  This varies a bit depending
+    on operating mode and image characteristics (particularly color vs.
+    grayscale), but it doesn't depend on image dimensions.
+ 2. Strip buffers (of size proportional to the image width) for IDCT and
+    upsampling results.  The worst case for commonly used sampling factors
+    is about 34 bytes * width in pixels for a color image.  A grayscale image
+    only needs about 8 bytes per pixel column.
+ 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
+    file (including progressive JPEGs), or whenever you select buffered-image
+    mode.  This takes 2 bytes/coefficient.  At typical 2x2 sampling, that's
+    3 bytes per pixel for a color image.  Worst case (1x1 sampling) requires
+    6 bytes/pixel.  For grayscale, figure 2 bytes/pixel.
+ 4. To perform 2-pass color quantization, the decompressor also needs a
+    128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
+This does not count any memory allocated by the application, such as a
+buffer to hold the final output image.
+
+The above figures are valid for 8-bit JPEG data precision and a machine with
+32-bit ints.  For 12-bit JPEG data, double the size of the strip buffers and
+quantization pixel buffer.  The "fixed-size" data will be somewhat smaller
+with 16-bit ints, larger with 64-bit ints.  Also, CMYK or other unusual
+color spaces will require different amounts of space.
+
+The full-image coefficient and pixel buffers, if needed at all, do not
+have to be fully RAM resident; you can have the library use temporary
+files instead when the total memory usage would exceed a limit you set.
+(But if your OS supports virtual memory, it's probably better to just use
+jmemnobs and let the OS do the swapping.)
+
+The compressor's memory requirements are similar, except that it has no need
+for color quantization.  Also, it needs a full-image DCT coefficient buffer
+if Huffman-table optimization is asked for, even if progressive mode is not
+requested.
+
+If you need more detailed information about memory usage in a particular
+situation, you can enable the MEM_STATS code in jmemmgr.c.
+
+
+Library compile-time options
+----------------------------
+
+A number of compile-time options are available by modifying jmorecfg.h.
+
+The JPEG standard provides for both the baseline 8-bit DCT process and
+a 12-bit DCT process.  The IJG code supports 12-bit lossy JPEG if you define
+BITS_IN_JSAMPLE as 12 rather than 8.  Note that this causes JSAMPLE to be
+larger than a char, so it affects the surrounding application's image data.
+The sample applications cjpeg and djpeg can support 12-bit mode only for PPM
+and GIF file formats; you must disable the other file formats to compile a
+12-bit cjpeg or djpeg.  (install.doc has more information about that.)
+At present, a 12-bit library can handle *only* 12-bit images, not both
+precisions.  (If you need to include both 8- and 12-bit libraries in a single
+application, you could probably do it by defining NEED_SHORT_EXTERNAL_NAMES
+for just one of the copies.  You'd have to access the 8-bit and 12-bit copies
+from separate application source files.  This is untested ... if you try it,
+we'd like to hear whether it works!)
+
+Note that a 12-bit library always compresses in Huffman optimization mode,
+in order to generate valid Huffman tables.  This is necessary because our
+default Huffman tables only cover 8-bit data.  If you need to output 12-bit
+files in one pass, you'll have to supply suitable default Huffman tables.
+You may also want to supply your own DCT quantization tables; the existing
+quality-scaling code has been developed for 8-bit use, and probably doesn't
+generate especially good tables for 12-bit.
+
+The maximum number of components (color channels) in the image is determined
+by MAX_COMPONENTS.  The JPEG standard allows up to 255 components, but we
+expect that few applications will need more than four or so.
+
+On machines with unusual data type sizes, you may be able to improve
+performance or reduce memory space by tweaking the various typedefs in
+jmorecfg.h.  In particular, on some RISC CPUs, access to arrays of "short"s
+is quite slow; consider trading memory for speed by making JCOEF, INT16, and
+UINT16 be "int" or "unsigned int".  UINT8 is also a candidate to become int.
+You probably don't want to make JSAMPLE be int unless you have lots of memory
+to burn.
+
+You can reduce the size of the library by compiling out various optional
+functions.  To do this, undefine xxx_SUPPORTED symbols as necessary.
+
+You can also save a few K by not having text error messages in the library;
+the standard error message table occupies about 5Kb.  This is particularly
+reasonable for embedded applications where there's no good way to display 
+a message anyway.  To do this, remove the creation of the message table
+(jpeg_std_message_table[]) from jerror.c, and alter format_message to do
+something reasonable without it.  You could output the numeric value of the
+message code number, for example.  If you do this, you can also save a couple
+more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
+you don't need trace capability anyway, right?
+
+
+Portability considerations
+--------------------------
+
+The JPEG library has been written to be extremely portable; the sample
+applications cjpeg and djpeg are slightly less so.  This section summarizes
+the design goals in this area.  (If you encounter any bugs that cause the
+library to be less portable than is claimed here, we'd appreciate hearing
+about them.)
+
+The code works fine on ANSI C, C++, and pre-ANSI C compilers, using any of
+the popular system include file setups, and some not-so-popular ones too.
+See install.doc for configuration procedures.
+
+The code is not dependent on the exact sizes of the C data types.  As
+distributed, we make the assumptions that
+	char	is at least 8 bits wide
+	short	is at least 16 bits wide
+	int	is at least 16 bits wide
+	long	is at least 32 bits wide
+(These are the minimum requirements of the ANSI C standard.)  Wider types will
+work fine, although memory may be used inefficiently if char is much larger
+than 8 bits or short is much bigger than 16 bits.  The code should work
+equally well with 16- or 32-bit ints.
+
+In a system where these assumptions are not met, you may be able to make the
+code work by modifying the typedefs in jmorecfg.h.  However, you will probably
+have difficulty if int is less than 16 bits wide, since references to plain
+int abound in the code.
+
+char can be either signed or unsigned, although the code runs faster if an
+unsigned char type is available.  If char is wider than 8 bits, you will need
+to redefine JOCTET and/or provide custom data source/destination managers so
+that JOCTET represents exactly 8 bits of data on external storage.
+
+The JPEG library proper does not assume ASCII representation of characters.
+But some of the image file I/O modules in cjpeg/djpeg do have ASCII
+dependencies in file-header manipulation; so does cjpeg's select_file_type()
+routine.
+
+The JPEG library does not rely heavily on the C library.  In particular, C
+stdio is used only by the data source/destination modules and the error
+handler, all of which are application-replaceable.  (cjpeg/djpeg are more
+heavily dependent on stdio.)  malloc and free are called only from the memory
+manager "back end" module, so you can use a different memory allocator by
+replacing that one file.
+
+The code generally assumes that C names must be unique in the first 15
+characters.  However, global function names can be made unique in the
+first 6 characters by defining NEED_SHORT_EXTERNAL_NAMES.
+
+More info about porting the code may be gleaned by reading jconfig.doc,
+jmorecfg.h, and jinclude.h.
+
+
+Notes for MS-DOS implementors
+-----------------------------
+
+The IJG code is designed to work efficiently in 80x86 "small" or "medium"
+memory models (i.e., data pointers are 16 bits unless explicitly declared
+"far"; code pointers can be either size).  You may be able to use small
+model to compile cjpeg or djpeg by itself, but you will probably have to use
+medium model for any larger application.  This won't make much difference in
+performance.  You *will* take a noticeable performance hit if you use a
+large-data memory model (perhaps 10%-25%), and you should avoid "huge" model
+if at all possible.
+
+The JPEG library typically needs 2Kb-3Kb of stack space.  It will also
+malloc about 20K-30K of near heap space while executing (and lots of far
+heap, but that doesn't count in this calculation).  This figure will vary
+depending on selected operating mode, and to a lesser extent on image size.
+There is also about 5Kb-6Kb of constant data which will be allocated in the
+near data segment (about 4Kb of this is the error message table).
+Thus you have perhaps 20K available for other modules' static data and near
+heap space before you need to go to a larger memory model.  The C library's
+static data will account for several K of this, but that still leaves a good
+deal for your needs.  (If you are tight on space, you could reduce the sizes
+of the I/O buffers allocated by jdatasrc.c and jdatadst.c, say from 4K to
+1K.  Another possibility is to move the error message table to far memory;
+this should be doable with only localized hacking on jerror.c.)
+
+About 2K of the near heap space is "permanent" memory that will not be
+released until you destroy the JPEG object.  This is only an issue if you
+save a JPEG object between compression or decompression operations.
+
+Far data space may also be a tight resource when you are dealing with large
+images.  The most memory-intensive case is decompression with two-pass color
+quantization, or single-pass quantization to an externally supplied color
+map.  This requires a 128Kb color lookup table plus strip buffers amounting
+to about 40 bytes per column for typical sampling ratios (eg, about 25600
+bytes for a 640-pixel-wide image).  You may not be able to process wide
+images if you have large data structures of your own.
+
+Of course, all of these concerns vanish if you use a 32-bit flat-memory-model
+compiler, such as DJGPP or Watcom C.  We highly recommend flat model if you
+can use it; the JPEG library is significantly faster in flat model.
diff --git a/jpeg/makedepend b/jpeg/makedepend
new file mode 100644
index 000000000..ecf2b024c
--- /dev/null
+++ b/jpeg/makedepend
@@ -0,0 +1,94 @@
+# DO NOT DELETE
+
+jmemnobs.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jmemnobs.o: jerror.h jmemsys.h
+jcapimin.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcapimin.o: jerror.h
+jcapistd.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcapistd.o: jerror.h
+jccoefct.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jccoefct.o: jerror.h
+jccolor.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jccolor.o: jerror.h
+jcdctmgr.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcdctmgr.o: jerror.h jdct.h
+jchuff.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jchuff.o: jerror.h jchuff.h
+jcinit.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcinit.o: jerror.h
+jcmainct.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcmainct.o: jerror.h
+jcmarker.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcmarker.o: jerror.h
+jcmaster.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcmaster.o: jerror.h
+jcomapi.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcomapi.o: jerror.h
+jcparam.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcparam.o: jerror.h
+jcphuff.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcphuff.o: jerror.h jchuff.h
+jcprepct.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcprepct.o: jerror.h
+jcsample.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jcsample.o: jerror.h
+jctrans.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jctrans.o: jerror.h
+jdapimin.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdapimin.o: jerror.h
+jdapistd.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdapistd.o: jerror.h
+jdatadst.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdatadst.o: jerror.h
+jdatasrc.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdatasrc.o: jerror.h
+jdcoefct.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdcoefct.o: jerror.h
+jdcolor.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdcolor.o: jerror.h
+jddctmgr.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jddctmgr.o: jerror.h jdct.h
+jdhuff.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdhuff.o: jerror.h jdhuff.h
+jdinput.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdinput.o: jerror.h
+jdmainct.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdmainct.o: jerror.h
+jdmarker.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdmarker.o: jerror.h
+jdmaster.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdmaster.o: jerror.h
+jdmerge.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdmerge.o: jerror.h
+jdphuff.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdphuff.o: jerror.h jdhuff.h
+jdpostct.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdpostct.o: jerror.h
+jdsample.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdsample.o: jerror.h
+jdtrans.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jdtrans.o: jerror.h
+jerror.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jerror.o: jerror.h jversion.h
+jfdctflt.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jfdctflt.o: jerror.h jdct.h
+jfdctfst.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jfdctfst.o: jerror.h jdct.h
+jfdctint.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jfdctint.o: jerror.h jdct.h
+jidctflt.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jidctflt.o: jerror.h jdct.h
+jidctfst.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jidctfst.o: jerror.h jdct.h
+jidctint.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jidctint.o: jerror.h jdct.h
+jidctred.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jidctred.o: jerror.h jdct.h
+jquant1.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jquant1.o: jerror.h
+jquant2.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jquant2.o: jerror.h
+jutils.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jutils.o: jerror.h
+jmemmgr.o: jinclude.h jconfig.h ../config.h jpeglib.h jmorecfg.h jpegint.h
+jmemmgr.o: jerror.h jmemsys.h
diff --git a/jpeg/structure.doc b/jpeg/structure.doc
new file mode 100644
index 000000000..51c9def7e
--- /dev/null
+++ b/jpeg/structure.doc
@@ -0,0 +1,948 @@
+IJG JPEG LIBRARY:  SYSTEM ARCHITECTURE
+
+Copyright (C) 1991-1995, Thomas G. Lane.
+This file is part of the Independent JPEG Group's software.
+For conditions of distribution and use, see the accompanying README file.
+
+
+This file provides an overview of the architecture of the IJG JPEG software;
+that is, the functions of the various modules in the system and the interfaces
+between modules.  For more precise details about any data structure or calling
+convention, see the include files and comments in the source code.
+
+We assume that the reader is already somewhat familiar with the JPEG standard.
+The README file includes references for learning about JPEG.  The file
+libjpeg.doc describes the library from the viewpoint of an application
+programmer using the library; it's best to read that file before this one.
+Also, the file coderules.doc describes the coding style conventions we use.
+
+In this document, JPEG-specific terminology follows the JPEG standard:
+  A "component" means a color channel, e.g., Red or Luminance.
+  A "sample" is a single component value (i.e., one number in the image data).
+  A "coefficient" is a frequency coefficient (a DCT transform output number).
+  A "block" is an 8x8 group of samples or coefficients.
+  An "MCU" (minimum coded unit) is an interleaved set of blocks of size
+	determined by the sampling factors, or a single block in a
+	noninterleaved scan.
+We do not use the terms "pixel" and "sample" interchangeably.  When we say
+pixel, we mean an element of the full-size image, while a sample is an element
+of the downsampled image.  Thus the number of samples may vary across
+components while the number of pixels does not.  (This terminology is not used
+rigorously throughout the code, but it is used in places where confusion would
+otherwise result.)
+
+
+*** System features ***
+
+The IJG distribution contains two parts:
+  * A subroutine library for JPEG compression and decompression.
+  * cjpeg/djpeg, two sample applications that use the library to transform
+    JFIF JPEG files to and from several other image formats.
+cjpeg/djpeg are of no great intellectual complexity: they merely add a simple
+command-line user interface and I/O routines for several uncompressed image
+formats.  This document concentrates on the library itself.
+
+We desire the library to be capable of supporting all JPEG baseline, extended
+sequential, and progressive DCT processes.  Hierarchical processes are not
+supported.
+
+The library does not support the lossless (spatial) JPEG process.  Lossless
+JPEG shares little or no code with lossy JPEG, and would normally be used
+without the extensive pre- and post-processing provided by this library.
+We feel that lossless JPEG is better handled by a separate library.
+
+Within these limits, any set of compression parameters allowed by the JPEG
+spec should be readable for decompression.  (We can be more restrictive about
+what formats we can generate.)  Although the system design allows for all
+parameter values, some uncommon settings are not yet implemented and may
+never be; nonintegral sampling ratios are the prime example.  Furthermore,
+we treat 8-bit vs. 12-bit data precision as a compile-time switch, not a
+run-time option, because most machines can store 8-bit pixels much more
+compactly than 12-bit.
+
+For legal reasons, JPEG arithmetic coding is not currently supported, but
+extending the library to include it would be straightforward.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, libtiff uses this
+library to implement JPEG compression within the TIFF file format.)
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  This code can be omitted if not needed.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+
+*** Portability issues ***
+
+Portability is an essential requirement for the library.  The key portability
+issues that show up at the level of system architecture are:
+
+1.  Memory usage.  We want the code to be able to run on PC-class machines
+with limited memory.  Images should therefore be processed sequentially (in
+strips), to avoid holding the whole image in memory at once.  Where a
+full-image buffer is necessary, we should be able to use either virtual memory
+or temporary files.
+
+2.  Near/far pointer distinction.  To run efficiently on 80x86 machines, the
+code should distinguish "small" objects (kept in near data space) from
+"large" ones (kept in far data space).  This is an annoying restriction, but
+fortunately it does not impact code quality for less brain-damaged machines,
+and the source code clutter turns out to be minimal with sufficient use of
+pointer typedefs.
+
+3. Data precision.  We assume that "char" is at least 8 bits, "short" and
+"int" at least 16, "long" at least 32.  The code will work fine with larger
+data sizes, although memory may be used inefficiently in some cases.  However,
+the JPEG compressed datastream must ultimately appear on external storage as a
+sequence of 8-bit bytes if it is to conform to the standard.  This may pose a
+problem on machines where char is wider than 8 bits.  The library represents
+compressed data as an array of values of typedef JOCTET.  If no data type
+exactly 8 bits wide is available, custom data source and data destination
+modules must be written to unpack and pack the chosen JOCTET datatype into
+8-bit external representation.
+
+
+*** System overview ***
+
+The compressor and decompressor are each divided into two main sections:
+the JPEG compressor or decompressor proper, and the preprocessing or
+postprocessing functions.  The interface between these two sections is the
+image data that the official JPEG spec regards as its input or output: this
+data is in the colorspace to be used for compression, and it is downsampled
+to the sampling factors to be used.  The preprocessing and postprocessing
+steps are responsible for converting a normal image representation to or from
+this form.  (Those few applications that want to deal with YCbCr downsampled
+data can skip the preprocessing or postprocessing step.)
+
+Looking more closely, the compressor library contains the following main
+elements:
+
+  Preprocessing:
+    * Color space conversion (e.g., RGB to YCbCr).
+    * Edge expansion and downsampling.  Optionally, this step can do simple
+      smoothing --- this is often helpful for low-quality source data.
+  JPEG proper:
+    * MCU assembly, DCT, quantization.
+    * Entropy coding (sequential or progressive, Huffman or arithmetic).
+
+In addition to these modules we need overall control, marker generation,
+and support code (memory management & error handling).  There is also a
+module responsible for physically writing the output data --- typically
+this is just an interface to fwrite(), but some applications may need to
+do something else with the data.
+
+The decompressor library contains the following main elements:
+
+  JPEG proper:
+    * Entropy decoding (sequential or progressive, Huffman or arithmetic).
+    * Dequantization, inverse DCT, MCU disassembly.
+  Postprocessing:
+    * Upsampling.  Optionally, this step may be able to do more general
+      rescaling of the image.
+    * Color space conversion (e.g., YCbCr to RGB).  This step may also
+      provide gamma adjustment [ currently it does not ].
+    * Optional color quantization (e.g., reduction to 256 colors).
+    * Optional color precision reduction (e.g., 24-bit to 15-bit color).
+      [This feature is not currently implemented.]
+
+We also need overall control, marker parsing, and a data source module.
+The support code (memory management & error handling) can be shared with
+the compression half of the library.
+
+There may be several implementations of each of these elements, particularly
+in the decompressor, where a wide range of speed/quality tradeoffs is very
+useful.  It must be understood that some of the best speedups involve
+merging adjacent steps in the pipeline.  For example, upsampling, color space
+conversion, and color quantization might all be done at once when using a
+low-quality ordered-dither technique.  The system architecture is designed to
+allow such merging where appropriate.
+
+
+Note: it is convenient to regard edge expansion (padding to block boundaries)
+as a preprocessing/postprocessing function, even though the JPEG spec includes
+it in compression/decompression.  We do this because downsampling/upsampling
+can be simplified a little if they work on padded data: it's not necessary to
+have special cases at the right and bottom edges.  Therefore the interface
+buffer is always an integral number of blocks wide and high, and we expect
+compression preprocessing to pad the source data properly.  Padding will occur
+only to the next block (8-sample) boundary.  In an interleaved-scan situation,
+additional dummy blocks may be used to fill out MCUs, but the MCU assembly and
+disassembly logic will create or discard these blocks internally.  (This is
+advantageous for speed reasons, since we avoid DCTing the dummy blocks.
+It also permits a small reduction in file size, because the compressor can
+choose dummy block contents so as to minimize their size in compressed form.
+Finally, it makes the interface buffer specification independent of whether
+the file is actually interleaved or not.)  Applications that wish to deal
+directly with the downsampled data must provide similar buffering and padding
+for odd-sized images.
+
+
+*** Poor man's object-oriented programming ***
+
+It should be clear by now that we have a lot of quasi-independent processing
+steps, many of which have several possible behaviors.  To avoid cluttering the
+code with lots of switch statements, we use a simple form of object-style
+programming to separate out the different possibilities.
+
+For example, two different color quantization algorithms could be implemented
+as two separate modules that present the same external interface; at runtime,
+the calling code will access the proper module indirectly through an "object".
+
+We can get the limited features we need while staying within portable C.
+The basic tool is a function pointer.  An "object" is just a struct
+containing one or more function pointer fields, each of which corresponds to
+a method name in real object-oriented languages.  During initialization we
+fill in the function pointers with references to whichever module we have
+determined we need to use in this run.  Then invocation of the module is done
+by indirecting through a function pointer; on most machines this is no more
+expensive than a switch statement, which would be the only other way of
+making the required run-time choice.  The really significant benefit, of
+course, is keeping the source code clean and well structured.
+
+We can also arrange to have private storage that varies between different
+implementations of the same kind of object.  We do this by making all the
+module-specific object structs be separately allocated entities, which will
+be accessed via pointers in the master compression or decompression struct.
+The "public" fields or methods for a given kind of object are specified by
+a commonly known struct.  But a module's initialization code can allocate
+a larger struct that contains the common struct as its first member, plus
+additional private fields.  With appropriate pointer casting, the module's
+internal functions can access these private fields.  (For a simple example,
+see jdatadst.c, which implements the external interface specified by struct
+jpeg_destination_mgr, but adds extra fields.)
+
+(Of course this would all be a lot easier if we were using C++, but we are
+not yet prepared to assume that everyone has a C++ compiler.)
+
+An important benefit of this scheme is that it is easy to provide multiple
+versions of any method, each tuned to a particular case.  While a lot of
+precalculation might be done to select an optimal implementation of a method,
+the cost per invocation is constant.  For example, the upsampling step might
+have a "generic" method, plus one or more "hardwired" methods for the most
+popular sampling factors; the hardwired methods would be faster because they'd
+use straight-line code instead of for-loops.  The cost to determine which
+method to use is paid only once, at startup, and the selection criteria are
+hidden from the callers of the method.
+
+This plan differs a little bit from usual object-oriented structures, in that
+only one instance of each object class will exist during execution.  The
+reason for having the class structure is that on different runs we may create
+different instances (choose to execute different modules).  You can think of
+the term "method" as denoting the common interface presented by a particular
+set of interchangeable functions, and "object" as denoting a group of related
+methods, or the total shared interface behavior of a group of modules.
+
+
+*** Overall control structure ***
+
+We previously mentioned the need for overall control logic in the compression
+and decompression libraries.  In IJG implementations prior to v5, overall
+control was mostly provided by "pipeline control" modules, which proved to be
+large, unwieldy, and hard to understand.  To improve the situation, the
+control logic has been subdivided into multiple modules.  The control modules
+consist of:
+
+1. Master control for module selection and initialization.  This has two
+responsibilities:
+
+   1A.  Startup initialization at the beginning of image processing.
+        The individual processing modules to be used in this run are selected
+        and given initialization calls.
+
+   1B.  Per-pass control.  This determines how many passes will be performed
+        and calls each active processing module to configure itself
+        appropriately at the beginning of each pass.  End-of-pass processing,
+	where necessary, is also invoked from the master control module.
+
+   Method selection is partially distributed, in that a particular processing
+   module may contain several possible implementations of a particular method,
+   which it will select among when given its initialization call.  The master
+   control code need only be concerned with decisions that affect more than
+   one module.
+ 
+2. Data buffering control.  A separate control module exists for each
+   inter-processing-step data buffer.  This module is responsible for
+   invoking the processing steps that write or read that data buffer.
+
+Each buffer controller sees the world as follows:
+
+input data => processing step A => buffer => processing step B => output data
+                      |              |               |
+              ------------------ controller ------------------
+
+The controller knows the dataflow requirements of steps A and B: how much data
+they want to accept in one chunk and how much they output in one chunk.  Its
+function is to manage its buffer and call A and B at the proper times.
+
+A data buffer control module may itself be viewed as a processing step by a
+higher-level control module; thus the control modules form a binary tree with
+elementary processing steps at the leaves of the tree.
+
+The control modules are objects.  A considerable amount of flexibility can
+be had by replacing implementations of a control module.  For example:
+* Merging of adjacent steps in the pipeline is done by replacing a control
+  module and its pair of processing-step modules with a single processing-
+  step module.  (Hence the possible merges are determined by the tree of
+  control modules.)
+* In some processing modes, a given interstep buffer need only be a "strip"
+  buffer large enough to accommodate the desired data chunk sizes.  In other
+  modes, a full-image buffer is needed and several passes are required.
+  The control module determines which kind of buffer is used and manipulates
+  virtual array buffers as needed.  One or both processing steps may be
+  unaware of the multi-pass behavior.
+
+In theory, we might be able to make all of the data buffer controllers
+interchangeable and provide just one set of implementations for all.  In
+practice, each one contains considerable special-case processing for its
+particular job.  The buffer controller concept should be regarded as an
+overall system structuring principle, not as a complete description of the
+task performed by any one controller.
+
+
+*** Compression object structure ***
+
+Here is a sketch of the logical structure of the JPEG compression library:
+
+                                                 |-- Colorspace conversion
+                  |-- Preprocessing controller --|
+                  |                              |-- Downsampling
+Main controller --|
+                  |                            |-- Forward DCT, quantize
+                  |-- Coefficient controller --|
+                                               |-- Entropy encoding
+
+This sketch also describes the flow of control (subroutine calls) during
+typical image data processing.  Each of the components shown in the diagram is
+an "object" which may have several different implementations available.  One
+or more source code files contain the actual implementation(s) of each object.
+
+The objects shown above are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the preprocessed input data.  This controller invokes preprocessing to
+  fill the subsampled-data buffer, and JPEG compression to empty it.  There is
+  usually no need for a full-image buffer here; a strip buffer is adequate.
+
+* Preprocessing controller: buffer controller for the downsampling input data
+  buffer, which lies between colorspace conversion and downsampling.  Note
+  that a unified conversion/downsampling module would probably replace this
+  controller entirely.
+
+* Colorspace conversion: converts application image data into the desired
+  JPEG color space; also changes the data from pixel-interleaved layout to
+  separate component planes.  Processes one pixel row at a time.
+
+* Downsampling: performs reduction of chroma components as required.
+  Optionally may perform pixel-level smoothing as well.  Processes a "row
+  group" at a time, where a row group is defined as Vmax pixel rows of each
+  component before downsampling, and Vk sample rows afterwards (remember Vk
+  differs across components).  Some downsampling or smoothing algorithms may
+  require context rows above and below the current row group; the
+  preprocessing controller is responsible for supplying these rows via proper
+  buffering.  The downsampler is responsible for edge expansion at the right
+  edge (i.e., extending each sample row to a multiple of 8 samples); but the
+  preprocessing controller is responsible for vertical edge expansion (i.e.,
+  duplicating the bottom sample row as needed to make a multiple of 8 rows).
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU assembly, including insertion of dummy DCT
+  blocks when needed at the right or bottom edge.  When performing
+  Huffman-code optimization or emitting a multiscan JPEG file, this
+  controller is responsible for buffering the full image.  The equivalent of
+  one fully interleaved MCU row of subsampled data is processed per call,
+  even when the JPEG file is noninterleaved.
+
+* Forward DCT and quantization: Perform DCT, quantize, and emit coefficients.
+  Works on one or more DCT blocks at a time.  (Note: the coefficients are now
+  emitted in normal array order, which the entropy encoder is expected to
+  convert to zigzag order as necessary.  Prior versions of the IJG code did
+  the conversion to zigzag order within the quantization step.)
+
+* Entropy encoding: Perform Huffman or arithmetic entropy coding and emit the
+  coded data to the data destination module.  Works on one MCU per call.
+  For progressive JPEG, the same DCT blocks are fed to the entropy coder
+  during each pass, and the coder must emit the appropriate subset of
+  coefficients.
+
+In addition to the above objects, the compression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.
+
+* Marker writing: generates JPEG markers (except for RSTn, which is emitted
+  by the entropy encoder when needed).
+
+* Data destination manager: writes the output JPEG datastream to its final
+  destination (e.g., a file).  The destination manager supplied with the
+  library knows how to write to a stdio stream; for other behaviors, the
+  surrounding application may provide its own destination manager.
+
+* Memory manager: allocates and releases memory, controls virtual arrays
+  (with backing store management, where required).
+
+* Error handler: performs formatting and output of error and trace messages;
+  determines handling of nonfatal errors.  The surrounding application may
+  override some or all of this object's methods to change error handling.
+
+* Progress monitor: supports output of "percent-done" progress reports.
+  This object represents an optional callback to the surrounding application:
+  if wanted, it must be supplied by the application.
+
+The error handler, destination manager, and progress monitor objects are
+defined as separate objects in order to simplify application-specific
+customization of the JPEG library.  A surrounding application may override
+individual methods or supply its own all-new implementation of one of these
+objects.  The object interfaces for these objects are therefore treated as
+part of the application interface of the library, whereas the other objects
+are internal to the library.
+
+The error handler and memory manager are shared by JPEG compression and
+decompression; the progress monitor, if used, may be shared as well.
+
+
+*** Decompression object structure ***
+
+Here is a sketch of the logical structure of the JPEG decompression library:
+
+                                               |-- Entropy decoding
+                  |-- Coefficient controller --|
+                  |                            |-- Dequantize, Inverse DCT
+Main controller --|
+                  |                               |-- Upsampling
+                  |-- Postprocessing controller --|   |-- Colorspace conversion
+                                                  |-- Color quantization
+                                                  |-- Color precision reduction
+
+As before, this diagram also represents typical control flow.  The objects
+shown are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the output of JPEG decompression proper.  This controller's primary
+  task is to feed the postprocessing procedure.  Some upsampling algorithms
+  may require context rows above and below the current row group; when this
+  is true, the main controller is responsible for managing its buffer so as
+  to make context rows available.  In the current design, the main buffer is
+  always a strip buffer; a full-image buffer is never required.
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU disassembly, including deletion of any dummy
+  DCT blocks at the right or bottom edge.  When reading a multiscan JPEG
+  file, this controller is responsible for buffering the full image.
+  (Buffering DCT coefficients, rather than samples, is necessary to support
+  progressive JPEG.)  The equivalent of one fully interleaved MCU row of
+  subsampled data is processed per call, even when the source JPEG file is
+  noninterleaved.
+
+* Entropy decoding: Read coded data from the data source module and perform
+  Huffman or arithmetic entropy decoding.  Works on one MCU per call.
+  For progressive JPEG decoding, the coefficient controller supplies the prior
+  coefficients of each MCU (initially all zeroes), which the entropy decoder
+  modifies in each scan.
+
+* Dequantization and inverse DCT: like it says.  Note that the coefficients
+  buffered by the coefficient controller have NOT been dequantized; we
+  merge dequantization and inverse DCT into a single step for speed reasons.
+  When scaled-down output is asked for, simplified DCT algorithms may be used
+  that emit only 1x1, 2x2, or 4x4 samples per DCT block, not the full 8x8.
+  Works on one DCT block at a time.
+
+* Postprocessing controller: buffer controller for the color quantization
+  input buffer, when quantization is in use.  (Without quantization, this
+  controller just calls the upsampler.)  For two-pass quantization, this
+  controller is responsible for buffering the full-image data.
+
+* Upsampling: restores chroma components to full size.  (May support more
+  general output rescaling, too.  Note that if undersized DCT outputs have
+  been emitted by the DCT module, this module must adjust so that properly
+  sized outputs are created.)  Works on one row group at a time.  This module
+  also calls the color conversion module, so its top level is effectively a
+  buffer controller for the upsampling->color conversion buffer.  However, in
+  all but the highest-quality operating modes, upsampling and color
+  conversion are likely to be merged into a single step.
+
+* Colorspace conversion: convert from JPEG color space to output color space,
+  and change data layout from separate component planes to pixel-interleaved.
+  Works on one pixel row at a time.
+
+* Color quantization: reduce the data to colormapped form, using either an
+  externally specified colormap or an internally generated one.  This module
+  is not used for full-color output.  Works on one pixel row at a time; may
+  require two passes to generate a color map.  Note that the output will
+  always be a single component representing colormap indexes.  In the current
+  design, the output values are JSAMPLEs, so an 8-bit compilation cannot
+  quantize to more than 256 colors.  This is unlikely to be a problem in
+  practice.
+
+* Color reduction: this module handles color precision reduction, e.g.,
+  generating 15-bit color (5 bits/primary) from JPEG's 24-bit output.
+  Not quite clear yet how this should be handled... should we merge it with
+  colorspace conversion???
+
+Note that some high-speed operating modes might condense the entire
+postprocessing sequence to a single module (upsample, color convert, and
+quantize in one step).
+
+In addition to the above objects, the decompression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.  This is subdivided into
+  input and output control: jdinput.c controls only input-side processing,
+  while jdmaster.c handles overall initialization and output-side control.
+
+* Marker reading: decodes JPEG markers (except for RSTn).
+
+* Data source manager: supplies the input JPEG datastream.  The source
+  manager supplied with the library knows how to read from a stdio stream;
+  for other behaviors, the surrounding application may provide its own source
+  manager.
+
+* Memory manager: same as for compression library.
+
+* Error handler: same as for compression library.
+
+* Progress monitor: same as for compression library.
+
+As with compression, the data source manager, error handler, and progress
+monitor are candidates for replacement by a surrounding application.
+
+
+*** Decompression input and output separation ***
+
+To support efficient incremental display of progressive JPEG files, the
+decompressor is divided into two sections that can run independently:
+
+1. Data input includes marker parsing, entropy decoding, and input into the
+   coefficient controller's DCT coefficient buffer.  Note that this
+   processing is relatively cheap and fast.
+
+2. Data output reads from the DCT coefficient buffer and performs the IDCT
+   and all postprocessing steps.
+
+For a progressive JPEG file, the data input processing is allowed to get
+arbitrarily far ahead of the data output processing.  (This occurs only
+if the application calls jpeg_consume_input(); otherwise input and output
+run in lockstep, since the input section is called only when the output
+section needs more data.)  In this way the application can avoid making
+extra display passes when data is arriving faster than the display pass
+can run.  Furthermore, it is possible to abort an output pass without
+losing anything, since the coefficient buffer is read-only as far as the
+output section is concerned.  See libjpeg.doc for more detail.
+
+A full-image coefficient array is only created if the JPEG file has multiple
+scans (or if the application specifies buffered-image mode anyway).  When
+reading a single-scan file, the coefficient controller normally creates only
+a one-MCU buffer, so input and output processing must run in lockstep in this
+case.  jpeg_consume_input() is effectively a no-op in this situation.
+
+The main impact of dividing the decompressor in this fashion is that we must
+be very careful with shared variables in the cinfo data structure.  Each
+variable that can change during the course of decompression must be
+classified as belonging to data input or data output, and each section must
+look only at its own variables.  For example, the data output section may not
+depend on any of the variables that describe the current scan in the JPEG
+file, because these may change as the data input section advances into a new
+scan.
+
+The progress monitor is (somewhat arbitrarily) defined to treat input of the
+file as one pass when buffered-image mode is not used, and to ignore data
+input work completely when buffered-image mode is used.  Note that the
+library has no reliable way to predict the number of passes when dealing
+with a progressive JPEG file, nor can it predict the number of output passes
+in buffered-image mode.  So the work estimate is inherently bogus anyway.
+
+No comparable division is currently made in the compression library, because
+there isn't any real need for it.
+
+
+*** Data formats ***
+
+Arrays of pixel sample values use the following data structure:
+
+    typedef something JSAMPLE;		a pixel component value, 0..MAXJSAMPLE
+    typedef JSAMPLE *JSAMPROW;		ptr to a row of samples
+    typedef JSAMPROW *JSAMPARRAY;	ptr to a list of rows
+    typedef JSAMPARRAY *JSAMPIMAGE;	ptr to a list of color-component arrays
+
+The basic element type JSAMPLE will typically be one of unsigned char,
+(signed) char, or short.  Short will be used if samples wider than 8 bits are
+to be supported (this is a compile-time option).  Otherwise, unsigned char is
+used if possible.  If the compiler only supports signed chars, then it is
+necessary to mask off the value when reading.  Thus, all reads of JSAMPLE
+values must be coded as "GETJSAMPLE(value)", where the macro will be defined
+as "((value) & 0xFF)" on signed-char machines and "((int) (value))" elsewhere.
+
+With these conventions, JSAMPLE values can be assumed to be >= 0.  This helps
+simplify correct rounding during downsampling, etc.  The JPEG standard's
+specification that sample values run from -128..127 is accommodated by
+subtracting 128 just as the sample value is copied into the source array for
+the DCT step (this will be an array of signed ints).  Similarly, during
+decompression the output of the IDCT step will be immediately shifted back to
+0..255.  (NB: different values are required when 12-bit samples are in use.
+The code is written in terms of MAXJSAMPLE and CENTERJSAMPLE, which will be
+defined as 255 and 128 respectively in an 8-bit implementation, and as 4095
+and 2048 in a 12-bit implementation.)
+
+We use a pointer per row, rather than a two-dimensional JSAMPLE array.  This
+choice costs only a small amount of memory and has several benefits:
+* Code using the data structure doesn't need to know the allocated width of
+  the rows.  This simplifies edge expansion/compression, since we can work
+  in an array that's wider than the logical picture width.
+* Indexing doesn't require multiplication; this is a performance win on many
+  machines.
+* Arrays with more than 64K total elements can be supported even on machines
+  where malloc() cannot allocate chunks larger than 64K.
+* The rows forming a component array may be allocated at different times
+  without extra copying.  This trick allows some speedups in smoothing steps
+  that need access to the previous and next rows.
+
+Note that each color component is stored in a separate array; we don't use the
+traditional layout in which the components of a pixel are stored together.
+This simplifies coding of modules that work on each component independently,
+because they don't need to know how many components there are.  Furthermore,
+we can read or write each component to a temporary file independently, which
+is helpful when dealing with noninterleaved JPEG files.
+
+In general, a specific sample value is accessed by code such as
+	GETJSAMPLE(image[colorcomponent][row][col])
+where col is measured from the image left edge, but row is measured from the
+first sample row currently in memory.  Either of the first two indexings can
+be precomputed by copying the relevant pointer.
+
+
+Since most image-processing applications prefer to work on images in which
+the components of a pixel are stored together, the data passed to or from the
+surrounding application uses the traditional convention: a single pixel is
+represented by N consecutive JSAMPLE values, and an image row is an array of
+(# of color components)*(image width) JSAMPLEs.  One or more rows of data can
+be represented by a pointer of type JSAMPARRAY in this scheme.  This scheme is
+converted to component-wise storage inside the JPEG library.  (Applications
+that want to skip JPEG preprocessing or postprocessing will have to contend
+with component-wise storage.)
+
+
+Arrays of DCT-coefficient values use the following data structure:
+
+    typedef short JCOEF;		a 16-bit signed integer
+    typedef JCOEF JBLOCK[DCTSIZE2];	an 8x8 block of coefficients
+    typedef JBLOCK *JBLOCKROW;		ptr to one horizontal row of 8x8 blocks
+    typedef JBLOCKROW *JBLOCKARRAY;	ptr to a list of such rows
+    typedef JBLOCKARRAY *JBLOCKIMAGE;	ptr to a list of color component arrays
+
+The underlying type is at least a 16-bit signed integer; while "short" is big
+enough on all machines of interest, on some machines it is preferable to use
+"int" for speed reasons, despite the storage cost.  Coefficients are grouped
+into 8x8 blocks (but we always use #defines DCTSIZE and DCTSIZE2 rather than
+"8" and "64").
+
+The contents of a coefficient block may be in either "natural" or zigzagged
+order, and may be true values or divided by the quantization coefficients,
+depending on where the block is in the processing pipeline.  In the current
+library, coefficient blocks are kept in natural order everywhere; the entropy
+codecs zigzag or dezigzag the data as it is written or read.  The blocks
+contain quantized coefficients everywhere outside the DCT/IDCT subsystems.
+(This latter decision may need to be revisited to support variable
+quantization a la JPEG Part 3.)
+
+Notice that the allocation unit is now a row of 8x8 blocks, corresponding to
+eight rows of samples.  Otherwise the structure is much the same as for
+samples, and for the same reasons.
+
+On machines where malloc() can't handle a request bigger than 64Kb, this data
+structure limits us to rows of less than 512 JBLOCKs, or a picture width of
+4000+ pixels.  This seems an acceptable restriction.
+
+
+On 80x86 machines, the bottom-level pointer types (JSAMPROW and JBLOCKROW)
+must be declared as "far" pointers, but the upper levels can be "near"
+(implying that the pointer lists are allocated in the DS segment).
+We use a #define symbol FAR, which expands to the "far" keyword when
+compiling on 80x86 machines and to nothing elsewhere.
+
+
+*** Suspendable processing ***
+
+In some applications it is desirable to use the JPEG library as an
+incremental, memory-to-memory filter.  In this situation the data source or
+destination may be a limited-size buffer, and we can't rely on being able to
+empty or refill the buffer at arbitrary times.  Instead the application would
+like to have control return from the library at buffer overflow/underrun, and
+then resume compression or decompression at a later time.
+
+This scenario is supported for simple cases.  (For anything more complex, we
+recommend that the application "bite the bullet" and develop real multitasking
+capability.)  The libjpeg.doc file goes into more detail about the usage and
+limitations of this capability; here we address the implications for library
+structure.
+
+The essence of the problem is that the entropy codec (coder or decoder) must
+be prepared to stop at arbitrary times.  In turn, the controllers that call
+the entropy codec must be able to stop before having produced or consumed all
+the data that they normally would handle in one call.  That part is reasonably
+straightforward: we make the controller call interfaces include "progress
+counters" which indicate the number of data chunks successfully processed, and
+we require callers to test the counter rather than just assume all of the data
+was processed.
+
+Rather than trying to restart at an arbitrary point, the current Huffman
+codecs are designed to restart at the beginning of the current MCU after a
+suspension due to buffer overflow/underrun.  At the start of each call, the
+codec's internal state is loaded from permanent storage (in the JPEG object
+structures) into local variables.  On successful completion of the MCU, the
+permanent state is updated.  (This copying is not very expensive, and may even
+lead to *improved* performance if the local variables can be registerized.)
+If a suspension occurs, the codec simply returns without updating the state,
+thus effectively reverting to the start of the MCU.  Note that this implies
+leaving some data unprocessed in the source/destination buffer (ie, the
+compressed partial MCU).  The data source/destination module interfaces are
+specified so as to make this possible.  This also implies that the data buffer
+must be large enough to hold a worst-case compressed MCU; a couple thousand
+bytes should be enough.
+
+In a successive-approximation AC refinement scan, the progressive Huffman
+decoder has to be able to undo assignments of newly nonzero coefficients if it
+suspends before the MCU is complete, since decoding requires distinguishing
+previously-zero and previously-nonzero coefficients.  This is a bit tedious
+but probably won't have much effect on performance.  Other variants of Huffman
+decoding need not worry about this, since they will just store the same values
+again if forced to repeat the MCU.
+
+This approach would probably not work for an arithmetic codec, since its
+modifiable state is quite large and couldn't be copied cheaply.  Instead it
+would have to suspend and resume exactly at the point of the buffer end.
+
+The JPEG marker reader is designed to cope with suspension at an arbitrary
+point.  It does so by backing up to the start of the marker parameter segment,
+so the data buffer must be big enough to hold the largest marker of interest.
+Again, a couple KB should be adequate.  (A special "skip" convention is used
+to bypass COM and APPn markers, so these can be larger than the buffer size
+without causing problems; otherwise a 64K buffer would be needed in the worst
+case.)
+
+The JPEG marker writer currently does *not* cope with suspension.  I feel that
+this is not necessary; it is much easier simply to require the application to
+ensure there is enough buffer space before starting.  (An empty 2K buffer is
+more than sufficient for the header markers; and ensuring there are a dozen or
+two bytes available before calling jpeg_finish_compress() will suffice for the
+trailer.)  This would not work for writing multi-scan JPEG files, but
+we simply do not intend to support that capability with suspension.
+
+
+*** Memory manager services ***
+
+The JPEG library's memory manager controls allocation and deallocation of
+memory, and it manages large "virtual" data arrays on machines where the
+operating system does not provide virtual memory.  Note that the same
+memory manager serves both compression and decompression operations.
+
+In all cases, allocated objects are tied to a particular compression or
+decompression master record, and they will be released when that master
+record is destroyed.
+
+The memory manager does not provide explicit deallocation of objects.
+Instead, objects are created in "pools" of free storage, and a whole pool
+can be freed at once.  This approach helps prevent storage-leak bugs, and
+it speeds up operations whenever malloc/free are slow (as they often are).
+The pools can be regarded as lifetime identifiers for objects.  Two
+pools/lifetimes are defined:
+  * JPOOL_PERMANENT	lasts until master record is destroyed
+  * JPOOL_IMAGE		lasts until done with image (JPEG datastream)
+Permanent lifetime is used for parameters and tables that should be carried
+across from one datastream to another; this includes all application-visible
+parameters.  Image lifetime is used for everything else.  (A third lifetime,
+JPOOL_PASS = one processing pass, was originally planned.  However it was
+dropped as not being worthwhile.  The actual usage patterns are such that the
+peak memory usage would be about the same anyway; and having per-pass storage
+substantially complicates the virtual memory allocation rules --- see below.)
+
+The memory manager deals with three kinds of object:
+1. "Small" objects.  Typically these require no more than 10K-20K total.
+2. "Large" objects.  These may require tens to hundreds of K depending on
+   image size.  Semantically they behave the same as small objects, but we
+   distinguish them for two reasons:
+     * On MS-DOS machines, large objects are referenced by FAR pointers,
+       small objects by NEAR pointers.
+     * Pool allocation heuristics may differ for large and small objects.
+   Note that individual "large" objects cannot exceed the size allowed by
+   type size_t, which may be 64K or less on some machines.
+3. "Virtual" objects.  These are large 2-D arrays of JSAMPLEs or JBLOCKs
+   (typically large enough for the entire image being processed).  The
+   memory manager provides stripwise access to these arrays.  On machines
+   without virtual memory, the rest of the array may be swapped out to a
+   temporary file.
+
+(Note: JSAMPARRAY and JBLOCKARRAY data structures are a combination of large
+objects for the data proper and small objects for the row pointers.  For
+convenience and speed, the memory manager provides single routines to create
+these structures.  Similarly, virtual arrays include a small control block
+and a JSAMPARRAY or JBLOCKARRAY working buffer, all created with one call.)
+
+In the present implementation, virtual arrays are only permitted to have image
+lifespan.  (Permanent lifespan would not be reasonable, and pass lifespan is
+not very useful since a virtual array's raison d'etre is to store data for
+multiple passes through the image.)  We also expect that only "small" objects
+will be given permanent lifespan, though this restriction is not required by
+the memory manager.
+
+In a non-virtual-memory machine, some performance benefit can be gained by
+making the in-memory buffers for virtual arrays be as large as possible.
+(For small images, the buffers might fit entirely in memory, so blind
+swapping would be very wasteful.)  The memory manager will adjust the height
+of the buffers to fit within a prespecified maximum memory usage.  In order
+to do this in a reasonably optimal fashion, the manager needs to allocate all
+of the virtual arrays at once.  Therefore, there isn't a one-step allocation
+routine for virtual arrays; instead, there is a "request" routine that simply
+allocates the control block, and a "realize" routine (called just once) that
+determines space allocation and creates all of the actual buffers.  The
+realize routine must allow for space occupied by non-virtual large objects.
+(We don't bother to factor in the space needed for small objects, on the
+grounds that it isn't worth the trouble.)
+
+To support all this, we establish the following protocol for doing business
+with the memory manager:
+  1. Modules must request virtual arrays (which may have only image lifespan)
+     during the initial setup phase, i.e., in their jinit_xxx routines.
+  2. All "large" objects (including JSAMPARRAYs and JBLOCKARRAYs) must also be
+     allocated during initial setup.
+  3. realize_virt_arrays will be called at the completion of initial setup.
+     The above conventions ensure that sufficient information is available
+     for it to choose a good size for virtual array buffers.
+Small objects of any lifespan may be allocated at any time.  We expect that
+the total space used for small objects will be small enough to be negligible
+in the realize_virt_arrays computation.
+
+In a virtual-memory machine, we simply pretend that the available space is
+infinite, thus causing realize_virt_arrays to decide that it can allocate all
+the virtual arrays as full-size in-memory buffers.  The overhead of the
+virtual-array access protocol is very small when no swapping occurs.
+
+A virtual array can be specified to be "pre-zeroed"; when this flag is set,
+never-yet-written sections of the array are set to zero before being made
+available to the caller.  If this flag is not set, never-written sections
+of the array contain garbage.  (This feature exists primarily because the
+equivalent logic would otherwise be needed in jdcoefct.c for progressive
+JPEG mode; we may as well make it available for possible other uses.)
+
+The first write pass on a virtual array is required to occur in top-to-bottom
+order; read passes, as well as any write passes after the first one, may
+access the array in any order.  This restriction exists partly to simplify
+the virtual array control logic, and partly because some file systems may not
+support seeking beyond the current end-of-file in a temporary file.  The main
+implication of this restriction is that rearrangement of rows (such as
+converting top-to-bottom data order to bottom-to-top) must be handled while
+reading data out of the virtual array, not while putting it in.
+
+
+*** Memory manager internal structure ***
+
+To isolate system dependencies as much as possible, we have broken the
+memory manager into two parts.  There is a reasonably system-independent
+"front end" (jmemmgr.c) and a "back end" that contains only the code
+likely to change across systems.  All of the memory management methods
+outlined above are implemented by the front end.  The back end provides
+the following routines for use by the front end (none of these routines
+are known to the rest of the JPEG code):
+
+jpeg_mem_init, jpeg_mem_term	system-dependent initialization/shutdown
+
+jpeg_get_small, jpeg_free_small	interface to malloc and free library routines
+				(or their equivalents)
+
+jpeg_get_large, jpeg_free_large	interface to FAR malloc/free in MSDOS machines;
+				else usually the same as
+				jpeg_get_small/jpeg_free_small
+
+jpeg_mem_available		estimate available memory
+
+jpeg_open_backing_store		create a backing-store object
+
+read_backing_store,		manipulate a backing-store object
+write_backing_store,
+close_backing_store
+
+On some systems there will be more than one type of backing-store object
+(specifically, in MS-DOS a backing store file might be an area of extended
+memory as well as a disk file).  jpeg_open_backing_store is responsible for
+choosing how to implement a given object.  The read/write/close routines
+are method pointers in the structure that describes a given object; this
+lets them be different for different object types.
+
+It may be necessary to ensure that backing store objects are explicitly
+released upon abnormal program termination.  For example, MS-DOS won't free
+extended memory by itself.  To support this, we will expect the main program
+or surrounding application to arrange to call self_destruct (typically via
+jpeg_destroy) upon abnormal termination.  This may require a SIGINT signal
+handler or equivalent.  We don't want to have the back end module install its
+own signal handler, because that would pre-empt the surrounding application's
+ability to control signal handling.
+
+The IJG distribution includes several memory manager back end implementations.
+Usually the same back end should be suitable for all applications on a given
+system, but it is possible for an application to supply its own back end at
+need.
+
+
+*** Implications of DNL marker ***
+
+Some JPEG files may use a DNL marker to postpone definition of the image
+height (this would be useful for a fax-like scanner's output, for instance).
+In these files the SOF marker claims the image height is 0, and you only
+find out the true image height at the end of the first scan.
+
+We could read these files as follows:
+1. Upon seeing zero image height, replace it by 65535 (the maximum allowed).
+2. When the DNL is found, update the image height in the global image
+   descriptor.
+This implies that control modules must avoid making copies of the image
+height, and must re-test for termination after each MCU row.  This would
+be easy enough to do.
+
+In cases where image-size data structures are allocated, this approach will
+result in very inefficient use of virtual memory or much-larger-than-necessary
+temporary files.  This seems acceptable for something that probably won't be a
+mainstream usage.  People might have to forgo use of memory-hogging options
+(such as two-pass color quantization or noninterleaved JPEG files) if they
+want efficient conversion of such files.  (One could improve efficiency by
+demanding a user-supplied upper bound for the height, less than 65536; in most
+cases it could be much less.)
+
+The standard also permits the SOF marker to overestimate the image height,
+with a DNL to give the true, smaller height at the end of the first scan.
+This would solve the space problems if the overestimate wasn't too great.
+However, it implies that you don't even know whether DNL will be used.
+
+This leads to a couple of very serious objections:
+1. Testing for a DNL marker must occur in the inner loop of the decompressor's
+   Huffman decoder; this implies a speed penalty whether the feature is used
+   or not.
+2. There is no way to hide the last-minute change in image height from an
+   application using the decoder.  Thus *every* application using the IJG
+   library would suffer a complexity penalty whether it cared about DNL or
+   not.
+We currently do not support DNL because of these problems.
+
+A different approach is to insist that DNL-using files be preprocessed by a
+separate program that reads ahead to the DNL, then goes back and fixes the SOF
+marker.  This is a much simpler solution and is probably far more efficient.
+Even if one wants piped input, buffering the first scan of the JPEG file needs
+a lot smaller temp file than is implied by the maximum-height method.  For
+this approach we'd simply treat DNL as a no-op in the decompressor (at most,
+check that it matches the SOF image height).
+
+We will not worry about making the compressor capable of outputting DNL.
+Something similar to the first scheme above could be applied if anyone ever
+wants to make that work.
diff --git a/jpeg/usage.doc b/jpeg/usage.doc
new file mode 100644
index 000000000..8c4970af0
--- /dev/null
+++ b/jpeg/usage.doc
@@ -0,0 +1,562 @@
+USAGE instructions for the Independent JPEG Group's JPEG software
+=================================================================
+
+This file describes usage of the JPEG conversion programs cjpeg and djpeg,
+as well as the utility programs jpegtran, rdjpgcom and wrjpgcom.  (See
+the other documentation files if you wish to use the JPEG library within
+your own programs.)
+
+If you are on a Unix machine you may prefer to read the Unix-style manual
+pages in files cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1.
+
+
+INTRODUCTION
+
+These programs implement JPEG image compression and decompression.  JPEG
+(pronounced "jay-peg") is a standardized compression method for full-color
+and gray-scale images.  JPEG is designed to handle "real-world" scenes,
+for example scanned photographs.  Cartoons, line drawings, and other
+non-realistic images are not JPEG's strong suit; on that sort of material
+you may get poor image quality and/or little compression.
+
+JPEG is lossy, meaning that the output image is not necessarily identical to
+the input image.  Hence you should not use JPEG if you have to have identical
+output bits.  However, on typical real-world images, very good compression
+levels can be obtained with no visible change, and amazingly high compression
+is possible if you can tolerate a low-quality image.  You can trade off image
+quality against file size by adjusting the compressor's "quality" setting.
+
+
+GENERAL USAGE
+
+We provide two programs, cjpeg to compress an image file into JPEG format,
+and djpeg to decompress a JPEG file back into a conventional image format.
+
+On Unix-like systems, you say:
+	cjpeg [switches] [imagefile] >jpegfile
+or
+	djpeg [switches] [jpegfile]  >imagefile
+The programs read the specified input file, or standard input if none is
+named.  They always write to standard output (with trace/error messages to
+standard error).  These conventions are handy for piping images between
+programs.
+
+On most non-Unix systems, you say:
+	cjpeg [switches] imagefile jpegfile
+or
+	djpeg [switches] jpegfile  imagefile
+i.e., both the input and output files are named on the command line.  This
+style is a little more foolproof, and it loses no functionality if you don't
+have pipes.  (You can get this style on Unix too, if you prefer, by defining
+TWO_FILE_COMMANDLINE when you compile the programs; see install.doc.)
+
+You can also say:
+	cjpeg [switches] -outfile jpegfile  imagefile
+or
+	djpeg [switches] -outfile imagefile  jpegfile
+This syntax works on all systems, so it is useful for scripts.
+
+The currently supported image file formats are: PPM (PBMPLUS color format),
+PGM (PBMPLUS gray-scale format), BMP, Targa, and RLE (Utah Raster Toolkit
+format).  (RLE is supported only if the URT library is available.)
+cjpeg recognizes the input image format automatically, with the exception
+of some Targa-format files.  You have to tell djpeg which format to generate.
+
+JPEG files are in the defacto standard JFIF file format.  There are other,
+less widely used JPEG-based file formats, but we don't support them.
+
+All switch names may be abbreviated; for example, -grayscale may be written
+-gray or -gr.  Most of the "basic" switches can be abbreviated to as little as
+one letter.  Upper and lower case are equivalent (-BMP is the same as -bmp).
+British spellings are also accepted (e.g., -greyscale), though for brevity
+these are not mentioned below.
+
+
+CJPEG DETAILS
+
+The basic command line switches for cjpeg are:
+
+	-quality N	Scale quantization tables to adjust image quality.
+			Quality is 0 (worst) to 100 (best); default is 75.
+			(See below for more info.)
+
+	-grayscale	Create monochrome JPEG file from color input.
+			Be sure to use this switch when compressing a grayscale
+			BMP file, because cjpeg isn't bright enough to notice
+			whether a BMP file uses only shades of gray.  By
+			saying -grayscale, you'll get a smaller JPEG file that
+			takes less time to process.
+
+	-optimize	Perform optimization of entropy encoding parameters.
+			Without this, default encoding parameters are used.
+			-optimize usually makes the JPEG file a little smaller,
+			but cjpeg runs somewhat slower and needs much more
+			memory.  Image quality and speed of decompression are
+			unaffected by -optimize.
+
+	-progressive	Create progressive JPEG file (see below).
+
+	-targa		Input file is Targa format.  Targa files that contain
+			an "identification" field will not be automatically
+			recognized by cjpeg; for such files you must specify
+			-targa to make cjpeg treat the input as Targa format.
+			For most Targa files, you won't need this switch.
+
+The -quality switch lets you trade off compressed file size against quality of
+the reconstructed image: the higher the quality setting, the larger the JPEG
+file, and the closer the output image will be to the original input.  Normally
+you want to use the lowest quality setting (smallest file) that decompresses
+into something visually indistinguishable from the original image.  For this
+purpose the quality setting should be between 50 and 95; the default of 75 is
+often about right.  If you see defects at -quality 75, then go up 5 or 10
+counts at a time until you are happy with the output image.  (The optimal
+setting will vary from one image to another.)
+
+-quality 100 will generate a quantization table of all 1's, minimizing loss
+in the quantization step (but there is still information loss in subsampling,
+as well as roundoff error).  This setting is mainly of interest for
+experimental purposes.  Quality values above about 95 are NOT recommended for
+normal use; the compressed file size goes up dramatically for hardly any gain
+in output image quality.
+
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try -quality 2 (or so) for some
+amusing Cubist effects.  (Note: quality values below about 25 generate 2-byte
+quantization tables, which are considered optional in the JPEG standard.
+cjpeg emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use -baseline
+if you need to ensure compatibility at low quality values.)
+
+The -progressive switch creates a "progressive JPEG" file.  In this type of
+JPEG file, the data is stored in multiple scans of increasing quality.  If the
+file is being transmitted over a slow communications link, the decoder can use
+the first scan to display a low-quality image very quickly, and can then
+improve the display with each subsequent scan.  The final image is exactly
+equivalent to a standard JPEG file of the same quality setting, and the total
+file size is about the same --- often a little smaller.  CAUTION: progressive
+JPEG is not yet widely implemented, so many decoders will be unable to view a
+progressive JPEG file at all.
+
+Switches for advanced users:
+
+	-dct int	Use integer DCT method (default).
+	-dct fast	Use fast integer DCT (less accurate).
+	-dct float	Use floating-point DCT method.
+			The float method is very slightly more accurate than
+			the int method, but is much slower unless your machine
+			has very fast floating-point hardware.  Also note that
+			results of the floating-point method may vary slightly
+			across machines, while the integer methods should give
+			the same results everywhere.  The fast integer method
+			is much less accurate than the other two.
+
+	-restart N	Emit a JPEG restart marker every N MCU rows, or every
+			N MCU blocks if "B" is attached to the number.
+			-restart 0 (the default) means no restart markers.
+
+	-smooth N	Smooth the input image to eliminate dithering noise.
+			N, ranging from 1 to 100, indicates the strength of
+			smoothing.  0 (the default) means no smoothing.
+
+	-maxmemory N	Set limit for amount of memory to use in processing
+			large images.  Value is in thousands of bytes, or
+			millions of bytes if "M" is attached to the number.
+			For example, -max 4m selects 4000000 bytes.  If more
+			space is needed, temporary files will be used.
+
+	-verbose	Enable debug printout.  More -v's give more printout.
+	or  -debug	Also, version information is printed at startup.
+
+The -restart option inserts extra markers that allow a JPEG decoder to
+resynchronize after a transmission error.  Without restart markers, any damage
+to a compressed file will usually ruin the image from the point of the error
+to the end of the image; with restart markers, the damage is usually confined
+to the portion of the image up to the next restart marker.  Of course, the
+restart markers occupy extra space.  We recommend -restart 1 for images that
+will be transmitted across unreliable networks such as Usenet.
+
+The -smooth option filters the input to eliminate fine-scale noise.  This is
+often useful when converting dithered images to JPEG: a moderate smoothing
+factor of 10 to 50 gets rid of dithering patterns in the input file, resulting
+in a smaller JPEG file and a better-looking image.  Too large a smoothing
+factor will visibly blur the image, however.
+
+Switches for wizards:
+
+	-baseline	Force baseline-compatible quantization tables to be
+			generated.  This clamps quantization values to 8 bits
+			even at low quality settings.  (This switch is poorly
+			named, since it does not ensure that the output is
+			actually baseline JPEG.  For example, you can use
+			-baseline and -progressive together.)
+
+	-qtables file	Use the quantization tables given in the specified
+			text file.
+
+	-qslots N[,...] Select which quantization table to use for each color
+			component.
+
+	-sample HxV[,...]  Set JPEG sampling factors for each color component.
+
+	-scans file	Use the scan script given in the specified text file.
+
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, DON'T USE THEM.  These switches are documented
+further in the file wizard.doc.
+
+
+DJPEG DETAILS
+
+The basic command line switches for djpeg are:
+
+	-colors N	Reduce image to at most N colors.  This reduces the
+	or -quantize N	number of colors used in the output image, so that it
+			can be displayed on a colormapped display or stored in
+			a colormapped file format.  For example, if you have
+			an 8-bit display, you'd need to reduce to 256 or fewer
+			colors.  (-colors is the recommended name, -quantize
+			is provided only for backwards compatibility.)
+
+	-fast		Select recommended processing options for fast, low
+			quality output.  (The default options are chosen for
+			highest quality output.)  Currently, this is equivalent
+			to "-dct fast -nosmooth -onepass -dither ordered".
+
+	-grayscale	Force gray-scale output even if JPEG file is color.
+			Useful for viewing on monochrome displays; also,
+			djpeg runs noticeably faster in this mode.
+
+	-scale M/N	Scale the output image by a factor M/N.  Currently
+			the scale factor must be 1/1, 1/2, 1/4, or 1/8.
+			Scaling is handy if the image is larger than your
+			screen; also, djpeg runs much faster when scaling
+			down the output.
+
+	-bmp		Select BMP output format (Windows flavor).  8-bit
+			colormapped format is emitted if -colors or -grayscale
+			is specified, or if the JPEG file is gray-scale;
+			otherwise, 24-bit full-color format is emitted.
+
+	-gif		Select GIF output format.  Since GIF does not support
+			more than 256 colors, -colors 256 is assumed (unless
+			you specify a smaller number of colors).  If you
+			specify -fast, the default number of colors is 216.
+
+	-os2		Select BMP output format (OS/2 1.x flavor).  8-bit
+			colormapped format is emitted if -colors or -grayscale
+			is specified, or if the JPEG file is gray-scale;
+			otherwise, 24-bit full-color format is emitted.
+
+	-pnm		Select PBMPLUS (PPM/PGM) output format (this is the
+			default format).  PGM is emitted if the JPEG file is
+			gray-scale or if -grayscale is specified; otherwise
+			PPM is emitted.
+
+	-rle		Select RLE output format.  (Requires URT library.)
+
+	-targa		Select Targa output format.  Gray-scale format is
+			emitted if the JPEG file is gray-scale or if
+			-grayscale is specified; otherwise, colormapped format
+			is emitted if -colors is specified; otherwise, 24-bit
+			full-color format is emitted.
+
+Switches for advanced users:
+
+	-dct int	Use integer DCT method (default).
+	-dct fast	Use fast integer DCT (less accurate).
+	-dct float	Use floating-point DCT method.
+			The float method is very slightly more accurate than
+			the int method, but is much slower unless your machine
+			has very fast floating-point hardware.  Also note that
+			results of the floating-point method may vary slightly
+			across machines, while the integer methods should give
+			the same results everywhere.  The fast integer method
+			is much less accurate than the other two.
+
+	-dither fs	Use Floyd-Steinberg dithering in color quantization.
+	-dither ordered	Use ordered dithering in color quantization.
+	-dither none	Do not use dithering in color quantization.
+			By default, Floyd-Steinberg dithering is applied when
+			quantizing colors; this is slow but usually produces
+			the best results.  Ordered dither is a compromise
+			between speed and quality; no dithering is fast but
+			usually looks awful.  Note that these switches have
+			no effect unless color quantization is being done.
+			Ordered dither is only available in -onepass mode.
+
+	-map FILE	Quantize to the colors used in the specified image
+			file.  This is useful for producing multiple files
+			with identical color maps, or for forcing a predefined
+			set of colors to be used.  The FILE must be a GIF
+			or PPM file.  This option overrides -colors and
+			-onepass.
+
+	-nosmooth	Use a faster, lower-quality upsampling routine.
+
+	-onepass	Use one-pass instead of two-pass color quantization.
+			The one-pass method is faster and needs less memory,
+			but it produces a lower-quality image.  -onepass is
+			ignored unless you also say -colors N.  Also,
+			the one-pass method is always used for gray-scale
+			output (the two-pass method is no improvement then).
+
+	-maxmemory N	Set limit for amount of memory to use in processing
+			large images.  Value is in thousands of bytes, or
+			millions of bytes if "M" is attached to the number.
+			For example, -max 4m selects 4000000 bytes.  If more
+			space is needed, temporary files will be used.
+
+	-verbose	Enable debug printout.  More -v's give more printout.
+	or  -debug	Also, version information is printed at startup.
+
+
+HINTS FOR CJPEG
+
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with cjpeg's -quality and -smooth options
+to get a satisfactory conversion.  -smooth 10 or so is often helpful.
+
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+
+The -optimize option to cjpeg is worth using when you are making a "final"
+version for posting or archiving.  It's also a win when you are using low
+quality settings to make very small JPEG files; the percentage improvement
+is often a lot more than it is on larger files.  (At present, -optimize
+mode is always selected when generating progressive JPEG files.)
+
+GIF input files are no longer supported, to avoid the Unisys LZW patent.
+Use a Unisys-licensed program if you need to read a GIF file.  (Conversion
+of GIF files to JPEG is usually a bad idea anyway.)
+
+
+HINTS FOR DJPEG
+
+To get a quick preview of an image, use the -grayscale and/or -scale switches.
+"-grayscale -scale 1/8" is the fastest case.
+
+Several options are available that trade off image quality to gain speed.
+"-fast" turns on the recommended settings.
+
+"-dct fast" and/or "-nosmooth" gain speed at a small sacrifice in quality.
+When producing a color-quantized image, "-onepass -dither ordered" is fast but
+much lower quality than the default behavior.  "-dither none" may give
+acceptable results in two-pass mode, but is seldom tolerable in one-pass mode.
+
+If you are fortunate enough to have very fast floating point hardware,
+"-dct float" may be even faster than "-dct fast".  But on most machines
+"-dct float" is slower than "-dct int"; in this case it is not worth using,
+because its theoretical accuracy advantage is too small to be significant
+in practice.
+
+Two-pass color quantization requires a good deal of memory; on MS-DOS machines
+it may run out of memory even with -maxmemory 0.  In that case you can still
+decompress, with some loss of image quality, by specifying -onepass for
+one-pass quantization.
+
+To avoid the Unisys LZW patent, djpeg produces uncompressed GIF files.  These
+are larger than they should be, but are readable by standard GIF decoders.
+
+
+HINTS FOR BOTH PROGRAMS
+
+If more space is needed than will fit in the available main memory (as
+determined by -maxmemory), temporary files will be used.  (MS-DOS versions
+will try to get extended or expanded memory first.)  The temporary files are
+often rather large: in typical cases they occupy three bytes per pixel, for
+example 3*800*600 = 1.44Mb for an 800x600 image.  If you don't have enough
+free disk space, leave out -progressive and -optimize (for cjpeg) or specify
+-onepass (for djpeg).
+
+On MS-DOS, the temporary files are created in the directory named by the TMP
+or TEMP environment variable, or in the current directory if neither of those
+exist.  Amiga implementations put the temp files in the directory named by
+JPEGTMP:, so be sure to assign JPEGTMP: to a disk partition with adequate free
+space.
+
+The default memory usage limit (-maxmemory) is set when the software is
+compiled.  If you get an "insufficient memory" error, try specifying a smaller
+-maxmemory value, even -maxmemory 0 to use the absolute minimum space.  You
+may want to recompile with a smaller default value if this happens often.
+
+On machines that have "environment" variables, you can define the environment
+variable JPEGMEM to set the default memory limit.  The value is specified as
+described for the -maxmemory switch.  JPEGMEM overrides the default value
+specified when the program was compiled, and itself is overridden by an
+explicit -maxmemory switch.
+
+On MS-DOS machines, -maxmemory is the amount of main (conventional) memory to
+use.  (Extended or expanded memory is also used if available.)  Most
+DOS-specific versions of this software do their own memory space estimation
+and do not need you to specify -maxmemory.
+
+
+JPEGTRAN
+
+jpegtran performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.
+
+jpegtran works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+djpeg followed by cjpeg to accomplish the same conversion.  But by the same
+token, jpegtran cannot perform lossy operations such as changing the image
+quality.
+
+jpegtran uses a command line syntax similar to cjpeg or djpeg.
+On Unix-like systems, you say:
+	jpegtran [switches] [inputfile] >outputfile
+On most non-Unix systems, you say:
+	jpegtran [switches] inputfile outputfile
+where both the input and output files are JPEG files.
+
+To specify the coded JPEG representation used in the output file,
+jpegtran accepts a subset of the switches recognized by cjpeg:
+	-optimize	Perform optimization of entropy encoding parameters.
+	-progressive	Create progressive JPEG file.
+	-restart N	Emit a JPEG restart marker every N MCU rows, or every
+			N MCU blocks if "B" is attached to the number.
+	-scans file	Use the scan script given in the specified text file.
+See the previous discussion of cjpeg for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+
+The image can be losslessly transformed by giving one of these switches:
+	-flip horizontal	Mirror image horizontally (left-right).
+	-flip vertical		Mirror image vertically (top-bottom).
+	-rotate 90		Rotate image 90 degrees clockwise.
+	-rotate 180		Rotate image 180 degrees.
+	-rotate 270		Rotate image 270 degrees clockwise (or 90 ccw).
+	-transpose		Transpose image (across UL-to-LR axis).
+	-transverse		Transverse transpose (across UR-to-LL axis).
+
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+
+jpegtran's default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the -trim switch:
+	-trim		Drop non-transformable edge blocks.
+Obviously, a transformation with -trim is not reversible, so strictly speaking
+jpegtran with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+"-rot 270 -trim" trims only the bottom edge, but "-rot 90 -trim" followed by
+"-rot 180 -trim" trims both edges.
+
+Another not-strictly-lossless transformation switch is:
+	-grayscale	Force grayscale output.
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+
+jpegtran also recognizes these switches that control what to do with "extra"
+markers, such as comment blocks:
+	-copy none	Copy no extra markers from source file.  This setting
+			suppresses all comments and other excess baggage
+			present in the source file.
+	-copy comments	Copy only comment markers.  This setting copies
+			comments from the source file, but discards
+			any other inessential data. 
+	-copy all	Copy all extra markers.  This setting preserves
+			miscellaneous markers found in the source file, such
+			as JFIF thumbnails and Photoshop settings.  In some
+			files these extra markers can be sizable.
+The default behavior is -copy comments.  (Note: in IJG releases v6 and v6a,
+jpegtran always did the equivalent of -copy none.)
+
+Additional switches recognized by jpegtran are:
+	-outfile filename
+	-maxmemory N
+	-verbose
+	-debug
+These work the same as in cjpeg or djpeg.
+
+
+THE COMMENT UTILITIES
+
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+
+We provide two utility programs to display COM block contents and add COM
+blocks to a JPEG file.
+
+rdjpgcom searches a JPEG file and prints the contents of any COM blocks on
+standard output.  The command line syntax is
+	rdjpgcom [-verbose] [inputfilename]
+The switch "-verbose" (or just "-v") causes rdjpgcom to also display the JPEG
+image dimensions.  If you omit the input file name from the command line,
+the JPEG file is read from standard input.  (This may not work on some
+operating systems, if binary data can't be read from stdin.)
+
+wrjpgcom adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks, but you
+can delete the old COM blocks if you wish.  wrjpgcom produces a new JPEG
+file; it does not modify the input file.  DO NOT try to overwrite the input
+file by directing wrjpgcom's output back into it; on most systems this will
+just destroy your file.
+
+The command line syntax for wrjpgcom is similar to cjpeg's.  On Unix-like
+systems, it is
+	wrjpgcom [switches] [inputfilename]
+The output file is written to standard output.  The input file comes from
+the named file, or from standard input if no input file is named.
+
+On most non-Unix systems, the syntax is
+	wrjpgcom [switches] inputfilename outputfilename
+where both input and output file names must be given explicitly.
+
+wrjpgcom understands three switches:
+	-replace		 Delete any existing COM blocks from the file.
+	-comment "Comment text"	 Supply new COM text on command line.
+	-cfile name		 Read text for new COM block from named file.
+(Switch names can be abbreviated.)  If you have only one line of comment text
+to add, you can provide it on the command line with -comment.  The comment
+text must be surrounded with quotes so that it is treated as a single
+argument.  Longer comments can be read from a text file.
+
+If you give neither -comment nor -cfile, then wrjpgcom will read the comment
+text from standard input.  (In this case an input image file name MUST be
+supplied, so that the source JPEG file comes from somewhere else.)  You can
+enter multiple lines, up to 64KB worth.  Type an end-of-file indicator
+(usually control-D or control-Z) to terminate the comment text entry.
+
+wrjpgcom will not add a COM block if the provided comment string is empty.
+Therefore -replace -comment "" can be used to delete all COM blocks from a
+file.
+
+These utility programs do not depend on the IJG JPEG library.  In
+particular, the source code for rdjpgcom is intended as an illustration of
+the minimum amount of code required to parse a JPEG file header correctly.
diff --git a/jpeg/wizard.doc b/jpeg/wizard.doc
new file mode 100644
index 000000000..54170b227
--- /dev/null
+++ b/jpeg/wizard.doc
@@ -0,0 +1,211 @@
+Advanced usage instructions for the Independent JPEG Group's JPEG software
+==========================================================================
+
+This file describes cjpeg's "switches for wizards".
+
+The "wizard" switches are intended for experimentation with JPEG by persons
+who are reasonably knowledgeable about the JPEG standard.  If you don't know
+what you are doing, DON'T USE THESE SWITCHES.  You'll likely produce files
+with worse image quality and/or poorer compression than you'd get from the
+default settings.  Furthermore, these switches must be used with caution
+when making files intended for general use, because not all JPEG decoders
+will support unusual JPEG parameter settings.
+
+
+Quantization Table Adjustment
+-----------------------------
+
+Ordinarily, cjpeg starts with a default set of tables (the same ones given
+as examples in the JPEG standard) and scales them up or down according to
+the -quality setting.  The details of the scaling algorithm can be found in
+jcparam.c.  At very low quality settings, some quantization table entries
+can get scaled up to values exceeding 255.  Although 2-byte quantization
+values are supported by the IJG software, this feature is not in baseline
+JPEG and is not supported by all implementations.  If you need to ensure
+wide compatibility of low-quality files, you can constrain the scaled
+quantization values to no more than 255 by giving the -baseline switch.
+Note that use of -baseline will result in poorer quality for the same file
+size, since more bits than necessary are expended on higher AC coefficients.
+
+You can substitute a different set of quantization values by using the
+-qtables switch:
+
+	-qtables file	Use the quantization tables given in the named file.
+
+The specified file should be a text file containing decimal quantization
+values.  The file should contain one to four tables, each of 64 elements.
+The tables are implicitly numbered 0,1,etc. in order of appearance.  Table
+entries appear in normal array order (NOT in the zigzag order in which they
+will be stored in the JPEG file).
+
+Quantization table files are free format, in that arbitrary whitespace can
+appear between numbers.  Also, comments can be included: a comment starts
+with '#' and extends to the end of the line.  Here is an example file that
+duplicates the default quantization tables:
+
+	# Quantization tables given in JPEG spec, section K.1
+
+	# This is table 0 (the luminance table):
+	  16  11  10  16  24  40  51  61
+	  12  12  14  19  26  58  60  55
+	  14  13  16  24  40  57  69  56
+	  14  17  22  29  51  87  80  62
+	  18  22  37  56  68 109 103  77
+	  24  35  55  64  81 104 113  92
+	  49  64  78  87 103 121 120 101
+	  72  92  95  98 112 100 103  99
+
+	# This is table 1 (the chrominance table):
+	  17  18  24  47  99  99  99  99
+	  18  21  26  66  99  99  99  99
+	  24  26  56  99  99  99  99  99
+	  47  66  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+	  99  99  99  99  99  99  99  99
+
+If the -qtables switch is used without -quality, then the specified tables
+are used exactly as-is.  If both -qtables and -quality are used, then the
+tables taken from the file are scaled in the same fashion that the default
+tables would be scaled for that quality setting.  If -baseline appears, then
+the quantization values are constrained to the range 1-255.
+
+By default, cjpeg will use quantization table 0 for luminance components and
+table 1 for chrominance components.  To override this choice, use the -qslots
+switch:
+
+	-qslots N[,...]		Select which quantization table to use for
+				each color component.
+
+The -qslots switch specifies a quantization table number for each color
+component, in the order in which the components appear in the JPEG SOF marker.
+For example, to create a separate table for each of Y,Cb,Cr, you could
+provide a -qtables file that defines three quantization tables and say
+"-qslots 0,1,2".  If -qslots gives fewer table numbers than there are color
+components, then the last table number is repeated as necessary.
+
+
+Sampling Factor Adjustment
+--------------------------
+
+By default, cjpeg uses 2:1 horizontal and vertical downsampling when
+compressing YCbCr data, and no downsampling for all other color spaces.
+You can override this default with the -sample switch:
+
+	-sample HxV[,...]	Set JPEG sampling factors for each color
+				component.
+
+The -sample switch specifies the JPEG sampling factors for each color
+component, in the order in which they appear in the JPEG SOF marker.
+If you specify fewer HxV pairs than there are components, the remaining
+components are set to 1x1 sampling.  For example, the default YCbCr setting
+is equivalent to "-sample 2x2,1x1,1x1", which can be abbreviated to
+"-sample 2x2".
+
+There are still some JPEG decoders in existence that support only 2x1
+sampling (also called 4:2:2 sampling).  Compatibility with such decoders can
+be achieved by specifying "-sample 2x1".  This is not recommended unless
+really necessary, since it increases file size and encoding/decoding time
+with very little quality gain.
+
+
+Multiple Scan / Progression Control
+-----------------------------------
+
+By default, cjpeg emits a single-scan sequential JPEG file.  The
+-progressive switch generates a progressive JPEG file using a default series
+of progression parameters.  You can create multiple-scan sequential JPEG
+files or progressive JPEG files with custom progression parameters by using
+the -scans switch:
+
+	-scans file	Use the scan sequence given in the named file.
+
+The specified file should be a text file containing a "scan script".
+The script specifies the contents and ordering of the scans to be emitted.
+Each entry in the script defines one scan.  A scan definition specifies
+the components to be included in the scan, and for progressive JPEG it also
+specifies the progression parameters Ss,Se,Ah,Al for the scan.  Scan
+definitions are separated by semicolons (';').  A semicolon after the last
+scan definition is optional.
+
+Each scan definition contains one to four component indexes, optionally
+followed by a colon (':') and the four progressive-JPEG parameters.  The
+component indexes denote which color component(s) are to be transmitted in
+the scan.  Components are numbered in the order in which they appear in the
+JPEG SOF marker, with the first component being numbered 0.  (Note that these
+indexes are not the "component ID" codes assigned to the components, just
+positional indexes.)
+
+The progression parameters for each scan are:
+	Ss	Zigzag index of first coefficient included in scan
+	Se	Zigzag index of last coefficient included in scan
+	Ah	Zero for first scan of a coefficient, else Al of prior scan
+	Al	Successive approximation low bit position for scan
+If the progression parameters are omitted, the values 0,63,0,0 are used,
+producing a sequential JPEG file.  cjpeg automatically determines whether
+the script represents a progressive or sequential file, by observing whether
+Ss and Se values other than 0 and 63 appear.  (The -progressive switch is
+not needed to specify this; in fact, it is ignored when -scans appears.)
+The scan script must meet the JPEG restrictions on progression sequences.
+(cjpeg checks that the spec's requirements are obeyed.)
+
+Scan script files are free format, in that arbitrary whitespace can appear
+between numbers and around punctuation.  Also, comments can be included: a
+comment starts with '#' and extends to the end of the line.  For additional
+legibility, commas or dashes can be placed between values.  (Actually, any
+single punctuation character other than ':' or ';' can be inserted.)  For
+example, the following two scan definitions are equivalent:
+	0 1 2: 0 63 0 0;
+	0,1,2 : 0-63, 0,0 ;
+
+Here is an example of a scan script that generates a partially interleaved
+sequential JPEG file:
+
+	0;			# Y only in first scan
+	1 2;			# Cb and Cr in second scan
+
+Here is an example of a progressive scan script using only spectral selection
+(no successive approximation):
+
+	# Interleaved DC scan for Y,Cb,Cr:
+	0,1,2: 0-0,   0, 0 ;
+	# AC scans:
+	0:     1-2,   0, 0 ;	# First two Y AC coefficients
+	0:     3-5,   0, 0 ;	# Three more
+	1:     1-63,  0, 0 ;	# All AC coefficients for Cb
+	2:     1-63,  0, 0 ;	# All AC coefficients for Cr
+	0:     6-9,   0, 0 ;	# More Y coefficients
+	0:     10-63, 0, 0 ;	# Remaining Y coefficients
+
+Here is an example of a successive-approximation script.  This is equivalent
+to the default script used by "cjpeg -progressive" for YCbCr images:
+
+	# Initial DC scan for Y,Cb,Cr (lowest bit not sent)
+	0,1,2: 0-0,   0, 1 ;
+	# First AC scan: send first 5 Y AC coefficients, minus 2 lowest bits:
+	0:     1-5,   0, 2 ;
+	# Send all Cr,Cb AC coefficients, minus lowest bit:
+	# (chroma data is usually too small to be worth subdividing further;
+	#  but note we send Cr first since eye is least sensitive to Cb)
+	2:     1-63,  0, 1 ;
+	1:     1-63,  0, 1 ;
+	# Send remaining Y AC coefficients, minus 2 lowest bits:
+	0:     6-63,  0, 2 ;
+	# Send next-to-lowest bit of all Y AC coefficients:
+	0:     1-63,  2, 1 ;
+	# At this point we've sent all but the lowest bit of all coefficients.
+	# Send lowest bit of DC coefficients
+	0,1,2: 0-0,   1, 0 ;
+	# Send lowest bit of AC coefficients
+	2:     1-63,  1, 0 ;
+	1:     1-63,  1, 0 ;
+	# Y AC lowest bit scan is last; it's usually the largest scan
+	0:     1-63,  1, 0 ;
+
+It may be worth pointing out that this script is tuned for quality settings
+of around 50 to 75.  For lower quality settings, you'd probably want to use
+a script with fewer stages of successive approximation (otherwise the
+initial scans will be really bad).  For higher quality settings, you might
+want to use more stages of successive approximation (so that the initial
+scans are not too large).
diff --git a/png/ANNOUNCE b/png/ANNOUNCE
new file mode 100644
index 000000000..5ec25799c
--- /dev/null
+++ b/png/ANNOUNCE
@@ -0,0 +1,31 @@
+
+Libpng 1.2.1 - December 12, 2001
+
+This is a public release of libpng, intended for use in production codes.
+
+Changes since the last public release (1.2.0):
+
+  Revised makefile.std in contrib/pngminus
+  Include background_1 in png_struct regardless of gamma support.
+  Revised makefile.netbsd and makefile.macosx, added makefile.darwin.
+  Revised example.c to provide more details about using row_callback().
+  Added makefile.so9.
+  Removed type casts from all NULLs, except for those appearing in function
+    calls when PNG_NO_TYPECAST_NULL is defined.
+  Simplified png_create_info_struct() and png_creat_struct_2().
+  Added error message if png_write_info() was omitted.
+  Changed typecast of "size" argument to png_size_t in pngmem.c calls to
+    the user malloc_fn, to agree with the prototype in png.h
+  Added a pop/push operation to pnggccrd.c, to preserve Eflag (Maxim Sobolev)
+  Updated makefile.sgi to recognize LIBPATH and INCPATH.
+  Updated various makefiles so "make clean" does not remove previous major
+    version of the shared library.
+  Added a pop/push operation to pngvcrd.c, to preserve Eflag.
+  Always allocate 256-entry internal palette, hist, and trans arrays, to
+    avoid out-of-bounds memory reference caused by invalid PNG datastreams.
+  Added a check for prefix_length > data_length in iCCP chunk handler.
+
+Send comments/corrections/commendations to
+png-implement@ccrc.wustl.edu or to randeg@alum.rpi.edu
+
+Glenn R-P
diff --git a/png/CHANGES b/png/CHANGES
new file mode 100644
index 000000000..9a81ab768
--- /dev/null
+++ b/png/CHANGES
@@ -0,0 +1,1021 @@
+
+CHANGES - changes for libpng
+
+version 0.2
+  added reader into png.h
+  fixed small problems in stub file
+version 0.3
+  added pull reader
+  split up pngwrite.c to several files
+  added pnglib.txt
+  added example.c
+  cleaned up writer, adding a few new tranformations
+  fixed some bugs in writer
+  interfaced with zlib 0.5
+  added K&R support
+  added check for 64 KB blocks for 16 bit machines
+version 0.4
+  cleaned up code and commented code
+  simplified time handling into png_time
+  created png_color_16 and png_color_8 to handle color needs
+  cleaned up color type defines
+  fixed various bugs
+  made various names more consistant
+  interfaced with zlib 0.71
+  cleaned up zTXt reader and writer (using zlib's Reset functions)
+  split transformations into pngrtran.c and pngwtran.c
+version 0.5
+  interfaced with zlib 0.8
+  fixed many reading and writing bugs
+  saved using 3 spaces instead of tabs
+version 0.6
+  added png_large_malloc() and png_large_free()
+  added png_size_t
+  cleaned up some compiler warnings
+  added png_start_read_image()
+version 0.7
+  cleaned up lots of bugs
+  finished dithering and other stuff
+  added test program
+  changed name from pnglib to libpng
+version 0.71 [June, 1995]
+  changed pngtest.png for zlib 0.93
+  fixed error in libpng.txt and example.c
+version 0.8
+  cleaned up some bugs
+  added png_set_filler()
+  split up pngstub.c into pngmem.c, pngio.c, and pngerror.c
+  added #define's to remove unwanted code
+  moved png_info_init() to png.c
+  added old_size into png_realloc()
+  added functions to manually set filtering and compression info
+  changed compression parameters based on image type
+  optimized filter selection code
+  added version info
+  changed external functions passing floats to doubles (k&r problems?)
+  put all the configurable stuff in pngconf.h
+  enabled png_set_shift to work with paletted images on read
+  added png_read_update_info() - updates info structure with
+     transformations
+version 0.81 [August, 1995]
+  incorporated Tim Wegner's medium model code (thanks, Tim)
+version 0.82 [September, 1995]
+  [unspecified changes]
+version 0.85 [December, 1995]
+  added more medium model code (almost everything's a far)
+  added i/o, error, and memory callback functions
+  fixed some bugs (16 bit, 4 bit interlaced, etc.)
+  added first run progressive reader (barely tested)
+version 0.86 [January, 1996]
+  fixed bugs
+  improved documentation
+version 0.87 [January, 1996]
+  fixed medium model bugs
+  fixed other bugs introduced in 0.85 and 0.86
+  added some minor documentation
+version 0.88 [January, 1996]
+  fixed progressive bugs
+  replaced tabs with spaces
+  cleaned up documentation
+  added callbacks for read/write and warning/error functions
+version 0.89 [July, 1996]
+  added new initialization API to make libpng work better with shared libs
+     we now have png_create_read_struct(), png_create_write_struct(),
+     png_create_info_struct(), png_destroy_read_struct(), and
+     png_destroy_write_struct() instead of the separate calls to
+     malloc and png_read_init(), png_info_init(), and png_write_init()
+  changed warning/error callback functions to fix bug - this means you
+     should use the new initialization API if you were using the old
+     png_set_message_fn() calls, and that the old API no longer exists
+     so that people are aware that they need to change their code
+  changed filter selection API to allow selection of multiple filters
+     since it didn't work in previous versions of libpng anyways
+  optimized filter selection code
+  fixed png_set_background() to allow using an arbitrary RGB color for
+     paletted images
+  fixed gamma and background correction for paletted images, so
+     png_correct_palette is not needed unless you are correcting an
+     external palette (you will need to #define PNG_CORRECT_PALETTE_SUPPORTED
+     in pngconf.h) - if nobody uses this, it may disappear in the future.
+  fixed bug with Borland 64K memory allocation (Alexander Lehmann)
+  fixed bug in interlace handling (Smarasderagd, I think)
+  added more error checking for writing and image to reduce invalid files
+  separated read and write functions so that they won't both be linked
+     into a binary when only reading or writing functionality is used
+  new pngtest image also has interlacing and zTXt
+  updated documentation to reflect new API
+version 0.90 [January, 1997]
+  made CRC errors/warnings on critical and ancillary chunks configurable
+  libpng will use the zlib CRC routines by (compile-time) default
+  changed DOS small/medium model memory support - needs zlib 1.04 (Tim Wegner)
+  added external C++ wrapper statements to png.h (Gilles Dauphin)
+  allow PNG file to be read when some or all of file signature has already
+     been read from the beginning of the stream.  ****This affects the size
+     of info_struct and invalidates all programs that use a shared libpng****
+  fixed png_filler() declarations
+  fixed? background color conversions
+  fixed order of error function pointers to match documentation
+  current chunk name is now available in png_struct to reduce the number
+     of nearly identical error messages (will simplify multi-lingual
+     support when available)
+  try to get ready for unknown-chunk callback functions:
+     - previously read critical chunks are flagged, so the chunk handling
+       routines can determine if the chunk is in the right place
+     - all chunk handling routines have the same prototypes, so we will
+       be able to handle all chunks via a callback mechanism
+  try to fix Linux "setjmp" buffer size problems
+  removed png_large_malloc, png_large_free, and png_realloc functions.
+version 0.95 [March, 1997]
+  fixed bug in pngwutil.c allocating "up_row" twice and "avg_row" never
+  fixed bug in PNG file signature compares when start != 0
+  changed parameter type of png_set_filler(...filler...) from png_byte
+     to png_uint_32
+  added test for MACOS to ensure that both math.h and fp.h are not #included
+  added macros for libpng to be compiled as a Windows DLL (Andreas Kupries)
+  added "packswap" transformation, which changes the endianness of
+     packed-pixel bytes (Kevin Bracey)
+  added "strip_alpha" transformation, which removes the alpha channel of
+     input images without using it (not neccesarily a good idea)
+  added "swap_alpha" transformation, which puts the alpha channel in front
+     of the color bytes instead of after
+  removed all implicit variable tests which assume NULL == 0 (I think)
+  changed several variables to "png_size_t" to show 16/32-bit limitations
+  added new pCAL chunk read/write support
+  added experimental filter selection weighting (Greg Roelofs)
+  removed old png_set_rgbx() and png_set_xrgb() functions that have been
+     obsolete for about 2 years now (use png_set_filler() instead)
+  added macros to read 16- and 32-bit ints directly from buffer, to be
+     used only on those systems that support it (namely PowerPC and 680x0)
+     With some testing, this may become the default for MACOS/PPC systems.
+  only calculate CRC on data if we are going to use it
+  added macros for zTXt compression type PNG_zTXt_COMPRESSION_???
+  added macros for simple libpng debugging output selectable at compile time
+  removed PNG_READ_END_MODE in progressive reader (Smarasderagd)
+  more description of info_struct in libpng.txt and png.h
+  more instructions in example.c
+  more chunk types tested in pngtest.c
+  renamed pngrcb.c to pngset.c, and all png_read_<chunk> functions to be
+     png_set_<chunk>.  We now have corresponding png_get_<chunk>
+     functions in pngget.c to get infomation in info_ptr.  This isolates
+     the application from the internal organization of png_info_struct
+     (good for shared library implementations).
+version 0.96 [May, 1997]
+  fixed serious bug with < 8bpp images introduced in 0.95
+  fixed 256-color transparency bug (Greg Roelofs)
+  fixed up documentation (Greg Roelofs, Laszlo Nyul)
+  fixed "error" in pngconf.h for Linux setjmp() behaviour
+  fixed DOS medium model support (Tim Wegner)
+  fixed png_check_keyword() for case with error in static string text
+  added read of CRC after IEND chunk for embedded PNGs (Laszlo Nyul)
+  added typecasts to quiet compiler errors
+  added more debugging info
+version 0.97 [January, 1998]
+  removed PNG_USE_OWN_CRC capability
+  relocated png_set_crc_action from pngrutil.c to pngrtran.c
+  fixed typecasts of "new_key", etc. (Andreas Dilger)
+  added RFC 1152 [sic] date support
+  fixed bug in gamma handling of 4-bit grayscale
+  added 2-bit grayscale gamma handling (Glenn R-P)
+  added more typecasts. 65536L becomes (png_uint_32)65536L, etc. (Glenn R-P)
+  minor corrections in libpng.txt
+  added simple sRGB support (Glenn R-P)
+  easier conditional compiling, e.g. define PNG_READ/WRITE_NOT_FULLY_SUPPORTED;
+     all configurable options can be selected from command-line instead
+     of having to edit pngconf.h (Glenn R-P)
+  fixed memory leak in pngwrite.c (free info_ptr->text) (Glenn R-P)
+  added more conditions for png_do_background, to avoid changing
+     black pixels to background when a background is supplied and
+     no pixels are transparent
+  repaired PNG_NO_STDIO behaviour
+  tested NODIV support and made it default behaviour (Greg Roelofs)
+  added "-m" option and PNGTEST_DEBUG_MEMORY to pngtest (John Bowler)
+  regularized version numbering scheme and bumped shared-library major
+     version number to 2 to avoid problems with libpng 0.89 apps (Greg Roelofs)
+version 0.98 [January, 1998]
+  cleaned up some typos in libpng.txt and in code documentation
+  fixed memory leaks in pCAL chunk processing (Glenn R-P and John Bowler)
+  cosmetic change "display_gamma" to "screen_gamma" in pngrtran.c
+  changed recommendation about file_gamma for PC images to .51 from .45,
+     in example.c and libpng.txt, added comments to distinguish between
+     screen_gamma, viewing_gamma, and display_gamma.
+  changed all references to RFC1152 to read RFC1123 and changed the
+     PNG_TIME_RFC1152_SUPPORTED macro to PNG_TIME_RFC1123_SUPPORTED
+  added png_invert_alpha capability (Glenn R-P -- suggestion by Jon Vincent)
+  changed srgb_intent from png_byte to int to avoid compiler bugs
+version 0.99 [January 30, 1998]
+  free info_ptr->text instead of end_info_ptr->text in pngread.c (John Bowler)
+  fixed a longstanding "packswap" bug in pngtrans.c
+  fixed some inconsistencies in pngconf.h that prevented compiling with
+     PNG_READ_GAMMA_SUPPORTED and PNG_READ_hIST_SUPPORTED undefined
+  fixed some typos and made other minor rearrangement of libpng.txt (Andreas)
+  changed recommendation about file_gamma for PC images to .50 from .51 in
+     example.c and libpng.txt, and changed file_gamma for sRGB images to .45
+  added a number of functions to access information from the png structure
+     png_get_image_height(), etc. (Glenn R-P, suggestion by Brad Pettit)
+  added TARGET_MACOS similar to zlib-1.0.8
+  define PNG_ALWAYS_EXTERN when __MWERKS__ && WIN32 are defined
+  added type casting to all png_malloc() function calls
+version 0.99a [January 31, 1998]
+  Added type casts and parentheses to all returns that return a value.(Tim W.)
+version 0.99b [February 4, 1998]
+  Added type cast png_uint_32 on malloc function calls where needed.
+  Changed type of num_hist from png_uint_32 to int (same as num_palette).
+  Added checks for rowbytes overflow, in case png_size_t is less than 32 bits.
+  Renamed makefile.elf to makefile.lnx.
+version 0.99c [February 7, 1998]
+  More type casting.  Removed erroneous overflow test in pngmem.c.
+  Added png_buffered_memcpy() and png_buffered_memset(), apply them to rowbytes.
+  Added UNIX manual pages libpng.3 (incorporating libpng.txt) and  png.5.
+version 0.99d [February 11, 1998]
+  Renamed "far_to_near()" "png_far_to_near()"
+  Revised libpng.3
+  Version 99c "buffered" operations didn't work as intended.  Replaced them
+    with png_memcpy_check() and png_memset_check().
+  Added many "if (png_ptr == NULL) return" to quell compiler warnings about
+    unused png_ptr, mostly in pngget.c and pngset.c.
+  Check for overlength tRNS chunk present when indexed-color PLTE is read.
+  Cleaned up spelling errors in libpng.3/libpng.txt
+  Corrected a problem with png_get_tRNS() which returned undefined trans array
+version 0.99e [February 28, 1998]
+  Corrected png_get_tRNS() again.
+  Add parentheses for easier reading of pngget.c, fixed "||" should be "&&".
+  Touched up example.c to make more of it compileable, although the entire
+    file still can't be compiled (Willem van Schaik)
+  Fixed a bug in png_do_shift() (Bryan Tsai)
+  Added a space in png.h prototype for png_write_chunk_start()
+  Replaced pngtest.png with one created with zlib 1.1.1
+  Changed pngtest to report PASS even when file size is different (Jean-loup G.)
+  Corrected some logic errors in png_do_invert_alpha() (Chris Patterson)
+version 0.99f [March 5, 1998]
+  Corrected a bug in pngpread() introduced in version 99c (Kevin Bracey)
+  Moved makefiles into a "scripts" directory, and added INSTALL instruction file
+  Added makefile.os2 and pngos2.def (A. Zabolotny) and makefile.s2x (W. Sebok)
+  Added pointers to "note on libpng versions" in makefile.lnx and README
+  Added row callback feature when reading and writing nonprogressive rows
+     and added a test of this feature in pngtest.c
+  Added user transform callbacks, with test of the feature in pngtest.c
+version 0.99g [March 6, 1998, morning]
+  Minor changes to pngtest.c to suppress compiler warnings.
+  Removed "beta" language from documentation.
+version 0.99h [March 6, 1998, evening]
+  Minor changes to previous minor changes to pngtest.c
+  Changed PNG_READ_NOT_FULLY_SUPPORTED to PNG_READ_TRANSFORMS_NOT_SUPPORTED
+  and added PNG_PROGRESSIVE_READ_NOT_SUPPORTED macro
+  Added user transform capability
+version 1.00 [March 7, 1998]
+  Changed several typedefs in pngrutil.c
+  Added makefile.wat (Pawel Mrochen), updated makefile.tc3 (Willem van Schaik)
+  replaced "while(1)" with "for(;;)"
+  added PNGARG() to prototypes in pngtest.c and removed some prototypes
+  updated some of the makefiles (Tom Lane)
+  changed some typedefs (s_start, etc.) in pngrutil.c
+  fixed dimensions of "short_months" array in pngwrite.c
+  Replaced ansi2knr.c with the one from jpeg-v6
+version 1.0.0 [March 8, 1998]
+  Changed name from 1.00 to 1.0.0 (Adam Costello)
+  Added smakefile.ppc (with SCOPTIONS.ppc) for Amiga PPC (Andreas Kleinert)
+version 1.0.0a [March 9, 1998]
+  Fixed three bugs in pngrtran.c to make gamma+background handling consistent
+  (Greg Roelofs)
+  Changed format of the PNG_LIBPNG_VER integer to xyyzz instead of xyz
+  for major, minor, and bugfix releases.  This is 10001. (Adam Costello,
+  Tom Lane)
+  Make months range from 1-12 in png_convert_to_rfc1123
+version 1.0.0b [March 13, 1998]
+  Quieted compiler complaints about two empty "for" loops in pngrutil.c
+  Minor changes to makefile.s2x
+  Removed #ifdef/#endif around a png_free() in pngread.c
+version 1.0.1 [March 14, 1998]
+  Changed makefile.s2x to reduce security risk of using a relative pathname
+  Fixed some typos in the documentation (Greg).
+  Fixed a problem with value of "channels" returned by png_read_update_info()
+version 1.0.1a [April 21, 1998]
+  Optimized Paeth calculations by replacing abs() function calls with intrinsics
+  plus other loop optimizations. Improves avg decoding speed by about 20%.
+  Commented out i386istic "align" compiler flags in makefile.lnx.
+  Reduced the default warning level in some makefiles, to make them consistent.
+  Removed references to IJG and JPEG in the ansi2knr.c copyright statement.
+  Fixed a bug in png_do_strip_filler with XXRRGGBB => RRGGBB transformation.
+  Added grayscale and 16-bit capability to png_do_read_filler().
+  Fixed a bug in pngset.c, introduced in version 0.99c, that sets rowbytes
+    too large when writing an image with bit_depth < 8 (Bob Dellaca).
+  Corrected some bugs in the experimental weighted filtering heuristics.
+  Moved a misplaced pngrutil code block that truncates tRNS if it has more
+    than num_palette entries -- test was done before num_palette was defined.
+  Fixed a png_convert_to_rfc1123() bug that converts day 31 to 0 (Steve Eddins).
+  Changed compiler flags in makefile.wat for better optimization (Pawel Mrochen).
+version 1.0.1b [May 2, 1998]
+  Relocated png_do_gray_to_rgb() within png_do_read_transformations() (Greg).
+  Relocated the png_composite macros from pngrtran.c to png.h (Greg).
+  Added makefile.sco (contributed by Mike Hopkirk).
+  Fixed two bugs (missing definitions of "istop") introduced in libpng-1.0.1a.
+  Fixed a bug in pngrtran.c that would set channels=5 under some circumstances.
+  More work on the Paeth-filtering, achieving imperceptible speedup (A Kleinert).
+  More work on loop optimization which may help when compiled with C++ compilers.
+  Added warnings when people try to use transforms they've defined out.
+  Collapsed 4 "i" and "c" loops into single "i" loops in pngrtran and pngwtran.
+  Revised paragraph about png_set_expand() in libpng.txt and libpng.3 (Greg)
+version 1.0.1c [May 11, 1998]
+  Fixed a bug in pngrtran.c (introduced in libpng-1.0.1a) where the masks for
+    filler bytes should have been 0xff instead of 0xf.
+  Added max_pixel_depth=32 in pngrutil.c when using FILLER with palette images.
+  Moved PNG_WRITE_WEIGHTED_FILTER_SUPPORTED and PNG_WRITE_FLUSH_SUPPORTED
+    out of the PNG_WRITE_TRANSFORMS_NOT_SUPPORTED block of pngconf.h
+  Added "PNG_NO_WRITE_TRANSFORMS" etc., as alternatives for *_NOT_SUPPORTED,
+    for consistency, in pngconf.h
+  Added individual "ifndef PNG_NO_[CAPABILITY]" in pngconf.h to make it easier
+    to remove unwanted capabilities via the compile line
+  Made some corrections to grammar (which, it's) in documentation (Greg).
+  Corrected example.c, use of row_pointers in png_write_image().
+version 1.0.1d [May 24, 1998]
+  Corrected several statements that used side effects illegally in pngrutil.c
+    and pngtrans.c, that were introduced in version 1.0.1b
+  Revised png_read_rows() to avoid repeated if-testing for NULL (A Kleinert)
+  More corrections to example.c, use of row_pointers in png_write_image()
+    and png_read_rows().
+  Added pngdll.mak and pngdef.pas to scripts directory, contributed by
+    Bob Dellaca, to make a png32bd.dll with Borland C++ 4.5
+  Fixed error in example.c with png_set_text: num_text is 3, not 2 (Guido V.)
+  Changed several loops from count-down to count-up, for consistency.
+version 1.0.1e [June 6, 1998]
+  Revised libpng.txt and libpng.3 description of png_set_read|write_fn(), and
+    added warnings when people try to set png_read_fn and png_write_fn in
+    the same structure.
+  Added a test such that png_do_gamma will be done when num_trans==0
+    for truecolor images that have defined a background.  This corrects an
+    error that was introduced in libpng-0.90 that can cause gamma processing
+    to be skipped.
+  Added tests in png.h to include "trans" and "trans_values" in structures
+    when PNG_READ_BACKGROUND_SUPPORTED or PNG_READ_EXPAND_SUPPORTED is defined.
+  Add png_free(png_ptr->time_buffer) in png_destroy_read_struct()
+  Moved png_convert_to_rfc_1123() from pngwrite.c to png.c
+  Added capability for user-provided malloc_fn() and free_fn() functions,
+    and revised pngtest.c to demonstrate their use, replacing the
+    PNGTEST_DEBUG_MEM feature.
+  Added makefile.w32, for Microsoft C++ 4.0 and later (Tim Wegner).
+version 1.0.2 [June 14, 1998]
+  Fixed two bugs in makefile.bor .
+version 1.0.2a [December 30, 1998]
+  Replaced and extended code that was removed from png_set_filler() in 1.0.1a.
+  Fixed a bug in png_do_filler() that made it fail to write filler bytes in
+    the left-most pixel of each row (Kevin Bracey).
+  Changed "static pngcharp tIME_string" to "static char tIME_string[30]"
+    in pngtest.c (Duncan Simpson).
+  Fixed a bug in pngtest.c that caused pngtest to try to write a tIME chunk
+    even when no tIME chunk was present in the source file.
+  Fixed a problem in pngrutil.c: gray_to_rgb didn't always work with 16-bit.
+  Fixed a problem in png_read_push_finish_row(), which would not skip some
+    passes that it should skip, for images that are less than 3 pixels high.
+  Interchanged the order of calls to png_do_swap() and png_do_shift()
+    in pngwtran.c (John Cromer).
+  Added #ifdef PNG_DEBUG/#endif surrounding use of PNG_DEBUG in png.h .
+  Changed "bad adaptive filter type" from error to warning in pngrutil.c .
+  Fixed a documentation error about default filtering with 8-bit indexed-color.
+  Separated the PNG_NO_STDIO macro into PNG_NO_STDIO and PNG_NO_CONSOLE_IO
+    (L. Peter Deutsch).
+  Added png_set_rgb_to_gray() and png_get_rgb_to_gray_status() functions.
+  Added png_get_copyright() and png_get_header_version() functions.
+  Revised comments on png_set_progressive_read_fn() in libpng.txt and example.c
+  Added information about debugging in libpng.txt and libpng.3 .
+  Changed "ln -sf" to "ln -s -f" in makefile.s2x, makefile.lnx, and makefile.sco.
+  Removed lines after Dynamic Dependencies" in makefile.aco .
+  Revised makefile.dec to make a shared library (Jeremie Petit).
+  Removed trailing blanks from all files.
+version 1.0.2a [January 6, 1999]
+  Removed misplaced #endif and #ifdef PNG_NO_EXTERN near the end of png.h
+  Added "if" tests to silence complaints about unused png_ptr in png.h and png.c
+  Changed "check_if_png" function in example.c to return true (nonzero) if PNG.
+  Changed libpng.txt to demonstrate png_sig_cmp() instead of png_check_sig()
+    which is obsolete.
+version 1.0.3 [January 14, 1999]
+  Added makefile.hux, for Hewlett Packard HPUX 10.20 and 11.00 (Jim Rice)
+  Added a statement of Y2K compliance in png.h, libpng.3, and Y2KINFO.
+version 1.0.3a [August 12, 1999]
+  Added check for PNG_READ_INTERLACE_SUPPORTED in pngread.c; issue a warning
+     if an attempt is made to read an interlaced image when it's not supported.
+  Added check if png_ptr->trans is defined before freeing it in pngread.c
+  Modified the Y2K statement to include versions back to version 0.71
+  Fixed a bug in the check for valid IHDR bit_depth/color_types in pngrutil.c
+  Modified makefile.wat (added -zp8 flag, ".symbolic", changed some comments)
+  Replaced leading blanks with tab characters in makefile.hux
+  Changed "dworkin.wustl.edu" to "ccrc.wustl.edu" in various documents.
+  Changed (float)red and (float)green to (double)red, (double)green
+     in png_set_rgb_to_gray() to avoid "promotion" problems in AIX.
+  Fixed a bug in pngconf.h that omitted <stdio.h> when PNG_DEBUG==0 (K Bracey).
+  Reformatted libpng.3 and libpngpf.3 with proper fonts (script by J. vanZandt).
+  Updated documentation to refer to the PNG-1.2 specification.
+  Removed ansi2knr.c and left pointers to the latest source for ansi2knr.c
+    in makefile.knr, INSTALL, and README (L. Peter Deutsch)
+  Fixed bugs in calculation of the length of rowbytes when adding alpha
+    channels to 16-bit images, in pngrtran.c (Chris Nokleberg)
+  Added function png_set_user_transform_info() to store user_transform_ptr,
+    user_depth, and user_channels into the png_struct, and a function
+    png_get_user_transform_ptr() to retrieve the pointer (Chris Nokleberg)
+  Added function png_set_empty_plte_permitted() to make libpng useable
+    in MNG applications.
+  Corrected the typedef for png_free_ptr in png.h (Jesse Jones).
+  Correct gamma with srgb is 45455 instead of 45000 in pngrutil.c, to be
+    consistent with PNG-1.2, and allow variance of 500 before complaining.
+  Added assembler code contributed by Intel in file pngvcrd.c and modified
+    makefile.w32 to use it (Nirav Chhatrapati, INTEL Corporation, Gilles Vollant)
+  Changed "ln -s -f" to "ln -f -s" in the makefiles to make Solaris happy.
+  Added some aliases for png_set_expand() in pngrtran.c, namely
+    png_set_expand_PLTE(), png_set_expand_depth(), and png_set_expand_tRNS()
+    (Greg Roelofs, in "PNG: The Definitive Guide").
+  Added makefile.beo for BEOS on X86, contributed by Sander Stok.
+version 1.0.3b [August 26, 1999]
+  Replaced 2147483647L several places with PNG_MAX_UINT macro, defined in png.h
+  Changed leading blanks to tabs in all makefiles.
+  Define PNG_USE_PNGVCRD in makefile.w32, to get MMX assembler code.
+  Made alternate versions of  png_set_expand() in pngrtran.c, namely
+    png_set_gray_1_2_4_to_8, png_set_palette_to_rgb, and png_set_tRNS_to_alpha
+    (Greg Roelofs, in "PNG: The Definitive Guide").  Deleted the 1.0.3a aliases.
+  Relocated start of 'extern "C"' block in png.h so it doesn't include pngconf.h
+  Revised calculation of num_blocks in pngmem.c to avoid a potentially
+    negative shift distance, whose results are undefined in the C language.
+  Added a check in pngset.c to prevent writing multiple tIME chunks.
+  Added a check in pngwrite.c to detect invalid small window_bits sizes.
+version 1.0.3d [September 4, 1999]
+  Fixed type casting of igamma in pngrutil.c
+  Added new png_expand functions to scripts/pngdef.pas and pngos2.def
+  Added a demo read_user_transform_fn that examines the row filters in pngtest.c
+version 1.0.4 [September 24, 1999]
+  Define PNG_ALWAYS_EXTERN in pngconf.h if __STDC__ is defined
+  Delete #define PNG_INTERNAL and include "png.h" from pngasmrd.h
+  Made several minor corrections to pngtest.c
+  Renamed the makefiles with longer but more user friendly extensions.
+  Copied the PNG copyright and license to a separate LICENSE file.
+  Revised documentation, png.h, and example.c to remove reference to
+    "viewing_gamma" which no longer appears in the PNG specification.
+  Revised pngvcrd.c to use MMX code for interlacing only on the final pass.
+  Updated pngvcrd.c to use the faster C filter algorithms from libpng-1.0.1a
+  Split makefile.win32vc into two versions, makefile.vcawin32 (uses MMX
+    assembler code) and makefile.vcwin32 (doesn't).
+  Added a CPU timing report to pngtest.c (enabled by defining PNGTEST_TIMING)
+  Added a copy of pngnow.png to the distribution.
+version 1.0.4a [September 25, 1999]
+  Increase max_pixel_depth in pngrutil.c if a user transform needs it.
+  Changed several division operations to right-shifts in pngvcrd.c
+version 1.0.4b [September 30, 1999]
+  Added parentheses in line 3732 of pngvcrd.c
+  Added a comment in makefile.linux warning about buggy -O3 in pgcc 2.95.1
+version 1.0.4c [October 1, 1999]
+  Added a "png_check_version" function in png.c and pngtest.c that will generate
+    a helpful compiler error if an old png.h is found in the search path.
+  Changed type of png_user_transform_depth|channels from int to png_byte.
+version 1.0.4d [October 6, 1999]
+  Changed 0.45 to 0.45455 in png_set_sRGB()
+  Removed unused PLTE entries from pngnow.png
+  Re-enabled some parts of pngvcrd.c (png_combine_row) that work properly.
+version 1.0.4e [October 10, 1999]
+  Fixed sign error in pngvcrd.c (Greg Roelofs)
+  Replaced some instances of memcpy with simple assignments in pngvcrd (GR-P)
+version 1.0.4f [October 15, 1999]
+  Surrounded example.c code with #if 0 .. #endif to prevent people from
+    inadvertently trying to compile it.
+  Changed png_get_header_version() from a function to a macro in png.h
+  Added type casting mostly in pngrtran.c and pngwtran.c
+  Removed some pointless "ptr = NULL" in pngmem.c
+  Added a "contrib" directory containing the source code from Greg's book.
+version 1.0.5 [October 15, 1999]
+  Minor editing of the INSTALL and README files.
+version 1.0.5a [October 23, 1999]
+  Added contrib/pngsuite and contrib/pngminus (Willem van Schaik)
+  Fixed a typo in the png_set_sRGB() function call in example.c (Jan Nijtmans)
+  Further optimization and bugfix of pngvcrd.c
+  Revised pngset.c so that it does not allocate or free memory in the user's
+    text_ptr structure.  Instead, it makes its own copy.
+  Created separate write_end_info_struct in pngtest.c for a more severe test.
+  Added code in pngwrite.c to free info_ptr->text[i].key to stop a memory leak.
+version 1.0.5b [November 23, 1999]
+  Moved PNG_FLAG_HAVE_CHUNK_HEADER, PNG_FLAG_BACKGROUND_IS_GRAY and
+    PNG_FLAG_WROTE_tIME from flags to mode.
+  Added png_write_info_before_PLTE() function.
+  Fixed some typecasting in contrib/gregbook/*.c
+  Updated scripts/makevms.com and added makevms.com to contrib/gregbook
+    and contrib/pngminus (Martin Zinser)
+version 1.0.5c [November 26, 1999]
+  Moved png_get_header_version from png.h to png.c, to accomodate ansi2knr.
+  Removed all global arrays (according to PNG_NO_GLOBAL_ARRAYS macro), to
+    accomodate making DLL's: Moved usr_png_ver from global variable to function
+    png_get_header_ver() in png.c.  Moved png_sig to png_sig_bytes in png.c and
+    eliminated use of png_sig in pngwutil.c.  Moved the various png_CHNK arrays
+    into pngtypes.h.  Eliminated use of global png_pass arrays.  Declared the
+    png_CHNK and png_pass arrays to be "const".  Made the global arrays
+    available to applications (although none are used in libpng itself) when
+    PNG_NO_GLOBAL_ARRAYS is not defined or when PNG_GLOBAL_ARRAYS is defined.
+  Removed some extraneous "-I" from contrib/pngminus/makefile.std
+  Changed the PNG_sRGB_INTENT macros in png.h to be consistent with PNG-1.2.
+  Change PNG_SRGB_INTENT to PNG_sRGB_INTENT in libpng.txt and libpng.3
+version 1.0.5d [November 29, 1999]
+  Add type cast (png_const_charp) two places in png.c
+  Eliminated pngtypes.h; use macros instead to declare PNG_CHNK arrays.
+  Renamed "PNG_GLOBAL_ARRAYS" to "PNG_USE_GLOBAL_ARRAYS" and made available
+    to applications a macro "PNG_USE_LOCAL_ARRAYS".
+  #ifdef out all the new declarations when PNG_USE_GLOBAL_ARRAYS is defined.
+  Added PNG_EXPORT_VAR macro to accommodate making DLL's.
+version 1.0.5e [November 30, 1999]
+  Added iCCP, iTXt, and sPLT support; added "lang" member to the png_text
+    structure; refactored the inflate/deflate support to make adding new chunks
+    with trailing compressed parts easier in the future, and added new functions
+    png_free_iCCP, png_free_pCAL, png_free_sPLT, png_free_text, png_get_iCCP,
+    png_get_spalettes, png_set_iCCP, png_set_spalettes (Eric S. Raymond).
+  NOTE: Applications that write text chunks MUST define png_text->lang
+    before calling png_set_text(). It must be set to NULL if you want to
+    write tEXt or zTXt chunks.  If you want your application to be able to
+    run with older versions of libpng, use
+
+      #ifdef PNG_iTXt_SUPPORTED
+         png_text[i].lang = NULL;
+      #endif
+
+  Changed png_get_oFFs() and png_set_oFFs() to use signed rather than unsigned
+    offsets (Eric S. Raymond).
+  Combined PNG_READ_cHNK_SUPPORTED and PNG_WRITE_cHNK_SUPPORTED macros into
+    PNG_cHNK_SUPPORTED and combined the three types of PNG_text_SUPPORTED
+    macros, leaving the separate macros also available.
+  Removed comments on #endifs at the end of many short, non-nested #if-blocks.
+version 1.0.5f [December 6, 1999]
+  Changed makefile.solaris to issue a warning about potential problems when
+    the ucb "ld" is in the path ahead of the ccs "ld".
+  Removed "- [date]" from the "synopsis" line in libpng.3 and libpngpf.3.
+  Added sCAL chunk support (Eric S. Raymond).
+version 1.0.5g [December 7, 1999]
+  Fixed "png_free_spallettes" typo in png.h
+  Added code to handle new chunks in pngpread.c
+  Moved PNG_CHNK string macro definitions outside of PNG_NO_EXTERN block
+  Added "translated_key" to png_text structure and png_write_iTXt().
+  Added code in pngwrite.c to work around a newly discovered zlib bug.
+version 1.0.5h [December 10, 1999]
+  NOTE: regarding the note for version 1.0.5e, the following must also
+    be included in your code:
+        png_text[i].translated_key = NULL;
+  Unknown chunk handling is now supported.
+  Option to eliminate all floating point support was added.  Some new
+    fixed-point functions such as png_set_gAMA_fixed() were added.
+  Expanded tabs and removed trailing blanks in source files.
+version 1.0.5i [December 13, 1999]
+  Added some type casts to silence compiler warnings.
+  Renamed "png_free_spalette" to "png_free_spalettes" for consistency.
+  Removed leading blanks from a #define in pngvcrd.c
+  Added some parameters to the new png_set_keep_unknown_chunks() function.
+  Added a test for up->location != 0 in the first instance of writing
+    unknown chunks in pngwrite.c
+  Changed "num" to "i" in png_free_spalettes() and png_free_unknowns() to
+    prevent recursion.
+  Added png_free_hIST() function.
+  Various patches to fix bugs in the sCAL and integer cHRM processing,
+    and to add some convenience macros for use with sCAL.
+version 1.0.5j [December 21, 1999]
+  Changed "unit" parameter of png_write_sCAL from png_byte to int, to work
+    around buggy compilers.
+  Added new type "png_fixed_point" for integers that hold float*100000 values
+  Restored backward compatibility of tEXt/zTXt chunk processing:
+    Restored the first four members of png_text to the same order as v.1.0.5d.
+    Added members "lang_key" and "itxt_length" to png_text struct.  Set
+    text_length=0 when "text" contains iTXt data.  Use the "compression"
+    member to distinguish among tEXt/zTXt/iTXt types.  Added
+    PNG_ITXT_COMPRESSION_NONE (1) and PNG_ITXT_COMPRESSION_zTXt(2) macros.
+    The "Note" above, about backward incompatibility of libpng-1.0.5e, no
+    longer applies.
+  Fixed png_read|write_iTXt() to read|write parameters in the right order,
+    and to write the iTXt chunk after IDAT if it appears in the end_ptr.
+  Added pnggccrd.c, version of pngvcrd.c Intel assembler for gcc (Greg Roelofs)
+  Reversed the order of trying to write floating-point and fixed-point gAMA.
+version 1.0.5k [December 27, 1999]
+  Added many parentheses, e.g., "if (a && b & c)" becomes "if (a && (b & c))"
+  Added png_handle_as_unknown() function (Glenn)
+  Added png_free_chunk_list() function and chunk_list and num_chunk_list members
+    of png_ptr.
+  Eliminated erroneous warnings about multiple sPLT chunks and sPLT-after-PLTE.
+  Fixed a libpng-1.0.5h bug in pngrutil.c that was issuing erroneous warnings
+    about ignoring incorrect gAMA with sRGB (gAMA was in fact not ignored)
+  Added png_free_tRNS(); png_set_tRNS() now malloc's its own trans array (ESR).
+  Define png_get_int_32 when oFFs chunk is supported as well as when pCAL is.
+  Changed type of proflen from png_int_32 to png_uint_32 in png_get_iCCP().
+version 1.0.5l [January 1, 2000]
+  Added functions png_set_read_user_chunk_fn() and png_get_user_chunk_ptr()
+    for setting a callback function to handle unknown chunks and for
+    retrieving the associated user pointer (Glenn).
+version 1.0.5m [January 7, 2000]
+  Added high-level functions png_read_png(), png_write_png(), png_free_pixels().
+version 1.0.5n [January 9, 2000]
+  Added png_free_PLTE() function, and modified png_set_PLTE() to malloc its
+    own memory for info_ptr->palette.  This makes it safe for the calling
+    application to free its copy of the palette any time after it calls
+    png_set_PLTE().
+version 1.0.5o [January 20, 2000]
+  Cosmetic changes only (removed some trailing blanks and TABs)
+version 1.0.5p [January 31, 2000]
+  Renamed pngdll.mak to makefile.bd32
+  Cosmetic changes in pngtest.c
+version 1.0.5q [February 5, 2000]
+  Relocated the makefile.solaris warning about PATH problems.
+  Fixed pngvcrd.c bug by pushing/popping registers in mmxsupport (Bruce Oberg)
+  Revised makefile.gcmmx
+  Added PNG_SETJMP_SUPPORTED, PNG_SETJMP_NOT_SUPPORTED, and PNG_ABORT() macros
+version 1.0.5r [February 7, 2000]
+  Removed superfluous prototype for png_get_itxt from png.h
+  Fixed a bug in pngrtran.c that improperly expanded the background color.
+  Return *num_text=0 from png_get_text() when appropriate, and fix documentation
+    of png_get_text() in libpng.txt/libpng.3.
+version 1.0.5s [February 18, 2000]
+  Added "png_jmp_env()" macro to pngconf.h, to help people migrate to the
+    new error handler that's planned for the next libpng release, and changed
+    example.c, pngtest.c, and contrib programs to use this macro.
+  Revised some of the DLL-export macros in pngconf.h (Greg Roelofs)
+  Fixed a bug in png_read_png() that caused it to fail to expand some images
+    that it should have expanded.
+  Fixed some mistakes in the unused and undocumented INCH_CONVERSIONS functions
+    in pngget.c
+  Changed the allocation of palette, history, and trans arrays back to
+    the version 1.0.5 method (linking instead of copying) which restores
+    backward compatibility with version 1.0.5.  Added some remarks about
+    that in example.c.  Added "free_me" member to info_ptr and png_ptr
+    and added png_free_data() function.
+  Updated makefile.linux and makefile.gccmmx to make directories conditionally.
+  Made cosmetic changes to pngasmrd.h
+  Added png_set_rows() and png_get_rows(), for use with png_read|write_png().
+  Modified png_read_png() to allocate info_ptr->row_pointers only if it
+    hasn't already been allocated.
+version 1.0.5t [March 4, 2000]
+  Changed png_jmp_env() migration aiding macro to png_jmpbuf().
+  Fixed "interlace" typo (should be "interlaced") in contrib/gregbook/read2-x.c
+  Fixed bug with use of PNG_BEFORE_IHDR bit in png_ptr->mode, introduced when
+    PNG_FLAG_HAVE_CHUNK_HEADER was moved into png_ptr->mode in version 1.0.5b
+  Files in contrib/gregbook were revised to use png_jmpbuf() and to select
+    a 24-bit visual if one is available, and to allow abbreviated options.
+  Files in contrib/pngminus were revised to use the png_jmpbuf() macro.
+  Removed spaces in makefile.linux and makefile.gcmmx, introduced in 1.0.5s
+version 1.0.5u [March 5, 2000]
+  Simplified the code that detects old png.h in png.c and pngtest.c
+  Renamed png_spalette (_p, _pp) to png_sPLT_t (_tp, _tpp)
+  Increased precision of rgb_to_gray calculations from 8 to 15 bits and
+    added png_set_rgb_to_gray_fixed() function.
+  Added makefile.bc32 (32-bit Borland C++, C mode)
+version 1.0.5v [March 11, 2000]
+  Added some parentheses to the png_jmpbuf macro definition.
+  Updated references to the zlib home page, which has moved to freesoftware.com.
+  Corrected bugs in documentation regarding png_read_row() and png_write_row().
+  Updated documentation of png_rgb_to_gray calculations in libpng.3/libpng.txt.
+  Renamed makefile.borland,turboc3 back to makefile.bor,tc3 as in version 1.0.3,
+    revised borland makefiles; added makefile.ibmvac3 and makefile.gcc (Cosmin)
+version 1.0.6 [March 20, 2000]
+  Minor revisions of makefile.bor, libpng.txt, and gregbook/rpng2-win.c
+  Added makefile.sggcc (SGI IRIX with gcc)
+version 1.0.6d [April 7, 2000]
+  Changed sprintf() to strcpy() in png_write_sCAL_s() to work without STDIO
+  Added data_length parameter to png_decompress_chunk() function
+  Revised documentation to remove reference to abandoned png_free_chnk functions
+  Fixed an error in png_rgb_to_gray_fixed()
+  Revised example.c, usage of png_destroy_write_struct().
+  Renamed makefile.ibmvac3 to makefile.ibmc, added libpng.icc IBM project file
+  Added a check for info_ptr->free_me&PNG_FREE_TEXT when freeing text in png.c
+  Simplify png_sig_bytes() function to remove use of non-ISO-C strdup().
+version 1.0.6e [April 9, 2000]
+  Added png_data_freer() function.
+  In the code that checks for over-length tRNS chunks, added check of
+    info_ptr->num_trans as well as png_ptr->num_trans (Matthias Benckmann)
+  Minor revisions of libpng.txt/libpng.3.
+  Check for existing data and free it if the free_me flag is set, in png_set_*()
+    and png_handle_*().
+  Only define PNG_WEIGHTED_FILTERS_SUPPORTED when PNG_FLOATING_POINT_SUPPORTED
+    is defined.
+  Changed several instances of PNG_NO_CONSOLE_ID to PNG_NO_STDIO in pngrutil.c
+    and mentioned the purposes of the two macros in libpng.txt/libpng.3.
+version 1.0.6f [April 14, 2000]
+  Revised png_set_iCCP() and png_set_rows() to avoid prematurely freeing data.
+  Add checks in png_set_text() for NULL members of the input text structure.
+  Revised libpng.txt/libpng.3.
+  Removed superfluous prototype for png_set_itxt from png.h
+  Removed "else" from pngread.c, after png_error(), and changed "0" to "length".
+  Changed several png_errors about malformed ancillary chunks to png_warnings.
+version 1.0.6g [April 24, 2000]
+  Added png_pass-* arrays to pnggccrd.c when PNG_USE_LOCAL_ARRAYS is defined.
+  Relocated paragraph about png_set_background() in libpng.3/libpng.txt
+    and other revisions (Matthias Benckmann)
+  Relocated info_ptr->free_me, png_ptr->free_me, and other info_ptr and
+    png_ptr members to restore binary compatibility with libpng-1.0.5
+    (breaks compatibility with libpng-1.0.6).
+version 1.0.6h [April 24, 2000]
+  Changed shared library so-number pattern from 2.x.y.z to xy.z (this builds
+    libpng.so.10 & libpng.so.10.6h instead of libpng.so.2 & libpng.so.2.1.0.6h)
+    This is a temporary change for test purposes.
+version 1.0.6i [May 2, 2000]
+  Rearranged some members at the end of png_info and png_struct, to put
+    unknown_chunks_num and free_me within the original size of the png_structs
+    and free_me, png_read_user_fn, and png_free_fn within the original png_info,
+    because some old applications allocate the structs directly instead of
+    using png_create_*().
+  Added documentation of user memory functions in libpng.txt/libpng.3
+  Modified png_read_png so that it will use user_allocated row_pointers
+    if present, unless free_me directs that it be freed, and added description
+    of the use of png_set_rows() and png_get_rows() in libpng.txt/libpng.3.
+  Added PNG_LEGACY_SUPPORTED macro, and #ifdef out all new (since version
+    1.00) members of png_struct and png_info, to regain binary compatibility
+    when you define this macro.  Capabilities lost in this event
+    are user transforms (new in version 1.0.0),the user transform pointer
+    (new in version 1.0.2), rgb_to_gray (new in 1.0.5), iCCP, sCAL, sPLT,
+    the high-level interface, and unknown chunks support (all new in 1.0.6).
+    This was necessary because of old applications that allocate the structs
+    directly as authors were instructed to do in libpng-0.88 and earlier,
+    instead of using png_create_*().
+  Added modes PNG_CREATED_READ_STRUCT and PNG_CREATED_WRITE_STRUCT which
+    can be used to detect codes that directly allocate the structs, and
+    code to check these modes in png_read_init() and png_write_init() and
+    generate a libpng error if the modes aren't set and PNG_LEGACY_SUPPORTED
+    was not defined.
+  Added makefile.intel and updated makefile.watcom (Pawel Mrochen)
+version 1.0.6j [May 3, 2000]
+  Overloaded png_read_init() and png_write_init() with macros that convert
+    calls to png_read_init_2() or png_write_init_2() that check the version
+    and structure sizes.
+version 1.0.7beta11 [May 7, 2000]
+  Removed the new PNG_CREATED_READ_STRUCT and PNG_CREATED_WRITE_STRUCT modes
+    which are no longer used.
+  Eliminated the three new members of png_text when PNG_LEGACY_SUPPORTED is
+    defined or when neither PNG_READ_iTXt_SUPPORTED nor PNG_WRITE_iTXT_SUPPORTED
+    is defined.
+  Made PNG_NO_READ|WRITE_iTXt the default setting, to avoid memory
+    overrun when old applications fill the info_ptr->text structure directly.
+  Added PNGAPI macro, and added it to the definitions of all exported functions.
+  Relocated version macro definitions ahead of the includes of zlib.h and
+    pngconf.h in png.h.
+version 1.0.7beta12 [May 12, 2000]
+  Revised pngset.c to avoid a problem with expanding the png_debug macro.
+  Deleted some extraneous defines from pngconf.h
+  Made PNG_NO_CONSOLE_IO the default condition when PNG_BUILD_DLL is defined.
+  Use MSC _RPTn debugging instead of fprintf if _MSC_VER is defined.
+  Added png_access_version_number() function.
+  Check for mask&PNG_FREE_CHNK (for TEXT, SCAL, PCAL) in png_free_data().
+  Expanded libpng.3/libpng.txt information about png_data_freer().
+version 1.0.7beta14 [May 17, 2000] (beta13 was not published)
+  Changed pnggccrd.c and pngvcrd.c to handle bad adaptive filter types as
+    warnings instead of errors, as pngrutil.c does.
+  Set the PNG_INFO_IDAT valid flag in png_set_rows() so png_write_png()
+    will actually write IDATs.
+  Made the default PNG_USE_LOCAL_ARRAYS depend on PNG_DLL instead of WIN32.
+  Make png_free_data() ignore its final parameter except when freeing data
+    that can have multiple instances (text, sPLT, unknowns).
+  Fixed a new bug in png_set_rows().
+  Removed info_ptr->valid tests from png_free_data(), as in version 1.0.5.
+  Added png_set_invalid() function.
+  Fixed incorrect illustrations of png_destroy_write_struct() in example.c.
+version 1.0.7beta15 [May 30, 2000]
+  Revised the deliberately erroneous Linux setjmp code in pngconf.h to produce
+    fewer error messages.
+  Rearranged checks for Z_OK to check the most likely path first in pngpread.c
+    and pngwutil.c.
+  Added checks in pngtest.c for png_create_*() returning NULL, and mentioned
+    in libpng.txt/libpng.3 the need for applications to check this.
+  Changed names of png_default_*() functions in pngtest to pngtest_*().
+  Changed return type of png_get_x|y_offset_*() from png_uint_32 to png_int_32.
+  Fixed some bugs in the unused PNG_INCH_CONVERSIONS functions in pngget.c
+  Set each pointer to NULL after freeing it in png_free_data().
+  Worked around a problem in pngconf.h; AIX's strings.h defines an "index"
+    macro that conflicts with libpng's png_color_16.index. (Dimitri Papadapoulos)
+  Added "msvc" directory with MSVC++ project files (Simon-Pierre Cadieux).
+version 1.0.7beta16 [June 4, 2000]
+  Revised the workaround of AIX string.h "index" bug.
+  Added a check for overlength PLTE chunk in pngrutil.c.
+  Added PNG_NO_POINTER_INDEXING macro to use array-indexing instead of pointer
+    indexing in pngrutil.c and pngwutil.c to accommodate a buggy compiler.
+  Added a warning in png_decompress_chunk() when it runs out of data, e.g.
+    when it tries to read an erroneous PhotoShop iCCP chunk.
+  Added PNG_USE_DLL macro.
+  Revised the copyright/disclaimer/license notice.
+  Added contrib/msvctest directory
+version 1.0.7rc1 [June 9, 2000]
+  Corrected the definition of PNG_TRANSFORM_INVERT_ALPHA  (0x0400 not 0x0200)
+  Added contrib/visupng directory (Willem van Schaik)
+version 1.0.7beta18 [June 23, 2000]
+  Revised PNGAPI definition, and pngvcrd.c to work with __GCC__
+    and do not redefine PNGAPI if it is passed in via a compiler directive.
+  Revised visupng/PngFile.c to remove returns from within the Try block.
+  Removed leading underscores from "_PNG_H" and "_PNG_SAVE_BSD_SOURCE" macros.
+  Updated contrib/visupng/cexcept.h to version 1.0.0.
+  Fixed bugs in pngwrite.c and pngwutil.c that prevented writing iCCP chunks.
+version 1.0.7rc2 [June 28, 2000]
+  Updated license to include disclaimers required by UCITA.
+  Fixed "DJBPP" typo in pnggccrd.c introduced in beta18.
+version 1.0.7 [July 1, 2000]
+  Revised the definition of "trans_values" in libpng.3/libpng.txt
+version 1.0.8beta1 [July 8, 2000]
+  Added png_free(png_ptr, key) two places in pngpread.c to stop memory leaks.
+  Changed PNG_NO_STDIO to PNG_NO_CONSOLE_IO, several places in pngrutil.c and
+     pngwutil.c.
+  Changed PNG_EXPORT_VAR to use PNG_IMPEXP, in pngconf.h.
+  Removed unused "#include <assert.h>" from png.c
+  Added WindowsCE support.
+  Revised pnggccrd.c to work with gcc-2.95.2 and in the Cygwin environment.
+version 1.0.8beta2 [July 10, 2000]
+  Added project files to the wince directory and made further revisions
+     of pngtest.c, pngrio.c, and pngwio.c in support of WindowsCE.
+version 1.0.8beta3 [July 11, 2000]
+  Only set the PNG_FLAG_FREE_TRNS or PNG_FREE_TRNS flag in png_handle_tRNS()
+     for indexed-color input files to avoid potential double-freeing trans array
+     under some unusual conditions; problem was introduced in version 1.0.6f.
+  Further revisions to pngtest.c and files in the wince subdirectory.
+version 1.0.8beta4 [July 14, 2000]
+  Added the files pngbar.png and pngbar.jpg to the distribution.
+  Added makefile.cygwin, and cygwin support in pngconf.h
+  Added PNG_NO_ZALLOC_ZERO macro (makes png_zalloc skip zeroing memory)
+version 1.0.8rc1 [July 16, 2000]
+  Revised png_debug() macros and statements to eliminate compiler warnings.
+version 1.0.8 [July 24, 2000]
+  Added png_flush() in pngwrite.c, after png_write_IEND().
+  Updated makefile.hpux to build a shared library.
+version 1.0.9beta1 [November 10, 2000]
+  Fixed typo in scripts/makefile.hpux
+  Updated makevms.com in scripts and contrib/* and contrib/* (Martin Zinser)
+  Fixed seqence-point bug in contrib/pngminus/png2pnm (Martin Zinser)
+  Changed "cdrom.com" in documentation to "libpng.org"
+  Revised pnggccrd.c to get it all working, and updated makefile.gcmmx (Greg).
+  Changed type of "params" from voidp to png_voidp in png_read|write_png().
+  Make sure PNGAPI and PNG_IMPEXP are defined in pngconf.h.
+  Revised the 3 instances of WRITEFILE in pngtest.c.
+  Relocated "msvc" and "wince" project subdirectories into "dll" subdirectory.
+  Updated png.rc in dll/msvc project
+  Revised makefile.dec to define and use LIBPATH and INCPATH
+  Increased size of global png_libpng_ver[] array from 12 to 18 chars.
+  Made global png_libpng_ver[], png_sig[] and png_pass_*[] arrays const.
+  Removed duplicate png_crc_finish() from png_handle_bKGD() function.
+  Added a warning when application calls png_read_update_info() multiple times.
+  Revised makefile.cygwin
+  Fixed bugs in iCCP support in pngrutil.c and pngwutil.c.
+  Replaced png_set_empty_plte_permitted() with png_permit_mng_features().
+version 1.0.9beta2 [November 19, 2000]
+  Renamed the "dll" subdirectory "projects".
+  Added borland project files to "projects" subdirectory.
+  Set VS_FF_PRERELEASE and VS_FF_PATCHED flags in msvc/png.rc when appropriate.
+  Add error message in png_set_compression_buffer_size() when malloc fails.
+version 1.0.9beta3 [November 23, 2000]
+  Revised PNG_LIBPNG_BUILD_TYPE macro in png.h, used in the msvc project.
+  Removed the png_flush() in pngwrite.c that crashes some applications
+    that don't set png_output_flush_fn.
+  Added makefile.macosx and makefile.aix to scripts directory.
+version 1.0.9beta4 [December 1, 2000]
+  Change png_chunk_warning to png_warning in png_check_keyword().
+  Increased the first part of msg buffer from 16 to 18 in png_chunk_error().
+version 1.0.9beta5 [December 15, 2000]
+  Added support for filter method 64 (for PNG datastreams embedded in MNG).
+version 1.0.9beta6 [December 18, 2000]
+  Revised png_set_filter() to accept filter method 64 when appropriate.
+  Added new PNG_HAVE_PNG_SIGNATURE bit to png_ptr->mode and use it to
+    help prevent applications from using MNG features in PNG datastreams.
+  Added png_permit_mng_features() function.
+  Revised libpng.3/libpng.txt.  Changed "filter type" to "filter method".
+version 1.0.9rc1 [December 23, 2000]
+  Revised test for PNG_HAVE_PNG_SIGNATURE in pngrutil.c
+  Fixed error handling of unknown compression type in png_decompress_chunk().
+  In pngconf.h, define __cdecl when _MSC_VER is defined.
+version 1.0.9beta7 [December 28, 2000]
+  Changed PNG_TEXT_COMPRESSION_zTXt to PNG_COMPRESSION_TYPE_BASE several places.
+  Revised memory management in png_set_hIST and png_handle_hIST in a backward
+    compatible manner.  PLTE and tRNS were revised similarly.
+  Revised the iCCP chunk reader to ignore trailing garbage.
+version 1.0.9beta8 [January 12, 2001]
+  Moved pngasmrd.h into pngconf.h.
+  Improved handling of out-of-spec garbage iCCP chunks generated by PhotoShop.
+version 1.0.9beta9 [January 15, 2001]
+  Added png_set_invalid, png_permit_mng_features, and png_mmx_supported to
+    wince and msvc project module definition files.
+  Minor revision of makefile.cygwin.
+  Fixed bug with progressive reading of narrow interlaced images in pngpread.c
+version 1.0.9beta10 [January 16, 2001]
+  Do not typedef png_FILE_p in pngconf.h when PNG_NO_STDIO is defined.
+  Fixed "png_mmx_supported" typo in project definition files.
+version 1.0.9beta11 [January 19, 2001]
+  Updated makefile.sgi to make shared library.
+  Removed png_mmx_support() function and disabled PNG_MNG_FEATURES_SUPPORTED
+    by default, for the benefit of DLL forward compatibility.  These will
+    be re-enabled in version 1.2.0.
+version 1.0.9rc2 [January 22, 2001]
+  Revised cygwin support.
+version 1.0.9 [January 31, 2001]
+  Added check of cygwin's ALL_STATIC in pngconf.h
+  Added "-nommx" parameter to contrib/gregbook/rpng2-win and rpng2-x demos.
+version 1.0.10beta1 [March 14, 2001]
+  Revised makefile.dec, makefile.sgi, and makefile.sggcc; added makefile.hpgcc.
+  Reformatted libpng.3 to eliminate bad line breaks.
+  Added checks for _mmx_supported in the read_filter_row function of pnggccrd.c
+  Added prototype for png_mmx_support() near the top of pnggccrd.c
+  Moved some error checking from png_handle_IHDR to png_set_IHDR.
+  Added PNG_NO_READ_SUPPORTED and PNG_NO_WRITE_SUPPORTED macros.
+  Revised png_mmx_support() function in pnggccrd.c
+  Restored version 1.0.8 PNG_WRITE_EMPTY_PLTE_SUPPORTED behavior in pngwutil.c
+  Fixed memory leak in contrib/visupng/PngFile.c
+  Fixed bugs in png_combine_row() in pnggccrd.c and pngvcrd.c (C version)
+  Added warnings when retrieving or setting gamma=0.
+  Increased the first part of msg buffer from 16 to 18 in png_chunk_warning().
+version 1.0.10rc1 [March 23, 2001]
+  Changed all instances of memcpy, strcpy, and strlen to png_memcpy, png_strcpy,
+    and png_strlen.
+  Revised png_mmx_supported() function in pnggccrd.c to return proper value.
+  Fixed bug in progressive reading (pngpread.c) with small images (height < 8).
+version 1.0.10 [March 30, 2001]
+  Deleted extraneous space (introduced in 1.0.9) from line 42 of makefile.cygwin
+  Added beos project files (Chris Herborth)
+version 1.0.11beta1 [April 3, 2001]
+  Added type casts on several png_malloc() calls (Dimitri Papadapoulos).
+  Removed a no-longer needed AIX work-around from pngconf.h
+  Changed several "//" single-line comments to C-style in pnggccrd.c
+version 1.0.11beta2 [April 11, 2001]
+  Removed PNGAPI from several functions whose prototypes did not have PNGAPI.
+  Updated scripts/pngos2.def
+version 1.0.11beta3 [April 14, 2001]
+  Added checking the results of many instances of png_malloc() for NULL
+version 1.0.11beta4 [April 20, 2001]
+  Undid the changes from version 1.0.11beta3.  Added a check for NULL return
+    from user's malloc_fn().
+  Removed some useless type casts of the NULL pointer.
+  Added makefile.netbsd
+version 1.0.11 [April 27, 2001]
+  Revised makefile.netbsd
+version 1.0.12beta1 [May 14, 2001]
+  Test for Windows platform in pngconf.h when including malloc.h (Emmanuel Blot)
+  Updated makefile.cygwin and handling of Cygwin's ALL_STATIC in pngconf.h
+  Added some never-to-be-executed code in pnggccrd.c to quiet compiler warnings.
+  Eliminated the png_error about apps using png_read|write_init().  Instead,
+    libpng will reallocate the png_struct and info_struct if they are too small.
+    This retains future binary compatibility for old applications written for
+    libpng-0.88 and earlier.
+version 1.2.0beta1 [May 6, 2001]
+  Bumped DLLNUM to 2.
+  Re-enabled PNG_MNG_FEATURES_SUPPORTED and enabled PNG_ASSEMBLER_CODE_SUPPORTED
+    by default.
+  Added runtime selection of MMX features.
+  Added png_set_strip_error_numbers function and related macros.
+version 1.2.0beta2 [May 7, 2001]
+  Finished merging 1.2.0beta1 with version 1.0.11
+  Added a check for attempts to read or write PLTE in grayscale PNG datastreams.
+version 1.2.0beta3 [May 17, 2001]
+  Enabled user memory function by default.
+  Modified png_create_struct so it passes user mem_ptr to user memory allocator.
+  Increased png_mng_features flag from png_byte to png_uint_32.
+  Bumped shared-library (so-number) and dll-number to 3.
+version 1.2.0beta4 [June 23, 2001]
+  Check for missing profile length field in iCCP chunk and free chunk_data
+     in case of truncated iCCP chunk.
+  Bumped shared-library number to 3 in makefile.sgi and makefile.sggcc
+  Bumped dll-number from 2 to 3 in makefile.cygwin
+  Revised contrib/gregbook/rpng*-x.c to avoid a memory leak and to exit cleanly
+     if user attempts to run it on an 8-bit display.
+  Updated contrib/gregbook
+  Use png_malloc instead of png_zalloc to allocate palette in pngset.c
+  Updated makefile.ibmc
+  Added some typecasts to eliminate gcc 3.0 warnings.  Changed prototypes
+     of png_write_oFFS width and height from png_uint_32 to png_int_32.
+  Updated example.c
+  Revised prototypes for png_debug_malloc and png_debug_free in pngtest.c
+version 1.2.0beta5 [August 8, 2001]
+  Revised contrib/gregbook
+  Revised makefile.gcmmx
+  Revised pnggccrd.c to conditionally compile some thread-unsafe code only
+     when PNG_THREAD_UNSAFE_OK is defined.
+  Added tests to prevent pngwutil.c from writing a bKGD or tRNS chunk with
+     value exceeding 2^bit_depth-1
+  Revised makefile.sgi and makefile.sggcc
+  Replaced calls to fprintf(stderr,...) with png_warning() in pnggccrd.c
+  Removed restriction that do_invert_mono only operate on 1-bit opaque files
+version 1.2.0 [September 1, 2001]
+  Changed a png_warning() to png_debug() in pnggccrd.c
+  Fixed contrib/gregbook/rpng-x.c, rpng2-x.c to avoid crash with XFreeGC().
+version 1.2.1beta1 [October 19, 2001]
+  Revised makefile.std in contrib/pngminus
+  Include background_1 in png_struct regardless of gamma support.
+  Revised makefile.netbsd and makefile.macosx, added makefile.darwin.
+  Revised example.c to provide more details about using row_callback().
+version 1.2.1beta2 [October 25, 2001]
+  Added type cast to each NULL appearing in a function call, except for
+    WINCE functions.
+  Added makefile.so9.
+version 1.2.1beta3 [October 27, 2001]
+  Removed type casts from all NULLs.
+  Simplified png_create_struct_2().
+version 1.2.1beta4 [November 7, 2001]
+  Revised png_create_info_struct() and png_creat_struct_2().
+  Added error message if png_write_info() was omitted.
+  Type cast NULLs appearing in function calls when _NO_PROTO or
+    PNG_TYPECAST_NULL is defined.
+version 1.2.1rc1 [November 24, 2001]
+  Type cast NULLs appearing in function calls except when PNG_NO_TYPECAST_NULL
+    is defined.
+  Changed typecast of "size" argument to png_size_t in pngmem.c calls to
+    the user malloc_fn, to agree with the prototype in png.h
+  Added a pop/push operation to pnggccrd.c, to preserve Eflag (Maxim Sobolev)
+  Updated makefile.sgi to recognize LIBPATH and INCPATH.
+  Updated various makefiles so "make clean" does not remove previous major
+    version of the shared library.
+version 1.2.1rc2 [December 4, 2001]
+  Added a pop/push operation to pngvcrd.c, to preserve Eflag.
+  Always allocate 256-entry internal palette, hist, and trans arrays, to
+    avoid out-of-bounds memory reference caused by invalid PNG datastreams.
+  Added a check for prefix_length > data_length in iCCP chunk handler.
+version 1.2.1 [December 12, 2001]
+  None.
+
+Send comments/corrections/commendations to
+png-implement@ccrc.wustl.edu or to randeg@alum.rpi.edu
+
+Glenn R-P
diff --git a/png/INSTALL b/png/INSTALL
new file mode 100644
index 000000000..5a7d4bba6
--- /dev/null
+++ b/png/INSTALL
@@ -0,0 +1,143 @@
+
+Installing libpng version 1.2.1 - December 12, 2001
+
+Before installing libpng, you must first install zlib.  zlib
+can usually be found wherever you got libpng.  zlib can be
+placed in another directory, at the same level as libpng.
+Note that your system might already have a preinstalled
+zlib, but you will still need to have access to the
+zlib.h and zconf.h include files that correspond to the
+version of zlib that's installed.
+
+You can rename the directories that you downloaded (they
+might be called "libpng-1.2.1" or "lpng109" and "zlib-1.1.3"
+or "zlib113") so that you have directories called "zlib" and "libpng".
+
+Your directory structure should look like this:
+
+   ..       (the parent directory)
+      libpng  (this directory)
+          INSTALL (this file)
+          README
+          *.h
+          *.c
+          contrib
+             gregbook
+             msvctest
+             pngminus
+             pngsuite
+             visupng
+          projects
+             beos
+             borland
+             msvc
+             netware.txt
+             wince.txt
+          scripts
+             makefile.*
+          pngtest.png
+          etc.
+      zlib
+          README
+          *.h
+          *.c
+          contrib
+          etc.
+
+If the line endings in the files look funny, you may wish to get the other
+distribution of libpng.  It is available in both tar.gz (UNIX style line
+endings) and zip (DOS style line endings) formats.
+
+If you are building libpng with MSVC, you can enter the libpng\msvc directory
+and follow the instructions in msvc\README.txt.
+
+You can build libpng for WindowsCE by entering the downloading and installing
+the libpng\wince directory as instructed in the projects\wince.txt file, and
+then following the instructions in the README* files.  Similarly, you can
+build libpng for Netware as instructed in projects\netware.txt.
+
+Else enter the zlib directory and follow the instructions in zlib/README,
+then come back here and choose the appropriate makefile.sys in the scripts
+directory.
+
+The files that are presently available in the scripts directory
+include
+
+ makefile.std      =>  Generic UNIX makefile (cc, creates static libpng.a)
+ makefile.linux    =>  Linux/ELF makefile (gcc, creates libpng.so.3.1.2.1)
+ makefile.gcmmx    =>  Linux/ELF makefile (gcc, creates libpng.so.3.1.2.1,
+                       uses assembler code tuned for Intel MMX platform)
+ makefile.gcc      =>  Generic makefile (gcc, creates static libpng.a)
+ makefile.knr      =>  Archaic UNIX Makefile that converts files with
+                       ansi2knr (Requires ansi2knr.c from
+                       ftp://ftp.cs.wisc.edu/ghost)
+ makefile.aix      =>  AIX makefile
+ makefile.cygwin   =>  Cygwin/gcc makefile
+ makefile.darwin   =>  Darwin makefile
+ makefile.dec      =>  DEC Alpha UNIX makefile
+ makefile.hpgcc    =>  HPUX makefile using gcc
+ makefile.hpux     =>  HPUX (10.20 and 11.00) makefile
+ makefile.ibmc     =>  IBM C/C++ version 3.x for Win32 and OS/2 (static)
+ makefile.intel    =>  Intel C/C++ version 4.0 and later
+ libpng.icc        =>  Project file for IBM VisualAge/C++ version 4.0 or later
+ makefile.macosx   =>  MACOS X Makefile
+ makefile.netbsd   =>  NetBSD/cc makefile, uses PNGGCCRD
+ makefile.sgi      =>  Silicon Graphics IRIX makefile (cc, creates static lib)
+ makefile.sggcc    =>  Silicon Graphics (gcc, creates libpng.so.3.1.2.1)
+ makefile.sunos    =>  Sun makefile
+ makefile.solaris  =>  Solaris 2.X makefile (gcc, creates libpng.so.3.1.2.1)
+ makefile.so9      =>  Solaris 9 makefile (gcc, creates libpng.so.3.1.2.1)
+ makefile.sco      =>  For SCO OSr5  ELF and Unixware 7 with Native cc
+ makefile.mips     =>  MIPS makefile
+ makefile.acorn    =>  Acorn makefile
+ makefile.amiga    =>  Amiga makefile
+ smakefile.ppc     =>  AMIGA smakefile for SAS C V6.58/7.00 PPC compiler
+                       (Requires SCOPTIONS, copied from scripts/SCOPTIONS.ppc)
+ makefile.atari    =>  Atari makefile
+ makefile.beos     =>  BEOS makefile for X86
+ makefile.bor      =>  Borland makefile (uses bcc)
+ makefile.bc32     =>  32-bit Borland C++ (all modules compiled in C mode)
+ makefile.bd32     =>  To make a png32bd.dll with Borland C++ 4.5
+ makefile.tc3      =>  Turbo C 3.0 makefile
+ makefile.dj2      =>  DJGPP 2 makefile
+ makefile.msc      =>  Microsoft C makefile
+ makefile.vcawin32 =>  makefile for Microsoft Visual C++ 5.0 and later (uses
+                       assembler code tuned for Intel MMX platform)
+ makefile.vcwin32  =>  makefile for Microsoft Visual C++ 4.0 and later (does
+                       not use assembler code)
+ makefile.os2      =>  OS/2 Makefile (gcc and emx, requires pngos2.def)
+ pngos2.def        =>  OS/2 module definition file used by makefile.os2
+ makefile.watcom   =>  Watcom 10a+ Makefile, 32-bit flat memory model
+ makevms.com       =>  VMS build script
+ descrip.mms       =>  VMS makefile for MMS or MMK
+ pngdef.pas        =>  Defines for a png32bd.dll with Borland C++ 4.5
+ SCOPTIONS.ppc     =>  Used with smakefile.ppc
+
+Copy the file (or files) that you need from the
+scripts directory into this directory, for example
+
+   MSDOS example: copy scripts\makefile.msc makefile
+   UNIX example:    cp scripts/makefile.std makefile
+
+Read the makefile to see if you need to change any source or
+target directories to match your preferences.
+
+Then read pngconf.h to see if you want to make any configuration
+changes.
+
+Then just run "make test" which will create the libpng library in
+this directory and run a quick test that reads the "pngtest.png"
+file and writes a "pngout.png" file that should be identical to it.
+Look for "9782 zero samples" in the output of the test.  For more
+confidence, you can run another test by typing "pngtest pngnow.png"
+and looking for "289 zero samples" in the output.  Also, you can
+run "pngtest -m *.png" in the "contrib/pngsuite" directory and compare
+your output with the result shown in contrib/pngsuite/README.
+
+Most of the makefiles will allow you to run "make install" to
+put the library in its final resting place (if you want to
+do that, run "make install" in the zlib directory first if necessary).
+
+Further information can be found in the README and libpng.txt
+files, in the individual makefiles, in png.h, in the README files in
+subdirectories of the LIB directory, and the manual pages libpng.3 and png.5.
diff --git a/png/KNOWNBUG b/png/KNOWNBUG
new file mode 100644
index 000000000..7844cbc01
--- /dev/null
+++ b/png/KNOWNBUG
@@ -0,0 +1,9 @@
+
+Known bugs in libpng version 1.2.1
+
+1. April 22, 2001: pnggccrd.c has been reported to crash on NetBSD when
+   reading interlaced PNG files, when assembler code is enabled.
+
+   STATUS: Under investigation.  The change to pnggccrd.c in libpng-1.2.1
+   fixed a problem under FreeBSD but not the problem with NetBSD.
+
diff --git a/png/LICENSE b/png/LICENSE
new file mode 100644
index 000000000..c11dd4fda
--- /dev/null
+++ b/png/LICENSE
@@ -0,0 +1,102 @@
+
+This copy of the libpng notices is provided for your convenience.  In case of
+any discrepancy between this copy and the notices in the file png.h that is
+included in the libpng distribution, the latter shall prevail.
+
+COPYRIGHT NOTICE, DISCLAIMER, and LICENSE:
+
+If you modify libpng you may insert additional notices immediately following
+this sentence.
+
+libpng versions 1.0.7, July 1, 2000, through 1.2.1, December 12, 2001, are
+Copyright (c) 2000 Glenn Randers-Pehrson
+and are distributed according to the same disclaimer and license as libpng-1.0.6
+with the following individuals added to the list of Contributing Authors
+
+   Simon-Pierre Cadieux
+   Eric S. Raymond
+   Gilles Vollant
+
+and with the following additions to the disclaimer:
+
+   There is no warranty against interference with your enjoyment of the
+   library or against infringement.  There is no warranty that our
+   efforts or the library will fulfill any of your particular purposes
+   or needs.  This library is provided with all faults, and the entire
+   risk of satisfactory quality, performance, accuracy, and effort is with
+   the user.
+
+libpng versions 0.97, January 1998, through 1.0.6, March 20, 2000, are
+Copyright (c) 1998, 1999 Glenn Randers-Pehrson, and are
+distributed according to the same disclaimer and license as libpng-0.96,
+with the following individuals added to the list of Contributing Authors:
+
+   Tom Lane
+   Glenn Randers-Pehrson
+   Willem van Schaik
+
+libpng versions 0.89, June 1996, through 0.96, May 1997, are
+Copyright (c) 1996, 1997 Andreas Dilger
+Distributed according to the same disclaimer and license as libpng-0.88,
+with the following individuals added to the list of Contributing Authors:
+
+   John Bowler
+   Kevin Bracey
+   Sam Bushell
+   Magnus Holmgren
+   Greg Roelofs
+   Tom Tanner
+
+libpng versions 0.5, May 1995, through 0.88, January 1996, are
+Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.
+
+For the purposes of this copyright and license, "Contributing Authors"
+is defined as the following set of individuals:
+
+   Andreas Dilger
+   Dave Martindale
+   Guy Eric Schalnat
+   Paul Schmidt
+   Tim Wegner
+
+The PNG Reference Library is supplied "AS IS".  The Contributing Authors
+and Group 42, Inc. disclaim all warranties, expressed or implied,
+including, without limitation, the warranties of merchantability and of
+fitness for any purpose.  The Contributing Authors and Group 42, Inc.
+assume no liability for direct, indirect, incidental, special, exemplary,
+or consequential damages, which may result from the use of the PNG
+Reference Library, even if advised of the possibility of such damage.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+source code, or portions hereof, for any purpose, without fee, subject
+to the following restrictions:
+
+1. The origin of this source code must not be misrepresented.
+
+2. Altered versions must be plainly marked as such and must not
+   be misrepresented as being the original source.
+
+3. This Copyright notice may not be removed or altered from any
+   source or altered source distribution.
+
+The Contributing Authors and Group 42, Inc. specifically permit, without
+fee, and encourage the use of this source code as a component to
+supporting the PNG file format in commercial products.  If you use this
+source code in a product, acknowledgment is not required but would be
+appreciated.
+
+
+A "png_get_copyright" function is available, for convenient use in "about"
+boxes and the like:
+
+   printf("%s",png_get_copyright(NULL));
+
+Also, the PNG logo (in PNG format, of course) is supplied in the
+files "pngbar.png" and "pngbar.jpg (88x31) and "pngnow.png" (98x31).
+
+Libpng is OSI Certified Open Source Software.  OSI Certified Open Source is a
+certification mark of the Open Source Initiative.
+
+Glenn Randers-Pehrson
+randeg@alum.rpi.edu
+December 12, 2001
diff --git a/png/Makefile b/png/Makefile
new file mode 100644
index 000000000..57ec21ed7
--- /dev/null
+++ b/png/Makefile
@@ -0,0 +1,78 @@
+#
+# "$Id: Makefile,v 1.1.2.1 2004/07/05 14:12:37 easysw Exp $"
+#
+# PNG library makefile for the Fast Light Toolkit (FLTK).
+#
+# Copyright 1997-2004 by Easy Software Products.
+#
+# This library is free software; you can redistribute it and/or
+# modify it under the terms of the GNU Library General Public
+# License as published by the Free Software Foundation; either
+# version 2 of the License, or (at your option) any later version.
+#
+# This library is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# Library General Public License for more details.
+#
+# You should have received a copy of the GNU Library General Public
+# License along with this library; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+# USA.
+#
+# Please report all bugs and problems to "fltk-bugs@fltk.org".
+#
+
+include ../makeinclude
+
+#
+# Object files...
+#
+
+OBJS	=	png.o pngset.o pngget.o pngrutil.o pngtrans.o pngwutil.o \
+		pngread.o pngrio.o pngwio.o pngwrite.o pngrtran.o \
+		pngwtran.o pngmem.o pngerror.o pngpread.o
+
+
+#
+# Make all of the targets...
+#
+
+all:	libpng.a
+
+
+#
+# Clean all of the targets and object files...
+#
+
+clean:
+	$(RM) $(OBJS)
+	$(RM) libpng.a
+
+
+#
+# libpng.a
+#
+
+libpng.a:	$(OBJS)
+	echo Archiving $@...
+	$(RM) $@
+	$(AR) $(ARFLAGS) $@  $(OBJS)
+	$(RANLIB) $@
+
+
+#
+# Make dependencies...
+#
+
+depend:	$(OBJS:.o=.c)
+	makedepend -Y -I.. -f makedepend $(OBJS:.o=.c)
+
+include makedepend
+
+$(OBJS):	../makeinclude
+
+
+#
+# End of "$Id: Makefile,v 1.1.2.1 2004/07/05 14:12:37 easysw Exp $".
+#
diff --git a/png/README b/png/README
new file mode 100644
index 000000000..ffaae47d5
--- /dev/null
+++ b/png/README
@@ -0,0 +1,254 @@
+README for libpng 1.2.1 - December 12, 2001 (shared library 2.1)
+See the note about version numbers near the top of png.h
+
+See INSTALL for instructions on how to install libpng.
+
+Libpng comes in two distribution formats.  Get libpng-*.tar.gz if you
+want UNIX-style line endings in the text files, or lpng*.zip if you want
+DOS-style line endings.
+
+Version 0.89 was the first official release of libpng.  Don't let the
+fact that it's the first release fool you.  The libpng library has been in
+extensive use and testing since mid-1995.  By late 1997 it had
+finally gotten to the stage where there hadn't been significant
+changes to the API in some time, and people have a bad feeling about
+libraries with versions < 1.0.  Version 1.0.0 was released in
+March 1998.
+
+****
+Note that some of the changes to the png_info structure render this
+version of the library binary incompatible with libpng-0.89 or
+earlier versions if you are using a shared library.  The type of the
+"filler" parameter for png_set_filler() has changed from png_byte to
+png_uint_32, which will affect shared-library applications that use
+this function.
+
+To avoid problems with changes to the internals of png_info_struct,
+new APIs have been made available in 0.95 to avoid direct application
+access to info_ptr.  These functions are the png_set_<chunk> and
+png_get_<chunk> functions.  These functions should be used when
+accessing/storing the info_struct data, rather than manipulating it
+directly, to avoid such problems in the future.
+
+It is important to note that the APIs do not make current programs
+that access the info struct directly incompatible with the new
+library.  However, it is strongly suggested that new programs use
+the new APIs (as shown in example.c and pngtest.c), and older programs
+be converted to the new format, to facilitate upgrades in the future.
+****
+
+Additions since 0.90 include the ability to compile libpng as a
+Windows DLL, and new APIs for accessing data in the info struct.
+Experimental functions include the ability to set weighting and cost
+factors for row filter selection, direct reads of integers from buffers
+on big-endian processors that support misaligned data access, faster
+methods of doing alpha composition, and more accurate 16->8 bit color
+conversion.
+
+The additions since 0.89 include the ability to read from a PNG stream
+which has had some (or all) of the signature bytes read by the calling
+application.  This also allows the reading of embedded PNG streams that
+do not have the PNG file signature.  As well, it is now possible to set
+the library action on the detection of chunk CRC errors.  It is possible
+to set different actions based on whether the CRC error occurred in a
+critical or an ancillary chunk.
+
+The changes made to the library, and bugs fixed are based on discussions
+on the PNG implementation mailing list <png-implement@ccrc.wustl.edu>
+and not on material submitted privately to Guy, Andreas, or Glenn.  They will
+forward any good suggestions to the list.
+
+For a detailed description on using libpng, read libpng.txt.  For
+examples of libpng in a program, see example.c and pngtest.c.  For usage
+information and restrictions (what little they are) on libpng, see
+png.h.  For a description on using zlib (the compression library used by
+libpng) and zlib's restrictions, see zlib.h
+
+I have included a general makefile, as well as several machine and
+compiler specific ones, but you may have to modify one for your own needs.
+
+You should use zlib 1.0.4 or later to run this, but it MAY work with
+versions as old as zlib 0.95.  Even so, there are bugs in older zlib
+versions which can cause the output of invalid compression streams for
+some images.  You will definitely need zlib 1.0.4 or later if you are
+taking advantage of the MS-DOS "far" structure allocation for the small
+and medium memory models.  You should also note that zlib is a
+compression library that is useful for more things than just PNG files.
+You can use zlib as a drop-in replacement for fread() and fwrite() if
+you are so inclined.
+
+zlib should be available at the same place that libpng is.
+If not, it should be at ftp.uu.net in /graphics/png
+Eventually, it will be at ftp.uu.net in /pub/archiving/zip/zlib
+
+You may also want a copy of the PNG specification.  It is available
+as an RFC and a W3C Recommendation.  Failing
+these resources you can try ftp.uu.net in the /graphics/png directory.
+
+This code is currently being archived at ftp.uu.net in the
+/graphics/png directory, and on CompuServe, Lib 20 (PNG SUPPORT)
+at GO GRAPHSUP.  If you can't find it in any of those places,
+e-mail me, and I'll help you find it.
+
+If you have any code changes, requests, problems, etc., please e-mail
+them to me.  Also, I'd appreciate any make files or project files,
+and any modifications you needed to make to get libpng to compile,
+along with a #define variable to tell what compiler/system you are on.
+If you needed to add transformations to libpng, or wish libpng would
+provide the image in a different way, drop me a note (and code, if
+possible), so I can consider supporting the transformation.
+Finally, if you get any warning messages when compiling libpng
+(note: not zlib), and they are easy to fix, I'd appreciate the
+fix.  Please mention "libpng" somewhere in the subject line.  Thanks.
+
+This release was created and will be supported by myself (of course
+based in a large way on Guy's and Andreas' earlier work), and the PNG group.
+
+randeg@alum.rpi.edu
+png-implement@ccrc.wustl.edu
+
+You can't reach Guy, the original libpng author, at the addresses
+given in previous versions of this document.  He and Andreas will read mail
+addressed to the png-implement list, however.
+
+Please do not send general questions about PNG.  Send them to
+the address in the specification (png-group@w3.org).  At the same
+time, please do not send libpng questions to that address, send them to me
+or to png-implement@ccrc.wustl.edu.  I'll
+get them in the end anyway.  If you have a question about something
+in the PNG specification that is related to using libpng, send it
+to me.  Send me any questions that start with "I was using libpng,
+and ...".  If in doubt, send questions to me.  I'll bounce them
+to others, if necessary.
+
+Please do not send suggestions on how to change PNG.  We have
+been discussing PNG for three years now, and it is official and
+finished.  If you have suggestions for libpng, however, I'll
+gladly listen.  Even if your suggestion is not used for version
+1.0, it may be used later.
+
+Files in this distribution:
+
+      ANNOUNCE      =>  Announcement of this version, with recent changes
+      CHANGES       =>  Description of changes between libpng versions
+      KNOWNBUG      =>  List of known bugs and deficiencies
+      LICENSE       =>  License to use and redistribute libpng
+      README        =>  This file
+      TODO          =>  Things not implemented in the current library
+      Y2KINFO       =>  Statement of Y2K compliance
+      example.c     =>  Example code for using libpng functions
+      libpng.3      =>  manual page for libpng (includes libpng.txt)
+      libpng.txt    =>  Description of libpng and its functions
+      libpngpf.3    =>  manual page for libpng's private functions
+      png.5         =>  manual page for the PNG format
+      png.c         =>  Basic interface functions common to library
+      png.h         =>  Library function and interface declarations
+      pngconf.h     =>  System specific library configuration
+      pngasmrd.h    =>  Header file for assembler-coded functions
+      pngerror.c    =>  Error/warning message I/O functions
+      pngget.c      =>  Functions for retrieving info from struct
+      pngmem.c      =>  Memory handling functions
+      pngbar.png    =>  PNG logo, 88x31
+      pngnow.png    =>  PNG logo, 98x31
+      pngpread.c    =>  Progressive reading functions
+      pngread.c     =>  Read data/helper high-level functions
+      pngrio.c      =>  Lowest-level data read I/O functions
+      pngrtran.c    =>  Read data transformation functions
+      pngrutil.c    =>  Read data utility functions
+      pngset.c      =>  Functions for storing data into the info_struct
+      pngtest.c     =>  Library test program
+      pngtest.png   =>  Library test sample image
+      pngtrans.c    =>  Common data transformation functions
+      pngwio.c      =>  Lowest-level write I/O functions
+      pngwrite.c    =>  High-level write functions
+      pngwtran.c    =>  Write data transformations
+      pngwutil.c    =>  Write utility functions
+      contrib       =>  Contributions
+       gregbook         =>  source code for PNG reading and writing, from
+                            Greg Roelofs' "PNG: The Definitive Guide",
+                            O'Reilly, 1999
+       msvctest     =>  Builds and runs pngtest using a MSVC workspace
+       pngminus     =>  Simple pnm2png and png2pnm programs
+       pngsuite     =>  Test images
+       visupng      =>  Contains a MSVC workspace for VisualPng
+      projects      =>  Contains project files and workspaces for building DLL
+       beos             =>  Contains a Beos workspace for building libpng
+       borland          =>  Contains a Borland workspace for building libpng
+                            and zlib
+       msvc             =>  Contains a Microsoft Visual C++ (MSVC) workspace
+                            for building libpng and zlib
+       netware.txt      =>  Contains instructions for downloading a set of
+                            project files for building libpng and zlib on
+                            Netware.
+       wince.txt        =>  Contains instructions for downloading a Microsoft
+                            Visual C++ (Windows CD Toolkit) workspace for
+                            building libpng and zlib on WindowsCE
+      scripts       =>  Directory containing scripts for building libpng:
+       descrip.mms      =>  VMS makefile for MMS or MMK
+       makefile.std     =>  Generic UNIX makefile (cc, creates static libpng.a)
+       makefile.linux   =>  Linux/ELF makefile
+                            (gcc, creates libpng.so.3.1.2.1)
+       makefile.gcmmx   =>  Linux/ELF makefile (gcc, creates
+                            libpng.so.3.1.2.1, uses assembler code
+                            tuned for Intel MMX platform)
+       makefile.gcc     =>  Generic makefile (gcc, creates static libpng.a)
+       makefile.knr     =>  Archaic UNIX Makefile that converts files with
+                            ansi2knr (Requires ansi2knr.c from
+                            ftp://ftp.cs.wisc.edu/ghost)
+       makefile.aix     =>  AIX makefile
+       makefile.cygwin  =>  Cygwin/gcc makefile
+       makefile.darwin  =>  Darwin makefile
+       makefile.dec     =>  DEC Alpha UNIX makefile
+       makefile.hpgcc   =>  HPUX makefile using gcc
+       makefile.hpux    =>  HPUX (10.20 and 11.00) makefile
+       makefile.ibmc    =>  IBM C/C++ version 3.x for Win32 and OS/2 (static)
+       makefile.intel   =>  Intel C/C++ version 4.0 and later
+       libpng.icc       =>  Project file, IBM VisualAge/C++ 4.0 or later
+       makefile.macosx  =>  MACOS X Makefile
+       makefile.netbsd  =>  NetBSD/cc makefile, uses PNGGCCRD
+       makefile.sgi     =>  Silicon Graphics IRIX (cc, creates static lib)
+       makefile.sggcc   =>  Silicon Graphics (gcc, creates libpng.so.3.1.2.1)
+       makefile.sunos   =>  Sun makefile
+       makefile.solaris =>  Solaris 2.X makefile
+                            (gcc, creates libpng.so.3.1.2.1)
+       makefile.so9     =>  Solaris 9 makefile
+                            (gcc, creates libpng.so.3.1.2.1)
+       makefile.sco     =>  For SCO OSr5  ELF and Unixware 7 with Native cc
+       makefile.mips    =>  MIPS makefile
+       makefile.acorn   =>  Acorn makefile
+       makefile.amiga   =>  Amiga makefile
+       smakefile.ppc    =>  AMIGA smakefile for SAS C V6.58/7.00 PPC
+                            compiler (Requires SCOPTIONS, copied from
+                            scripts/SCOPTIONS.ppc)
+       makefile.atari   =>  Atari makefile
+       makefile.beos    =>  BEOS makefile for X86
+       makefile.bor     =>  Borland makefile (uses bcc)
+       makefile.bc32    =>  32-bit Borland C++ (all modules compiled in C mode)
+       makefile.bd32    =>  To make a png32bd.dll with Borland C++ 4.5
+       makefile.tc3     =>  Turbo C 3.0 makefile
+       makefile.dj2     =>  DJGPP 2 makefile
+       makefile.msc     =>  Microsoft C makefile
+       makefile.vcawin32 => makefile for Microsoft Visual C++ 5.0 and
+                            later (uses assembler code tuned for Intel MMX
+                            platform)
+       makefile.vcwin32 =>  makefile for Microsoft Visual C++ 4.0 and
+                            later (does not use assembler code)
+       makefile.os2     =>  OS/2 Makefile (gcc and emx, requires pngos2.def)
+       pngos2.def       =>  OS/2 module definition file used by makefile.os2
+       makefile.watcom  =>  Watcom 10a+ Makefile, 32-bit flat memory model
+       makevms.com      =>  VMS build script
+       pngdef.pas       =>  Defines for a png32bd.dll with Borland C++ 4.5
+       SCOPTIONS.ppc    =>  Used with smakefile.ppc
+
+Good luck, and happy coding.
+
+-Glenn Randers-Pehrson
+ Internet: randeg@alum.rpi.edu
+
+-Andreas Eric Dilger
+ Internet: adilger@enel.ucalgary.ca
+ Web: http://www-mddsp.enel.ucalgary.ca/People/adilger/
+
+-Guy Eric Schalnat
+ (formerly of Group 42, Inc)
+ Internet: gschal@infinet.com
diff --git a/png/TODO b/png/TODO
new file mode 100644
index 000000000..a5f639577
--- /dev/null
+++ b/png/TODO
@@ -0,0 +1,24 @@
+TODO - list of things to do for libpng:
+
+Final bug fixes.
+Improve API by hiding the png_struct and png_info structs.
+Finish work on the no-floating-point version (including gamma compensation)
+Better C++ wrapper/full C++ implementation?
+Fix problem with C++ and EXTERN "C".
+cHRM transformation.
+Improve setjmp/longjmp usage or remove it in favor of returning error codes.
+Add "grayscale->palette" transformation and "palette->grayscale" detection.
+Improved dithering.
+Multi-lingual error and warning message support.
+Complete sRGB transformation (presently it simply uses gamma=0.45455).
+Man pages for function calls.
+Better documentation.
+Better filter selection
+   (counting huffman bits/precompression?  filter inertia?  filter costs?).
+Histogram creation.
+Text conversion between different code pages (Latin-1 -> Mac and DOS).
+Should we always malloc 2^bit_depth PLTE/tRNS/hIST entries for safety?
+Build gamma tables using fixed point (and do away with floating point entirely).
+Use greater precision when changing to linear gamma for compositing against
+  background and doing rgb-to-gray transformation.
+Investigate pre-incremented loop counters and other loop constructions.
diff --git a/png/Y2KINFO b/png/Y2KINFO
new file mode 100644
index 000000000..6e6ab7882
--- /dev/null
+++ b/png/Y2KINFO
@@ -0,0 +1,55 @@
+   Y2K compliance in libpng:
+   =========================
+
+      December 12, 2001
+
+      Since the PNG Development group is an ad-hoc body, we can't make
+      an official declaration.
+
+      This is your unofficial assurance that libpng from version 0.71 and
+      upward through 1.2.1 are Y2K compliant.  It is my belief that earlier
+      versions were also Y2K compliant.
+
+      Libpng only has three year fields.  One is a 2-byte unsigned integer
+      that will hold years up to 65535.  The other two hold the date in text
+      format, and will hold years up to 9999.
+
+      The integer is
+          "png_uint_16 year" in png_time_struct.
+
+      The strings are
+          "png_charp time_buffer" in png_struct and
+          "near_time_buffer", which is a local character string in png.c.
+
+      There are seven time-related functions:
+
+          png_convert_to_rfc_1123() in png.c
+            (formerly png_convert_to_rfc_1152() in error)
+          png_convert_from_struct_tm() in pngwrite.c, called in pngwrite.c
+          png_convert_from_time_t() in pngwrite.c
+          png_get_tIME() in pngget.c
+          png_handle_tIME() in pngrutil.c, called in pngread.c
+          png_set_tIME() in pngset.c
+          png_write_tIME() in pngwutil.c, called in pngwrite.c
+
+      All appear to handle dates properly in a Y2K environment.  The
+      png_convert_from_time_t() function calls gmtime() to convert from system
+      clock time, which returns (year - 1900), which we properly convert to
+      the full 4-digit year.  There is a possibility that applications using
+      libpng are not passing 4-digit years into the png_convert_to_rfc_1123()
+      function, or that they are incorrectly passing only a 2-digit year
+      instead of "year - 1900" into the png_convert_from_struct_tm() function,
+      but this is not under our control.  The libpng documentation has always
+      stated that it works with 4-digit years, and the APIs have been
+      documented as such.
+
+      The tIME chunk itself is also Y2K compliant.  It uses a 2-byte unsigned
+      integer to hold the year, and can hold years as large as 65535.
+
+      zlib, upon which libpng depends, is also Y2K compliant.  It contains
+      no date-related code.
+
+
+         Glenn Randers-Pehrson
+         libpng maintainer
+         PNG Development Group
diff --git a/png/libpng.3 b/png/libpng.3
new file mode 100644
index 000000000..3c97c96cc
--- /dev/null
+++ b/png/libpng.3
@@ -0,0 +1,3867 @@
+.TH LIBPNG 3 "December 12, 2001"
+.SH NAME
+libpng \- Portable Network Graphics (PNG) Reference Library 1.2.1
+.SH SYNOPSIS
+\fI\fB
+
+\fB#include <png.h>\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_access_version_number \fI(void\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_check_sig (png_bytep \fP\fIsig\fP\fB, int \fInum\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_chunk_error (png_structp \fP\fIpng_ptr\fP\fB, png_const_charp \fIerror\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_chunk_warning (png_structp \fP\fIpng_ptr\fP\fB, png_const_charp \fImessage\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_convert_from_struct_tm (png_timep \fP\fIptime\fP\fB, struct tm FAR * \fIttime\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_convert_from_time_t (png_timep \fP\fIptime\fP\fB, time_t \fIttime\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_charp png_convert_to_rfc1123 (png_structp \fP\fIpng_ptr\fP\fB, png_timep \fIptime\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_infop png_create_info_struct (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_structp png_create_read_struct (png_const_charp \fP\fIuser_png_ver\fP\fB, png_voidp \fP\fIerror_ptr\fP\fB, png_error_ptr \fP\fIerror_fn\fP\fB, png_error_ptr \fIwarn_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_structp png_create_read_struct_2(png_const_charp \fP\fIuser_png_ver\fP\fB, png_voidp \fP\fIerror_ptr\fP\fB, png_error_ptr \fP\fIerror_fn\fP\fB, png_error_ptr \fP\fIwarn_fn\fP\fB, png_voidp \fP\fImem_ptr\fP\fB, png_malloc_ptr \fP\fImalloc_fn\fP\fB, png_free_ptr \fIfree_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_structp png_create_write_struct (png_const_charp \fP\fIuser_png_ver\fP\fB, png_voidp \fP\fIerror_ptr\fP\fB, png_error_ptr \fP\fIerror_fn\fP\fB, png_error_ptr \fIwarn_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_structp png_create_write_struct_2(png_const_charp \fP\fIuser_png_ver\fP\fB, png_voidp \fP\fIerror_ptr\fP\fB, png_error_ptr \fP\fIerror_fn\fP\fB, png_error_ptr \fP\fIwarn_fn\fP\fB, png_voidp \fP\fImem_ptr\fP\fB, png_malloc_ptr \fP\fImalloc_fn\fP\fB, png_free_ptr \fIfree_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_debug(int \fP\fIlevel\fP\fB, png_const_charp \fImessage\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_debug1(int \fP\fIlevel\fP\fB, png_const_charp \fP\fImessage\fP\fB, \fIp1\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_debug2(int \fP\fIlevel\fP\fB, png_const_charp \fP\fImessage\fP\fB, \fP\fIp1\fP\fB, \fIp2\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_destroy_info_struct (png_structp \fP\fIpng_ptr\fP\fB, png_infopp \fIinfo_ptr_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_destroy_read_struct (png_structpp \fP\fIpng_ptr_ptr\fP\fB, png_infopp \fP\fIinfo_ptr_ptr\fP\fB, png_infopp \fIend_info_ptr_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_destroy_write_struct (png_structpp \fP\fIpng_ptr_ptr\fP\fB, png_infopp \fIinfo_ptr_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_error (png_structp \fP\fIpng_ptr\fP\fB, png_const_charp \fIerror\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_free (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fIptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_free_chunk_list (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_free_default(png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fIptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_free_data (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fInum\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_asm_flags (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_bit_depth (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_bKGD (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_color_16p \fI*background\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_channels (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_cHRM (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, double \fP\fI*white_x\fP\fB, double \fP\fI*white_y\fP\fB, double \fP\fI*red_x\fP\fB, double \fP\fI*red_y\fP\fB, double \fP\fI*green_x\fP\fB, double \fP\fI*green_y\fP\fB, double \fP\fI*blue_x\fP\fB, double \fI*blue_y\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_cHRM_fixed (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fI*white_x\fP\fB, png_uint_32 \fP\fI*white_y\fP\fB, png_uint_32 \fP\fI*red_x\fP\fB, png_uint_32 \fP\fI*red_y\fP\fB, png_uint_32 \fP\fI*green_x\fP\fB, png_uint_32 \fP\fI*green_y\fP\fB, png_uint_32 \fP\fI*blue_x\fP\fB, png_uint_32 \fI*blue_y\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_color_type (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_compression_type (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_copyright (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_get_error_ptr (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_filter_type (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_gAMA (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, double \fI*file_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_gAMA_fixed (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fI*int_file_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_header_ver (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_header_version (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_hIST (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_16p \fI*hist\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_iCCP (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_charpp \fP\fIname\fP\fB, int \fP\fI*compression_type\fP\fB, png_charpp \fP\fIprofile\fP\fB, png_uint_32 \fI*proflen\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_IHDR (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fI*width\fP\fB, png_uint_32 \fP\fI*height\fP\fB, int \fP\fI*bit_depth\fP\fB, int \fP\fI*color_type\fP\fB, int \fP\fI*interlace_type\fP\fB, int \fP\fI*compression_type\fP\fB, int \fI*filter_type\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_image_height (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_image_width (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_interlace_type (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_get_io_ptr (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_libpng_ver (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_get_mem_ptr(png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_mmx_bitdepth_threshold (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_mmx_flagmask (int \fP\fIflag_select\fP\fB, int \fI*compilerID\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_mmx_rowbytes_threshold (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_oFFs (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fI*offset_x\fP\fB, png_uint_32 \fP\fI*offset_y\fP\fB, int \fI*unit_type\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_pCAL (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_charp \fP\fI*purpose\fP\fB, png_int_32 \fP\fI*X0\fP\fB, png_int_32 \fP\fI*X1\fP\fB, int \fP\fI*type\fP\fB, int \fP\fI*nparams\fP\fB, png_charp \fP\fI*units\fP\fB, png_charpp \fI*params\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_pHYs (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fI*res_x\fP\fB, png_uint_32 \fP\fI*res_y\fP\fB, int \fI*unit_type\fP\fB);\fP
+
+\fI\fB
+
+\fBfloat png_get_pixel_aspect_ratio (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_pixels_per_meter (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_get_progressive_ptr (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_PLTE (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_colorp \fP\fI*palette\fP\fB, int \fI*num_palette\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_byte png_get_rgb_to_gray_status (png_structp \fIpng_ptr)
+
+\fBpng_uint_32 png_get_rowbytes (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_bytepp png_get_rows (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_sBIT (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_color_8p \fI*sig_bit\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_bytep png_get_signature (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_sPLT (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_spalette_p \fI*splt_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_sRGB (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fI*intent\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_text (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_textp \fP\fI*text_ptr\fP\fB, int \fI*num_text\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_tIME (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_timep \fI*mod_time\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_tRNS (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_bytep \fP\fI*trans\fP\fB, int \fP\fI*num_trans\fP\fB, png_color_16p \fI*trans_values\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_unknown_chunks (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_unknown_chunkpp \fIunknowns\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_get_user_chunk_ptr (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_get_user_transform_ptr (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_valid (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIflag\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_int_32 png_get_x_offset_microns (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_int_32 png_get_x_offset_pixels (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_x_pixels_per_meter (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_int_32 png_get_y_offset_microns (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_int_32 png_get_y_offset_pixels (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_y_pixels_per_meter (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_compression_buffer_size (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_init_io (png_structp \fP\fIpng_ptr\fP\fB, FILE \fI*fp\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_info_init (png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_info_init_2 (png_infopp \fP\fIptr_ptr\fP\fB, png_size_t \fIpng_info_struct_size\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_malloc (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_malloc_default(png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBvoidp png_memcpy (png_voidp \fP\fIs1\fP\fB, png_voidp \fP\fIs2\fP\fB, png_size_t \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_memcpy_check (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIs1\fP\fB, png_voidp \fP\fIs2\fP\fB, png_uint_32 \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBvoidp png_memset (png_voidp \fP\fIs1\fP\fB, int \fP\fIvalue\fP\fB, png_size_t \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_memset_check (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIs1\fP\fB, int \fP\fIvalue\fP\fB, png_uint_32 \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_mmx_support \fI(void\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_permit_empty_plte (png_structp \fP\fIpng_ptr\fP\fB, int \fIempty_plte_permitted\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_process_data (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_bytep \fP\fIbuffer\fP\fB, png_size_t \fIbuffer_size\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_progressive_combine_row (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIold_row\fP\fB, png_bytep \fInew_row\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_destroy (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_infop \fIend_info_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_end (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_image (png_structp \fP\fIpng_ptr\fP\fB, png_bytepp \fIimage\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_read_init (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_read_init_2 (png_structpp \fP\fIptr_ptr\fP\fB, png_const_charp \fP\fIuser_png_ver\fP\fB, png_size_t \fP\fIpng_struct_size\fP\fB, png_size_t \fIpng_info_size\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_info (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_png (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fP\fItransforms\fP\fB, png_voidp \fIparams\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_row (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIrow\fP\fB, png_bytep \fIdisplay_row\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_rows (png_structp \fP\fIpng_ptr\fP\fB, png_bytepp \fP\fIrow\fP\fB, png_bytepp \fP\fIdisplay_row\fP\fB, png_uint_32 \fInum_rows\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_update_info (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_set_asm_flags (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIasm_flags\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_background (png_structp \fP\fIpng_ptr\fP\fB, png_color_16p \fP\fIbackground_color\fP\fB, int \fP\fIbackground_gamma_code\fP\fB, int \fP\fIneed_expand\fP\fB, double \fIbackground_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_bgr (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_bKGD (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_color_16p \fIbackground\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_cHRM (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, double \fP\fIwhite_x\fP\fB, double \fP\fIwhite_y\fP\fB, double \fP\fIred_x\fP\fB, double \fP\fIred_y\fP\fB, double \fP\fIgreen_x\fP\fB, double \fP\fIgreen_y\fP\fB, double \fP\fIblue_x\fP\fB, double \fIblue_y\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_cHRM_fixed (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fIwhite_x\fP\fB, png_uint_32 \fP\fIwhite_y\fP\fB, png_uint_32 \fP\fIred_x\fP\fB, png_uint_32 \fP\fIred_y\fP\fB, png_uint_32 \fP\fIgreen_x\fP\fB, png_uint_32 \fP\fIgreen_y\fP\fB, png_uint_32 \fP\fIblue_x\fP\fB, png_uint_32 \fIblue_y\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_compression_level (png_structp \fP\fIpng_ptr\fP\fB, int \fIlevel\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_compression_mem_level (png_structp \fP\fIpng_ptr\fP\fB, int \fImem_level\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_compression_method (png_structp \fP\fIpng_ptr\fP\fB, int \fImethod\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_compression_strategy (png_structp \fP\fIpng_ptr\fP\fB, int \fIstrategy\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_compression_window_bits (png_structp \fP\fIpng_ptr\fP\fB, int \fIwindow_bits\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_crc_action (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fIcrit_action\fP\fB, int \fIancil_action\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_dither (png_structp \fP\fIpng_ptr\fP\fB, png_colorp \fP\fIpalette\fP\fB, int \fP\fInum_palette\fP\fB, int \fP\fImaximum_colors\fP\fB, png_uint_16p \fP\fIhistogram\fP\fB, int \fIfull_dither\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_error_fn (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIerror_ptr\fP\fB, png_error_ptr \fP\fIerror_fn\fP\fB, png_error_ptr \fIwarning_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_expand (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_filler (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fP\fIfiller\fP\fB, int \fIflags\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_filter (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fImethod\fP\fB, int \fIfilters\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_filter_heuristics (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fIheuristic_method\fP\fB, int \fP\fInum_weights\fP\fB, png_doublep \fP\fIfilter_weights\fP\fB, png_doublep \fIfilter_costs\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_flush (png_structp \fP\fIpng_ptr\fP\fB, int \fInrows\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_gamma (png_structp \fP\fIpng_ptr\fP\fB, double \fP\fIscreen_gamma\fP\fB, double \fIdefault_file_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_gAMA (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, double \fIfile_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_gAMA_fixed (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIfile_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_gray_1_2_4_to_8(png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_gray_to_rgb (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_hIST (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_16p \fIhist\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_iCCP (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_charp \fP\fIname\fP\fB, int \fP\fIcompression_type\fP\fB, png_charp \fP\fIprofile\fP\fB, png_uint_32 \fIproflen\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_set_interlace_handling (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_invalid (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fImask\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_invert_alpha (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_invert_mono (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_IHDR (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fIwidth\fP\fB, png_uint_32 \fP\fIheight\fP\fB, int \fP\fIbit_depth\fP\fB, int \fP\fIcolor_type\fP\fB, int \fP\fIinterlace_type\fP\fB, int \fP\fIcompression_type\fP\fB, int \fIfilter_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_keep_unknown_chunks (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fIkeep\fP\fB, png_bytep \fP\fIchunk_list\fP\fB, int \fInum_chunks\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_mem_fn(png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fImem_ptr\fP\fB, png_malloc_ptr \fP\fImalloc_fn\fP\fB, png_free_ptr \fIfree_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_set_mmx_thresholds (png_structp \fP\fIpng_ptr\fP\fB, png_byte \fP\fImmx_bitdepth_threshold\fP\fB, png_uint_32 \fImmx_rowbytes_threshold\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_oFFs (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fIoffset_x\fP\fB, png_uint_32 \fP\fIoffset_y\fP\fB, int \fIunit_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_packing (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_packswap (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_palette_to_rgb(png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_pCAL (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_charp \fP\fIpurpose\fP\fB, png_int_32 \fP\fIX0\fP\fB, png_int_32 \fP\fIX1\fP\fB, int \fP\fItype\fP\fB, int \fP\fInparams\fP\fB, png_charp \fP\fIunits\fP\fB, png_charpp \fIparams\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_pHYs (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fP\fIres_x\fP\fB, png_uint_32 \fP\fIres_y\fP\fB, int \fIunit_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_progressive_read_fn (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIprogressive_ptr\fP\fB, png_progressive_info_ptr \fP\fIinfo_fn\fP\fB, png_progressive_row_ptr \fP\fIrow_fn\fP\fB, png_progressive_end_ptr \fIend_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_PLTE (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_colorp \fP\fIpalette\fP\fB, int \fInum_palette\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_read_fn (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIio_ptr\fP\fB, png_rw_ptr \fIread_data_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_read_status_fn (png_structp \fP\fIpng_ptr\fP\fB, png_read_status_ptr \fIread_row_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_read_user_transform_fn (png_structp \fP\fIpng_ptr\fP\fB, png_user_transform_ptr \fIread_user_transform_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_rgb_to_gray (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fIerror_action\fP\fB, double \fP\fIred\fP\fB, double \fIgreen\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_rgb_to_gray_fixed (png_structp \fP\fIpng_ptr\fP\fB, int error_action png_fixed_point \fP\fIred\fP\fB, png_fixed_point \fIgreen\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_rows (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_bytepp \fIrow_pointers\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_sBIT (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_color_8p \fIsig_bit\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_sCAL (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_charp \fP\fIunit\fP\fB, double \fP\fIwidth\fP\fB, double \fIheight\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_shift (png_structp \fP\fIpng_ptr\fP\fB, png_color_8p \fItrue_bits\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_sig_bytes (png_structp \fP\fIpng_ptr\fP\fB, int \fInum_bytes\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_sPLT (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_spalette_p \fP\fIsplt_ptr\fP\fB, int \fInum_spalettes\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_sRGB (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fIintent\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_sRGB_gAMA_and_cHRM (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fIintent\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_strip_16 (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_strip_alpha (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_strip_error_numbers (png_structp \fIpng_ptr,
+
+\fBpng_uint_32 \fIstrip_mode\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_swap (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_swap_alpha (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_text (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_textp \fP\fItext_ptr\fP\fB, int \fInum_text\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_tIME (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_timep \fImod_time\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_tRNS (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_bytep \fP\fItrans\fP\fB, int \fP\fInum_trans\fP\fB, png_color_16p \fItrans_values\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_tRNS_to_alpha(png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_set_unknown_chunks (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_unknown_chunkp \fP\fIunknowns\fP\fB, int \fP\fInum\fP\fB, int \fIlocation\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_unknown_chunk_location(png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fP\fIchunk\fP\fB, int \fIlocation\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_read_user_chunk_fn (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIuser_chunk_ptr\fP\fB, png_user_chunk_ptr \fIread_user_chunk_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_user_transform_info (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIuser_transform_ptr\fP\fB, int \fP\fIuser_transform_depth\fP\fB, int \fIuser_transform_channels\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_write_fn (png_structp \fP\fIpng_ptr\fP\fB, png_voidp \fP\fIio_ptr\fP\fB, png_rw_ptr \fP\fIwrite_data_fn\fP\fB, png_flush_ptr \fIoutput_flush_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_write_status_fn (png_structp \fP\fIpng_ptr\fP\fB, png_write_status_ptr \fIwrite_row_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_write_user_transform_fn (png_structp \fP\fIpng_ptr\fP\fB, png_user_transform_ptr \fIwrite_user_transform_fn\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_set_compression_buffer_size(png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_sig_cmp (png_bytep \fP\fIsig\fP\fB, png_size_t \fP\fIstart\fP\fB, png_size_t \fInum_to_check\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_start_read_image (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_warning (png_structp \fP\fIpng_ptr\fP\fB, png_const_charp \fImessage\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_chunk (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIchunk_name\fP\fB, png_bytep \fP\fIdata\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_chunk_data (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIdata\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_chunk_end (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_chunk_start (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIchunk_name\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_destroy (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_destroy_info (png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_end (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_flush (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_image (png_structp \fP\fIpng_ptr\fP\fB, png_bytepp \fIimage\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_write_init (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBDEPRECATED: void png_write_init_2 (png_structpp \fP\fIptr_ptr\fP\fB, png_const_charp \fP\fIuser_png_ver\fP\fB, png_size_t \fP\fIpng_struct_size\fP\fB, png_size_t \fIpng_info_size\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_info (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_info_before_PLTE (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_png (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, int \fP\fItransforms\fP\fB, png_voidp \fIparams\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_row (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_rows (png_structp \fP\fIpng_ptr\fP\fB, png_bytepp \fP\fIrow\fP\fB, png_uint_32 \fInum_rows\fP\fB);\fP
+
+\fI\fB
+
+.SH DESCRIPTION
+The
+.I libpng
+library supports encoding, decoding, and various manipulations of
+the Portable Network Graphics (PNG) format image files.  It uses the
+.IR zlib(3)
+compression library.
+Following is a copy of the libpng.txt file that accompanies libpng.
+.SH LIBPNG.TXT
+libpng.txt - A description on how to use and modify libpng
+
+ libpng version 1.2.1 - December 12, 2001
+ Updated and distributed by Glenn Randers-Pehrson
+ <randeg@alum.rpi.edu>
+ Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ For conditions of distribution and use, see copyright
+ notice in png.h.
+
+ based on:
+
+ libpng 1.0 beta 6  version 0.96 May 28, 1997
+ Updated and distributed by Andreas Dilger
+ Copyright (c) 1996, 1997 Andreas Dilger
+
+ libpng 1.0 beta 2 - version 0.88  January 26, 1996
+ For conditions of distribution and use, see copyright
+ notice in png.h. Copyright (c) 1995, 1996 Guy Eric
+ Schalnat, Group 42, Inc.
+
+ Updated/rewritten per request in the libpng FAQ
+ Copyright (c) 1995, 1996 Frank J. T. Wojcik
+ December 18, 1995 & January 20, 1996
+
+.SH I. Introduction
+
+This file describes how to use and modify the PNG reference library
+(known as libpng) for your own use.  There are five sections to this
+file: introduction, structures, reading, writing, and modification and
+configuration notes for various special platforms.  In addition to this
+file, example.c is a good starting point for using the library, as
+it is heavily commented and should include everything most people
+will need.  We assume that libpng is already installed; see the
+INSTALL file for instructions on how to install libpng.
+
+Libpng was written as a companion to the PNG specification, as a way
+of reducing the amount of time and effort it takes to support the PNG
+file format in application programs.
+
+The PNG-1.2 specification is available at <http://www.libpng.org/pub/png>
+and at <ftp://ftp.uu.net/graphics/png/documents/>.
+
+The PNG-1.0 specification is available
+as RFC 2083 <ftp://ftp.uu.net/graphics/png/documents/> and as a
+W3C Recommendation <http://www.w3.org/TR/REC.png.html>. Some
+additional chunks are described in the special-purpose public chunks
+documents at <ftp://ftp.uu.net/graphics/png/documents/>.
+
+Other information
+about PNG, and the latest version of libpng, can be found at the PNG home
+page, <http://www.libpng.org/pub/png/>
+and at <ftp://ftp.uu.net/graphics/png/>.
+
+Most users will not have to modify the library significantly; advanced
+users may want to modify it more.  All attempts were made to make it as
+complete as possible, while keeping the code easy to understand.
+Currently, this library only supports C.  Support for other languages
+is being considered.
+
+Libpng has been designed to handle multiple sessions at one time,
+to be easily modifiable, to be portable to the vast majority of
+machines (ANSI, K&R, 16-, 32-, and 64-bit) available, and to be easy
+to use.  The ultimate goal of libpng is to promote the acceptance of
+the PNG file format in whatever way possible.  While there is still
+work to be done (see the TODO file), libpng should cover the
+majority of the needs of its users.
+
+Libpng uses zlib for its compression and decompression of PNG files.
+Further information about zlib, and the latest version of zlib, can
+be found at the zlib home page, <http://www.info-zip.org/pub/infozip/zlib/>.
+The zlib compression utility is a general purpose utility that is
+useful for more than PNG files, and can be used without libpng.
+See the documentation delivered with zlib for more details.
+You can usually find the source files for the zlib utility wherever you
+find the libpng source files.
+
+Libpng is thread safe, provided the threads are using different
+instances of the structures.  Each thread should have its own
+png_struct and png_info instances, and thus its own image.
+Libpng does not protect itself against two threads using the
+same instance of a structure.  Note: thread safety may be defeated
+by use of some of the MMX assembler code in pnggccrd.c, which is only
+compiled when the user defines PNG_THREAD_UNSAFE_OK.
+
+
+.SH II. Structures
+
+There are two main structures that are important to libpng, png_struct
+and png_info.  The first, png_struct, is an internal structure that
+will not, for the most part, be used by a user except as the first
+variable passed to every libpng function call.
+
+The png_info structure is designed to provide information about the
+PNG file.  At one time, the fields of png_info were intended to be
+directly accessible to the user.  However, this tended to cause problems
+with applications using dynamically loaded libraries, and as a result
+a set of interface functions for png_info (the png_get_*() and png_set_*()
+functions) was developed.  The fields of png_info are still available for
+older applications, but it is suggested that applications use the new
+interfaces if at all possible.
+
+Applications that do make direct access to the members of png_struct (except
+for png_ptr->jmpbuf) must be recompiled whenever the library is updated,
+and applications that make direct access to the members of png_info must
+be recompiled if they were compiled or loaded with libpng version 1.0.6,
+in which the members were in a different order.  In version 1.0.7, the
+members of the png_info structure reverted to the old order, as they were
+in versions 0.97c through 1.0.5.  Starting with version 2.0.0, both
+structures are going to be hidden, and the contents of the structures will
+only be accessible through the png_get/png_set functions.
+
+The png.h header file is an invaluable reference for programming with libpng.
+And while I'm on the topic, make sure you include the libpng header file:
+
+#include <png.h>
+
+.SH III. Reading
+
+We'll now walk you through the possible functions to call when reading
+in a PNG file sequentially, briefly explaining the syntax and purpose
+of each one.  See example.c and png.h for more detail.  While
+progressive reading is covered in the next section, you will still
+need some of the functions discussed in this section to read a PNG
+file.
+
+.SS Setup
+
+You will want to do the I/O initialization(*) before you get into libpng,
+so if it doesn't work, you don't have much to undo.  Of course, you
+will also want to insure that you are, in fact, dealing with a PNG
+file.  Libpng provides a simple check to see if a file is a PNG file.
+To use it, pass in the first 1 to 8 bytes of the file to the function
+png_sig_cmp(), and it will return 0 if the bytes match the corresponding
+bytes of the PNG signature, or nonzero otherwise.  Of course, the more bytes
+you pass in, the greater the accuracy of the prediction.
+
+If you are intending to keep the file pointer open for use in libpng,
+you must ensure you don't read more than 8 bytes from the beginning
+of the file, and you also have to make a call to png_set_sig_bytes_read()
+with the number of bytes you read from the beginning.  Libpng will
+then only check the bytes (if any) that your program didn't read.
+
+(*): If you are not using the standard I/O functions, you will need
+to replace them with custom functions.  See the discussion under
+Customizing libpng.
+
+
+    FILE *fp = fopen(file_name, "rb");
+    if (!fp)
+    {
+        return (ERROR);
+    }
+    fread(header, 1, number, fp);
+    is_png = !png_sig_cmp(header, 0, number);
+    if (!is_png)
+    {
+        return (NOT_PNG);
+    }
+
+
+Next, png_struct and png_info need to be allocated and initialized.  In
+order to ensure that the size of these structures is correct even with a
+dynamically linked libpng, there are functions to initialize and
+allocate the structures.  We also pass the library version, optional
+pointers to error handling functions, and a pointer to a data struct for
+use by the error functions, if necessary (the pointer and functions can
+be NULL if the default error handlers are to be used).  See the section
+on Changes to Libpng below regarding the old initialization functions.
+The structure allocation functions quietly return NULL if they fail to
+create the structure, so your application should check for that.
+
+    png_structp png_ptr = png_create_read_struct
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn);
+    if (!png_ptr)
+        return (ERROR);
+
+    png_infop info_ptr = png_create_info_struct(png_ptr);
+    if (!info_ptr)
+    {
+        png_destroy_read_struct(&png_ptr,
+           (png_infopp)NULL, (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    png_infop end_info = png_create_info_struct(png_ptr);
+    if (!end_info)
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+          (png_infopp)NULL);
+        return (ERROR);
+    }
+
+If you want to use your own memory allocation routines,
+define PNG_USER_MEM_SUPPORTED and use
+png_create_read_struct_2() instead of png_create_read_struct():
+
+    png_structp png_ptr = png_create_read_struct_2
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn, (png_voidp)
+        user_mem_ptr, user_malloc_fn, user_free_fn);
+
+The error handling routines passed to png_create_read_struct()
+and the memory alloc/free routines passed to png_create_struct_2()
+are only necessary if you are not using the libpng supplied error
+handling and memory alloc/free functions.
+
+When libpng encounters an error, it expects to longjmp back
+to your routine.  Therefore, you will need to call setjmp and pass
+your png_jmpbuf(png_ptr).  If you read the file from different
+routines, you will need to update the jmpbuf field every time you enter
+a new routine that will call a png_*() function.
+
+See your documentation of setjmp/longjmp for your compiler for more
+information on setjmp/longjmp.  See the discussion on libpng error
+handling in the Customizing Libpng section below for more information
+on the libpng error handling.  If an error occurs, and libpng longjmp's
+back to your setjmp, you will want to call png_destroy_read_struct() to
+free any memory.
+
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+           &end_info);
+        fclose(fp);
+        return (ERROR);
+    }
+
+If you would rather avoid the complexity of setjmp/longjmp issues,
+you can compile libpng with PNG_SETJMP_NOT_SUPPORTED, in which case
+errors will result in a call to PNG_ABORT() which defaults to abort().
+
+Now you need to set up the input code.  The default for libpng is to
+use the C function fread().  If you use this, you will need to pass a
+valid FILE * in the function png_init_io().  Be sure that the file is
+opened in binary mode.  If you wish to handle reading data in another
+way, you need not call the png_init_io() function, but you must then
+implement the libpng I/O methods discussed in the Customizing Libpng
+section below.
+
+    png_init_io(png_ptr, fp);
+
+If you had previously opened the file and read any of the signature from
+the beginning in order to see if this was a PNG file, you need to let
+libpng know that there are some bytes missing from the start of the file.
+
+    png_set_sig_bytes(png_ptr, number);
+
+.SS Setting up callback code
+
+You can set up a callback function to handle any unknown chunks in the
+input stream. You must supply the function
+
+    read_chunk_callback(png_ptr ptr,
+         png_unknown_chunkp chunk);
+    {
+       /* The unknown chunk structure contains your
+          chunk data: */
+           png_byte name[5];
+           png_byte *data;
+           png_size_t size;
+       /* Note that libpng has already taken care of
+          the CRC handling */
+
+       /* put your code here.  Return one of the
+          following: */
+
+       return (-n); /* chunk had an error */
+       return (0); /* did not recognize */
+       return (n); /* success */
+    }
+
+(You can give your function another name that you like instead of
+"read_chunk_callback")
+
+To inform libpng about your function, use
+
+    png_set_read_user_chunk_fn(png_ptr, user_chunk_ptr,
+        read_chunk_callback);
+
+This names not only the callback function, but also a user pointer that
+you can retrieve with
+
+    png_get_user_chunk_ptr(png_ptr);
+
+At this point, you can set up a callback function that will be
+called after each row has been read, which you can use to control
+a progress meter or the like.  It's demonstrated in pngtest.c.
+You must supply a function
+
+    void read_row_callback(png_ptr ptr, png_uint_32 row,
+       int pass);
+    {
+      /* put your code here */
+    }
+
+(You can give it another name that you like instead of "read_row_callback")
+
+To inform libpng about your function, use
+
+    png_set_read_status_fn(png_ptr, read_row_callback);
+
+.SS Unknown-chunk handling
+
+Now you get to set the way the library processes unknown chunks in the
+input PNG stream. Both known and unknown chunks will be read.  Normal
+behavior is that known chunks will be parsed into information in
+various info_ptr members; unknown chunks will be discarded. To change
+this, you can call:
+
+    png_set_keep_unknown_chunks(png_ptr, info_ptr, keep,
+        chunk_list, num_chunks);
+    keep       - 0: do not keep
+                 1: keep only if safe-to-copy
+                 2: keep even if unsafe-to-copy
+    chunk_list - list of chunks affected (a byte string,
+                 five bytes per chunk, NULL or '\0' if
+                 num_chunks is 0)
+    num_chunks - number of chunks affected; if 0, all
+                 unknown chunks are affected
+
+Unknown chunks declared in this way will be saved as raw data onto a
+list of png_unknown_chunk structures.  If a chunk that is normally
+known to libpng is named in the list, it will be handled as unknown,
+according to the "keep" directive.  If a chunk is named in successive
+instances of png_set_keep_unknown_chunks(), the final instance will
+take precedence.
+
+.SS The high-level read interface
+
+At this point there are two ways to proceed; through the high-level
+read interface, or through a sequence of low-level read operations.
+You can use the high-level interface if (a) you are willing to read
+the entire image into memory, and (b) the input transformations
+you want to do are limited to the following set:
+
+    PNG_TRANSFORM_IDENTITY      No transformation
+    PNG_TRANSFORM_STRIP_16      Strip 16-bit samples to
+                                8 bits
+    PNG_TRANSFORM_STRIP_ALPHA   Discard the alpha channel
+    PNG_TRANSFORM_PACKING       Expand 1, 2 and 4-bit
+                                samples to bytes
+    PNG_TRANSFORM_PACKSWAP      Change order of packed
+                                pixels to LSB first
+    PNG_TRANSFORM_EXPAND        Perform set_expand()
+    PNG_TRANSFORM_INVERT_MONO   Invert monochrome images
+    PNG_TRANSFORM_SHIFT         Normalize pixels to the
+                                sBIT depth
+    PNG_TRANSFORM_BGR           Flip RGB to BGR, RGBA
+                                to BGRA
+    PNG_TRANSFORM_SWAP_ALPHA    Flip RGBA to ARGB or GA
+                                to AG
+    PNG_TRANSFORM_INVERT_ALPHA  Change alpha from opacity
+                                to transparency
+    PNG_TRANSFORM_SWAP_ENDIAN   Byte-swap 16-bit samples
+
+(This excludes setting a background color, doing gamma transformation,
+dithering, and setting filler.)  If this is the case, simply do this:
+
+    png_read_png(png_ptr, info_ptr, png_transforms, NULL)
+
+where png_transforms is an integer containing the logical OR of
+some set of transformation flags.  This call is equivalent to png_read_info(),
+followed the set of transformations indicated by the transform mask,
+then png_read_image(), and finally png_read_end().
+
+(The final parameter of this call is not yet used.  Someday it might point
+to transformation parameters required by some future input transform.)
+
+After you have called png_read_png(), you can retrieve the image data
+with
+
+   row_pointers = png_get_rows(png_ptr, info_ptr);
+
+where row_pointers is an array of pointers to the pixel data for each row:
+
+   png_bytep row_pointers[height];
+
+If you know your image size and pixel size ahead of time, you can allocate
+row_pointers prior to calling png_read_png() with
+
+   row_pointers = png_malloc(png_ptr,
+      height*sizeof(png_bytep));
+   for (int i=0; i<height, i++)
+      row_pointers[i]=png_malloc(png_ptr,
+         width*pixel_size);
+   png_set_rows(png_ptr, info_ptr, &row_pointers);
+
+Alternatively you could allocate your image in one big block and define
+row_pointers[i] to point into the proper places in your block.
+
+If you use png_set_rows(), the application is responsible for freeing
+row_pointers (and row_pointers[i], if they were separately allocated).
+
+If you don't allocate row_pointers ahead of time, png_read_png() will
+do it, and it'll be free'ed when you call png_destroy_*().
+
+.SS The low-level read interface
+
+If you are going the low-level route, you are now ready to read all
+the file information up to the actual image data.  You do this with a
+call to png_read_info().
+
+    png_read_info(png_ptr, info_ptr);
+
+This will process all chunks up to but not including the image data.
+
+.SS Querying the info structure
+
+Functions are used to get the information from the info_ptr once it
+has been read.  Note that these fields may not be completely filled
+in until png_read_end() has read the chunk data following the image.
+
+    png_get_IHDR(png_ptr, info_ptr, &width, &height,
+       &bit_depth, &color_type, &interlace_type,
+       &compression_type, &filter_method);
+
+    width          - holds the width of the image
+                     in pixels (up to 2^31).
+    height         - holds the height of the image
+                     in pixels (up to 2^31).
+    bit_depth      - holds the bit depth of one of the
+                     image channels.  (valid values are
+                     1, 2, 4, 8, 16 and depend also on
+                     the color_type.  See also
+                     significant bits (sBIT) below).
+    color_type     - describes which color/alpha channels
+                         are present.
+                     PNG_COLOR_TYPE_GRAY
+                        (bit depths 1, 2, 4, 8, 16)
+                     PNG_COLOR_TYPE_GRAY_ALPHA
+                        (bit depths 8, 16)
+                     PNG_COLOR_TYPE_PALETTE
+                        (bit depths 1, 2, 4, 8)
+                     PNG_COLOR_TYPE_RGB
+                        (bit_depths 8, 16)
+                     PNG_COLOR_TYPE_RGB_ALPHA
+                        (bit_depths 8, 16)
+
+                     PNG_COLOR_MASK_PALETTE
+                     PNG_COLOR_MASK_COLOR
+                     PNG_COLOR_MASK_ALPHA
+
+    filter_method  - (must be PNG_FILTER_TYPE_BASE
+                     for PNG 1.0, and can also be
+                     PNG_INTRAPIXEL_DIFFERENCING if
+                     the PNG datastream is embedded in
+                     a MNG-1.0 datastream)
+    compression_type - (must be PNG_COMPRESSION_TYPE_BASE
+                     for PNG 1.0)
+    interlace_type - (PNG_INTERLACE_NONE or
+                     PNG_INTERLACE_ADAM7)
+    Any or all of interlace_type, compression_type, of
+    filter_method can be NULL if you are
+    not interested in their values.
+
+    channels = png_get_channels(png_ptr, info_ptr);
+    channels       - number of channels of info for the
+                     color type (valid values are 1 (GRAY,
+                     PALETTE), 2 (GRAY_ALPHA), 3 (RGB),
+                     4 (RGB_ALPHA or RGB + filler byte))
+    rowbytes = png_get_rowbytes(png_ptr, info_ptr);
+    rowbytes       - number of bytes needed to hold a row
+
+    signature = png_get_signature(png_ptr, info_ptr);
+    signature      - holds the signature read from the
+                     file (if any).  The data is kept in
+                     the same offset it would be if the
+                     whole signature were read (i.e. if an
+                     application had already read in 4
+                     bytes of signature before starting
+                     libpng, the remaining 4 bytes would
+                     be in signature[4] through signature[7]
+                     (see png_set_sig_bytes())).
+
+
+    width            = png_get_image_width(png_ptr,
+                         info_ptr);
+    height           = png_get_image_height(png_ptr,
+                         info_ptr);
+    bit_depth        = png_get_bit_depth(png_ptr,
+                         info_ptr);
+    color_type       = png_get_color_type(png_ptr,
+                         info_ptr);
+    filter_method    = png_get_filter_type(png_ptr,
+                         info_ptr);
+    compression_type = png_get_compression_type(png_ptr,
+                         info_ptr);
+    interlace_type   = png_get_interlace_type(png_ptr,
+                         info_ptr);
+
+
+These are also important, but their validity depends on whether the chunk
+has been read.  The png_get_valid(png_ptr, info_ptr, PNG_INFO_<chunk>) and
+png_get_<chunk>(png_ptr, info_ptr, ...) functions return non-zero if the
+data has been read, or zero if it is missing.  The parameters to the
+png_get_<chunk> are set directly if they are simple data types, or a pointer
+into the info_ptr is returned for any complex types.
+
+    png_get_PLTE(png_ptr, info_ptr, &palette,
+                     &num_palette);
+    palette        - the palette for the file
+                     (array of png_color)
+    num_palette    - number of entries in the palette
+
+    png_get_gAMA(png_ptr, info_ptr, &gamma);
+    gamma          - the gamma the file is written
+                     at (PNG_INFO_gAMA)
+
+    png_get_sRGB(png_ptr, info_ptr, &srgb_intent);
+    srgb_intent    - the rendering intent (PNG_INFO_sRGB)
+                     The presence of the sRGB chunk
+                     means that the pixel data is in the
+                     sRGB color space.  This chunk also
+                     implies specific values of gAMA and
+                     cHRM.
+
+    png_get_iCCP(png_ptr, info_ptr, &name,
+       &compression_type, &profile, &proflen);
+    name            - The profile name.
+    compression     - The compression type; always
+                      PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
+                      You may give NULL to this argument to
+                      ignore it.
+    profile         - International Color Consortium color
+                      profile data. May contain NULs.
+    proflen         - length of profile data in bytes.
+
+    png_get_sBIT(png_ptr, info_ptr, &sig_bit);
+    sig_bit        - the number of significant bits for
+                     (PNG_INFO_sBIT) each of the gray,
+                     red, green, and blue channels,
+                     whichever are appropriate for the
+                     given color type (png_color_16)
+
+    png_get_tRNS(png_ptr, info_ptr, &trans, &num_trans,
+                     &trans_values);
+    trans          - array of transparent entries for
+                     palette (PNG_INFO_tRNS)
+    trans_values   - graylevel or color sample values of
+                     the single transparent color for
+                     non-paletted images (PNG_INFO_tRNS)
+    num_trans      - number of transparent entries
+                     (PNG_INFO_tRNS)
+
+    png_get_hIST(png_ptr, info_ptr, &hist);
+                     (PNG_INFO_hIST)
+    hist           - histogram of palette (array of
+                     png_uint_16)
+
+    png_get_tIME(png_ptr, info_ptr, &mod_time);
+    mod_time       - time image was last modified
+                    (PNG_VALID_tIME)
+
+    png_get_bKGD(png_ptr, info_ptr, &background);
+    background     - background color (PNG_VALID_bKGD)
+                     valid 16-bit red, green and blue
+                     values, regardless of color_type
+
+    num_comments   = png_get_text(png_ptr, info_ptr,
+                     &text_ptr, &num_text);
+    num_comments   - number of comments
+    text_ptr       - array of png_text holding image
+                     comments
+    text_ptr[i].compression - type of compression used
+                 on "text" PNG_TEXT_COMPRESSION_NONE
+                           PNG_TEXT_COMPRESSION_zTXt
+                           PNG_ITXT_COMPRESSION_NONE
+                           PNG_ITXT_COMPRESSION_zTXt
+    text_ptr[i].key   - keyword for comment.  Must contain
+                         1-79 characters.
+    text_ptr[i].text  - text comments for current
+                         keyword.  Can be empty.
+    text_ptr[i].text_length - length of text string,
+                 after decompression, 0 for iTXt
+    text_ptr[i].itxt_length - length of itxt string,
+                 after decompression, 0 for tEXt/zTXt
+    text_ptr[i].lang  - language of comment (empty
+                         string for unknown).
+    text_ptr[i].translated_keyword  - keyword in UTF-8
+                         (empty string for unknown).
+    num_text       - number of comments (same as
+                     num_comments; you can put NULL here
+                     to avoid the duplication)
+    Note while png_set_text() will accept text, language,
+    and translated keywords that can be NULL pointers, the
+    structure returned by png_get_text will always contain
+    regular zero-terminated C strings.  They might be
+    empty strings but they will never be NULL pointers.
+
+    num_spalettes = png_get_sPLT(png_ptr, info_ptr,
+       &palette_ptr);
+    palette_ptr    - array of palette structures holding
+                     contents of one or more sPLT chunks
+                     read.
+    num_spalettes  - number of sPLT chunks read.
+
+    png_get_oFFs(png_ptr, info_ptr, &offset_x, &offset_y,
+       &unit_type);
+    offset_x       - positive offset from the left edge
+                     of the screen
+    offset_y       - positive offset from the top edge
+                     of the screen
+    unit_type      - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
+
+    png_get_pHYs(png_ptr, info_ptr, &res_x, &res_y,
+       &unit_type);
+    res_x          - pixels/unit physical resolution in
+                     x direction
+    res_y          - pixels/unit physical resolution in
+                     x direction
+    unit_type      - PNG_RESOLUTION_UNKNOWN,
+                     PNG_RESOLUTION_METER
+
+    png_get_sCAL(png_ptr, info_ptr, &unit, &width,
+       &height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                 (width and height are doubles)
+
+    png_get_sCAL_s(png_ptr, info_ptr, &unit, &width,
+       &height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                 (width and height are strings like "2.54")
+
+    num_unknown_chunks = png_get_unknown_chunks(png_ptr,
+       info_ptr, &unknowns)
+    unknowns          - array of png_unknown_chunk
+                        structures holding unknown chunks
+    unknowns[i].name  - name of unknown chunk
+    unknowns[i].data  - data of unknown chunk
+    unknowns[i].size  - size of unknown chunk's data
+    unknowns[i].location - position of chunk in file
+
+    The value of "i" corresponds to the order in which the
+    chunks were read from the PNG file or inserted with the
+    png_set_unknown_chunks() function.
+
+The data from the pHYs chunk can be retrieved in several convenient
+forms:
+
+    res_x = png_get_x_pixels_per_meter(png_ptr,
+       info_ptr)
+    res_y = png_get_y_pixels_per_meter(png_ptr,
+       info_ptr)
+    res_x_and_y = png_get_pixels_per_meter(png_ptr,
+       info_ptr)
+    res_x = png_get_x_pixels_per_inch(png_ptr,
+       info_ptr)
+    res_y = png_get_y_pixels_per_inch(png_ptr,
+       info_ptr)
+    res_x_and_y = png_get_pixels_per_inch(png_ptr,
+       info_ptr)
+    aspect_ratio = png_get_pixel_aspect_ratio(png_ptr,
+       info_ptr)
+
+   (Each of these returns 0 [signifying "unknown"] if
+       the data is not present or if res_x is 0;
+       res_x_and_y is 0 if res_x != res_y)
+
+The data from the oFFs chunk can be retrieved in several convenient
+forms:
+
+    x_offset = png_get_x_offset_microns(png_ptr, info_ptr);
+    y_offset = png_get_y_offset_microns(png_ptr, info_ptr);
+    x_offset = png_get_x_offset_inches(png_ptr, info_ptr);
+    y_offset = png_get_y_offset_inches(png_ptr, info_ptr);
+
+   (Each of these returns 0 [signifying "unknown" if both
+       x and y are 0] if the data is not present or if the
+       chunk is present but the unit is the pixel)
+
+For more information, see the png_info definition in png.h and the
+PNG specification for chunk contents.  Be careful with trusting
+rowbytes, as some of the transformations could increase the space
+needed to hold a row (expand, filler, gray_to_rgb, etc.).
+See png_read_update_info(), below.
+
+A quick word about text_ptr and num_text.  PNG stores comments in
+keyword/text pairs, one pair per chunk, with no limit on the number
+of text chunks, and a 2^31 byte limit on their size.  While there are
+suggested keywords, there is no requirement to restrict the use to these
+strings.  It is strongly suggested that keywords and text be sensible
+to humans (that's the point), so don't use abbreviations.  Non-printing
+symbols are not allowed.  See the PNG specification for more details.
+There is also no requirement to have text after the keyword.
+
+Keywords should be limited to 79 Latin-1 characters without leading or
+trailing spaces, but non-consecutive spaces are allowed within the
+keyword.  It is possible to have the same keyword any number of times.
+The text_ptr is an array of png_text structures, each holding a
+pointer to a language string, a pointer to a keyword and a pointer to
+a text string.  The text string, language code, and translated
+keyword may be empty or NULL pointers.  The keyword/text
+pairs are put into the array in the order that they are received.
+However, some or all of the text chunks may be after the image, so, to
+make sure you have read all the text chunks, don't mess with these
+until after you read the stuff after the image.  This will be
+mentioned again below in the discussion that goes with png_read_end().
+
+.SS Input transformations
+
+After you've read the header information, you can set up the library
+to handle any special transformations of the image data.  The various
+ways to transform the data will be described in the order that they
+should occur.  This is important, as some of these change the color
+type and/or bit depth of the data, and some others only work on
+certain color types and bit depths.  Even though each transformation
+checks to see if it has data that it can do something with, you should
+make sure to only enable a transformation if it will be valid for the
+data.  For example, don't swap red and blue on grayscale data.
+
+The colors used for the background and transparency values should be
+supplied in the same format/depth as the current image data.  They
+are stored in the same format/depth as the image data in a bKGD or tRNS
+chunk, so this is what libpng expects for this data.  The colors are
+transformed to keep in sync with the image data when an application
+calls the png_read_update_info() routine (see below).
+
+Data will be decoded into the supplied row buffers packed into bytes
+unless the library has been told to transform it into another format.
+For example, 4 bit/pixel paletted or grayscale data will be returned
+2 pixels/byte with the leftmost pixel in the high-order bits of the
+byte, unless png_set_packing() is called.  8-bit RGB data will be stored
+in RGB RGB RGB format unless png_set_filler() is called to insert filler
+bytes, either before or after each RGB triplet.  16-bit RGB data will
+be returned RRGGBB RRGGBB, with the most significant byte of the color
+value first, unless png_set_strip_16() is called to transform it to
+regular RGB RGB triplets, or png_set_filler() is called to insert
+filler bytes, either before or after each RRGGBB triplet.  Similarly,
+8-bit or 16-bit grayscale data can be modified with png_set_filler()
+or png_set_strip_16().
+
+The following code transforms grayscale images of less than 8 to 8 bits,
+changes paletted images to RGB, and adds a full alpha channel if there is
+transparency information in a tRNS chunk.  This is most useful on
+grayscale images with bit depths of 2 or 4 or if there is a multiple-image
+viewing application that wishes to treat all images in the same way.
+
+    if (color_type == PNG_COLOR_TYPE_PALETTE)
+        png_set_palette_to_rgb(png_ptr);
+
+    if (color_type == PNG_COLOR_TYPE_GRAY &&
+        bit_depth < 8) png_set_gray_1_2_4_to_8(png_ptr);
+
+    if (png_get_valid(png_ptr, info_ptr,
+        PNG_INFO_tRNS)) png_set_tRNS_to_alpha(png_ptr);
+
+These three functions are actually aliases for png_set_expand(), added
+in libpng version 1.0.4, with the function names expanded to improve code
+readability.  In some future version they may actually do different
+things.
+
+PNG can have files with 16 bits per channel.  If you only can handle
+8 bits per channel, this will strip the pixels down to 8 bit.
+
+    if (bit_depth == 16)
+        png_set_strip_16(png_ptr);
+
+If, for some reason, you don't need the alpha channel on an image,
+and you want to remove it rather than combining it with the background
+(but the image author certainly had in mind that you *would* combine
+it with the background, so that's what you should probably do):
+
+    if (color_type & PNG_COLOR_MASK_ALPHA)
+        png_set_strip_alpha(png_ptr);
+
+In PNG files, the alpha channel in an image
+is the level of opacity.  If you need the alpha channel in an image to
+be the level of transparency instead of opacity, you can invert the
+alpha channel (or the tRNS chunk data) after it's read, so that 0 is
+fully opaque and 255 (in 8-bit or paletted images) or 65535 (in 16-bit
+images) is fully transparent, with
+
+    png_set_invert_alpha(png_ptr);
+
+PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
+they can, resulting in, for example, 8 pixels per byte for 1 bit
+files.  This code expands to 1 pixel per byte without changing the
+values of the pixels:
+
+    if (bit_depth < 8)
+        png_set_packing(png_ptr);
+
+PNG files have possible bit depths of 1, 2, 4, 8, and 16.  All pixels
+stored in a PNG image have been "scaled" or "shifted" up to the next
+higher possible bit depth (e.g. from 5 bits/sample in the range [0,31] to
+8 bits/sample in the range [0, 255]).  However, it is also possible to
+convert the PNG pixel data back to the original bit depth of the image.
+This call reduces the pixels back down to the original bit depth:
+
+    png_color_8p sig_bit;
+
+    if (png_get_sBIT(png_ptr, info_ptr, &sig_bit))
+        png_set_shift(png_ptr, sig_bit);
+
+PNG files store 3-color pixels in red, green, blue order.  This code
+changes the storage of the pixels to blue, green, red:
+
+    if (color_type == PNG_COLOR_TYPE_RGB ||
+        color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+        png_set_bgr(png_ptr);
+
+PNG files store RGB pixels packed into 3 or 6 bytes. This code expands them
+into 4 or 8 bytes for windowing systems that need them in this format:
+
+    if (color_type == PNG_COLOR_TYPE_RGB)
+        png_set_filler(png_ptr, filler, PNG_FILLER_BEFORE);
+
+where "filler" is the 8 or 16-bit number to fill with, and the location is
+either PNG_FILLER_BEFORE or PNG_FILLER_AFTER, depending upon whether
+you want the filler before the RGB or after.  This transformation
+does not affect images that already have full alpha channels.  To add an
+opaque alpha channel, use filler=0xff or 0xffff and PNG_FILLER_AFTER which
+will generate RGBA pixels.
+
+If you are reading an image with an alpha channel, and you need the
+data as ARGB instead of the normal PNG format RGBA:
+
+    if (color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+        png_set_swap_alpha(png_ptr);
+
+For some uses, you may want a grayscale image to be represented as
+RGB.  This code will do that conversion:
+
+    if (color_type == PNG_COLOR_TYPE_GRAY ||
+        color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+          png_set_gray_to_rgb(png_ptr);
+
+Conversely, you can convert an RGB or RGBA image to grayscale or grayscale
+with alpha.
+
+    if (color_type == PNG_COLOR_TYPE_RGB ||
+        color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+          png_set_rgb_to_gray_fixed(png_ptr, error_action,
+             int red_weight, int green_weight);
+
+    error_action = 1: silently do the conversion
+    error_action = 2: issue a warning if the original
+                      image has any pixel where
+                      red != green or red != blue
+    error_action = 3: issue an error and abort the
+                      conversion if the original
+                      image has any pixel where
+                      red != green or red != blue
+
+    red_weight:       weight of red component times 100000
+    green_weight:     weight of green component times 100000
+                      If either weight is negative, default
+                      weights (21268, 71514) are used.
+
+If you have set error_action = 1 or 2, you can
+later check whether the image really was gray, after processing
+the image rows, with the png_get_rgb_to_gray_status(png_ptr) function.
+It will return a png_byte that is zero if the image was gray or
+1 if there were any non-gray pixels.  bKGD and sBIT data
+will be silently converted to grayscale, using the green channel
+data, regardless of the error_action setting.
+
+With red_weight+green_weight<=100000,
+the normalized graylevel is computed:
+
+    int rw = red_weight * 65536;
+    int gw = green_weight * 65536;
+    int bw = 65536 - (rw + gw);
+    gray = (rw*red + gw*green + bw*blue)/65536;
+
+The default values approximate those recommended in the Charles
+Poynton's Color FAQ, <http://www.inforamp.net/~poynton/>
+Copyright (c) 1998-01-04 Charles Poynton poynton@inforamp.net
+
+    Y = 0.212671 * R + 0.715160 * G + 0.072169 * B
+
+Libpng approximates this with
+
+    Y = 0.21268 * R    + 0.7151 * G    + 0.07217 * B
+
+which can be expressed with integers as
+
+    Y = (6969 * R + 23434 * G + 2365 * B)/32768
+
+The calculation is done in a linear colorspace, if the image gamma
+is known.
+
+If you have a grayscale and you are using png_set_expand_depth() or
+png_set_expand() to change to
+a higher bit-depth, you must either supply the background color as a gray
+value at the original file bit-depth (need_expand = 1) or else supply the
+background color as an RGB triplet at the final, expanded bit depth
+(need_expand = 0).  Similarly, if you are reading a paletted image, you
+must either supply the background color as a palette index (need_expand = 1)
+or as an RGB triplet that may or may not be in the palette (need_expand = 0).
+
+    png_color_16 my_background;
+    png_color_16p image_background;
+
+    if (png_get_bKGD(png_ptr, info_ptr, &image_background))
+        png_set_background(png_ptr, image_background,
+          PNG_BACKGROUND_GAMMA_FILE, 1, 1.0);
+    else
+        png_set_background(png_ptr, &my_background,
+          PNG_BACKGROUND_GAMMA_SCREEN, 0, 1.0);
+
+The png_set_background() function tells libpng to composite images
+with alpha or simple transparency against the supplied background
+color.  If the PNG file contains a bKGD chunk (PNG_INFO_bKGD valid),
+you may use this color, or supply another color more suitable for
+the current display (e.g., the background color from a web page).  You
+need to tell libpng whether the color is in the gamma space of the
+display (PNG_BACKGROUND_GAMMA_SCREEN for colors you supply), the file
+(PNG_BACKGROUND_GAMMA_FILE for colors from the bKGD chunk), or one
+that is neither of these gammas (PNG_BACKGROUND_GAMMA_UNIQUE - I don't
+know why anyone would use this, but it's here).
+
+To properly display PNG images on any kind of system, the application needs
+to know what the display gamma is.  Ideally, the user will know this, and
+the application will allow them to set it.  One method of allowing the user
+to set the display gamma separately for each system is to check for a
+SCREEN_GAMMA or DISPLAY_GAMMA environment variable, which will hopefully be
+correctly set.
+
+Note that display_gamma is the overall gamma correction required to produce
+pleasing results, which depends on the lighting conditions in the surrounding
+environment.  In a dim or brightly lit room, no compensation other than
+the physical gamma exponent of the monitor is needed, while in a dark room
+a slightly smaller exponent is better.
+
+   double gamma, screen_gamma;
+
+   if (/* We have a user-defined screen
+       gamma value */)
+   {
+      screen_gamma = user_defined_screen_gamma;
+   }
+   /* One way that applications can share the same
+      screen gamma value */
+   else if ((gamma_str = getenv("SCREEN_GAMMA"))
+      != NULL)
+   {
+      screen_gamma = (double)atof(gamma_str);
+   }
+   /* If we don't have another value */
+   else
+   {
+      screen_gamma = 2.2; /* A good guess for a
+           PC monitor in a bright office or a dim room */
+      screen_gamma = 2.0; /* A good guess for a
+           PC monitor in a dark room */
+      screen_gamma = 1.7 or 1.0;  /* A good
+           guess for Mac systems */
+   }
+
+The png_set_gamma() function handles gamma transformations of the data.
+Pass both the file gamma and the current screen_gamma.  If the file does
+not have a gamma value, you can pass one anyway if you have an idea what
+it is (usually 0.45455 is a good guess for GIF images on PCs).  Note
+that file gammas are inverted from screen gammas.  See the discussions
+on gamma in the PNG specification for an excellent description of what
+gamma is, and why all applications should support it.  It is strongly
+recommended that PNG viewers support gamma correction.
+
+   if (png_get_gAMA(png_ptr, info_ptr, &gamma))
+      png_set_gamma(png_ptr, screen_gamma, gamma);
+   else
+      png_set_gamma(png_ptr, screen_gamma, 0.45455);
+
+If you need to reduce an RGB file to a paletted file, or if a paletted
+file has more entries then will fit on your screen, png_set_dither()
+will do that.  Note that this is a simple match dither that merely
+finds the closest color available.  This should work fairly well with
+optimized palettes, and fairly badly with linear color cubes.  If you
+pass a palette that is larger then maximum_colors, the file will
+reduce the number of colors in the palette so it will fit into
+maximum_colors.  If there is a histogram, it will use it to make
+more intelligent choices when reducing the palette.  If there is no
+histogram, it may not do as good a job.
+
+   if (color_type & PNG_COLOR_MASK_COLOR)
+   {
+      if (png_get_valid(png_ptr, info_ptr,
+         PNG_INFO_PLTE))
+      {
+         png_uint_16p histogram;
+
+         png_get_hIST(png_ptr, info_ptr,
+            &histogram);
+         png_set_dither(png_ptr, palette, num_palette,
+            max_screen_colors, histogram, 1);
+      }
+      else
+      {
+         png_color std_color_cube[MAX_SCREEN_COLORS] =
+            { ... colors ... };
+
+         png_set_dither(png_ptr, std_color_cube,
+            MAX_SCREEN_COLORS, MAX_SCREEN_COLORS,
+            NULL,0);
+      }
+   }
+
+PNG files describe monochrome as black being zero and white being one.
+The following code will reverse this (make black be one and white be
+zero):
+
+   if (bit_depth == 1 && color_type == PNG_COLOR_TYPE_GRAY)
+      png_set_invert_mono(png_ptr);
+
+This function can also be used to invert grayscale and gray-alpha images:
+
+   if (color_type == PNG_COLOR_TYPE_GRAY ||
+        color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      png_set_invert_mono(png_ptr);
+
+PNG files store 16 bit pixels in network byte order (big-endian,
+ie. most significant bits first).  This code changes the storage to the
+other way (little-endian, i.e. least significant bits first, the
+way PCs store them):
+
+    if (bit_depth == 16)
+        png_set_swap(png_ptr);
+
+If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
+need to change the order the pixels are packed into bytes, you can use:
+
+    if (bit_depth < 8)
+       png_set_packswap(png_ptr);
+
+Finally, you can write your own transformation function if none of
+the existing ones meets your needs.  This is done by setting a callback
+with
+
+    png_set_read_user_transform_fn(png_ptr,
+       read_transform_fn);
+
+You must supply the function
+
+    void read_transform_fn(png_ptr ptr, row_info_ptr
+       row_info, png_bytep data)
+
+See pngtest.c for a working example.  Your function will be called
+after all of the other transformations have been processed.
+
+You can also set up a pointer to a user structure for use by your
+callback function, and you can inform libpng that your transform
+function will change the number of channels or bit depth with the
+function
+
+    png_set_user_transform_info(png_ptr, user_ptr,
+       user_depth, user_channels);
+
+The user's application, not libpng, is responsible for allocating and
+freeing any memory required for the user structure.
+
+You can retrieve the pointer via the function
+png_get_user_transform_ptr().  For example:
+
+    voidp read_user_transform_ptr =
+       png_get_user_transform_ptr(png_ptr);
+
+The last thing to handle is interlacing; this is covered in detail below,
+but you must call the function here if you want libpng to handle expansion
+of the interlaced image.
+
+    number_of_passes = png_set_interlace_handling(png_ptr);
+
+After setting the transformations, libpng can update your png_info
+structure to reflect any transformations you've requested with this
+call.  This is most useful to update the info structure's rowbytes
+field so you can use it to allocate your image memory.  This function
+will also update your palette with the correct screen_gamma and
+background if these have been given with the calls above.
+
+    png_read_update_info(png_ptr, info_ptr);
+
+After you call png_read_update_info(), you can allocate any
+memory you need to hold the image.  The row data is simply
+raw byte data for all forms of images.  As the actual allocation
+varies among applications, no example will be given.  If you
+are allocating one large chunk, you will need to build an
+array of pointers to each row, as it will be needed for some
+of the functions below.
+
+.SS Reading image data
+
+After you've allocated memory, you can read the image data.
+The simplest way to do this is in one function call.  If you are
+allocating enough memory to hold the whole image, you can just
+call png_read_image() and libpng will read in all the image data
+and put it in the memory area supplied.  You will need to pass in
+an array of pointers to each row.
+
+This function automatically handles interlacing, so you don't need
+to call png_set_interlace_handling() or call this function multiple
+times, or any of that other stuff necessary with png_read_rows().
+
+   png_read_image(png_ptr, row_pointers);
+
+where row_pointers is:
+
+   png_bytep row_pointers[height];
+
+You can point to void or char or whatever you use for pixels.
+
+If you don't want to read in the whole image at once, you can
+use png_read_rows() instead.  If there is no interlacing (check
+interlace_type == PNG_INTERLACE_NONE), this is simple:
+
+    png_read_rows(png_ptr, row_pointers, NULL,
+       number_of_rows);
+
+where row_pointers is the same as in the png_read_image() call.
+
+If you are doing this just one row at a time, you can do this with
+a single row_pointer instead of an array of row_pointers:
+
+    png_bytep row_pointer = row;
+    png_read_row(png_ptr, row_pointer, NULL);
+
+If the file is interlaced (interlace_type != 0 in the IHDR chunk), things
+get somewhat harder.  The only current (PNG Specification version 1.2)
+interlacing type for PNG is (interlace_type == PNG_INTERLACE_ADAM7)
+is a somewhat complicated 2D interlace scheme, known as Adam7, that
+breaks down an image into seven smaller images of varying size, based
+on an 8x8 grid.
+
+libpng can fill out those images or it can give them to you "as is".
+If you want them filled out, there are two ways to do that.  The one
+mentioned in the PNG specification is to expand each pixel to cover
+those pixels that have not been read yet (the "rectangle" method).
+This results in a blocky image for the first pass, which gradually
+smooths out as more pixels are read.  The other method is the "sparkle"
+method, where pixels are drawn only in their final locations, with the
+rest of the image remaining whatever colors they were initialized to
+before the start of the read.  The first method usually looks better,
+but tends to be slower, as there are more pixels to put in the rows.
+
+If you don't want libpng to handle the interlacing details, just call
+png_read_rows() seven times to read in all seven images.  Each of the
+images is a valid image by itself, or they can all be combined on an
+8x8 grid to form a single image (although if you intend to combine them
+you would be far better off using the libpng interlace handling).
+
+The first pass will return an image 1/8 as wide as the entire image
+(every 8th column starting in column 0) and 1/8 as high as the original
+(every 8th row starting in row 0), the second will be 1/8 as wide
+(starting in column 4) and 1/8 as high (also starting in row 0).  The
+third pass will be 1/4 as wide (every 4th pixel starting in column 0) and
+1/8 as high (every 8th row starting in row 4), and the fourth pass will
+be 1/4 as wide and 1/4 as high (every 4th column starting in column 2,
+and every 4th row starting in row 0).  The fifth pass will return an
+image 1/2 as wide, and 1/4 as high (starting at column 0 and row 2),
+while the sixth pass will be 1/2 as wide and 1/2 as high as the original
+(starting in column 1 and row 0).  The seventh and final pass will be as
+wide as the original, and 1/2 as high, containing all of the odd
+numbered scanlines.  Phew!
+
+If you want libpng to expand the images, call this before calling
+png_start_read_image() or png_read_update_info():
+
+    if (interlace_type == PNG_INTERLACE_ADAM7)
+        number_of_passes
+           = png_set_interlace_handling(png_ptr);
+
+This will return the number of passes needed.  Currently, this
+is seven, but may change if another interlace type is added.
+This function can be called even if the file is not interlaced,
+where it will return one pass.
+
+If you are not going to display the image after each pass, but are
+going to wait until the entire image is read in, use the sparkle
+effect.  This effect is faster and the end result of either method
+is exactly the same.  If you are planning on displaying the image
+after each pass, the "rectangle" effect is generally considered the
+better looking one.
+
+If you only want the "sparkle" effect, just call png_read_rows() as
+normal, with the third parameter NULL.  Make sure you make pass over
+the image number_of_passes times, and you don't change the data in the
+rows between calls.  You can change the locations of the data, just
+not the data.  Each pass only writes the pixels appropriate for that
+pass, and assumes the data from previous passes is still valid.
+
+    png_read_rows(png_ptr, row_pointers, NULL,
+       number_of_rows);
+
+If you only want the first effect (the rectangles), do the same as
+before except pass the row buffer in the third parameter, and leave
+the second parameter NULL.
+
+    png_read_rows(png_ptr, NULL, row_pointers,
+       number_of_rows);
+
+.SS Finishing a sequential read
+
+After you are finished reading the image through either the high- or
+low-level interfaces, you can finish reading the file.  If you are
+interested in comments or time, which may be stored either before or
+after the image data, you should pass the separate png_info struct if
+you want to keep the comments from before and after the image
+separate.  If you are not interested, you can pass NULL.
+
+   png_read_end(png_ptr, end_info);
+
+When you are done, you can free all memory allocated by libpng like this:
+
+   png_destroy_read_struct(&png_ptr, &info_ptr,
+       &end_info);
+
+It is also possible to individually free the info_ptr members that
+point to libpng-allocated storage with the following function:
+
+    png_free_data(png_ptr, info_ptr, mask, seq)
+    mask - identifies data to be freed, a mask
+           containing the logical OR of one or
+           more of
+             PNG_FREE_PLTE, PNG_FREE_TRNS,
+             PNG_FREE_HIST, PNG_FREE_ICCP,
+             PNG_FREE_PCAL, PNG_FREE_ROWS,
+             PNG_FREE_SCAL, PNG_FREE_SPLT,
+             PNG_FREE_TEXT, PNG_FREE_UNKN,
+           or simply PNG_FREE_ALL
+    seq  - sequence number of item to be freed
+           (-1 for all items)
+
+This function may be safely called when the relevant storage has
+already been freed, or has not yet been allocated, or was allocated
+by the user and not by libpng,  and will in those
+cases do nothing.  The "seq" parameter is ignored if only one item
+of the selected data type, such as PLTE, is allowed.  If "seq" is not
+-1, and multiple items are allowed for the data type identified in
+the mask, such as text or sPLT, only the n'th item in the structure
+is freed, where n is "seq".
+
+The default behavior is only to free data that was allocated internally
+by libpng.  This can be changed, so that libpng will not free the data,
+or so that it will free data that was allocated by the user with png_malloc()
+or png_zalloc() and passed in via a png_set_*() function, with
+
+    png_data_freer(png_ptr, info_ptr, freer, mask)
+    mask   - which data elements are affected
+             same choices as in png_free_data()
+    freer  - one of
+               PNG_DESTROY_WILL_FREE_DATA
+               PNG_SET_WILL_FREE_DATA
+               PNG_USER_WILL_FREE_DATA
+
+This function only affects data that has already been allocated.
+You can call this function after reading the PNG data but before calling
+any png_set_*() functions, to control whether the user or the png_set_*()
+function is responsible for freeing any existing data that might be present,
+and again after the png_set_*() functions to control whether the user
+or png_destroy_*() is supposed to free the data.  When the user assumes
+responsibility for libpng-allocated data, the application must use
+png_free() to free it, and when the user transfers responsibility to libpng
+for data that the user has allocated, the user must have used png_malloc()
+or png_zalloc() to allocate it.
+
+If you allocated your row_pointers in a single block, as suggested above in
+the description of the high level read interface, you must not transfer
+responsibility for freeing it to the png_set_rows or png_read_destroy function,
+because they would also try to free the individual row_pointers[i].
+
+If you allocated text_ptr.text, text_ptr.lang, and text_ptr.translated_keyword
+separately, do not transfer responsibility for freeing text_ptr to libpng,
+because when libpng fills a png_text structure it combines these members with
+the key member, and png_free_data() will free only text_ptr.key.  Similarly,
+if you transfer responsibility for free'ing text_ptr from libpng to your
+application, your application must not separately free those members.
+
+The png_free_data() function will turn off the "valid" flag for anything
+it frees.  If you need to turn the flag off for a chunk that was freed by your
+application instead of by libpng, you can use
+
+    png_set_invalid(png_ptr, info_ptr, mask);
+    mask - identifies the chunks to be made invalid,
+           containing the logical OR of one or
+           more of
+             PNG_INFO_gAMA, PNG_INFO_sBIT,
+             PNG_INFO_cHRM, PNG_INFO_PLTE,
+             PNG_INFO_tRNS, PNG_INFO_bKGD,
+             PNG_INFO_hIST, PNG_INFO_pHYs,
+             PNG_INFO_oFFs, PNG_INFO_tIME,
+             PNG_INFO_pCAL, PNG_INFO_sRGB,
+             PNG_INFO_iCCP, PNG_INFO_sPLT,
+             PNG_INFO_sCAL, PNG_INFO_IDAT
+
+For a more compact example of reading a PNG image, see the file example.c.
+
+.SS Reading PNG files progressively
+
+The progressive reader is slightly different then the non-progressive
+reader.  Instead of calling png_read_info(), png_read_rows(), and
+png_read_end(), you make one call to png_process_data(), which calls
+callbacks when it has the info, a row, or the end of the image.  You
+set up these callbacks with png_set_progressive_read_fn().  You don't
+have to worry about the input/output functions of libpng, as you are
+giving the library the data directly in png_process_data().  I will
+assume that you have read the section on reading PNG files above,
+so I will only highlight the differences (although I will show
+all of the code).
+
+png_structp png_ptr;
+png_infop info_ptr;
+
+ /*  An example code fragment of how you would
+     initialize the progressive reader in your
+     application. */
+ int
+ initialize_png_reader()
+ {
+    png_ptr = png_create_read_struct
+        (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+         user_error_fn, user_warning_fn);
+    if (!png_ptr)
+        return (ERROR);
+    info_ptr = png_create_info_struct(png_ptr);
+    if (!info_ptr)
+    {
+        png_destroy_read_struct(&png_ptr, (png_infopp)NULL,
+           (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+           (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    /* This one's new.  You can provide functions
+       to be called when the header info is valid,
+       when each row is completed, and when the image
+       is finished.  If you aren't using all functions,
+       you can specify NULL parameters.  Even when all
+       three functions are NULL, you need to call
+       png_set_progressive_read_fn().  You can use
+       any struct as the user_ptr (cast to a void pointer
+       for the function call), and retrieve the pointer
+       from inside the callbacks using the function
+
+          png_get_progressive_ptr(png_ptr);
+
+       which will return a void pointer, which you have
+       to cast appropriately.
+     */
+    png_set_progressive_read_fn(png_ptr, (void *)user_ptr,
+        info_callback, row_callback, end_callback);
+
+    return 0;
+ }
+
+ /* A code fragment that you call as you receive blocks
+   of data */
+ int
+ process_data(png_bytep buffer, png_uint_32 length)
+ {
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+           (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    /* This one's new also.  Simply give it a chunk
+       of data from the file stream (in order, of
+       course).  On machines with segmented memory
+       models machines, don't give it any more than
+       64K.  The library seems to run fine with sizes
+       of 4K. Although you can give it much less if
+       necessary (I assume you can give it chunks of
+       1 byte, I haven't tried less then 256 bytes
+       yet).  When this function returns, you may
+       want to display any rows that were generated
+       in the row callback if you don't already do
+       so there.
+     */
+    png_process_data(png_ptr, info_ptr, buffer, length);
+    return 0;
+ }
+
+ /* This function is called (as set by
+    png_set_progressive_read_fn() above) when enough data
+    has been supplied so all of the header has been
+    read.
+ */
+ void
+ info_callback(png_structp png_ptr, png_infop info)
+ {
+    /* Do any setup here, including setting any of
+       the transformations mentioned in the Reading
+       PNG files section.  For now, you _must_ call
+       either png_start_read_image() or
+       png_read_update_info() after all the
+       transformations are set (even if you don't set
+       any).  You may start getting rows before
+       png_process_data() returns, so this is your
+       last chance to prepare for that.
+     */
+ }
+
+ /* This function is called when each row of image
+    data is complete */
+ void
+ row_callback(png_structp png_ptr, png_bytep new_row,
+    png_uint_32 row_num, int pass)
+ {
+    /* If the image is interlaced, and you turned
+       on the interlace handler, this function will
+       be called for every row in every pass.  Some
+       of these rows will not be changed from the
+       previous pass.  When the row is not changed,
+       the new_row variable will be NULL.  The rows
+       and passes are called in order, so you don't
+       really need the row_num and pass, but I'm
+       supplying them because it may make your life
+       easier.
+
+       For the non-NULL rows of interlaced images,
+       you must call png_progressive_combine_row()
+       passing in the row and the old row.  You can
+       call this function for NULL rows (it will just
+       return) and for non-interlaced images (it just
+       does the memcpy for you) if it will make the
+       code easier.  Thus, you can just do this for
+       all cases:
+     */
+
+        png_progressive_combine_row(png_ptr, old_row,
+          new_row);
+
+    /* where old_row is what was displayed for
+       previously for the row.  Note that the first
+       pass (pass == 0, really) will completely cover
+       the old row, so the rows do not have to be
+       initialized.  After the first pass (and only
+       for interlaced images), you will have to pass
+       the current row, and the function will combine
+       the old row and the new row.
+    */
+ }
+
+ void
+ end_callback(png_structp png_ptr, png_infop info)
+ {
+    /* This function is called after the whole image
+       has been read, including any chunks after the
+       image (up to and including the IEND).  You
+       will usually have the same info chunk as you
+       had in the header, although some data may have
+       been added to the comments and time fields.
+
+       Most people won't do much here, perhaps setting
+       a flag that marks the image as finished.
+     */
+ }
+
+
+
+.SH IV. Writing
+
+Much of this is very similar to reading.  However, everything of
+importance is repeated here, so you won't have to constantly look
+back up in the reading section to understand writing.
+
+.SS Setup
+
+You will want to do the I/O initialization before you get into libpng,
+so if it doesn't work, you don't have anything to undo. If you are not
+using the standard I/O functions, you will need to replace them with
+custom writing functions.  See the discussion under Customizing libpng.
+
+    FILE *fp = fopen(file_name, "wb");
+    if (!fp)
+    {
+       return (ERROR);
+    }
+
+Next, png_struct and png_info need to be allocated and initialized.
+As these can be both relatively large, you may not want to store these
+on the stack, unless you have stack space to spare.  Of course, you
+will want to check if they return NULL.  If you are also reading,
+you won't want to name your read structure and your write structure
+both "png_ptr"; you can call them anything you like, such as
+"read_ptr" and "write_ptr".  Look at pngtest.c, for example.
+
+    png_structp png_ptr = png_create_write_struct
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn);
+    if (!png_ptr)
+       return (ERROR);
+
+    png_infop info_ptr = png_create_info_struct(png_ptr);
+    if (!info_ptr)
+    {
+       png_destroy_write_struct(&png_ptr,
+         (png_infopp)NULL);
+       return (ERROR);
+    }
+
+If you want to use your own memory allocation routines,
+define PNG_USER_MEM_SUPPORTED and use
+png_create_write_struct_2() instead of png_create_write_struct():
+
+    png_structp png_ptr = png_create_write_struct_2
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn, (png_voidp)
+        user_mem_ptr, user_malloc_fn, user_free_fn);
+
+After you have these structures, you will need to set up the
+error handling.  When libpng encounters an error, it expects to
+longjmp() back to your routine.  Therefore, you will need to call
+setjmp() and pass the png_jmpbuf(png_ptr).  If you
+write the file from different routines, you will need to update
+the png_jmpbuf(png_ptr) every time you enter a new routine that will
+call a png_*() function.  See your documentation of setjmp/longjmp
+for your compiler for more information on setjmp/longjmp.  See
+the discussion on libpng error handling in the Customizing Libpng
+section below for more information on the libpng error handling.
+
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+       png_destroy_write_struct(&png_ptr, &info_ptr);
+       fclose(fp);
+       return (ERROR);
+    }
+    ...
+    return;
+
+If you would rather avoid the complexity of setjmp/longjmp issues,
+you can compile libpng with PNG_SETJMP_NOT_SUPPORTED, in which case
+errors will result in a call to PNG_ABORT() which defaults to abort().
+
+Now you need to set up the output code.  The default for libpng is to
+use the C function fwrite().  If you use this, you will need to pass a
+valid FILE * in the function png_init_io().  Be sure that the file is
+opened in binary mode.  Again, if you wish to handle writing data in
+another way, see the discussion on libpng I/O handling in the Customizing
+Libpng section below.
+
+    png_init_io(png_ptr, fp);
+
+.SS Write callbacks
+
+At this point, you can set up a callback function that will be
+called after each row has been written, which you can use to control
+a progress meter or the like.  It's demonstrated in pngtest.c.
+You must supply a function
+
+    void write_row_callback(png_ptr, png_uint_32 row,
+       int pass);
+    {
+      /* put your code here */
+    }
+
+(You can give it another name that you like instead of "write_row_callback")
+
+To inform libpng about your function, use
+
+    png_set_write_status_fn(png_ptr, write_row_callback);
+
+You now have the option of modifying how the compression library will
+run.  The following functions are mainly for testing, but may be useful
+in some cases, like if you need to write PNG files extremely fast and
+are willing to give up some compression, or if you want to get the
+maximum possible compression at the expense of slower writing.  If you
+have no special needs in this area, let the library do what it wants by
+not calling this function at all, as it has been tuned to deliver a good
+speed/compression ratio. The second parameter to png_set_filter() is
+the filter method, for which the only valid values are 0 (as of the
+July 1999 PNG specification, version 1.2) or 64 (if you are writing
+a PNG datastream that is to be embedded in a MNG datastream).  The third
+parameter is a flag that indicates which filter type(s) are to be tested
+for each scanline.  See the PNG specification for details on the specific filter
+types.
+
+
+    /* turn on or off filtering, and/or choose
+       specific filters.  You can use either a single
+       PNG_FILTER_VALUE_NAME or the logical OR of one
+       or more PNG_FILTER_NAME masks. */
+    png_set_filter(png_ptr, 0,
+       PNG_FILTER_NONE  | PNG_FILTER_VALUE_NONE |
+       PNG_FILTER_SUB   | PNG_FILTER_VALUE_SUB  |
+       PNG_FILTER_UP    | PNG_FILTER_VALUE_UP   |
+       PNG_FILTER_AVE   | PNG_FILTER_VALUE_AVE  |
+       PNG_FILTER_PAETH | PNG_FILTER_VALUE_PAETH|
+       PNG_ALL_FILTERS);
+
+If an application
+wants to start and stop using particular filters during compression,
+it should start out with all of the filters (to ensure that the previous
+row of pixels will be stored in case it's needed later), and then add
+and remove them after the start of compression.
+
+If you are writing a PNG datastream that is to be embedded in a MNG
+datastream, the second parameter can be either 0 or 64.
+
+The png_set_compression_*() functions interface to the zlib compression
+library, and should mostly be ignored unless you really know what you are
+doing.  The only generally useful call is png_set_compression_level()
+which changes how much time zlib spends on trying to compress the image
+data.  See the Compression Library (zlib.h and algorithm.txt, distributed
+with zlib) for details on the compression levels.
+
+    /* set the zlib compression level */
+    png_set_compression_level(png_ptr,
+        Z_BEST_COMPRESSION);
+
+    /* set other zlib parameters */
+    png_set_compression_mem_level(png_ptr, 8);
+    png_set_compression_strategy(png_ptr,
+        Z_DEFAULT_STRATEGY);
+    png_set_compression_window_bits(png_ptr, 15);
+    png_set_compression_method(png_ptr, 8);
+    png_set_compression_buffer_size(png_ptr, 8192)
+
+extern PNG_EXPORT(void,png_set_zbuf_size)
+
+.SS Setting the contents of info for output
+
+You now need to fill in the png_info structure with all the data you
+wish to write before the actual image.  Note that the only thing you
+are allowed to write after the image is the text chunks and the time
+chunk (as of PNG Specification 1.2, anyway).  See png_write_end() and
+the latest PNG specification for more information on that.  If you
+wish to write them before the image, fill them in now, and flag that
+data as being valid.  If you want to wait until after the data, don't
+fill them until png_write_end().  For all the fields in png_info and
+their data types, see png.h.  For explanations of what the fields
+contain, see the PNG specification.
+
+Some of the more important parts of the png_info are:
+
+    png_set_IHDR(png_ptr, info_ptr, width, height,
+       bit_depth, color_type, interlace_type,
+       compression_type, filter_method)
+    width          - holds the width of the image
+                     in pixels (up to 2^31).
+    height         - holds the height of the image
+                     in pixels (up to 2^31).
+    bit_depth      - holds the bit depth of one of the
+                     image channels.
+                     (valid values are 1, 2, 4, 8, 16
+                     and depend also on the
+                     color_type.  See also significant
+                     bits (sBIT) below).
+    color_type     - describes which color/alpha
+                     channels are present.
+                     PNG_COLOR_TYPE_GRAY
+                        (bit depths 1, 2, 4, 8, 16)
+                     PNG_COLOR_TYPE_GRAY_ALPHA
+                        (bit depths 8, 16)
+                     PNG_COLOR_TYPE_PALETTE
+                        (bit depths 1, 2, 4, 8)
+                     PNG_COLOR_TYPE_RGB
+                        (bit_depths 8, 16)
+                     PNG_COLOR_TYPE_RGB_ALPHA
+                        (bit_depths 8, 16)
+
+                     PNG_COLOR_MASK_PALETTE
+                     PNG_COLOR_MASK_COLOR
+                     PNG_COLOR_MASK_ALPHA
+
+    interlace_type - PNG_INTERLACE_NONE or
+                     PNG_INTERLACE_ADAM7
+    compression_type - (must be
+                     PNG_COMPRESSION_TYPE_DEFAULT)
+    filter_method  - (must be PNG_FILTER_TYPE_DEFAULT
+                     or, if you are writing a PNG to
+                     be embedded in a MNG datastream,
+                     can also be
+                     PNG_INTRAPIXEL_DIFFERENCING)
+
+    png_set_PLTE(png_ptr, info_ptr, palette,
+       num_palette);
+    palette        - the palette for the file
+                     (array of png_color)
+    num_palette    - number of entries in the palette
+
+    png_set_gAMA(png_ptr, info_ptr, gamma);
+    gamma          - the gamma the image was created
+                     at (PNG_INFO_gAMA)
+
+    png_set_sRGB(png_ptr, info_ptr, srgb_intent);
+    srgb_intent    - the rendering intent
+                     (PNG_INFO_sRGB) The presence of
+                     the sRGB chunk means that the pixel
+                     data is in the sRGB color space.
+                     This chunk also implies specific
+                     values of gAMA and cHRM.  Rendering
+                     intent is the CSS-1 property that
+                     has been defined by the International
+                     Color Consortium
+                     (http://www.color.org).
+                     It can be one of
+                     PNG_sRGB_INTENT_SATURATION,
+                     PNG_sRGB_INTENT_PERCEPTUAL,
+                     PNG_sRGB_INTENT_ABSOLUTE, or
+                     PNG_sRGB_INTENT_RELATIVE.
+
+
+    png_set_sRGB_gAMA_and_cHRM(png_ptr, info_ptr,
+       srgb_intent);
+    srgb_intent    - the rendering intent
+                     (PNG_INFO_sRGB) The presence of the
+                     sRGB chunk means that the pixel
+                     data is in the sRGB color space.
+                     This function also causes gAMA and
+                     cHRM chunks with the specific values
+                     that are consistent with sRGB to be
+                     written.
+
+    png_set_iCCP(png_ptr, info_ptr, name, compression_type,
+                      profile, proflen);
+    name            - The profile name.
+    compression     - The compression type; always
+                      PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
+                      You may give NULL to this argument to
+                      ignore it.
+    profile         - International Color Consortium color
+                      profile data. May contain NULs.
+    proflen         - length of profile data in bytes.
+
+    png_set_sBIT(png_ptr, info_ptr, sig_bit);
+    sig_bit        - the number of significant bits for
+                     (PNG_INFO_sBIT) each of the gray, red,
+                     green, and blue channels, whichever are
+                     appropriate for the given color type
+                     (png_color_16)
+
+    png_set_tRNS(png_ptr, info_ptr, trans, num_trans,
+       trans_values);
+    trans          - array of transparent entries for
+                     palette (PNG_INFO_tRNS)
+    trans_values   - graylevel or color sample values of
+                     the single transparent color for
+                     non-paletted images (PNG_INFO_tRNS)
+    num_trans      - number of transparent entries
+                     (PNG_INFO_tRNS)
+
+    png_set_hIST(png_ptr, info_ptr, hist);
+                    (PNG_INFO_hIST)
+    hist           - histogram of palette (array of
+                     png_uint_16)
+
+    png_set_tIME(png_ptr, info_ptr, mod_time);
+    mod_time       - time image was last modified
+                     (PNG_VALID_tIME)
+
+    png_set_bKGD(png_ptr, info_ptr, background);
+    background     - background color (PNG_VALID_bKGD)
+
+    png_set_text(png_ptr, info_ptr, text_ptr, num_text);
+    text_ptr       - array of png_text holding image
+                     comments
+    text_ptr[i].compression - type of compression used
+                 on "text" PNG_TEXT_COMPRESSION_NONE
+                           PNG_TEXT_COMPRESSION_zTXt
+                           PNG_ITXT_COMPRESSION_NONE
+                           PNG_ITXT_COMPRESSION_zTXt
+    text_ptr[i].key   - keyword for comment.  Must contain
+                 1-79 characters.
+    text_ptr[i].text  - text comments for current
+                         keyword.  Can be NULL or empty.
+    text_ptr[i].text_length - length of text string,
+                 after decompression, 0 for iTXt
+    text_ptr[i].itxt_length - length of itxt string,
+                 after decompression, 0 for tEXt/zTXt
+    text_ptr[i].lang  - language of comment (NULL or
+                         empty for unknown).
+    text_ptr[i].translated_keyword  - keyword in UTF-8 (NULL
+                         or empty for unknown).
+    num_text       - number of comments
+
+    png_set_sPLT(png_ptr, info_ptr, &palette_ptr,
+       num_spalettes);
+    palette_ptr    - array of png_sPLT_struct structures
+                     to be added to the list of palettes
+                     in the info structure.
+    num_spalettes  - number of palette structures to be
+                     added.
+
+    png_set_oFFs(png_ptr, info_ptr, offset_x, offset_y,
+        unit_type);
+    offset_x  - positive offset from the left
+                     edge of the screen
+    offset_y  - positive offset from the top
+                     edge of the screen
+    unit_type - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
+
+    png_set_pHYs(png_ptr, info_ptr, res_x, res_y,
+        unit_type);
+    res_x       - pixels/unit physical resolution
+                  in x direction
+    res_y       - pixels/unit physical resolution
+                  in y direction
+    unit_type   - PNG_RESOLUTION_UNKNOWN,
+                  PNG_RESOLUTION_METER
+
+    png_set_sCAL(png_ptr, info_ptr, unit, width, height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                  (width and height are doubles)
+
+    png_set_sCAL_s(png_ptr, info_ptr, unit, width, height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                 (width and height are strings like "2.54")
+
+    png_set_unknown_chunks(png_ptr, info_ptr, &unknowns,
+       num_unknowns)
+    unknowns          - array of png_unknown_chunk
+                        structures holding unknown chunks
+    unknowns[i].name  - name of unknown chunk
+    unknowns[i].data  - data of unknown chunk
+    unknowns[i].size  - size of unknown chunk's data
+    unknowns[i].location - position to write chunk in file
+                           0: do not write chunk
+                           PNG_HAVE_IHDR: before PLTE
+                           PNG_HAVE_PLTE: before IDAT
+                           PNG_AFTER_IDAT: after IDAT
+
+The "location" member is set automatically according to
+what part of the output file has already been written.
+You can change its value after calling png_set_unknown_chunks()
+as demonstrated in pngtest.c.  Within each of the "locations",
+the chunks are sequenced according to their position in the
+structure (that is, the value of "i", which is the order in which
+the chunk was either read from the input file or defined with
+png_set_unknown_chunks).
+
+A quick word about text and num_text.  text is an array of png_text
+structures.  num_text is the number of valid structures in the array.
+Each png_text structure holds a language code, a keyword, a text value,
+and a compression type.
+
+The compression types have the same valid numbers as the compression
+types of the image data.  Currently, the only valid number is zero.
+However, you can store text either compressed or uncompressed, unlike
+images, which always have to be compressed.  So if you don't want the
+text compressed, set the compression type to PNG_TEXT_COMPRESSION_NONE.
+Because tEXt and zTXt chunks don't have a language field, if you
+specify PNG_TEXT_COMPRESSION_NONE or PNG_TEXT_COMPRESSION_zTXt
+any language code or translated keyword will not be written out.
+
+Until text gets around 1000 bytes, it is not worth compressing it.
+After the text has been written out to the file, the compression type
+is set to PNG_TEXT_COMPRESSION_NONE_WR or PNG_TEXT_COMPRESSION_zTXt_WR,
+so that it isn't written out again at the end (in case you are calling
+png_write_end() with the same struct.
+
+The keywords that are given in the PNG Specification are:
+
+    Title            Short (one line) title or
+                     caption for image
+    Author           Name of image's creator
+    Description      Description of image (possibly long)
+    Copyright        Copyright notice
+    Creation Time    Time of original image creation
+                     (usually RFC 1123 format, see below)
+    Software         Software used to create the image
+    Disclaimer       Legal disclaimer
+    Warning          Warning of nature of content
+    Source           Device used to create the image
+    Comment          Miscellaneous comment; conversion
+                     from other image format
+
+The keyword-text pairs work like this.  Keywords should be short
+simple descriptions of what the comment is about.  Some typical
+keywords are found in the PNG specification, as is some recommendations
+on keywords.  You can repeat keywords in a file.  You can even write
+some text before the image and some after.  For example, you may want
+to put a description of the image before the image, but leave the
+disclaimer until after, so viewers working over modem connections
+don't have to wait for the disclaimer to go over the modem before
+they start seeing the image.  Finally, keywords should be full
+words, not abbreviations.  Keywords and text are in the ISO 8859-1
+(Latin-1) character set (a superset of regular ASCII) and can not
+contain NUL characters, and should not contain control or other
+unprintable characters.  To make the comments widely readable, stick
+with basic ASCII, and avoid machine specific character set extensions
+like the IBM-PC character set.  The keyword must be present, but
+you can leave off the text string on non-compressed pairs.
+Compressed pairs must have a text string, as only the text string
+is compressed anyway, so the compression would be meaningless.
+
+PNG supports modification time via the png_time structure.  Two
+conversion routines are provided, png_convert_from_time_t() for
+time_t and png_convert_from_struct_tm() for struct tm.  The
+time_t routine uses gmtime().  You don't have to use either of
+these, but if you wish to fill in the png_time structure directly,
+you should provide the time in universal time (GMT) if possible
+instead of your local time.  Note that the year number is the full
+year (e.g. 1998, rather than 98 - PNG is year 2000 compliant!), and
+that months start with 1.
+
+If you want to store the time of the original image creation, you should
+use a plain tEXt chunk with the "Creation Time" keyword.  This is
+necessary because the "creation time" of a PNG image is somewhat vague,
+depending on whether you mean the PNG file, the time the image was
+created in a non-PNG format, a still photo from which the image was
+scanned, or possibly the subject matter itself.  In order to facilitate
+machine-readable dates, it is recommended that the "Creation Time"
+tEXt chunk use RFC 1123 format dates (e.g. "22 May 1997 18:07:10 GMT"),
+although this isn't a requirement.  Unlike the tIME chunk, the
+"Creation Time" tEXt chunk is not expected to be automatically changed
+by the software.  To facilitate the use of RFC 1123 dates, a function
+png_convert_to_rfc1123(png_timep) is provided to convert from PNG
+time to an RFC 1123 format string.
+
+.SS Writing unknown chunks
+
+You can use the png_set_unknown_chunks function to queue up chunks
+for writing.  You give it a chunk name, raw data, and a size; that's
+all there is to it.  The chunks will be written by the next following
+png_write_info_before_PLTE, png_write_info, or png_write_end function.
+Any chunks previously read into the info structure's unknown-chunk
+list will also be written out in a sequence that satisfies the PNG
+specification's ordering rules.
+
+.SS The high-level write interface
+
+At this point there are two ways to proceed; through the high-level
+write interface, or through a sequence of low-level write operations.
+You can use the high-level interface if your image data is present
+in the info structure.  All defined output
+transformations are permitted, enabled by the following masks.
+
+    PNG_TRANSFORM_IDENTITY      No transformation
+    PNG_TRANSFORM_PACKING       Pack 1, 2 and 4-bit samples
+    PNG_TRANSFORM_PACKSWAP      Change order of packed
+                                pixels to LSB first
+    PNG_TRANSFORM_INVERT_MONO   Invert monochrome images
+    PNG_TRANSFORM_SHIFT         Normalize pixels to the
+                                sBIT depth
+    PNG_TRANSFORM_BGR           Flip RGB to BGR, RGBA
+                                to BGRA
+    PNG_TRANSFORM_SWAP_ALPHA    Flip RGBA to ARGB or GA
+                                to AG
+    PNG_TRANSFORM_INVERT_ALPHA  Change alpha from opacity
+                                to transparency
+    PNG_TRANSFORM_SWAP_ENDIAN   Byte-swap 16-bit samples
+    PNG_TRANSFORM_STRIP_FILLER  Strip out filler bytes.
+
+If you have valid image data in the info structure (you can use
+png_set_rows() to put image data in the info structure), simply do this:
+
+    png_write_png(png_ptr, info_ptr, png_transforms, NULL)
+
+where png_transforms is an integer containing the logical OR of some set of
+transformation flags.  This call is equivalent to png_write_info(),
+followed the set of transformations indicated by the transform mask,
+then png_write_image(), and finally png_write_end().
+
+(The final parameter of this call is not yet used.  Someday it might point
+to transformation parameters required by some future output transform.)
+
+.SS The low-level write interface
+
+If you are going the low-level route instead, you are now ready to
+write all the file information up to the actual image data.  You do
+this with a call to png_write_info().
+
+    png_write_info(png_ptr, info_ptr);
+
+Note that there is one transformation you may need to do before
+png_write_info().  In PNG files, the alpha channel in an image is the
+level of opacity.  If your data is supplied as a level of
+transparency, you can invert the alpha channel before you write it, so
+that 0 is fully transparent and 255 (in 8-bit or paletted images) or
+65535 (in 16-bit images) is fully opaque, with
+
+    png_set_invert_alpha(png_ptr);
+
+This must appear before png_write_info() instead of later with the
+other transformations because in the case of paletted images the tRNS
+chunk data has to be inverted before the tRNS chunk is written.  If
+your image is not a paletted image, the tRNS data (which in such cases
+represents a single color to be rendered as transparent) won't need to
+be changed, and you can safely do this transformation after your
+png_write_info() call.
+
+If you need to write a private chunk that you want to appear before
+the PLTE chunk when PLTE is present, you can write the PNG info in
+two steps, and insert code to write your own chunk between them:
+
+    png_write_info_before_PLTE(png_ptr, info_ptr);
+    png_set_unknown_chunks(png_ptr, info_ptr, ...);
+    png_write_info(png_ptr, info_ptr);
+
+After you've written the file information, you can set up the library
+to handle any special transformations of the image data.  The various
+ways to transform the data will be described in the order that they
+should occur.  This is important, as some of these change the color
+type and/or bit depth of the data, and some others only work on
+certain color types and bit depths.  Even though each transformation
+checks to see if it has data that it can do something with, you should
+make sure to only enable a transformation if it will be valid for the
+data.  For example, don't swap red and blue on grayscale data.
+
+PNG files store RGB pixels packed into 3 or 6 bytes.  This code tells
+the library to strip input data that has 4 or 8 bytes per pixel down
+to 3 or 6 bytes (or strip 2 or 4-byte grayscale+filler data to 1 or 2
+bytes per pixel).
+
+    png_set_filler(png_ptr, 0, PNG_FILLER_BEFORE);
+
+where the 0 is unused, and the location is either PNG_FILLER_BEFORE or
+PNG_FILLER_AFTER, depending upon whether the filler byte in the pixel
+is stored XRGB or RGBX.
+
+PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
+they can, resulting in, for example, 8 pixels per byte for 1 bit files.
+If the data is supplied at 1 pixel per byte, use this code, which will
+correctly pack the pixels into a single byte:
+
+    png_set_packing(png_ptr);
+
+PNG files reduce possible bit depths to 1, 2, 4, 8, and 16.  If your
+data is of another bit depth, you can write an sBIT chunk into the
+file so that decoders can recover the original data if desired.
+
+    /* Set the true bit depth of the image data */
+    if (color_type & PNG_COLOR_MASK_COLOR)
+    {
+        sig_bit.red = true_bit_depth;
+        sig_bit.green = true_bit_depth;
+        sig_bit.blue = true_bit_depth;
+    }
+    else
+    {
+        sig_bit.gray = true_bit_depth;
+    }
+    if (color_type & PNG_COLOR_MASK_ALPHA)
+    {
+        sig_bit.alpha = true_bit_depth;
+    }
+
+    png_set_sBIT(png_ptr, info_ptr, &sig_bit);
+
+If the data is stored in the row buffer in a bit depth other than
+one supported by PNG (e.g. 3 bit data in the range 0-7 for a 4-bit PNG),
+this will scale the values to appear to be the correct bit depth as
+is required by PNG.
+
+    png_set_shift(png_ptr, &sig_bit);
+
+PNG files store 16 bit pixels in network byte order (big-endian,
+ie. most significant bits first).  This code would be used if they are
+supplied the other way (little-endian, i.e. least significant bits
+first, the way PCs store them):
+
+    if (bit_depth > 8)
+       png_set_swap(png_ptr);
+
+If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
+need to change the order the pixels are packed into bytes, you can use:
+
+    if (bit_depth < 8)
+       png_set_packswap(png_ptr);
+
+PNG files store 3 color pixels in red, green, blue order.  This code
+would be used if they are supplied as blue, green, red:
+
+    png_set_bgr(png_ptr);
+
+PNG files describe monochrome as black being zero and white being
+one. This code would be used if the pixels are supplied with this reversed
+(black being one and white being zero):
+
+    png_set_invert_mono(png_ptr);
+
+Finally, you can write your own transformation function if none of
+the existing ones meets your needs.  This is done by setting a callback
+with
+
+    png_set_write_user_transform_fn(png_ptr,
+       write_transform_fn);
+
+You must supply the function
+
+    void write_transform_fn(png_ptr ptr, row_info_ptr
+       row_info, png_bytep data)
+
+See pngtest.c for a working example.  Your function will be called
+before any of the other transformations are processed.
+
+You can also set up a pointer to a user structure for use by your
+callback function.
+
+    png_set_user_transform_info(png_ptr, user_ptr, 0, 0);
+
+The user_channels and user_depth parameters of this function are ignored
+when writing; you can set them to zero as shown.
+
+You can retrieve the pointer via the function png_get_user_transform_ptr().
+For example:
+
+    voidp write_user_transform_ptr =
+       png_get_user_transform_ptr(png_ptr);
+
+It is possible to have libpng flush any pending output, either manually,
+or automatically after a certain number of lines have been written.  To
+flush the output stream a single time call:
+
+    png_write_flush(png_ptr);
+
+and to have libpng flush the output stream periodically after a certain
+number of scanlines have been written, call:
+
+    png_set_flush(png_ptr, nrows);
+
+Note that the distance between rows is from the last time png_write_flush()
+was called, or the first row of the image if it has never been called.
+So if you write 50 lines, and then png_set_flush 25, it will flush the
+output on the next scanline, and every 25 lines thereafter, unless
+png_write_flush() is called before 25 more lines have been written.
+If nrows is too small (less than about 10 lines for a 640 pixel wide
+RGB image) the image compression may decrease noticeably (although this
+may be acceptable for real-time applications).  Infrequent flushing will
+only degrade the compression performance by a few percent over images
+that do not use flushing.
+
+.SS Writing the image data
+
+That's it for the transformations.  Now you can write the image data.
+The simplest way to do this is in one function call.  If you have the
+whole image in memory, you can just call png_write_image() and libpng
+will write the image.  You will need to pass in an array of pointers to
+each row.  This function automatically handles interlacing, so you don't
+need to call png_set_interlace_handling() or call this function multiple
+times, or any of that other stuff necessary with png_write_rows().
+
+    png_write_image(png_ptr, row_pointers);
+
+where row_pointers is:
+
+    png_byte *row_pointers[height];
+
+You can point to void or char or whatever you use for pixels.
+
+If you don't want to write the whole image at once, you can
+use png_write_rows() instead.  If the file is not interlaced,
+this is simple:
+
+    png_write_rows(png_ptr, row_pointers,
+       number_of_rows);
+
+row_pointers is the same as in the png_write_image() call.
+
+If you are just writing one row at a time, you can do this with
+a single row_pointer instead of an array of row_pointers:
+
+    png_bytep row_pointer = row;
+
+    png_write_row(png_ptr, row_pointer);
+
+When the file is interlaced, things can get a good deal more
+complicated.  The only currently (as of the PNG Specification
+version 1.2, dated July 1999) defined interlacing scheme for PNG files
+is the "Adam7" interlace scheme, that breaks down an
+image into seven smaller images of varying size.  libpng will build
+these images for you, or you can do them yourself.  If you want to
+build them yourself, see the PNG specification for details of which
+pixels to write when.
+
+If you don't want libpng to handle the interlacing details, just
+use png_set_interlace_handling() and call png_write_rows() the
+correct number of times to write all seven sub-images.
+
+If you want libpng to build the sub-images, call this before you start
+writing any rows:
+
+    number_of_passes =
+       png_set_interlace_handling(png_ptr);
+
+This will return the number of passes needed.  Currently, this
+is seven, but may change if another interlace type is added.
+
+Then write the complete image number_of_passes times.
+
+    png_write_rows(png_ptr, row_pointers,
+       number_of_rows);
+
+As some of these rows are not used, and thus return immediately,
+you may want to read about interlacing in the PNG specification,
+and only update the rows that are actually used.
+
+.SS Finishing a sequential write
+
+After you are finished writing the image, you should finish writing
+the file.  If you are interested in writing comments or time, you should
+pass an appropriately filled png_info pointer.  If you are not interested,
+you can pass NULL.
+
+    png_write_end(png_ptr, info_ptr);
+
+When you are done, you can free all memory used by libpng like this:
+
+    png_destroy_write_struct(&png_ptr, &info_ptr);
+
+It is also possible to individually free the info_ptr members that
+point to libpng-allocated storage with the following functions:
+
+    png_free_data(png_ptr, info_ptr, mask, n)
+    mask         - identifies data to be freed, a mask
+                   made up by the OR one or more of
+                   PNG_FREE_PLTE, PNG_FREE_TRNS,
+                   PNG_FREE_HIST, PNG_FREE_ICCP,
+                   PNG_FREE_SPLT, PNG_FREE_ROWS,
+                   PNG_FREE_PCAL, PNG_FREE_SCAL,
+                   PNG_FREE_TEXT, PNG_FREE_UNKN,
+                   or simply PNG_FREE_ALL
+    n            - sequence number of item to be freed
+                   (-1 for all items)
+
+These functions may be safely called when the relevant storage has
+already been freed, or has not yet been allocated, and will in that
+case do nothing.  The "n" parameter is ignored if only one item
+of the selected data type, such as PLTE, is allowed.  If "n" is not
+-1, and multiple items are allowed for the data type identified in
+the mask, such as text or splt, only the n'th item is freed.
+
+If you allocated data such as a palette that you passed in to libpng with
+png_set_*, you must not free it until just before the call to
+png_destroy_write_struct().
+
+For a more compact example of writing a PNG image, see the file example.c.
+
+.SH V. Modifying/Customizing libpng:
+
+There are two issues here.  The first is changing how libpng does
+standard things like memory allocation, input/output, and error handling.
+The second deals with more complicated things like adding new chunks,
+adding new transformations, and generally changing how libpng works.
+
+All of the memory allocation, input/output, and error handling in libpng
+goes through callbacks that are user settable.  The default routines are
+in pngmem.c, pngrio.c, pngwio.c, and pngerror.c respectively.  To change
+these functions, call the appropriate png_set_*_fn() function.
+
+Memory allocation is done through the functions png_large_malloc(),
+png_malloc(), png_realloc(), png_large_free(), and png_free().  These
+currently just call the standard C functions.  The large functions must
+handle exactly 64K, but they don't have to handle more than that.  If
+your pointers can't access more then 64K at a time, you will want to set
+MAXSEG_64K in zlib.h.  Since it is unlikely that the method of handling
+memory allocation on a platform will change between applications, these
+functions must be modified in the library at compile time.
+
+Input/Output in libpng is done through png_read() and png_write(),
+which currently just call fread() and fwrite().  The FILE * is stored in
+png_struct and is initialized via png_init_io().  If you wish to change
+the method of I/O, the library supplies callbacks that you can set
+through the function png_set_read_fn() and png_set_write_fn() at run
+time, instead of calling the png_init_io() function.
+These functions
+also provide a void pointer that can be retrieved via the function
+png_get_io_ptr().  For example:
+
+    png_set_read_fn(png_structp read_ptr,
+        voidp read_io_ptr, png_rw_ptr read_data_fn)
+
+    png_set_write_fn(png_structp write_ptr,
+        voidp write_io_ptr, png_rw_ptr write_data_fn,
+        png_flush_ptr output_flush_fn);
+
+    voidp read_io_ptr = png_get_io_ptr(read_ptr);
+    voidp write_io_ptr = png_get_io_ptr(write_ptr);
+
+The replacement I/O functions must have prototypes as follows:
+
+    void user_read_data(png_structp png_ptr,
+        png_bytep data, png_uint_32 length);
+    void user_write_data(png_structp png_ptr,
+        png_bytep data, png_uint_32 length);
+    void user_flush_data(png_structp png_ptr);
+
+Supplying NULL for the read, write, or flush functions sets them back
+to using the default C stream functions.  It is an error to read from
+a write stream, and vice versa.
+
+Error handling in libpng is done through png_error() and png_warning().
+Errors handled through png_error() are fatal, meaning that png_error()
+should never return to its caller.  Currently, this is handled via
+setjmp() and longjmp() (unless you have compiled libpng with
+PNG_SETJMP_NOT_SUPPORTED, in which case it is handled via PNG_ABORT()),
+but you could change this to do things like exit() if you should wish.
+
+On non-fatal errors, png_warning() is called
+to print a warning message, and then control returns to the calling code.
+By default png_error() and png_warning() print a message on stderr via
+fprintf() unless the library is compiled with PNG_NO_CONSOLE_IO defined
+(because you don't want the messages) or PNG_NO_STDIO defined (because
+fprintf() isn't available).  If you wish to change the behavior of the error
+functions, you will need to set up your own message callbacks.  These
+functions are normally supplied at the time that the png_struct is created.
+It is also possible to redirect errors and warnings to your own replacement
+functions after png_create_*_struct() has been called by calling:
+
+    png_set_error_fn(png_structp png_ptr,
+        png_voidp error_ptr, png_error_ptr error_fn,
+        png_error_ptr warning_fn);
+
+    png_voidp error_ptr = png_get_error_ptr(png_ptr);
+
+If NULL is supplied for either error_fn or warning_fn, then the libpng
+default function will be used, calling fprintf() and/or longjmp() if a
+problem is encountered.  The replacement error functions should have
+parameters as follows:
+
+    void user_error_fn(png_structp png_ptr,
+        png_const_charp error_msg);
+    void user_warning_fn(png_structp png_ptr,
+        png_const_charp warning_msg);
+
+The motivation behind using setjmp() and longjmp() is the C++ throw and
+catch exception handling methods.  This makes the code much easier to write,
+as there is no need to check every return code of every function call.
+However, there are some uncertainties about the status of local variables
+after a longjmp, so the user may want to be careful about doing anything after
+setjmp returns non-zero besides returning itself.  Consult your compiler
+documentation for more details.  For an alternative approach, you may wish
+to use the "cexcept" facility (see http://cexcept.sourceforge.net).
+
+.SS Custom chunks
+
+If you need to read or write custom chunks, you may need to get deeper
+into the libpng code.  The library now has mechanisms for storing
+and writing chunks of unknown type; you can even declare callbacks
+for custom chunks.  Hoewver, this may not be good enough if the
+library code itself needs to know about interactions between your
+chunk and existing `intrinsic' chunks.
+
+If you need to write a new intrinsic chunk, first read the PNG
+specification. Acquire a first level of
+understanding of how it works.  Pay particular attention to the
+sections that describe chunk names, and look at how other chunks were
+designed, so you can do things similarly.  Second, check out the
+sections of libpng that read and write chunks.  Try to find a chunk
+that is similar to yours and use it as a template.  More details can
+be found in the comments inside the code.  It is best to handle unknown
+chunks in a generic method, via callback functions, instead of by
+modifying libpng functions.
+
+If you wish to write your own transformation for the data, look through
+the part of the code that does the transformations, and check out some of
+the simpler ones to get an idea of how they work.  Try to find a similar
+transformation to the one you want to add and copy off of it.  More details
+can be found in the comments inside the code itself.
+
+.SS Configuring for 16 bit platforms
+
+You will want to look into zconf.h to tell zlib (and thus libpng) that
+it cannot allocate more then 64K at a time.  Even if you can, the memory
+won't be accessible.  So limit zlib and libpng to 64K by defining MAXSEG_64K.
+
+.SS Configuring for DOS
+
+For DOS users who only have access to the lower 640K, you will
+have to limit zlib's memory usage via a png_set_compression_mem_level()
+call.  See zlib.h or zconf.h in the zlib library for more information.
+
+.SS Configuring for Medium Model
+
+Libpng's support for medium model has been tested on most of the popular
+compilers.  Make sure MAXSEG_64K gets defined, USE_FAR_KEYWORD gets
+defined, and FAR gets defined to far in pngconf.h, and you should be
+all set.  Everything in the library (except for zlib's structure) is
+expecting far data.  You must use the typedefs with the p or pp on
+the end for pointers (or at least look at them and be careful).  Make
+note that the rows of data are defined as png_bytepp, which is an
+unsigned char far * far *.
+
+.SS Configuring for gui/windowing platforms:
+
+You will need to write new error and warning functions that use the GUI
+interface, as described previously, and set them to be the error and
+warning functions at the time that png_create_*_struct() is called,
+in order to have them available during the structure initialization.
+They can be changed later via png_set_error_fn().  On some compilers,
+you may also have to change the memory allocators (png_malloc, etc.).
+
+.SS Configuring for compiler xxx:
+
+All includes for libpng are in pngconf.h.  If you need to add/change/delete
+an include, this is the place to do it.  The includes that are not
+needed outside libpng are protected by the PNG_INTERNAL definition,
+which is only defined for those routines inside libpng itself.  The
+files in libpng proper only include png.h, which includes pngconf.h.
+
+.SS Configuring zlib:
+
+There are special functions to configure the compression.  Perhaps the
+most useful one changes the compression level, which currently uses
+input compression values in the range 0 - 9.  The library normally
+uses the default compression level (Z_DEFAULT_COMPRESSION = 6).  Tests
+have shown that for a large majority of images, compression values in
+the range 3-6 compress nearly as well as higher levels, and do so much
+faster.  For online applications it may be desirable to have maximum speed
+(Z_BEST_SPEED = 1).  With versions of zlib after v0.99, you can also
+specify no compression (Z_NO_COMPRESSION = 0), but this would create
+files larger than just storing the raw bitmap.  You can specify the
+compression level by calling:
+
+    png_set_compression_level(png_ptr, level);
+
+Another useful one is to reduce the memory level used by the library.
+The memory level defaults to 8, but it can be lowered if you are
+short on memory (running DOS, for example, where you only have 640K).
+
+    png_set_compression_mem_level(png_ptr, level);
+
+The other functions are for configuring zlib.  They are not recommended
+for normal use and may result in writing an invalid PNG file.  See
+zlib.h for more information on what these mean.
+
+    png_set_compression_strategy(png_ptr,
+        strategy);
+    png_set_compression_window_bits(png_ptr,
+        window_bits);
+    png_set_compression_method(png_ptr, method);
+    png_set_compression_buffer_size(png_ptr, size);
+
+.SS Controlling row filtering
+
+If you want to control whether libpng uses filtering or not, which
+filters are used, and how it goes about picking row filters, you
+can call one of these functions.  The selection and configuration
+of row filters can have a significant impact on the size and
+encoding speed and a somewhat lesser impact on the decoding speed
+of an image.  Filtering is enabled by default for RGB and grayscale
+images (with and without alpha), but not for paletted images nor
+for any images with bit depths less than 8 bits/pixel.
+
+The 'method' parameter sets the main filtering method, which is
+currently only '0' in the PNG 1.2 specification.  The 'filters'
+parameter sets which filter(s), if any, should be used for each
+scanline.  Possible values are PNG_ALL_FILTERS and PNG_NO_FILTERS
+to turn filtering on and off, respectively.
+
+Individual filter types are PNG_FILTER_NONE, PNG_FILTER_SUB,
+PNG_FILTER_UP, PNG_FILTER_AVG, PNG_FILTER_PAETH, which can be bitwise
+ORed together with '|' to specify one or more filters to use.
+These filters are described in more detail in the PNG specification.  If
+you intend to change the filter type during the course of writing
+the image, you should start with flags set for all of the filters
+you intend to use so that libpng can initialize its internal
+structures appropriately for all of the filter types.
+
+    filters = PNG_FILTER_NONE | PNG_FILTER_SUB
+              PNG_FILTER_UP | PNG_FILTER_AVE |
+              PNG_FILTER_PAETH | PNG_ALL_FILTERS;
+    or
+    filters = one of PNG_FILTER_VALUE_NONE,
+              PNG_FILTER_VALUE_SUB, PNG_FILTER_VALUE_UP,
+              PNG_FILTER_VALUE_AVE, PNG_FILTER_VALUE_PAETH
+
+    png_set_filter(png_ptr, PNG_FILTER_TYPE_BASE,
+       filters);
+              The second parameter can also be
+              PNG_INTRAPIXEL_DIFFERENCING if you are
+              writing a PNG to be embedded in a MNG
+              datastream.  This parameter must be the
+              same as the value of filter_method used
+              in png_set_IHDR().
+
+It is also possible to influence how libpng chooses from among the
+available filters.  This is done in two ways - by telling it how
+important it is to keep the same filter for successive rows, and
+by telling it the relative computational costs of the filters.
+
+    double weights[3] = {1.5, 1.3, 1.1},
+       costs[PNG_FILTER_VALUE_LAST] =
+       {1.0, 1.3, 1.3, 1.5, 1.7};
+
+    png_set_filter_selection(png_ptr,
+       PNG_FILTER_SELECTION_WEIGHTED, 3,
+       weights, costs);
+
+The weights are multiplying factors that indicate to libpng that the
+row filter should be the same for successive rows unless another row filter
+is that many times better than the previous filter.  In the above example,
+if the previous 3 filters were SUB, SUB, NONE, the SUB filter could have a
+"sum of absolute differences" 1.5 x 1.3 times higher than other filters
+and still be chosen, while the NONE filter could have a sum 1.1 times
+higher than other filters and still be chosen.  Unspecified weights are
+taken to be 1.0, and the specified weights should probably be declining
+like those above in order to emphasize recent filters over older filters.
+
+The filter costs specify for each filter type a relative decoding cost
+to be considered when selecting row filters.  This means that filters
+with higher costs are less likely to be chosen over filters with lower
+costs, unless their "sum of absolute differences" is that much smaller.
+The costs do not necessarily reflect the exact computational speeds of
+the various filters, since this would unduly influence the final image
+size.
+
+Note that the numbers above were invented purely for this example and
+are given only to help explain the function usage.  Little testing has
+been done to find optimum values for either the costs or the weights.
+
+.SS Removing unwanted object code
+
+There are a bunch of #define's in pngconf.h that control what parts of
+libpng are compiled.  All the defines end in _SUPPORTED.  If you are
+never going to use a capability, you can change the #define to #undef
+before recompiling libpng and save yourself code and data space, or
+you can turn off individual capabilities with defines that begin with
+PNG_NO_.
+
+You can also turn all of the transforms and ancillary chunk capabilities
+off en masse with compiler directives that define
+PNG_NO_READ[or WRITE]_TRANSFORMS, or PNG_NO_READ[or WRITE]_ANCILLARY_CHUNKS,
+or all four,
+along with directives to turn on any of the capabilities that you do
+want.  The PNG_NO_READ[or WRITE]_TRANSFORMS directives disable
+the extra transformations but still leave the library fully capable of reading
+and writing PNG files with all known public chunks
+Use of the PNG_NO_READ[or WRITE]_ANCILLARY_CHUNKS directive
+produces a library that is incapable of reading or writing ancillary chunks.
+If you are not using the progressive reading capability, you can
+turn that off with PNG_NO_PROGRESSIVE_READ (don't confuse
+this with the INTERLACING capability, which you'll still have).
+
+All the reading and writing specific code are in separate files, so the
+linker should only grab the files it needs.  However, if you want to
+make sure, or if you are building a stand alone library, all the
+reading files start with pngr and all the writing files start with
+pngw.  The files that don't match either (like png.c, pngtrans.c, etc.)
+are used for both reading and writing, and always need to be included.
+The progressive reader is in pngpread.c
+
+If you are creating or distributing a dynamically linked library (a .so
+or DLL file), you should not remove or disable any parts of the library,
+as this will cause applications linked with different versions of the
+library to fail if they call functions not available in your library.
+The size of the library itself should not be an issue, because only
+those sections that are actually used will be loaded into memory.
+
+.SS Requesting debug printout
+
+The macro definition PNG_DEBUG can be used to request debugging
+printout.  Set it to an integer value in the range 0 to 3.  Higher
+numbers result in increasing amounts of debugging information.  The
+information is printed to the "stderr" file, unless another file
+name is specified in the PNG_DEBUG_FILE macro definition.
+
+When PNG_DEBUG > 0, the following functions (macros) become available:
+
+   png_debug(level, message)
+   png_debug1(level, message, p1)
+   png_debug2(level, message, p1, p2)
+
+in which "level" is compared to PNG_DEBUG to decide whether to print
+the message, "message" is the formatted string to be printed,
+and p1 and p2 are parameters that are to be embedded in the string
+according to printf-style formatting directives.  For example,
+
+   png_debug1(2, "foo=%d\n", foo);
+
+is expanded to
+
+   if(PNG_DEBUG > 2)
+     fprintf(PNG_DEBUG_FILE, "foo=%d\n", foo);
+
+When PNG_DEBUG is defined but is zero, the macros aren't defined, but you
+can still use PNG_DEBUG to control your own debugging:
+
+   #ifdef PNG_DEBUG
+       fprintf(stderr, ...
+   #endif
+
+When PNG_DEBUG = 1, the macros are defined, but only png_debug statements
+having level = 0 will be printed.  There aren't any such statements in
+this version of libpng, but if you insert some they will be printed.
+
+.SH VI.  Runtime optimization
+
+A new feature in libpng 1.2.0 is the ability to dynamically switch between
+standard and optimized versions of some routines.  Currently these are
+limited to three computationally intensive tasks when reading PNG files:
+decoding row filters, expanding interlacing, and combining interlaced or
+transparent row data with previous row data.  Currently the optimized
+versions are available only for x86 (Intel, AMD, etc.) platforms with
+MMX support, though this may change in future versions.  (For example,
+the non-MMX assembler optimizations for zlib might become similarly
+runtime-selectable in future releases, in which case libpng could be
+extended to support them.  Alternatively, the compile-time choice of
+floating-point versus integer routines for gamma correction might become
+runtime-selectable.)
+
+Because such optimizations tend to be very platform- and compiler-dependent,
+both in how they are written and in how they perform, the new runtime code
+in libpng has been written to allow programs to query, enable, and disable
+either specific optimizations or all such optimizations.  For example, to
+enable all possible optimizations (bearing in mind that some "optimizations"
+may actually run more slowly in rare cases):
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       png_uint_32 mask, flags;
+
+       flags = png_get_asm_flags(png_ptr);
+       mask = png_get_asm_flagmask(PNG_SELECT_READ | PNG_SELECT_WRITE);
+       png_set_asm_flags(png_ptr, flags | mask);
+    #endif
+
+To enable only optimizations relevant to reading PNGs, use PNG_SELECT_READ
+by itself when calling png_get_asm_flagmask(); similarly for optimizing
+only writing.  To disable all optimizations:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       flags = png_get_asm_flags(png_ptr);
+       mask = png_get_asm_flagmask(PNG_SELECT_READ | PNG_SELECT_WRITE);
+       png_set_asm_flags(png_ptr, flags & ~mask);
+    #endif
+
+To enable or disable only MMX-related features, use png_get_mmx_flagmask()
+in place of png_get_asm_flagmask().  The mmx version takes one additional
+parameter:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       int selection = PNG_SELECT_READ | PNG_SELECT_WRITE;
+       int compilerID;
+
+       mask = png_get_mmx_flagmask(selection, &compilerID);
+    #endif
+
+On return, compilerID will indicate which version of the MMX assembler
+optimizations was compiled.  Currently two flavors exist:  Microsoft
+Visual C++ (compilerID == 1) and GNU C (a.k.a. gcc/gas, compilerID == 2).
+On non-x86 platforms or on systems compiled without MMX optimizations, a
+value of -1 is used.
+
+Note that both png_get_asm_flagmask() and png_get_mmx_flagmask() return
+all valid, settable optimization bits for the version of the library that's
+currently in use.  In the case of shared (dynamically linked) libraries,
+this may include optimizations that did not exist at the time the code was
+written and compiled.  It is also possible, of course, to enable only known,
+specific optimizations; for example:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       flags = PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  \
+             | PNG_ASM_FLAG_MMX_READ_INTERLACE    \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_SUB   \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_UP    \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_AVG   \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_PAETH ;
+       png_set_asm_flags(png_ptr, flags);
+    #endif
+
+This method would enable only the MMX read-optimizations available at the
+time of libpng 1.2.0's release, regardless of whether a later version of
+the DLL were actually being used.  (Also note that these functions did not
+exist in versions older than 1.2.0, so any attempt to run a dynamically
+linked app on such an older version would fail.)
+
+To determine whether the processor supports MMX instructions at all, use
+the png_mmx_support() function:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       mmxsupport = png_mmx_support();
+    #endif
+
+It returns -1 if MMX support is not compiled into libpng, 0 if MMX code
+is compiled but MMX is not supported by the processor, or 1 if MMX support
+is fully available.  Note that png_mmx_support(), png_get_mmx_flagmask(),
+and png_get_asm_flagmask() all may be called without allocating and ini-
+tializing any PNG structures (for example, as part of a usage screen or
+"about" box).
+
+The following code can be used to prevent an application from using the
+thread_unsafe features, even if libpng was built with PNG_THREAD_UNSAFE_OK
+defined:
+
+#if defined(PNG_USE_PNGGCCRD) && defined(PNG_ASSEMBLER_CODE_SUPPORTED) \
+  && defined(PNG_THREAD_UNSAFE_OK)
+    /* Disable thread-unsafe features of pnggccrd */
+    if (png_access_version() >= 10200)
+    {
+      png_uint_32 mmx_disable_mask = 0;
+      png_uint_32 asm_flags;
+
+      mmx_disable_mask |= ( PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  \
+                          | PNG_ASM_FLAG_MMX_READ_FILTER_SUB   \
+                          | PNG_ASM_FLAG_MMX_READ_FILTER_AVG   \
+                          | PNG_ASM_FLAG_MMX_READ_FILTER_PAETH );
+      asm_flags = png_get_asm_flags(png_ptr);
+      png_set_asm_flags(png_ptr, asm_flags & ~mmx_disable_mask);
+    }
+#endif
+
+For more extensive examples of runtime querying, enabling and disabling
+of optimized features, see contrib/gregbook/readpng2.c in the libpng
+source-code distribution.
+
+
+.SH VII.  MNG support
+
+The MNG specification (available at http://www.libpng.org/pub/mng) allows
+certain extensions to PNG for PNG images that are embedded in MNG datastreams.
+Libpng can support some of these extensions.  To enable them, use the
+png_permit_mng_features() function:
+
+   feature_set = png_permit_mng_features(png_ptr, mask)
+   mask is a png_uint_32 containing the logical OR of the
+        features you want to enable.  These include
+        PNG_FLAG_MNG_EMPTY_PLTE
+        PNG_FLAG_MNG_FILTER_64
+        PNG_ALL_MNG_FEATURES
+   feature_set is a png_32_uint that is the logical AND of
+      your mask with the set of MNG features that is
+      supported by the version of libpng that you are using.
+
+It is an error to use this function when reading or writing a standalone
+PNG file with the PNG 8-byte signature.  The PNG datastream must be wrapped
+in a MNG datastream.  As a minimum, it must have the MNG 8-byte signature
+and the MHDR and MEND chunks.  Libpng does not provide support for these
+or any other MNG chunks; your application must provide its own support for
+them.  You may wish to consider using libmng (available at
+http://www.libmng.com) instead.
+
+.SH VIII.  Changes to Libpng from version 0.88
+
+It should be noted that versions of libpng later than 0.96 are not
+distributed by the original libpng author, Guy Schalnat, nor by
+Andreas Dilger, who had taken over from Guy during 1996 and 1997, and
+distributed versions 0.89 through 0.96, but rather by another member
+of the original PNG Group, Glenn Randers-Pehrson.  Guy and Andreas are
+still alive and well, but they have moved on to other things.
+
+The old libpng functions png_read_init(), png_write_init(),
+png_info_init(), png_read_destroy(), and png_write_destroy() have been
+moved to PNG_INTERNAL in version 0.95 to discourage their use.  These
+functions will be removed from libpng version 2.0.0.
+
+The preferred method of creating and initializing the libpng structures is
+via the png_create_read_struct(), png_create_write_struct(), and
+png_create_info_struct() because they isolate the size of the structures
+from the application, allow version error checking, and also allow the
+use of custom error handling routines during the initialization, which
+the old functions do not.  The functions png_read_destroy() and
+png_write_destroy() do not actually free the memory that libpng
+allocated for these structs, but just reset the data structures, so they
+can be used instead of png_destroy_read_struct() and
+png_destroy_write_struct() if you feel there is too much system overhead
+allocating and freeing the png_struct for each image read.
+
+Setting the error callbacks via png_set_message_fn() before
+png_read_init() as was suggested in libpng-0.88 is no longer supported
+because this caused applications that do not use custom error functions
+to fail if the png_ptr was not initialized to zero.  It is still possible
+to set the error callbacks AFTER png_read_init(), or to change them with
+png_set_error_fn(), which is essentially the same function, but with a new
+name to force compilation errors with applications that try to use the old
+method.
+
+Starting with version 1.0.7, you can find out which version of the library
+you are using at run-time:
+
+   png_uint_32 libpng_vn = png_access_version_number();
+
+The number libpng_vn is constructed from the major version, minor
+version with leading zero, and release number with leading zero,
+(e.g., libpng_vn for version 1.0.7 is 10007).
+
+You can also check which version of png.h you used when compiling your
+application:
+
+   png_uint_32 application_vn = PNG_LIBPNG_VER;
+
+.SH IX. Y2K Compliance in libpng
+
+December 12, 2001
+
+Since the PNG Development group is an ad-hoc body, we can't make
+an official declaration.
+
+This is your unofficial assurance that libpng from version 0.71 and
+upward through 1.2.1 are Y2K compliant.  It is my belief that earlier
+versions were also Y2K compliant.
+
+Libpng only has three year fields.  One is a 2-byte unsigned integer that
+will hold years up to 65535.  The other two hold the date in text
+format, and will hold years up to 9999.
+
+The integer is
+    "png_uint_16 year" in png_time_struct.
+
+The strings are
+    "png_charp time_buffer" in png_struct and
+    "near_time_buffer", which is a local character string in png.c.
+
+There are seven time-related functions:
+
+    png_convert_to_rfc_1123() in png.c
+      (formerly png_convert_to_rfc_1152() in error)
+    png_convert_from_struct_tm() in pngwrite.c, called
+      in pngwrite.c
+    png_convert_from_time_t() in pngwrite.c
+    png_get_tIME() in pngget.c
+    png_handle_tIME() in pngrutil.c, called in pngread.c
+    png_set_tIME() in pngset.c
+    png_write_tIME() in pngwutil.c, called in pngwrite.c
+
+All appear to handle dates properly in a Y2K environment.  The
+png_convert_from_time_t() function calls gmtime() to convert from system
+clock time, which returns (year - 1900), which we properly convert to
+the full 4-digit year.  There is a possibility that applications using
+libpng are not passing 4-digit years into the png_convert_to_rfc_1123()
+function, or that they are incorrectly passing only a 2-digit year
+instead of "year - 1900" into the png_convert_from_struct_tm() function,
+but this is not under our control.  The libpng documentation has always
+stated that it works with 4-digit years, and the APIs have been
+documented as such.
+
+The tIME chunk itself is also Y2K compliant.  It uses a 2-byte unsigned
+integer to hold the year, and can hold years as large as 65535.
+
+zlib, upon which libpng depends, is also Y2K compliant.  It contains
+no date-related code.
+
+
+   Glenn Randers-Pehrson
+   libpng maintainer
+   PNG Development Group
+
+.SH NOTE
+
+Note about libpng version numbers:
+
+Due to various miscommunications, unforeseen code incompatibilities
+and occasional factors outside the authors' control, version numbering
+on the library has not always been consistent and straightforward.
+The following table summarizes matters since version 0.89c, which was
+the first widely used release:
+
+ source             png.h  png.h  shared-lib
+ version            string   int  version
+ -------            ------  ----- ----------
+ 0.89c ("beta 3")  0.89       89  1.0.89
+ 0.90  ("beta 4")  0.90       90  0.90
+ 0.95  ("beta 5")  0.95       95  0.95
+ 0.96  ("beta 6")  0.96       96  0.96
+ 0.97b ("beta 7")  1.00.97    97  1.0.1
+ 0.97c             0.97       97  2.0.97
+ 0.98              0.98       98  2.0.98
+ 0.99              0.99       98  2.0.99
+ 0.99a-m           0.99       99  2.0.99
+ 1.00              1.00      100  2.1.0
+ 1.0.0             1.0.0     100  2.1.0
+ 1.0.0   (from here on, the  100  2.1.0
+ 1.0.1    png.h string is  10001  2.1.0
+ 1.0.1a-e identical to the 10002  from here on, the
+ 1.0.2    source version)  10002  shared library is 2.V
+ 1.0.2a-b                  10003  where V is the source
+ 1.0.1                     10001  code version except as
+ 1.0.1a-e                  10002  2.1.0.1a-e   noted.
+ 1.0.2                     10002  2.1.0.2
+ 1.0.2a-b                  10003  2.1.0.2a-b
+ 1.0.3                     10003  2.1.0.3
+ 1.0.3a-d                  10004  2.1.0.3a-d
+ 1.0.4                     10004  2.1.0.4
+ 1.0.4a-f                  10005  2.1.0.4a-f
+ 1.0.5 (+ 2 patches)       10005  2.1.0.5
+ 1.0.5a-d                  10006  2.1.0.5a-d
+ 1.0.5e-r                  10100  2.1.0.5e-r
+ 1.0.5s-v                  10006  2.1.0.5s-v
+ 1.0.6 (+ 3 patches)       10006  2.1.0.6
+ 1.0.6d-g                  10007  2.1.0.6d-g
+ 1.0.6h                    10007  10.6h
+ 1.0.6i                    10007  10.6i
+ 1.0.6j                    10007  2.1.0.6j
+ 1.0.7beta11-14    DLLNUM  10007  2.1.0.7beta11-14
+ 1.0.7beta15-18       1    10007  2.1.0.7beta15-18
+ 1.0.7rc1-2           1    10007  2.1.0.7rc1-2
+ 1.0.7                1    10007  2.1.0.7
+ 1.0.8beta1-4         1    10008  2.1.0.8beta1-4
+ 1.0.8rc1             1    10008  2.1.0.8rc1
+ 1.0.8                1    10008  2.1.0.8
+ 1.0.9beta1-6         1    10009  2.1.0.9beta1-6
+ 1.0.9rc1             1    10009  2.1.0.9rc1
+ 1.0.9beta7-10        1    10009  2.1.0.9beta7-10
+ 1.0.9rc2             1    10009  2.1.0.9rc2
+ 1.0.9                1    10009  2.1.0.9
+ 1.0.10beta1          1    10010  2.1.0.10beta1
+ 1.0.10rc1            1    10010  2.1.0.10rc1
+ 1.0.10               1    10010  2.1.0.10
+ 1.0.11beta1-3        1    10011  2.1.0.11beta1-3
+ 1.0.11rc1            1    10011  2.1.0.11rc1
+ 1.0.11               1    10011  2.1.0.11
+ 1.0.12beta1-2        2    10012  2.1.0.12beta1-2
+ 1.0.12rc1            2    10012  2.1.0.12rc1
+ 1.0.12               2    10012  2.1.0.12
+ 1.1.0a-f             -    10100  2.1.1.0a-f abandoned
+ 1.2.0beta1-2         2    10200  2.1.2.0beta1-2
+ 1.2.0beta3-5         3    10200  3.1.2.0beta3-5
+ 1.2.0rc1             3    10200  3.1.2.0rc1
+ 1.2.0                3    10200  3.1.2.0
+ 1.2.1beta-4          3    10201  3.1.2.1beta1-4
+ 1.2.1rc1-2           3    10201  3.1.2.1rc1-2
+ 1.2.1                3    10201  3.1.2.1
+
+Henceforth the source version will match the shared-library minor
+and patch numbers; the shared-library major version number will be
+used for changes in backward compatibility, as it is intended.  The
+PNG_PNGLIB_VER macro, which is not used within libpng but is available
+for applications, is an unsigned integer of the form xyyzz corresponding
+to the source version x.y.z (leading zeros in y and z).  Beta versions
+were given the previous public release number plus a letter, until
+version 1.0.6j; from then on they were given the upcoming public
+release number plus "betaNN" or "rcN".
+
+.SH "SEE ALSO"
+.BR libpngpf (3),
+.BR png (5).
+.LP
+.IR libpng:
+.IP
+ftp://ftp.uu.net/graphics/png
+http://www.libpng.org/pub/png
+
+.LP
+.IR zlib:
+.IP
+(generally) at the same location as
+.I libpng
+or at
+.br
+ftp://ftp.uu.net/pub/archiving/zip/zlib
+.br
+ftp://ftp.info-zip.org/pub/infozip/zlib
+
+.LP
+.IR "PNG specification":
+RFC 2083
+.IP
+(generally) at the same location as
+.I libpng
+or at
+.br
+ftp://ds.internic.net/rfc/rfc2083.txt
+.br
+or (as a W3C Recommendation) at
+.br
+http://www.w3.org/TR/REC-png.html
+
+.LP
+In the case of any inconsistency between the PNG specification
+and this library, the specification takes precedence.
+
+.SH AUTHORS
+This man page: Glenn Randers-Pehrson
+<randeg@alum.rpi.edu>
+
+The contributing authors would like to thank all those who helped
+with testing, bug fixes, and patience.  This wouldn't have been
+possible without all of you.
+
+Thanks to Frank J. T. Wojcik for helping with the documentation.
+
+Libpng version 1.2.1 - December 12, 2001:
+Initially created in 1995 by Guy Eric Schalnat, then of Group 42, Inc.
+Currently maintained by Glenn Randers-Pehrson (randeg@alum.rpi.edu).
+
+Supported by the PNG development group
+.br
+(png-implement@ccrc.wustl.edu).
+
+.SH COPYRIGHT NOTICE, DISCLAIMER, and LICENSE:
+
+(This copy of the libpng notices is provided for your convenience.  In case of
+any discrepancy between this copy and the notices in the file png.h that is
+included in the libpng distribution, the latter shall prevail.)
+
+If you modify libpng you may insert additional notices immediately following
+this sentence.
+
+libpng versions 1.0.7, July 1, 2000, through  1.2.1, December 12, 2001, are
+Copyright (c) 2000-2001 Glenn Randers-Pehrson, and are
+distributed according to the same disclaimer and license as libpng-1.0.6
+with the following individuals added to the list of Contributing Authors
+
+   Simon-Pierre Cadieux
+   Eric S. Raymond
+   Gilles Vollant
+
+and with the following additions to the disclaimer:
+
+   There is no warranty against interference with your
+   enjoyment of the library or against infringement.
+   There is no warranty that our efforts or the library
+   will fulfill any of your particular purposes or needs.
+   This library is provided with all faults, and the entire
+   risk of satisfactory quality, performance, accuracy, and
+   effort is with the user.
+
+libpng versions 0.97, January 1998, through 1.0.6, March 20, 2000, are
+Copyright (c) 1998, 1999 Glenn Randers-Pehrson
+Distributed according to the same disclaimer and license as libpng-0.96,
+with the following individuals added to the list of Contributing Authors:
+
+   Tom Lane
+   Glenn Randers-Pehrson
+   Willem van Schaik
+
+libpng versions 0.89, June 1996, through 0.96, May 1997, are
+Copyright (c) 1996, 1997 Andreas Dilger
+Distributed according to the same disclaimer and license as libpng-0.88,
+with the following individuals added to the list of Contributing Authors:
+
+   John Bowler
+   Kevin Bracey
+   Sam Bushell
+   Magnus Holmgren
+   Greg Roelofs
+   Tom Tanner
+
+libpng versions 0.5, May 1995, through 0.88, January 1996, are
+Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.
+
+For the purposes of this copyright and license, "Contributing Authors"
+is defined as the following set of individuals:
+
+   Andreas Dilger
+   Dave Martindale
+   Guy Eric Schalnat
+   Paul Schmidt
+   Tim Wegner
+
+The PNG Reference Library is supplied "AS IS".  The Contributing Authors
+and Group 42, Inc. disclaim all warranties, expressed or implied,
+including, without limitation, the warranties of merchantability and of
+fitness for any purpose.  The Contributing Authors and Group 42, Inc.
+assume no liability for direct, indirect, incidental, special, exemplary,
+or consequential damages, which may result from the use of the PNG
+Reference Library, even if advised of the possibility of such damage.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+source code, or portions hereof, for any purpose, without fee, subject
+to the following restrictions:
+
+1. The origin of this source code must not be misrepresented.
+
+2. Altered versions must be plainly marked as such and
+   must not be misrepresented as being the original source.
+
+3. This Copyright notice may not be removed or altered from
+   any source or altered source distribution.
+
+The Contributing Authors and Group 42, Inc. specifically permit, without
+fee, and encourage the use of this source code as a component to
+supporting the PNG file format in commercial products.  If you use this
+source code in a product, acknowledgment is not required but would be
+appreciated.
+
+
+A "png_get_copyright" function is available, for convenient use in "about"
+boxes and the like:
+
+   printf("%s",png_get_copyright(NULL));
+
+Also, the PNG logo (in PNG format, of course) is supplied in the
+files "pngbar.png" and "pngbar.jpg (88x31) and "pngnow.png" (98x31).
+
+Libpng is OSI Certified Open Source Software.  OSI Certified Open Source is a
+certification mark of the Open Source Initiative.
+
+Glenn Randers-Pehrson
+randeg@alum.rpi.edu
+December 12, 2001
+
+.\" end of man page
+
diff --git a/png/libpng.txt b/png/libpng.txt
new file mode 100644
index 000000000..a5f1be4b1
--- /dev/null
+++ b/png/libpng.txt
@@ -0,0 +1,2832 @@
+libpng.txt - A description on how to use and modify libpng
+
+ libpng version 1.2.1 - December 12, 2001
+ Updated and distributed by Glenn Randers-Pehrson
+ <randeg@alum.rpi.edu>
+ Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ For conditions of distribution and use, see copyright
+ notice in png.h.
+
+ based on:
+
+ libpng 1.0 beta 6  version 0.96 May 28, 1997
+ Updated and distributed by Andreas Dilger
+ Copyright (c) 1996, 1997 Andreas Dilger
+
+ libpng 1.0 beta 2 - version 0.88  January 26, 1996
+ For conditions of distribution and use, see copyright
+ notice in png.h. Copyright (c) 1995, 1996 Guy Eric
+ Schalnat, Group 42, Inc.
+
+ Updated/rewritten per request in the libpng FAQ
+ Copyright (c) 1995, 1996 Frank J. T. Wojcik
+ December 18, 1995 & January 20, 1996
+
+I. Introduction
+
+This file describes how to use and modify the PNG reference library
+(known as libpng) for your own use.  There are five sections to this
+file: introduction, structures, reading, writing, and modification and
+configuration notes for various special platforms.  In addition to this
+file, example.c is a good starting point for using the library, as
+it is heavily commented and should include everything most people
+will need.  We assume that libpng is already installed; see the
+INSTALL file for instructions on how to install libpng.
+
+Libpng was written as a companion to the PNG specification, as a way
+of reducing the amount of time and effort it takes to support the PNG
+file format in application programs.
+
+The PNG-1.2 specification is available at <http://www.libpng.org/pub/png>
+and at <ftp://ftp.uu.net/graphics/png/documents/>.
+
+The PNG-1.0 specification is available
+as RFC 2083 <ftp://ftp.uu.net/graphics/png/documents/> and as a
+W3C Recommendation <http://www.w3.org/TR/REC.png.html>. Some
+additional chunks are described in the special-purpose public chunks
+documents at <ftp://ftp.uu.net/graphics/png/documents/>.
+
+Other information
+about PNG, and the latest version of libpng, can be found at the PNG home
+page, <http://www.libpng.org/pub/png/>
+and at <ftp://ftp.uu.net/graphics/png/>.
+
+Most users will not have to modify the library significantly; advanced
+users may want to modify it more.  All attempts were made to make it as
+complete as possible, while keeping the code easy to understand.
+Currently, this library only supports C.  Support for other languages
+is being considered.
+
+Libpng has been designed to handle multiple sessions at one time,
+to be easily modifiable, to be portable to the vast majority of
+machines (ANSI, K&R, 16-, 32-, and 64-bit) available, and to be easy
+to use.  The ultimate goal of libpng is to promote the acceptance of
+the PNG file format in whatever way possible.  While there is still
+work to be done (see the TODO file), libpng should cover the
+majority of the needs of its users.
+
+Libpng uses zlib for its compression and decompression of PNG files.
+Further information about zlib, and the latest version of zlib, can
+be found at the zlib home page, <http://www.info-zip.org/pub/infozip/zlib/>.
+The zlib compression utility is a general purpose utility that is
+useful for more than PNG files, and can be used without libpng.
+See the documentation delivered with zlib for more details.
+You can usually find the source files for the zlib utility wherever you
+find the libpng source files.
+
+Libpng is thread safe, provided the threads are using different
+instances of the structures.  Each thread should have its own
+png_struct and png_info instances, and thus its own image.
+Libpng does not protect itself against two threads using the
+same instance of a structure.  Note: thread safety may be defeated
+by use of some of the MMX assembler code in pnggccrd.c, which is only
+compiled when the user defines PNG_THREAD_UNSAFE_OK.
+
+
+II. Structures
+
+There are two main structures that are important to libpng, png_struct
+and png_info.  The first, png_struct, is an internal structure that
+will not, for the most part, be used by a user except as the first
+variable passed to every libpng function call.
+
+The png_info structure is designed to provide information about the
+PNG file.  At one time, the fields of png_info were intended to be
+directly accessible to the user.  However, this tended to cause problems
+with applications using dynamically loaded libraries, and as a result
+a set of interface functions for png_info (the png_get_*() and png_set_*()
+functions) was developed.  The fields of png_info are still available for
+older applications, but it is suggested that applications use the new
+interfaces if at all possible.
+
+Applications that do make direct access to the members of png_struct (except
+for png_ptr->jmpbuf) must be recompiled whenever the library is updated,
+and applications that make direct access to the members of png_info must
+be recompiled if they were compiled or loaded with libpng version 1.0.6,
+in which the members were in a different order.  In version 1.0.7, the
+members of the png_info structure reverted to the old order, as they were
+in versions 0.97c through 1.0.5.  Starting with version 2.0.0, both
+structures are going to be hidden, and the contents of the structures will
+only be accessible through the png_get/png_set functions.
+
+The png.h header file is an invaluable reference for programming with libpng.
+And while I'm on the topic, make sure you include the libpng header file:
+
+#include <png.h>
+
+III. Reading
+
+We'll now walk you through the possible functions to call when reading
+in a PNG file sequentially, briefly explaining the syntax and purpose
+of each one.  See example.c and png.h for more detail.  While
+progressive reading is covered in the next section, you will still
+need some of the functions discussed in this section to read a PNG
+file.
+
+Setup
+
+You will want to do the I/O initialization(*) before you get into libpng,
+so if it doesn't work, you don't have much to undo.  Of course, you
+will also want to insure that you are, in fact, dealing with a PNG
+file.  Libpng provides a simple check to see if a file is a PNG file.
+To use it, pass in the first 1 to 8 bytes of the file to the function
+png_sig_cmp(), and it will return 0 if the bytes match the corresponding
+bytes of the PNG signature, or nonzero otherwise.  Of course, the more bytes
+you pass in, the greater the accuracy of the prediction.
+
+If you are intending to keep the file pointer open for use in libpng,
+you must ensure you don't read more than 8 bytes from the beginning
+of the file, and you also have to make a call to png_set_sig_bytes_read()
+with the number of bytes you read from the beginning.  Libpng will
+then only check the bytes (if any) that your program didn't read.
+
+(*): If you are not using the standard I/O functions, you will need
+to replace them with custom functions.  See the discussion under
+Customizing libpng.
+
+
+    FILE *fp = fopen(file_name, "rb");
+    if (!fp)
+    {
+        return (ERROR);
+    }
+    fread(header, 1, number, fp);
+    is_png = !png_sig_cmp(header, 0, number);
+    if (!is_png)
+    {
+        return (NOT_PNG);
+    }
+
+
+Next, png_struct and png_info need to be allocated and initialized.  In
+order to ensure that the size of these structures is correct even with a
+dynamically linked libpng, there are functions to initialize and
+allocate the structures.  We also pass the library version, optional
+pointers to error handling functions, and a pointer to a data struct for
+use by the error functions, if necessary (the pointer and functions can
+be NULL if the default error handlers are to be used).  See the section
+on Changes to Libpng below regarding the old initialization functions.
+The structure allocation functions quietly return NULL if they fail to
+create the structure, so your application should check for that.
+
+    png_structp png_ptr = png_create_read_struct
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn);
+    if (!png_ptr)
+        return (ERROR);
+
+    png_infop info_ptr = png_create_info_struct(png_ptr);
+    if (!info_ptr)
+    {
+        png_destroy_read_struct(&png_ptr,
+           (png_infopp)NULL, (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    png_infop end_info = png_create_info_struct(png_ptr);
+    if (!end_info)
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+          (png_infopp)NULL);
+        return (ERROR);
+    }
+
+If you want to use your own memory allocation routines,
+define PNG_USER_MEM_SUPPORTED and use
+png_create_read_struct_2() instead of png_create_read_struct():
+
+    png_structp png_ptr = png_create_read_struct_2
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn, (png_voidp)
+        user_mem_ptr, user_malloc_fn, user_free_fn);
+
+The error handling routines passed to png_create_read_struct()
+and the memory alloc/free routines passed to png_create_struct_2()
+are only necessary if you are not using the libpng supplied error
+handling and memory alloc/free functions.
+
+When libpng encounters an error, it expects to longjmp back
+to your routine.  Therefore, you will need to call setjmp and pass
+your png_jmpbuf(png_ptr).  If you read the file from different
+routines, you will need to update the jmpbuf field every time you enter
+a new routine that will call a png_*() function.
+
+See your documentation of setjmp/longjmp for your compiler for more
+information on setjmp/longjmp.  See the discussion on libpng error
+handling in the Customizing Libpng section below for more information
+on the libpng error handling.  If an error occurs, and libpng longjmp's
+back to your setjmp, you will want to call png_destroy_read_struct() to
+free any memory.
+
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+           &end_info);
+        fclose(fp);
+        return (ERROR);
+    }
+
+If you would rather avoid the complexity of setjmp/longjmp issues,
+you can compile libpng with PNG_SETJMP_NOT_SUPPORTED, in which case
+errors will result in a call to PNG_ABORT() which defaults to abort().
+
+Now you need to set up the input code.  The default for libpng is to
+use the C function fread().  If you use this, you will need to pass a
+valid FILE * in the function png_init_io().  Be sure that the file is
+opened in binary mode.  If you wish to handle reading data in another
+way, you need not call the png_init_io() function, but you must then
+implement the libpng I/O methods discussed in the Customizing Libpng
+section below.
+
+    png_init_io(png_ptr, fp);
+
+If you had previously opened the file and read any of the signature from
+the beginning in order to see if this was a PNG file, you need to let
+libpng know that there are some bytes missing from the start of the file.
+
+    png_set_sig_bytes(png_ptr, number);
+
+Setting up callback code
+
+You can set up a callback function to handle any unknown chunks in the
+input stream. You must supply the function
+
+    read_chunk_callback(png_ptr ptr,
+         png_unknown_chunkp chunk);
+    {
+       /* The unknown chunk structure contains your
+          chunk data: */
+           png_byte name[5];
+           png_byte *data;
+           png_size_t size;
+       /* Note that libpng has already taken care of
+          the CRC handling */
+
+       /* put your code here.  Return one of the
+          following: */
+
+       return (-n); /* chunk had an error */
+       return (0); /* did not recognize */
+       return (n); /* success */
+    }
+
+(You can give your function another name that you like instead of
+"read_chunk_callback")
+
+To inform libpng about your function, use
+
+    png_set_read_user_chunk_fn(png_ptr, user_chunk_ptr,
+        read_chunk_callback);
+
+This names not only the callback function, but also a user pointer that
+you can retrieve with
+
+    png_get_user_chunk_ptr(png_ptr);
+
+At this point, you can set up a callback function that will be
+called after each row has been read, which you can use to control
+a progress meter or the like.  It's demonstrated in pngtest.c.
+You must supply a function
+
+    void read_row_callback(png_ptr ptr, png_uint_32 row,
+       int pass);
+    {
+      /* put your code here */
+    }
+
+(You can give it another name that you like instead of "read_row_callback")
+
+To inform libpng about your function, use
+
+    png_set_read_status_fn(png_ptr, read_row_callback);
+
+Unknown-chunk handling
+
+Now you get to set the way the library processes unknown chunks in the
+input PNG stream. Both known and unknown chunks will be read.  Normal
+behavior is that known chunks will be parsed into information in
+various info_ptr members; unknown chunks will be discarded. To change
+this, you can call:
+
+    png_set_keep_unknown_chunks(png_ptr, info_ptr, keep,
+        chunk_list, num_chunks);
+    keep       - 0: do not keep
+                 1: keep only if safe-to-copy
+                 2: keep even if unsafe-to-copy
+    chunk_list - list of chunks affected (a byte string,
+                 five bytes per chunk, NULL or '\0' if
+                 num_chunks is 0)
+    num_chunks - number of chunks affected; if 0, all
+                 unknown chunks are affected
+
+Unknown chunks declared in this way will be saved as raw data onto a
+list of png_unknown_chunk structures.  If a chunk that is normally
+known to libpng is named in the list, it will be handled as unknown,
+according to the "keep" directive.  If a chunk is named in successive
+instances of png_set_keep_unknown_chunks(), the final instance will
+take precedence.
+
+The high-level read interface
+
+At this point there are two ways to proceed; through the high-level
+read interface, or through a sequence of low-level read operations.
+You can use the high-level interface if (a) you are willing to read
+the entire image into memory, and (b) the input transformations
+you want to do are limited to the following set:
+
+    PNG_TRANSFORM_IDENTITY      No transformation
+    PNG_TRANSFORM_STRIP_16      Strip 16-bit samples to
+                                8 bits
+    PNG_TRANSFORM_STRIP_ALPHA   Discard the alpha channel
+    PNG_TRANSFORM_PACKING       Expand 1, 2 and 4-bit
+                                samples to bytes
+    PNG_TRANSFORM_PACKSWAP      Change order of packed
+                                pixels to LSB first
+    PNG_TRANSFORM_EXPAND        Perform set_expand()
+    PNG_TRANSFORM_INVERT_MONO   Invert monochrome images
+    PNG_TRANSFORM_SHIFT         Normalize pixels to the
+                                sBIT depth
+    PNG_TRANSFORM_BGR           Flip RGB to BGR, RGBA
+                                to BGRA
+    PNG_TRANSFORM_SWAP_ALPHA    Flip RGBA to ARGB or GA
+                                to AG
+    PNG_TRANSFORM_INVERT_ALPHA  Change alpha from opacity
+                                to transparency
+    PNG_TRANSFORM_SWAP_ENDIAN   Byte-swap 16-bit samples
+
+(This excludes setting a background color, doing gamma transformation,
+dithering, and setting filler.)  If this is the case, simply do this:
+
+    png_read_png(png_ptr, info_ptr, png_transforms, NULL)
+
+where png_transforms is an integer containing the logical OR of
+some set of transformation flags.  This call is equivalent to png_read_info(),
+followed the set of transformations indicated by the transform mask,
+then png_read_image(), and finally png_read_end().
+
+(The final parameter of this call is not yet used.  Someday it might point
+to transformation parameters required by some future input transform.)
+
+After you have called png_read_png(), you can retrieve the image data
+with
+
+   row_pointers = png_get_rows(png_ptr, info_ptr);
+
+where row_pointers is an array of pointers to the pixel data for each row:
+
+   png_bytep row_pointers[height];
+
+If you know your image size and pixel size ahead of time, you can allocate
+row_pointers prior to calling png_read_png() with
+
+   row_pointers = png_malloc(png_ptr,
+      height*sizeof(png_bytep));
+   for (int i=0; i<height, i++)
+      row_pointers[i]=png_malloc(png_ptr,
+         width*pixel_size);
+   png_set_rows(png_ptr, info_ptr, &row_pointers);
+
+Alternatively you could allocate your image in one big block and define
+row_pointers[i] to point into the proper places in your block.
+
+If you use png_set_rows(), the application is responsible for freeing
+row_pointers (and row_pointers[i], if they were separately allocated).
+
+If you don't allocate row_pointers ahead of time, png_read_png() will
+do it, and it'll be free'ed when you call png_destroy_*().
+
+The low-level read interface
+
+If you are going the low-level route, you are now ready to read all
+the file information up to the actual image data.  You do this with a
+call to png_read_info().
+
+    png_read_info(png_ptr, info_ptr);
+
+This will process all chunks up to but not including the image data.
+
+Querying the info structure
+
+Functions are used to get the information from the info_ptr once it
+has been read.  Note that these fields may not be completely filled
+in until png_read_end() has read the chunk data following the image.
+
+    png_get_IHDR(png_ptr, info_ptr, &width, &height,
+       &bit_depth, &color_type, &interlace_type,
+       &compression_type, &filter_method);
+
+    width          - holds the width of the image
+                     in pixels (up to 2^31).
+    height         - holds the height of the image
+                     in pixels (up to 2^31).
+    bit_depth      - holds the bit depth of one of the
+                     image channels.  (valid values are
+                     1, 2, 4, 8, 16 and depend also on
+                     the color_type.  See also
+                     significant bits (sBIT) below).
+    color_type     - describes which color/alpha channels
+                         are present.
+                     PNG_COLOR_TYPE_GRAY
+                        (bit depths 1, 2, 4, 8, 16)
+                     PNG_COLOR_TYPE_GRAY_ALPHA
+                        (bit depths 8, 16)
+                     PNG_COLOR_TYPE_PALETTE
+                        (bit depths 1, 2, 4, 8)
+                     PNG_COLOR_TYPE_RGB
+                        (bit_depths 8, 16)
+                     PNG_COLOR_TYPE_RGB_ALPHA
+                        (bit_depths 8, 16)
+
+                     PNG_COLOR_MASK_PALETTE
+                     PNG_COLOR_MASK_COLOR
+                     PNG_COLOR_MASK_ALPHA
+
+    filter_method  - (must be PNG_FILTER_TYPE_BASE
+                     for PNG 1.0, and can also be
+                     PNG_INTRAPIXEL_DIFFERENCING if
+                     the PNG datastream is embedded in
+                     a MNG-1.0 datastream)
+    compression_type - (must be PNG_COMPRESSION_TYPE_BASE
+                     for PNG 1.0)
+    interlace_type - (PNG_INTERLACE_NONE or
+                     PNG_INTERLACE_ADAM7)
+    Any or all of interlace_type, compression_type, of
+    filter_method can be NULL if you are
+    not interested in their values.
+
+    channels = png_get_channels(png_ptr, info_ptr);
+    channels       - number of channels of info for the
+                     color type (valid values are 1 (GRAY,
+                     PALETTE), 2 (GRAY_ALPHA), 3 (RGB),
+                     4 (RGB_ALPHA or RGB + filler byte))
+    rowbytes = png_get_rowbytes(png_ptr, info_ptr);
+    rowbytes       - number of bytes needed to hold a row
+
+    signature = png_get_signature(png_ptr, info_ptr);
+    signature      - holds the signature read from the
+                     file (if any).  The data is kept in
+                     the same offset it would be if the
+                     whole signature were read (i.e. if an
+                     application had already read in 4
+                     bytes of signature before starting
+                     libpng, the remaining 4 bytes would
+                     be in signature[4] through signature[7]
+                     (see png_set_sig_bytes())).
+
+
+    width            = png_get_image_width(png_ptr,
+                         info_ptr);
+    height           = png_get_image_height(png_ptr,
+                         info_ptr);
+    bit_depth        = png_get_bit_depth(png_ptr,
+                         info_ptr);
+    color_type       = png_get_color_type(png_ptr,
+                         info_ptr);
+    filter_method    = png_get_filter_type(png_ptr,
+                         info_ptr);
+    compression_type = png_get_compression_type(png_ptr,
+                         info_ptr);
+    interlace_type   = png_get_interlace_type(png_ptr,
+                         info_ptr);
+
+
+These are also important, but their validity depends on whether the chunk
+has been read.  The png_get_valid(png_ptr, info_ptr, PNG_INFO_<chunk>) and
+png_get_<chunk>(png_ptr, info_ptr, ...) functions return non-zero if the
+data has been read, or zero if it is missing.  The parameters to the
+png_get_<chunk> are set directly if they are simple data types, or a pointer
+into the info_ptr is returned for any complex types.
+
+    png_get_PLTE(png_ptr, info_ptr, &palette,
+                     &num_palette);
+    palette        - the palette for the file
+                     (array of png_color)
+    num_palette    - number of entries in the palette
+
+    png_get_gAMA(png_ptr, info_ptr, &gamma);
+    gamma          - the gamma the file is written
+                     at (PNG_INFO_gAMA)
+
+    png_get_sRGB(png_ptr, info_ptr, &srgb_intent);
+    srgb_intent    - the rendering intent (PNG_INFO_sRGB)
+                     The presence of the sRGB chunk
+                     means that the pixel data is in the
+                     sRGB color space.  This chunk also
+                     implies specific values of gAMA and
+                     cHRM.
+
+    png_get_iCCP(png_ptr, info_ptr, &name,
+       &compression_type, &profile, &proflen);
+    name            - The profile name.
+    compression     - The compression type; always
+                      PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
+                      You may give NULL to this argument to
+                      ignore it.
+    profile         - International Color Consortium color
+                      profile data. May contain NULs.
+    proflen         - length of profile data in bytes.
+
+    png_get_sBIT(png_ptr, info_ptr, &sig_bit);
+    sig_bit        - the number of significant bits for
+                     (PNG_INFO_sBIT) each of the gray,
+                     red, green, and blue channels,
+                     whichever are appropriate for the
+                     given color type (png_color_16)
+
+    png_get_tRNS(png_ptr, info_ptr, &trans, &num_trans,
+                     &trans_values);
+    trans          - array of transparent entries for
+                     palette (PNG_INFO_tRNS)
+    trans_values   - graylevel or color sample values of
+                     the single transparent color for
+                     non-paletted images (PNG_INFO_tRNS)
+    num_trans      - number of transparent entries
+                     (PNG_INFO_tRNS)
+
+    png_get_hIST(png_ptr, info_ptr, &hist);
+                     (PNG_INFO_hIST)
+    hist           - histogram of palette (array of
+                     png_uint_16)
+
+    png_get_tIME(png_ptr, info_ptr, &mod_time);
+    mod_time       - time image was last modified
+                    (PNG_VALID_tIME)
+
+    png_get_bKGD(png_ptr, info_ptr, &background);
+    background     - background color (PNG_VALID_bKGD)
+                     valid 16-bit red, green and blue
+                     values, regardless of color_type
+
+    num_comments   = png_get_text(png_ptr, info_ptr,
+                     &text_ptr, &num_text);
+    num_comments   - number of comments
+    text_ptr       - array of png_text holding image
+                     comments
+    text_ptr[i].compression - type of compression used
+                 on "text" PNG_TEXT_COMPRESSION_NONE
+                           PNG_TEXT_COMPRESSION_zTXt
+                           PNG_ITXT_COMPRESSION_NONE
+                           PNG_ITXT_COMPRESSION_zTXt
+    text_ptr[i].key   - keyword for comment.  Must contain
+                         1-79 characters.
+    text_ptr[i].text  - text comments for current
+                         keyword.  Can be empty.
+    text_ptr[i].text_length - length of text string,
+                 after decompression, 0 for iTXt
+    text_ptr[i].itxt_length - length of itxt string,
+                 after decompression, 0 for tEXt/zTXt
+    text_ptr[i].lang  - language of comment (empty
+                         string for unknown).
+    text_ptr[i].translated_keyword  - keyword in UTF-8
+                         (empty string for unknown).
+    num_text       - number of comments (same as
+                     num_comments; you can put NULL here
+                     to avoid the duplication)
+    Note while png_set_text() will accept text, language,
+    and translated keywords that can be NULL pointers, the
+    structure returned by png_get_text will always contain
+    regular zero-terminated C strings.  They might be
+    empty strings but they will never be NULL pointers.
+
+    num_spalettes = png_get_sPLT(png_ptr, info_ptr,
+       &palette_ptr);
+    palette_ptr    - array of palette structures holding
+                     contents of one or more sPLT chunks
+                     read.
+    num_spalettes  - number of sPLT chunks read.
+
+    png_get_oFFs(png_ptr, info_ptr, &offset_x, &offset_y,
+       &unit_type);
+    offset_x       - positive offset from the left edge
+                     of the screen
+    offset_y       - positive offset from the top edge
+                     of the screen
+    unit_type      - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
+
+    png_get_pHYs(png_ptr, info_ptr, &res_x, &res_y,
+       &unit_type);
+    res_x          - pixels/unit physical resolution in
+                     x direction
+    res_y          - pixels/unit physical resolution in
+                     x direction
+    unit_type      - PNG_RESOLUTION_UNKNOWN,
+                     PNG_RESOLUTION_METER
+
+    png_get_sCAL(png_ptr, info_ptr, &unit, &width,
+       &height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                 (width and height are doubles)
+
+    png_get_sCAL_s(png_ptr, info_ptr, &unit, &width,
+       &height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                 (width and height are strings like "2.54")
+
+    num_unknown_chunks = png_get_unknown_chunks(png_ptr,
+       info_ptr, &unknowns)
+    unknowns          - array of png_unknown_chunk
+                        structures holding unknown chunks
+    unknowns[i].name  - name of unknown chunk
+    unknowns[i].data  - data of unknown chunk
+    unknowns[i].size  - size of unknown chunk's data
+    unknowns[i].location - position of chunk in file
+
+    The value of "i" corresponds to the order in which the
+    chunks were read from the PNG file or inserted with the
+    png_set_unknown_chunks() function.
+
+The data from the pHYs chunk can be retrieved in several convenient
+forms:
+
+    res_x = png_get_x_pixels_per_meter(png_ptr,
+       info_ptr)
+    res_y = png_get_y_pixels_per_meter(png_ptr,
+       info_ptr)
+    res_x_and_y = png_get_pixels_per_meter(png_ptr,
+       info_ptr)
+    res_x = png_get_x_pixels_per_inch(png_ptr,
+       info_ptr)
+    res_y = png_get_y_pixels_per_inch(png_ptr,
+       info_ptr)
+    res_x_and_y = png_get_pixels_per_inch(png_ptr,
+       info_ptr)
+    aspect_ratio = png_get_pixel_aspect_ratio(png_ptr,
+       info_ptr)
+
+   (Each of these returns 0 [signifying "unknown"] if
+       the data is not present or if res_x is 0;
+       res_x_and_y is 0 if res_x != res_y)
+
+The data from the oFFs chunk can be retrieved in several convenient
+forms:
+
+    x_offset = png_get_x_offset_microns(png_ptr, info_ptr);
+    y_offset = png_get_y_offset_microns(png_ptr, info_ptr);
+    x_offset = png_get_x_offset_inches(png_ptr, info_ptr);
+    y_offset = png_get_y_offset_inches(png_ptr, info_ptr);
+
+   (Each of these returns 0 [signifying "unknown" if both
+       x and y are 0] if the data is not present or if the
+       chunk is present but the unit is the pixel)
+
+For more information, see the png_info definition in png.h and the
+PNG specification for chunk contents.  Be careful with trusting
+rowbytes, as some of the transformations could increase the space
+needed to hold a row (expand, filler, gray_to_rgb, etc.).
+See png_read_update_info(), below.
+
+A quick word about text_ptr and num_text.  PNG stores comments in
+keyword/text pairs, one pair per chunk, with no limit on the number
+of text chunks, and a 2^31 byte limit on their size.  While there are
+suggested keywords, there is no requirement to restrict the use to these
+strings.  It is strongly suggested that keywords and text be sensible
+to humans (that's the point), so don't use abbreviations.  Non-printing
+symbols are not allowed.  See the PNG specification for more details.
+There is also no requirement to have text after the keyword.
+
+Keywords should be limited to 79 Latin-1 characters without leading or
+trailing spaces, but non-consecutive spaces are allowed within the
+keyword.  It is possible to have the same keyword any number of times.
+The text_ptr is an array of png_text structures, each holding a
+pointer to a language string, a pointer to a keyword and a pointer to
+a text string.  The text string, language code, and translated
+keyword may be empty or NULL pointers.  The keyword/text
+pairs are put into the array in the order that they are received.
+However, some or all of the text chunks may be after the image, so, to
+make sure you have read all the text chunks, don't mess with these
+until after you read the stuff after the image.  This will be
+mentioned again below in the discussion that goes with png_read_end().
+
+Input transformations
+
+After you've read the header information, you can set up the library
+to handle any special transformations of the image data.  The various
+ways to transform the data will be described in the order that they
+should occur.  This is important, as some of these change the color
+type and/or bit depth of the data, and some others only work on
+certain color types and bit depths.  Even though each transformation
+checks to see if it has data that it can do something with, you should
+make sure to only enable a transformation if it will be valid for the
+data.  For example, don't swap red and blue on grayscale data.
+
+The colors used for the background and transparency values should be
+supplied in the same format/depth as the current image data.  They
+are stored in the same format/depth as the image data in a bKGD or tRNS
+chunk, so this is what libpng expects for this data.  The colors are
+transformed to keep in sync with the image data when an application
+calls the png_read_update_info() routine (see below).
+
+Data will be decoded into the supplied row buffers packed into bytes
+unless the library has been told to transform it into another format.
+For example, 4 bit/pixel paletted or grayscale data will be returned
+2 pixels/byte with the leftmost pixel in the high-order bits of the
+byte, unless png_set_packing() is called.  8-bit RGB data will be stored
+in RGB RGB RGB format unless png_set_filler() is called to insert filler
+bytes, either before or after each RGB triplet.  16-bit RGB data will
+be returned RRGGBB RRGGBB, with the most significant byte of the color
+value first, unless png_set_strip_16() is called to transform it to
+regular RGB RGB triplets, or png_set_filler() is called to insert
+filler bytes, either before or after each RRGGBB triplet.  Similarly,
+8-bit or 16-bit grayscale data can be modified with png_set_filler()
+or png_set_strip_16().
+
+The following code transforms grayscale images of less than 8 to 8 bits,
+changes paletted images to RGB, and adds a full alpha channel if there is
+transparency information in a tRNS chunk.  This is most useful on
+grayscale images with bit depths of 2 or 4 or if there is a multiple-image
+viewing application that wishes to treat all images in the same way.
+
+    if (color_type == PNG_COLOR_TYPE_PALETTE)
+        png_set_palette_to_rgb(png_ptr);
+
+    if (color_type == PNG_COLOR_TYPE_GRAY &&
+        bit_depth < 8) png_set_gray_1_2_4_to_8(png_ptr);
+
+    if (png_get_valid(png_ptr, info_ptr,
+        PNG_INFO_tRNS)) png_set_tRNS_to_alpha(png_ptr);
+
+These three functions are actually aliases for png_set_expand(), added
+in libpng version 1.0.4, with the function names expanded to improve code
+readability.  In some future version they may actually do different
+things.
+
+PNG can have files with 16 bits per channel.  If you only can handle
+8 bits per channel, this will strip the pixels down to 8 bit.
+
+    if (bit_depth == 16)
+        png_set_strip_16(png_ptr);
+
+If, for some reason, you don't need the alpha channel on an image,
+and you want to remove it rather than combining it with the background
+(but the image author certainly had in mind that you *would* combine
+it with the background, so that's what you should probably do):
+
+    if (color_type & PNG_COLOR_MASK_ALPHA)
+        png_set_strip_alpha(png_ptr);
+
+In PNG files, the alpha channel in an image
+is the level of opacity.  If you need the alpha channel in an image to
+be the level of transparency instead of opacity, you can invert the
+alpha channel (or the tRNS chunk data) after it's read, so that 0 is
+fully opaque and 255 (in 8-bit or paletted images) or 65535 (in 16-bit
+images) is fully transparent, with
+
+    png_set_invert_alpha(png_ptr);
+
+PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
+they can, resulting in, for example, 8 pixels per byte for 1 bit
+files.  This code expands to 1 pixel per byte without changing the
+values of the pixels:
+
+    if (bit_depth < 8)
+        png_set_packing(png_ptr);
+
+PNG files have possible bit depths of 1, 2, 4, 8, and 16.  All pixels
+stored in a PNG image have been "scaled" or "shifted" up to the next
+higher possible bit depth (e.g. from 5 bits/sample in the range [0,31] to
+8 bits/sample in the range [0, 255]).  However, it is also possible to
+convert the PNG pixel data back to the original bit depth of the image.
+This call reduces the pixels back down to the original bit depth:
+
+    png_color_8p sig_bit;
+
+    if (png_get_sBIT(png_ptr, info_ptr, &sig_bit))
+        png_set_shift(png_ptr, sig_bit);
+
+PNG files store 3-color pixels in red, green, blue order.  This code
+changes the storage of the pixels to blue, green, red:
+
+    if (color_type == PNG_COLOR_TYPE_RGB ||
+        color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+        png_set_bgr(png_ptr);
+
+PNG files store RGB pixels packed into 3 or 6 bytes. This code expands them
+into 4 or 8 bytes for windowing systems that need them in this format:
+
+    if (color_type == PNG_COLOR_TYPE_RGB)
+        png_set_filler(png_ptr, filler, PNG_FILLER_BEFORE);
+
+where "filler" is the 8 or 16-bit number to fill with, and the location is
+either PNG_FILLER_BEFORE or PNG_FILLER_AFTER, depending upon whether
+you want the filler before the RGB or after.  This transformation
+does not affect images that already have full alpha channels.  To add an
+opaque alpha channel, use filler=0xff or 0xffff and PNG_FILLER_AFTER which
+will generate RGBA pixels.
+
+If you are reading an image with an alpha channel, and you need the
+data as ARGB instead of the normal PNG format RGBA:
+
+    if (color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+        png_set_swap_alpha(png_ptr);
+
+For some uses, you may want a grayscale image to be represented as
+RGB.  This code will do that conversion:
+
+    if (color_type == PNG_COLOR_TYPE_GRAY ||
+        color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+          png_set_gray_to_rgb(png_ptr);
+
+Conversely, you can convert an RGB or RGBA image to grayscale or grayscale
+with alpha.
+
+    if (color_type == PNG_COLOR_TYPE_RGB ||
+        color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+          png_set_rgb_to_gray_fixed(png_ptr, error_action,
+             int red_weight, int green_weight);
+
+    error_action = 1: silently do the conversion
+    error_action = 2: issue a warning if the original
+                      image has any pixel where
+                      red != green or red != blue
+    error_action = 3: issue an error and abort the
+                      conversion if the original
+                      image has any pixel where
+                      red != green or red != blue
+
+    red_weight:       weight of red component times 100000
+    green_weight:     weight of green component times 100000
+                      If either weight is negative, default
+                      weights (21268, 71514) are used.
+
+If you have set error_action = 1 or 2, you can
+later check whether the image really was gray, after processing
+the image rows, with the png_get_rgb_to_gray_status(png_ptr) function.
+It will return a png_byte that is zero if the image was gray or
+1 if there were any non-gray pixels.  bKGD and sBIT data
+will be silently converted to grayscale, using the green channel
+data, regardless of the error_action setting.
+
+With red_weight+green_weight<=100000,
+the normalized graylevel is computed:
+
+    int rw = red_weight * 65536;
+    int gw = green_weight * 65536;
+    int bw = 65536 - (rw + gw);
+    gray = (rw*red + gw*green + bw*blue)/65536;
+
+The default values approximate those recommended in the Charles
+Poynton's Color FAQ, <http://www.inforamp.net/~poynton/>
+Copyright (c) 1998-01-04 Charles Poynton poynton@inforamp.net
+
+    Y = 0.212671 * R + 0.715160 * G + 0.072169 * B
+
+Libpng approximates this with
+
+    Y = 0.21268 * R    + 0.7151 * G    + 0.07217 * B
+
+which can be expressed with integers as
+
+    Y = (6969 * R + 23434 * G + 2365 * B)/32768
+
+The calculation is done in a linear colorspace, if the image gamma
+is known.
+
+If you have a grayscale and you are using png_set_expand_depth() or
+png_set_expand() to change to
+a higher bit-depth, you must either supply the background color as a gray
+value at the original file bit-depth (need_expand = 1) or else supply the
+background color as an RGB triplet at the final, expanded bit depth
+(need_expand = 0).  Similarly, if you are reading a paletted image, you
+must either supply the background color as a palette index (need_expand = 1)
+or as an RGB triplet that may or may not be in the palette (need_expand = 0).
+
+    png_color_16 my_background;
+    png_color_16p image_background;
+
+    if (png_get_bKGD(png_ptr, info_ptr, &image_background))
+        png_set_background(png_ptr, image_background,
+          PNG_BACKGROUND_GAMMA_FILE, 1, 1.0);
+    else
+        png_set_background(png_ptr, &my_background,
+          PNG_BACKGROUND_GAMMA_SCREEN, 0, 1.0);
+
+The png_set_background() function tells libpng to composite images
+with alpha or simple transparency against the supplied background
+color.  If the PNG file contains a bKGD chunk (PNG_INFO_bKGD valid),
+you may use this color, or supply another color more suitable for
+the current display (e.g., the background color from a web page).  You
+need to tell libpng whether the color is in the gamma space of the
+display (PNG_BACKGROUND_GAMMA_SCREEN for colors you supply), the file
+(PNG_BACKGROUND_GAMMA_FILE for colors from the bKGD chunk), or one
+that is neither of these gammas (PNG_BACKGROUND_GAMMA_UNIQUE - I don't
+know why anyone would use this, but it's here).
+
+To properly display PNG images on any kind of system, the application needs
+to know what the display gamma is.  Ideally, the user will know this, and
+the application will allow them to set it.  One method of allowing the user
+to set the display gamma separately for each system is to check for a
+SCREEN_GAMMA or DISPLAY_GAMMA environment variable, which will hopefully be
+correctly set.
+
+Note that display_gamma is the overall gamma correction required to produce
+pleasing results, which depends on the lighting conditions in the surrounding
+environment.  In a dim or brightly lit room, no compensation other than
+the physical gamma exponent of the monitor is needed, while in a dark room
+a slightly smaller exponent is better.
+
+   double gamma, screen_gamma;
+
+   if (/* We have a user-defined screen
+       gamma value */)
+   {
+      screen_gamma = user_defined_screen_gamma;
+   }
+   /* One way that applications can share the same
+      screen gamma value */
+   else if ((gamma_str = getenv("SCREEN_GAMMA"))
+      != NULL)
+   {
+      screen_gamma = (double)atof(gamma_str);
+   }
+   /* If we don't have another value */
+   else
+   {
+      screen_gamma = 2.2; /* A good guess for a
+           PC monitor in a bright office or a dim room */
+      screen_gamma = 2.0; /* A good guess for a
+           PC monitor in a dark room */
+      screen_gamma = 1.7 or 1.0;  /* A good
+           guess for Mac systems */
+   }
+
+The png_set_gamma() function handles gamma transformations of the data.
+Pass both the file gamma and the current screen_gamma.  If the file does
+not have a gamma value, you can pass one anyway if you have an idea what
+it is (usually 0.45455 is a good guess for GIF images on PCs).  Note
+that file gammas are inverted from screen gammas.  See the discussions
+on gamma in the PNG specification for an excellent description of what
+gamma is, and why all applications should support it.  It is strongly
+recommended that PNG viewers support gamma correction.
+
+   if (png_get_gAMA(png_ptr, info_ptr, &gamma))
+      png_set_gamma(png_ptr, screen_gamma, gamma);
+   else
+      png_set_gamma(png_ptr, screen_gamma, 0.45455);
+
+If you need to reduce an RGB file to a paletted file, or if a paletted
+file has more entries then will fit on your screen, png_set_dither()
+will do that.  Note that this is a simple match dither that merely
+finds the closest color available.  This should work fairly well with
+optimized palettes, and fairly badly with linear color cubes.  If you
+pass a palette that is larger then maximum_colors, the file will
+reduce the number of colors in the palette so it will fit into
+maximum_colors.  If there is a histogram, it will use it to make
+more intelligent choices when reducing the palette.  If there is no
+histogram, it may not do as good a job.
+
+   if (color_type & PNG_COLOR_MASK_COLOR)
+   {
+      if (png_get_valid(png_ptr, info_ptr,
+         PNG_INFO_PLTE))
+      {
+         png_uint_16p histogram;
+
+         png_get_hIST(png_ptr, info_ptr,
+            &histogram);
+         png_set_dither(png_ptr, palette, num_palette,
+            max_screen_colors, histogram, 1);
+      }
+      else
+      {
+         png_color std_color_cube[MAX_SCREEN_COLORS] =
+            { ... colors ... };
+
+         png_set_dither(png_ptr, std_color_cube,
+            MAX_SCREEN_COLORS, MAX_SCREEN_COLORS,
+            NULL,0);
+      }
+   }
+
+PNG files describe monochrome as black being zero and white being one.
+The following code will reverse this (make black be one and white be
+zero):
+
+   if (bit_depth == 1 && color_type == PNG_COLOR_TYPE_GRAY)
+      png_set_invert_mono(png_ptr);
+
+This function can also be used to invert grayscale and gray-alpha images:
+
+   if (color_type == PNG_COLOR_TYPE_GRAY ||
+        color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      png_set_invert_mono(png_ptr);
+
+PNG files store 16 bit pixels in network byte order (big-endian,
+ie. most significant bits first).  This code changes the storage to the
+other way (little-endian, i.e. least significant bits first, the
+way PCs store them):
+
+    if (bit_depth == 16)
+        png_set_swap(png_ptr);
+
+If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
+need to change the order the pixels are packed into bytes, you can use:
+
+    if (bit_depth < 8)
+       png_set_packswap(png_ptr);
+
+Finally, you can write your own transformation function if none of
+the existing ones meets your needs.  This is done by setting a callback
+with
+
+    png_set_read_user_transform_fn(png_ptr,
+       read_transform_fn);
+
+You must supply the function
+
+    void read_transform_fn(png_ptr ptr, row_info_ptr
+       row_info, png_bytep data)
+
+See pngtest.c for a working example.  Your function will be called
+after all of the other transformations have been processed.
+
+You can also set up a pointer to a user structure for use by your
+callback function, and you can inform libpng that your transform
+function will change the number of channels or bit depth with the
+function
+
+    png_set_user_transform_info(png_ptr, user_ptr,
+       user_depth, user_channels);
+
+The user's application, not libpng, is responsible for allocating and
+freeing any memory required for the user structure.
+
+You can retrieve the pointer via the function
+png_get_user_transform_ptr().  For example:
+
+    voidp read_user_transform_ptr =
+       png_get_user_transform_ptr(png_ptr);
+
+The last thing to handle is interlacing; this is covered in detail below,
+but you must call the function here if you want libpng to handle expansion
+of the interlaced image.
+
+    number_of_passes = png_set_interlace_handling(png_ptr);
+
+After setting the transformations, libpng can update your png_info
+structure to reflect any transformations you've requested with this
+call.  This is most useful to update the info structure's rowbytes
+field so you can use it to allocate your image memory.  This function
+will also update your palette with the correct screen_gamma and
+background if these have been given with the calls above.
+
+    png_read_update_info(png_ptr, info_ptr);
+
+After you call png_read_update_info(), you can allocate any
+memory you need to hold the image.  The row data is simply
+raw byte data for all forms of images.  As the actual allocation
+varies among applications, no example will be given.  If you
+are allocating one large chunk, you will need to build an
+array of pointers to each row, as it will be needed for some
+of the functions below.
+
+Reading image data
+
+After you've allocated memory, you can read the image data.
+The simplest way to do this is in one function call.  If you are
+allocating enough memory to hold the whole image, you can just
+call png_read_image() and libpng will read in all the image data
+and put it in the memory area supplied.  You will need to pass in
+an array of pointers to each row.
+
+This function automatically handles interlacing, so you don't need
+to call png_set_interlace_handling() or call this function multiple
+times, or any of that other stuff necessary with png_read_rows().
+
+   png_read_image(png_ptr, row_pointers);
+
+where row_pointers is:
+
+   png_bytep row_pointers[height];
+
+You can point to void or char or whatever you use for pixels.
+
+If you don't want to read in the whole image at once, you can
+use png_read_rows() instead.  If there is no interlacing (check
+interlace_type == PNG_INTERLACE_NONE), this is simple:
+
+    png_read_rows(png_ptr, row_pointers, NULL,
+       number_of_rows);
+
+where row_pointers is the same as in the png_read_image() call.
+
+If you are doing this just one row at a time, you can do this with
+a single row_pointer instead of an array of row_pointers:
+
+    png_bytep row_pointer = row;
+    png_read_row(png_ptr, row_pointer, NULL);
+
+If the file is interlaced (interlace_type != 0 in the IHDR chunk), things
+get somewhat harder.  The only current (PNG Specification version 1.2)
+interlacing type for PNG is (interlace_type == PNG_INTERLACE_ADAM7)
+is a somewhat complicated 2D interlace scheme, known as Adam7, that
+breaks down an image into seven smaller images of varying size, based
+on an 8x8 grid.
+
+libpng can fill out those images or it can give them to you "as is".
+If you want them filled out, there are two ways to do that.  The one
+mentioned in the PNG specification is to expand each pixel to cover
+those pixels that have not been read yet (the "rectangle" method).
+This results in a blocky image for the first pass, which gradually
+smooths out as more pixels are read.  The other method is the "sparkle"
+method, where pixels are drawn only in their final locations, with the
+rest of the image remaining whatever colors they were initialized to
+before the start of the read.  The first method usually looks better,
+but tends to be slower, as there are more pixels to put in the rows.
+
+If you don't want libpng to handle the interlacing details, just call
+png_read_rows() seven times to read in all seven images.  Each of the
+images is a valid image by itself, or they can all be combined on an
+8x8 grid to form a single image (although if you intend to combine them
+you would be far better off using the libpng interlace handling).
+
+The first pass will return an image 1/8 as wide as the entire image
+(every 8th column starting in column 0) and 1/8 as high as the original
+(every 8th row starting in row 0), the second will be 1/8 as wide
+(starting in column 4) and 1/8 as high (also starting in row 0).  The
+third pass will be 1/4 as wide (every 4th pixel starting in column 0) and
+1/8 as high (every 8th row starting in row 4), and the fourth pass will
+be 1/4 as wide and 1/4 as high (every 4th column starting in column 2,
+and every 4th row starting in row 0).  The fifth pass will return an
+image 1/2 as wide, and 1/4 as high (starting at column 0 and row 2),
+while the sixth pass will be 1/2 as wide and 1/2 as high as the original
+(starting in column 1 and row 0).  The seventh and final pass will be as
+wide as the original, and 1/2 as high, containing all of the odd
+numbered scanlines.  Phew!
+
+If you want libpng to expand the images, call this before calling
+png_start_read_image() or png_read_update_info():
+
+    if (interlace_type == PNG_INTERLACE_ADAM7)
+        number_of_passes
+           = png_set_interlace_handling(png_ptr);
+
+This will return the number of passes needed.  Currently, this
+is seven, but may change if another interlace type is added.
+This function can be called even if the file is not interlaced,
+where it will return one pass.
+
+If you are not going to display the image after each pass, but are
+going to wait until the entire image is read in, use the sparkle
+effect.  This effect is faster and the end result of either method
+is exactly the same.  If you are planning on displaying the image
+after each pass, the "rectangle" effect is generally considered the
+better looking one.
+
+If you only want the "sparkle" effect, just call png_read_rows() as
+normal, with the third parameter NULL.  Make sure you make pass over
+the image number_of_passes times, and you don't change the data in the
+rows between calls.  You can change the locations of the data, just
+not the data.  Each pass only writes the pixels appropriate for that
+pass, and assumes the data from previous passes is still valid.
+
+    png_read_rows(png_ptr, row_pointers, NULL,
+       number_of_rows);
+
+If you only want the first effect (the rectangles), do the same as
+before except pass the row buffer in the third parameter, and leave
+the second parameter NULL.
+
+    png_read_rows(png_ptr, NULL, row_pointers,
+       number_of_rows);
+
+Finishing a sequential read
+
+After you are finished reading the image through either the high- or
+low-level interfaces, you can finish reading the file.  If you are
+interested in comments or time, which may be stored either before or
+after the image data, you should pass the separate png_info struct if
+you want to keep the comments from before and after the image
+separate.  If you are not interested, you can pass NULL.
+
+   png_read_end(png_ptr, end_info);
+
+When you are done, you can free all memory allocated by libpng like this:
+
+   png_destroy_read_struct(&png_ptr, &info_ptr,
+       &end_info);
+
+It is also possible to individually free the info_ptr members that
+point to libpng-allocated storage with the following function:
+
+    png_free_data(png_ptr, info_ptr, mask, seq)
+    mask - identifies data to be freed, a mask
+           containing the logical OR of one or
+           more of
+             PNG_FREE_PLTE, PNG_FREE_TRNS,
+             PNG_FREE_HIST, PNG_FREE_ICCP,
+             PNG_FREE_PCAL, PNG_FREE_ROWS,
+             PNG_FREE_SCAL, PNG_FREE_SPLT,
+             PNG_FREE_TEXT, PNG_FREE_UNKN,
+           or simply PNG_FREE_ALL
+    seq  - sequence number of item to be freed
+           (-1 for all items)
+
+This function may be safely called when the relevant storage has
+already been freed, or has not yet been allocated, or was allocated
+by the user and not by libpng,  and will in those
+cases do nothing.  The "seq" parameter is ignored if only one item
+of the selected data type, such as PLTE, is allowed.  If "seq" is not
+-1, and multiple items are allowed for the data type identified in
+the mask, such as text or sPLT, only the n'th item in the structure
+is freed, where n is "seq".
+
+The default behavior is only to free data that was allocated internally
+by libpng.  This can be changed, so that libpng will not free the data,
+or so that it will free data that was allocated by the user with png_malloc()
+or png_zalloc() and passed in via a png_set_*() function, with
+
+    png_data_freer(png_ptr, info_ptr, freer, mask)
+    mask   - which data elements are affected
+             same choices as in png_free_data()
+    freer  - one of
+               PNG_DESTROY_WILL_FREE_DATA
+               PNG_SET_WILL_FREE_DATA
+               PNG_USER_WILL_FREE_DATA
+
+This function only affects data that has already been allocated.
+You can call this function after reading the PNG data but before calling
+any png_set_*() functions, to control whether the user or the png_set_*()
+function is responsible for freeing any existing data that might be present,
+and again after the png_set_*() functions to control whether the user
+or png_destroy_*() is supposed to free the data.  When the user assumes
+responsibility for libpng-allocated data, the application must use
+png_free() to free it, and when the user transfers responsibility to libpng
+for data that the user has allocated, the user must have used png_malloc()
+or png_zalloc() to allocate it.
+
+If you allocated your row_pointers in a single block, as suggested above in
+the description of the high level read interface, you must not transfer
+responsibility for freeing it to the png_set_rows or png_read_destroy function,
+because they would also try to free the individual row_pointers[i].
+
+If you allocated text_ptr.text, text_ptr.lang, and text_ptr.translated_keyword
+separately, do not transfer responsibility for freeing text_ptr to libpng,
+because when libpng fills a png_text structure it combines these members with
+the key member, and png_free_data() will free only text_ptr.key.  Similarly,
+if you transfer responsibility for free'ing text_ptr from libpng to your
+application, your application must not separately free those members.
+
+The png_free_data() function will turn off the "valid" flag for anything
+it frees.  If you need to turn the flag off for a chunk that was freed by your
+application instead of by libpng, you can use
+
+    png_set_invalid(png_ptr, info_ptr, mask);
+    mask - identifies the chunks to be made invalid,
+           containing the logical OR of one or
+           more of
+             PNG_INFO_gAMA, PNG_INFO_sBIT,
+             PNG_INFO_cHRM, PNG_INFO_PLTE,
+             PNG_INFO_tRNS, PNG_INFO_bKGD,
+             PNG_INFO_hIST, PNG_INFO_pHYs,
+             PNG_INFO_oFFs, PNG_INFO_tIME,
+             PNG_INFO_pCAL, PNG_INFO_sRGB,
+             PNG_INFO_iCCP, PNG_INFO_sPLT,
+             PNG_INFO_sCAL, PNG_INFO_IDAT
+
+For a more compact example of reading a PNG image, see the file example.c.
+
+Reading PNG files progressively
+
+The progressive reader is slightly different then the non-progressive
+reader.  Instead of calling png_read_info(), png_read_rows(), and
+png_read_end(), you make one call to png_process_data(), which calls
+callbacks when it has the info, a row, or the end of the image.  You
+set up these callbacks with png_set_progressive_read_fn().  You don't
+have to worry about the input/output functions of libpng, as you are
+giving the library the data directly in png_process_data().  I will
+assume that you have read the section on reading PNG files above,
+so I will only highlight the differences (although I will show
+all of the code).
+
+png_structp png_ptr;
+png_infop info_ptr;
+
+ /*  An example code fragment of how you would
+     initialize the progressive reader in your
+     application. */
+ int
+ initialize_png_reader()
+ {
+    png_ptr = png_create_read_struct
+        (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+         user_error_fn, user_warning_fn);
+    if (!png_ptr)
+        return (ERROR);
+    info_ptr = png_create_info_struct(png_ptr);
+    if (!info_ptr)
+    {
+        png_destroy_read_struct(&png_ptr, (png_infopp)NULL,
+           (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+           (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    /* This one's new.  You can provide functions
+       to be called when the header info is valid,
+       when each row is completed, and when the image
+       is finished.  If you aren't using all functions,
+       you can specify NULL parameters.  Even when all
+       three functions are NULL, you need to call
+       png_set_progressive_read_fn().  You can use
+       any struct as the user_ptr (cast to a void pointer
+       for the function call), and retrieve the pointer
+       from inside the callbacks using the function
+
+          png_get_progressive_ptr(png_ptr);
+
+       which will return a void pointer, which you have
+       to cast appropriately.
+     */
+    png_set_progressive_read_fn(png_ptr, (void *)user_ptr,
+        info_callback, row_callback, end_callback);
+
+    return 0;
+ }
+
+ /* A code fragment that you call as you receive blocks
+   of data */
+ int
+ process_data(png_bytep buffer, png_uint_32 length)
+ {
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+        png_destroy_read_struct(&png_ptr, &info_ptr,
+           (png_infopp)NULL);
+        return (ERROR);
+    }
+
+    /* This one's new also.  Simply give it a chunk
+       of data from the file stream (in order, of
+       course).  On machines with segmented memory
+       models machines, don't give it any more than
+       64K.  The library seems to run fine with sizes
+       of 4K. Although you can give it much less if
+       necessary (I assume you can give it chunks of
+       1 byte, I haven't tried less then 256 bytes
+       yet).  When this function returns, you may
+       want to display any rows that were generated
+       in the row callback if you don't already do
+       so there.
+     */
+    png_process_data(png_ptr, info_ptr, buffer, length);
+    return 0;
+ }
+
+ /* This function is called (as set by
+    png_set_progressive_read_fn() above) when enough data
+    has been supplied so all of the header has been
+    read.
+ */
+ void
+ info_callback(png_structp png_ptr, png_infop info)
+ {
+    /* Do any setup here, including setting any of
+       the transformations mentioned in the Reading
+       PNG files section.  For now, you _must_ call
+       either png_start_read_image() or
+       png_read_update_info() after all the
+       transformations are set (even if you don't set
+       any).  You may start getting rows before
+       png_process_data() returns, so this is your
+       last chance to prepare for that.
+     */
+ }
+
+ /* This function is called when each row of image
+    data is complete */
+ void
+ row_callback(png_structp png_ptr, png_bytep new_row,
+    png_uint_32 row_num, int pass)
+ {
+    /* If the image is interlaced, and you turned
+       on the interlace handler, this function will
+       be called for every row in every pass.  Some
+       of these rows will not be changed from the
+       previous pass.  When the row is not changed,
+       the new_row variable will be NULL.  The rows
+       and passes are called in order, so you don't
+       really need the row_num and pass, but I'm
+       supplying them because it may make your life
+       easier.
+
+       For the non-NULL rows of interlaced images,
+       you must call png_progressive_combine_row()
+       passing in the row and the old row.  You can
+       call this function for NULL rows (it will just
+       return) and for non-interlaced images (it just
+       does the memcpy for you) if it will make the
+       code easier.  Thus, you can just do this for
+       all cases:
+     */
+
+        png_progressive_combine_row(png_ptr, old_row,
+          new_row);
+
+    /* where old_row is what was displayed for
+       previously for the row.  Note that the first
+       pass (pass == 0, really) will completely cover
+       the old row, so the rows do not have to be
+       initialized.  After the first pass (and only
+       for interlaced images), you will have to pass
+       the current row, and the function will combine
+       the old row and the new row.
+    */
+ }
+
+ void
+ end_callback(png_structp png_ptr, png_infop info)
+ {
+    /* This function is called after the whole image
+       has been read, including any chunks after the
+       image (up to and including the IEND).  You
+       will usually have the same info chunk as you
+       had in the header, although some data may have
+       been added to the comments and time fields.
+
+       Most people won't do much here, perhaps setting
+       a flag that marks the image as finished.
+     */
+ }
+
+
+
+IV. Writing
+
+Much of this is very similar to reading.  However, everything of
+importance is repeated here, so you won't have to constantly look
+back up in the reading section to understand writing.
+
+Setup
+
+You will want to do the I/O initialization before you get into libpng,
+so if it doesn't work, you don't have anything to undo. If you are not
+using the standard I/O functions, you will need to replace them with
+custom writing functions.  See the discussion under Customizing libpng.
+
+    FILE *fp = fopen(file_name, "wb");
+    if (!fp)
+    {
+       return (ERROR);
+    }
+
+Next, png_struct and png_info need to be allocated and initialized.
+As these can be both relatively large, you may not want to store these
+on the stack, unless you have stack space to spare.  Of course, you
+will want to check if they return NULL.  If you are also reading,
+you won't want to name your read structure and your write structure
+both "png_ptr"; you can call them anything you like, such as
+"read_ptr" and "write_ptr".  Look at pngtest.c, for example.
+
+    png_structp png_ptr = png_create_write_struct
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn);
+    if (!png_ptr)
+       return (ERROR);
+
+    png_infop info_ptr = png_create_info_struct(png_ptr);
+    if (!info_ptr)
+    {
+       png_destroy_write_struct(&png_ptr,
+         (png_infopp)NULL);
+       return (ERROR);
+    }
+
+If you want to use your own memory allocation routines,
+define PNG_USER_MEM_SUPPORTED and use
+png_create_write_struct_2() instead of png_create_write_struct():
+
+    png_structp png_ptr = png_create_write_struct_2
+       (PNG_LIBPNG_VER_STRING, (png_voidp)user_error_ptr,
+        user_error_fn, user_warning_fn, (png_voidp)
+        user_mem_ptr, user_malloc_fn, user_free_fn);
+
+After you have these structures, you will need to set up the
+error handling.  When libpng encounters an error, it expects to
+longjmp() back to your routine.  Therefore, you will need to call
+setjmp() and pass the png_jmpbuf(png_ptr).  If you
+write the file from different routines, you will need to update
+the png_jmpbuf(png_ptr) every time you enter a new routine that will
+call a png_*() function.  See your documentation of setjmp/longjmp
+for your compiler for more information on setjmp/longjmp.  See
+the discussion on libpng error handling in the Customizing Libpng
+section below for more information on the libpng error handling.
+
+    if (setjmp(png_jmpbuf(png_ptr)))
+    {
+       png_destroy_write_struct(&png_ptr, &info_ptr);
+       fclose(fp);
+       return (ERROR);
+    }
+    ...
+    return;
+
+If you would rather avoid the complexity of setjmp/longjmp issues,
+you can compile libpng with PNG_SETJMP_NOT_SUPPORTED, in which case
+errors will result in a call to PNG_ABORT() which defaults to abort().
+
+Now you need to set up the output code.  The default for libpng is to
+use the C function fwrite().  If you use this, you will need to pass a
+valid FILE * in the function png_init_io().  Be sure that the file is
+opened in binary mode.  Again, if you wish to handle writing data in
+another way, see the discussion on libpng I/O handling in the Customizing
+Libpng section below.
+
+    png_init_io(png_ptr, fp);
+
+Write callbacks
+
+At this point, you can set up a callback function that will be
+called after each row has been written, which you can use to control
+a progress meter or the like.  It's demonstrated in pngtest.c.
+You must supply a function
+
+    void write_row_callback(png_ptr, png_uint_32 row,
+       int pass);
+    {
+      /* put your code here */
+    }
+
+(You can give it another name that you like instead of "write_row_callback")
+
+To inform libpng about your function, use
+
+    png_set_write_status_fn(png_ptr, write_row_callback);
+
+You now have the option of modifying how the compression library will
+run.  The following functions are mainly for testing, but may be useful
+in some cases, like if you need to write PNG files extremely fast and
+are willing to give up some compression, or if you want to get the
+maximum possible compression at the expense of slower writing.  If you
+have no special needs in this area, let the library do what it wants by
+not calling this function at all, as it has been tuned to deliver a good
+speed/compression ratio. The second parameter to png_set_filter() is
+the filter method, for which the only valid values are 0 (as of the
+July 1999 PNG specification, version 1.2) or 64 (if you are writing
+a PNG datastream that is to be embedded in a MNG datastream).  The third
+parameter is a flag that indicates which filter type(s) are to be tested
+for each scanline.  See the PNG specification for details on the specific filter
+types.
+
+
+    /* turn on or off filtering, and/or choose
+       specific filters.  You can use either a single
+       PNG_FILTER_VALUE_NAME or the logical OR of one
+       or more PNG_FILTER_NAME masks. */
+    png_set_filter(png_ptr, 0,
+       PNG_FILTER_NONE  | PNG_FILTER_VALUE_NONE |
+       PNG_FILTER_SUB   | PNG_FILTER_VALUE_SUB  |
+       PNG_FILTER_UP    | PNG_FILTER_VALUE_UP   |
+       PNG_FILTER_AVE   | PNG_FILTER_VALUE_AVE  |
+       PNG_FILTER_PAETH | PNG_FILTER_VALUE_PAETH|
+       PNG_ALL_FILTERS);
+
+If an application
+wants to start and stop using particular filters during compression,
+it should start out with all of the filters (to ensure that the previous
+row of pixels will be stored in case it's needed later), and then add
+and remove them after the start of compression.
+
+If you are writing a PNG datastream that is to be embedded in a MNG
+datastream, the second parameter can be either 0 or 64.
+
+The png_set_compression_*() functions interface to the zlib compression
+library, and should mostly be ignored unless you really know what you are
+doing.  The only generally useful call is png_set_compression_level()
+which changes how much time zlib spends on trying to compress the image
+data.  See the Compression Library (zlib.h and algorithm.txt, distributed
+with zlib) for details on the compression levels.
+
+    /* set the zlib compression level */
+    png_set_compression_level(png_ptr,
+        Z_BEST_COMPRESSION);
+
+    /* set other zlib parameters */
+    png_set_compression_mem_level(png_ptr, 8);
+    png_set_compression_strategy(png_ptr,
+        Z_DEFAULT_STRATEGY);
+    png_set_compression_window_bits(png_ptr, 15);
+    png_set_compression_method(png_ptr, 8);
+    png_set_compression_buffer_size(png_ptr, 8192)
+
+extern PNG_EXPORT(void,png_set_zbuf_size)
+
+Setting the contents of info for output
+
+You now need to fill in the png_info structure with all the data you
+wish to write before the actual image.  Note that the only thing you
+are allowed to write after the image is the text chunks and the time
+chunk (as of PNG Specification 1.2, anyway).  See png_write_end() and
+the latest PNG specification for more information on that.  If you
+wish to write them before the image, fill them in now, and flag that
+data as being valid.  If you want to wait until after the data, don't
+fill them until png_write_end().  For all the fields in png_info and
+their data types, see png.h.  For explanations of what the fields
+contain, see the PNG specification.
+
+Some of the more important parts of the png_info are:
+
+    png_set_IHDR(png_ptr, info_ptr, width, height,
+       bit_depth, color_type, interlace_type,
+       compression_type, filter_method)
+    width          - holds the width of the image
+                     in pixels (up to 2^31).
+    height         - holds the height of the image
+                     in pixels (up to 2^31).
+    bit_depth      - holds the bit depth of one of the
+                     image channels.
+                     (valid values are 1, 2, 4, 8, 16
+                     and depend also on the
+                     color_type.  See also significant
+                     bits (sBIT) below).
+    color_type     - describes which color/alpha
+                     channels are present.
+                     PNG_COLOR_TYPE_GRAY
+                        (bit depths 1, 2, 4, 8, 16)
+                     PNG_COLOR_TYPE_GRAY_ALPHA
+                        (bit depths 8, 16)
+                     PNG_COLOR_TYPE_PALETTE
+                        (bit depths 1, 2, 4, 8)
+                     PNG_COLOR_TYPE_RGB
+                        (bit_depths 8, 16)
+                     PNG_COLOR_TYPE_RGB_ALPHA
+                        (bit_depths 8, 16)
+
+                     PNG_COLOR_MASK_PALETTE
+                     PNG_COLOR_MASK_COLOR
+                     PNG_COLOR_MASK_ALPHA
+
+    interlace_type - PNG_INTERLACE_NONE or
+                     PNG_INTERLACE_ADAM7
+    compression_type - (must be
+                     PNG_COMPRESSION_TYPE_DEFAULT)
+    filter_method  - (must be PNG_FILTER_TYPE_DEFAULT
+                     or, if you are writing a PNG to
+                     be embedded in a MNG datastream,
+                     can also be
+                     PNG_INTRAPIXEL_DIFFERENCING)
+
+    png_set_PLTE(png_ptr, info_ptr, palette,
+       num_palette);
+    palette        - the palette for the file
+                     (array of png_color)
+    num_palette    - number of entries in the palette
+
+    png_set_gAMA(png_ptr, info_ptr, gamma);
+    gamma          - the gamma the image was created
+                     at (PNG_INFO_gAMA)
+
+    png_set_sRGB(png_ptr, info_ptr, srgb_intent);
+    srgb_intent    - the rendering intent
+                     (PNG_INFO_sRGB) The presence of
+                     the sRGB chunk means that the pixel
+                     data is in the sRGB color space.
+                     This chunk also implies specific
+                     values of gAMA and cHRM.  Rendering
+                     intent is the CSS-1 property that
+                     has been defined by the International
+                     Color Consortium
+                     (http://www.color.org).
+                     It can be one of
+                     PNG_sRGB_INTENT_SATURATION,
+                     PNG_sRGB_INTENT_PERCEPTUAL,
+                     PNG_sRGB_INTENT_ABSOLUTE, or
+                     PNG_sRGB_INTENT_RELATIVE.
+
+
+    png_set_sRGB_gAMA_and_cHRM(png_ptr, info_ptr,
+       srgb_intent);
+    srgb_intent    - the rendering intent
+                     (PNG_INFO_sRGB) The presence of the
+                     sRGB chunk means that the pixel
+                     data is in the sRGB color space.
+                     This function also causes gAMA and
+                     cHRM chunks with the specific values
+                     that are consistent with sRGB to be
+                     written.
+
+    png_set_iCCP(png_ptr, info_ptr, name, compression_type,
+                      profile, proflen);
+    name            - The profile name.
+    compression     - The compression type; always
+                      PNG_COMPRESSION_TYPE_BASE for PNG 1.0.
+                      You may give NULL to this argument to
+                      ignore it.
+    profile         - International Color Consortium color
+                      profile data. May contain NULs.
+    proflen         - length of profile data in bytes.
+
+    png_set_sBIT(png_ptr, info_ptr, sig_bit);
+    sig_bit        - the number of significant bits for
+                     (PNG_INFO_sBIT) each of the gray, red,
+                     green, and blue channels, whichever are
+                     appropriate for the given color type
+                     (png_color_16)
+
+    png_set_tRNS(png_ptr, info_ptr, trans, num_trans,
+       trans_values);
+    trans          - array of transparent entries for
+                     palette (PNG_INFO_tRNS)
+    trans_values   - graylevel or color sample values of
+                     the single transparent color for
+                     non-paletted images (PNG_INFO_tRNS)
+    num_trans      - number of transparent entries
+                     (PNG_INFO_tRNS)
+
+    png_set_hIST(png_ptr, info_ptr, hist);
+                    (PNG_INFO_hIST)
+    hist           - histogram of palette (array of
+                     png_uint_16)
+
+    png_set_tIME(png_ptr, info_ptr, mod_time);
+    mod_time       - time image was last modified
+                     (PNG_VALID_tIME)
+
+    png_set_bKGD(png_ptr, info_ptr, background);
+    background     - background color (PNG_VALID_bKGD)
+
+    png_set_text(png_ptr, info_ptr, text_ptr, num_text);
+    text_ptr       - array of png_text holding image
+                     comments
+    text_ptr[i].compression - type of compression used
+                 on "text" PNG_TEXT_COMPRESSION_NONE
+                           PNG_TEXT_COMPRESSION_zTXt
+                           PNG_ITXT_COMPRESSION_NONE
+                           PNG_ITXT_COMPRESSION_zTXt
+    text_ptr[i].key   - keyword for comment.  Must contain
+                 1-79 characters.
+    text_ptr[i].text  - text comments for current
+                         keyword.  Can be NULL or empty.
+    text_ptr[i].text_length - length of text string,
+                 after decompression, 0 for iTXt
+    text_ptr[i].itxt_length - length of itxt string,
+                 after decompression, 0 for tEXt/zTXt
+    text_ptr[i].lang  - language of comment (NULL or
+                         empty for unknown).
+    text_ptr[i].translated_keyword  - keyword in UTF-8 (NULL
+                         or empty for unknown).
+    num_text       - number of comments
+
+    png_set_sPLT(png_ptr, info_ptr, &palette_ptr,
+       num_spalettes);
+    palette_ptr    - array of png_sPLT_struct structures
+                     to be added to the list of palettes
+                     in the info structure.
+    num_spalettes  - number of palette structures to be
+                     added.
+
+    png_set_oFFs(png_ptr, info_ptr, offset_x, offset_y,
+        unit_type);
+    offset_x  - positive offset from the left
+                     edge of the screen
+    offset_y  - positive offset from the top
+                     edge of the screen
+    unit_type - PNG_OFFSET_PIXEL, PNG_OFFSET_MICROMETER
+
+    png_set_pHYs(png_ptr, info_ptr, res_x, res_y,
+        unit_type);
+    res_x       - pixels/unit physical resolution
+                  in x direction
+    res_y       - pixels/unit physical resolution
+                  in y direction
+    unit_type   - PNG_RESOLUTION_UNKNOWN,
+                  PNG_RESOLUTION_METER
+
+    png_set_sCAL(png_ptr, info_ptr, unit, width, height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                  (width and height are doubles)
+
+    png_set_sCAL_s(png_ptr, info_ptr, unit, width, height)
+    unit        - physical scale units (an integer)
+    width       - width of a pixel in physical scale units
+    height      - height of a pixel in physical scale units
+                 (width and height are strings like "2.54")
+
+    png_set_unknown_chunks(png_ptr, info_ptr, &unknowns,
+       num_unknowns)
+    unknowns          - array of png_unknown_chunk
+                        structures holding unknown chunks
+    unknowns[i].name  - name of unknown chunk
+    unknowns[i].data  - data of unknown chunk
+    unknowns[i].size  - size of unknown chunk's data
+    unknowns[i].location - position to write chunk in file
+                           0: do not write chunk
+                           PNG_HAVE_IHDR: before PLTE
+                           PNG_HAVE_PLTE: before IDAT
+                           PNG_AFTER_IDAT: after IDAT
+
+The "location" member is set automatically according to
+what part of the output file has already been written.
+You can change its value after calling png_set_unknown_chunks()
+as demonstrated in pngtest.c.  Within each of the "locations",
+the chunks are sequenced according to their position in the
+structure (that is, the value of "i", which is the order in which
+the chunk was either read from the input file or defined with
+png_set_unknown_chunks).
+
+A quick word about text and num_text.  text is an array of png_text
+structures.  num_text is the number of valid structures in the array.
+Each png_text structure holds a language code, a keyword, a text value,
+and a compression type.
+
+The compression types have the same valid numbers as the compression
+types of the image data.  Currently, the only valid number is zero.
+However, you can store text either compressed or uncompressed, unlike
+images, which always have to be compressed.  So if you don't want the
+text compressed, set the compression type to PNG_TEXT_COMPRESSION_NONE.
+Because tEXt and zTXt chunks don't have a language field, if you
+specify PNG_TEXT_COMPRESSION_NONE or PNG_TEXT_COMPRESSION_zTXt
+any language code or translated keyword will not be written out.
+
+Until text gets around 1000 bytes, it is not worth compressing it.
+After the text has been written out to the file, the compression type
+is set to PNG_TEXT_COMPRESSION_NONE_WR or PNG_TEXT_COMPRESSION_zTXt_WR,
+so that it isn't written out again at the end (in case you are calling
+png_write_end() with the same struct.
+
+The keywords that are given in the PNG Specification are:
+
+    Title            Short (one line) title or
+                     caption for image
+    Author           Name of image's creator
+    Description      Description of image (possibly long)
+    Copyright        Copyright notice
+    Creation Time    Time of original image creation
+                     (usually RFC 1123 format, see below)
+    Software         Software used to create the image
+    Disclaimer       Legal disclaimer
+    Warning          Warning of nature of content
+    Source           Device used to create the image
+    Comment          Miscellaneous comment; conversion
+                     from other image format
+
+The keyword-text pairs work like this.  Keywords should be short
+simple descriptions of what the comment is about.  Some typical
+keywords are found in the PNG specification, as is some recommendations
+on keywords.  You can repeat keywords in a file.  You can even write
+some text before the image and some after.  For example, you may want
+to put a description of the image before the image, but leave the
+disclaimer until after, so viewers working over modem connections
+don't have to wait for the disclaimer to go over the modem before
+they start seeing the image.  Finally, keywords should be full
+words, not abbreviations.  Keywords and text are in the ISO 8859-1
+(Latin-1) character set (a superset of regular ASCII) and can not
+contain NUL characters, and should not contain control or other
+unprintable characters.  To make the comments widely readable, stick
+with basic ASCII, and avoid machine specific character set extensions
+like the IBM-PC character set.  The keyword must be present, but
+you can leave off the text string on non-compressed pairs.
+Compressed pairs must have a text string, as only the text string
+is compressed anyway, so the compression would be meaningless.
+
+PNG supports modification time via the png_time structure.  Two
+conversion routines are provided, png_convert_from_time_t() for
+time_t and png_convert_from_struct_tm() for struct tm.  The
+time_t routine uses gmtime().  You don't have to use either of
+these, but if you wish to fill in the png_time structure directly,
+you should provide the time in universal time (GMT) if possible
+instead of your local time.  Note that the year number is the full
+year (e.g. 1998, rather than 98 - PNG is year 2000 compliant!), and
+that months start with 1.
+
+If you want to store the time of the original image creation, you should
+use a plain tEXt chunk with the "Creation Time" keyword.  This is
+necessary because the "creation time" of a PNG image is somewhat vague,
+depending on whether you mean the PNG file, the time the image was
+created in a non-PNG format, a still photo from which the image was
+scanned, or possibly the subject matter itself.  In order to facilitate
+machine-readable dates, it is recommended that the "Creation Time"
+tEXt chunk use RFC 1123 format dates (e.g. "22 May 1997 18:07:10 GMT"),
+although this isn't a requirement.  Unlike the tIME chunk, the
+"Creation Time" tEXt chunk is not expected to be automatically changed
+by the software.  To facilitate the use of RFC 1123 dates, a function
+png_convert_to_rfc1123(png_timep) is provided to convert from PNG
+time to an RFC 1123 format string.
+
+Writing unknown chunks
+
+You can use the png_set_unknown_chunks function to queue up chunks
+for writing.  You give it a chunk name, raw data, and a size; that's
+all there is to it.  The chunks will be written by the next following
+png_write_info_before_PLTE, png_write_info, or png_write_end function.
+Any chunks previously read into the info structure's unknown-chunk
+list will also be written out in a sequence that satisfies the PNG
+specification's ordering rules.
+
+The high-level write interface
+
+At this point there are two ways to proceed; through the high-level
+write interface, or through a sequence of low-level write operations.
+You can use the high-level interface if your image data is present
+in the info structure.  All defined output
+transformations are permitted, enabled by the following masks.
+
+    PNG_TRANSFORM_IDENTITY      No transformation
+    PNG_TRANSFORM_PACKING       Pack 1, 2 and 4-bit samples
+    PNG_TRANSFORM_PACKSWAP      Change order of packed
+                                pixels to LSB first
+    PNG_TRANSFORM_INVERT_MONO   Invert monochrome images
+    PNG_TRANSFORM_SHIFT         Normalize pixels to the
+                                sBIT depth
+    PNG_TRANSFORM_BGR           Flip RGB to BGR, RGBA
+                                to BGRA
+    PNG_TRANSFORM_SWAP_ALPHA    Flip RGBA to ARGB or GA
+                                to AG
+    PNG_TRANSFORM_INVERT_ALPHA  Change alpha from opacity
+                                to transparency
+    PNG_TRANSFORM_SWAP_ENDIAN   Byte-swap 16-bit samples
+    PNG_TRANSFORM_STRIP_FILLER  Strip out filler bytes.
+
+If you have valid image data in the info structure (you can use
+png_set_rows() to put image data in the info structure), simply do this:
+
+    png_write_png(png_ptr, info_ptr, png_transforms, NULL)
+
+where png_transforms is an integer containing the logical OR of some set of
+transformation flags.  This call is equivalent to png_write_info(),
+followed the set of transformations indicated by the transform mask,
+then png_write_image(), and finally png_write_end().
+
+(The final parameter of this call is not yet used.  Someday it might point
+to transformation parameters required by some future output transform.)
+
+The low-level write interface
+
+If you are going the low-level route instead, you are now ready to
+write all the file information up to the actual image data.  You do
+this with a call to png_write_info().
+
+    png_write_info(png_ptr, info_ptr);
+
+Note that there is one transformation you may need to do before
+png_write_info().  In PNG files, the alpha channel in an image is the
+level of opacity.  If your data is supplied as a level of
+transparency, you can invert the alpha channel before you write it, so
+that 0 is fully transparent and 255 (in 8-bit or paletted images) or
+65535 (in 16-bit images) is fully opaque, with
+
+    png_set_invert_alpha(png_ptr);
+
+This must appear before png_write_info() instead of later with the
+other transformations because in the case of paletted images the tRNS
+chunk data has to be inverted before the tRNS chunk is written.  If
+your image is not a paletted image, the tRNS data (which in such cases
+represents a single color to be rendered as transparent) won't need to
+be changed, and you can safely do this transformation after your
+png_write_info() call.
+
+If you need to write a private chunk that you want to appear before
+the PLTE chunk when PLTE is present, you can write the PNG info in
+two steps, and insert code to write your own chunk between them:
+
+    png_write_info_before_PLTE(png_ptr, info_ptr);
+    png_set_unknown_chunks(png_ptr, info_ptr, ...);
+    png_write_info(png_ptr, info_ptr);
+
+After you've written the file information, you can set up the library
+to handle any special transformations of the image data.  The various
+ways to transform the data will be described in the order that they
+should occur.  This is important, as some of these change the color
+type and/or bit depth of the data, and some others only work on
+certain color types and bit depths.  Even though each transformation
+checks to see if it has data that it can do something with, you should
+make sure to only enable a transformation if it will be valid for the
+data.  For example, don't swap red and blue on grayscale data.
+
+PNG files store RGB pixels packed into 3 or 6 bytes.  This code tells
+the library to strip input data that has 4 or 8 bytes per pixel down
+to 3 or 6 bytes (or strip 2 or 4-byte grayscale+filler data to 1 or 2
+bytes per pixel).
+
+    png_set_filler(png_ptr, 0, PNG_FILLER_BEFORE);
+
+where the 0 is unused, and the location is either PNG_FILLER_BEFORE or
+PNG_FILLER_AFTER, depending upon whether the filler byte in the pixel
+is stored XRGB or RGBX.
+
+PNG files pack pixels of bit depths 1, 2, and 4 into bytes as small as
+they can, resulting in, for example, 8 pixels per byte for 1 bit files.
+If the data is supplied at 1 pixel per byte, use this code, which will
+correctly pack the pixels into a single byte:
+
+    png_set_packing(png_ptr);
+
+PNG files reduce possible bit depths to 1, 2, 4, 8, and 16.  If your
+data is of another bit depth, you can write an sBIT chunk into the
+file so that decoders can recover the original data if desired.
+
+    /* Set the true bit depth of the image data */
+    if (color_type & PNG_COLOR_MASK_COLOR)
+    {
+        sig_bit.red = true_bit_depth;
+        sig_bit.green = true_bit_depth;
+        sig_bit.blue = true_bit_depth;
+    }
+    else
+    {
+        sig_bit.gray = true_bit_depth;
+    }
+    if (color_type & PNG_COLOR_MASK_ALPHA)
+    {
+        sig_bit.alpha = true_bit_depth;
+    }
+
+    png_set_sBIT(png_ptr, info_ptr, &sig_bit);
+
+If the data is stored in the row buffer in a bit depth other than
+one supported by PNG (e.g. 3 bit data in the range 0-7 for a 4-bit PNG),
+this will scale the values to appear to be the correct bit depth as
+is required by PNG.
+
+    png_set_shift(png_ptr, &sig_bit);
+
+PNG files store 16 bit pixels in network byte order (big-endian,
+ie. most significant bits first).  This code would be used if they are
+supplied the other way (little-endian, i.e. least significant bits
+first, the way PCs store them):
+
+    if (bit_depth > 8)
+       png_set_swap(png_ptr);
+
+If you are using packed-pixel images (1, 2, or 4 bits/pixel), and you
+need to change the order the pixels are packed into bytes, you can use:
+
+    if (bit_depth < 8)
+       png_set_packswap(png_ptr);
+
+PNG files store 3 color pixels in red, green, blue order.  This code
+would be used if they are supplied as blue, green, red:
+
+    png_set_bgr(png_ptr);
+
+PNG files describe monochrome as black being zero and white being
+one. This code would be used if the pixels are supplied with this reversed
+(black being one and white being zero):
+
+    png_set_invert_mono(png_ptr);
+
+Finally, you can write your own transformation function if none of
+the existing ones meets your needs.  This is done by setting a callback
+with
+
+    png_set_write_user_transform_fn(png_ptr,
+       write_transform_fn);
+
+You must supply the function
+
+    void write_transform_fn(png_ptr ptr, row_info_ptr
+       row_info, png_bytep data)
+
+See pngtest.c for a working example.  Your function will be called
+before any of the other transformations are processed.
+
+You can also set up a pointer to a user structure for use by your
+callback function.
+
+    png_set_user_transform_info(png_ptr, user_ptr, 0, 0);
+
+The user_channels and user_depth parameters of this function are ignored
+when writing; you can set them to zero as shown.
+
+You can retrieve the pointer via the function png_get_user_transform_ptr().
+For example:
+
+    voidp write_user_transform_ptr =
+       png_get_user_transform_ptr(png_ptr);
+
+It is possible to have libpng flush any pending output, either manually,
+or automatically after a certain number of lines have been written.  To
+flush the output stream a single time call:
+
+    png_write_flush(png_ptr);
+
+and to have libpng flush the output stream periodically after a certain
+number of scanlines have been written, call:
+
+    png_set_flush(png_ptr, nrows);
+
+Note that the distance between rows is from the last time png_write_flush()
+was called, or the first row of the image if it has never been called.
+So if you write 50 lines, and then png_set_flush 25, it will flush the
+output on the next scanline, and every 25 lines thereafter, unless
+png_write_flush() is called before 25 more lines have been written.
+If nrows is too small (less than about 10 lines for a 640 pixel wide
+RGB image) the image compression may decrease noticeably (although this
+may be acceptable for real-time applications).  Infrequent flushing will
+only degrade the compression performance by a few percent over images
+that do not use flushing.
+
+Writing the image data
+
+That's it for the transformations.  Now you can write the image data.
+The simplest way to do this is in one function call.  If you have the
+whole image in memory, you can just call png_write_image() and libpng
+will write the image.  You will need to pass in an array of pointers to
+each row.  This function automatically handles interlacing, so you don't
+need to call png_set_interlace_handling() or call this function multiple
+times, or any of that other stuff necessary with png_write_rows().
+
+    png_write_image(png_ptr, row_pointers);
+
+where row_pointers is:
+
+    png_byte *row_pointers[height];
+
+You can point to void or char or whatever you use for pixels.
+
+If you don't want to write the whole image at once, you can
+use png_write_rows() instead.  If the file is not interlaced,
+this is simple:
+
+    png_write_rows(png_ptr, row_pointers,
+       number_of_rows);
+
+row_pointers is the same as in the png_write_image() call.
+
+If you are just writing one row at a time, you can do this with
+a single row_pointer instead of an array of row_pointers:
+
+    png_bytep row_pointer = row;
+
+    png_write_row(png_ptr, row_pointer);
+
+When the file is interlaced, things can get a good deal more
+complicated.  The only currently (as of the PNG Specification
+version 1.2, dated July 1999) defined interlacing scheme for PNG files
+is the "Adam7" interlace scheme, that breaks down an
+image into seven smaller images of varying size.  libpng will build
+these images for you, or you can do them yourself.  If you want to
+build them yourself, see the PNG specification for details of which
+pixels to write when.
+
+If you don't want libpng to handle the interlacing details, just
+use png_set_interlace_handling() and call png_write_rows() the
+correct number of times to write all seven sub-images.
+
+If you want libpng to build the sub-images, call this before you start
+writing any rows:
+
+    number_of_passes =
+       png_set_interlace_handling(png_ptr);
+
+This will return the number of passes needed.  Currently, this
+is seven, but may change if another interlace type is added.
+
+Then write the complete image number_of_passes times.
+
+    png_write_rows(png_ptr, row_pointers,
+       number_of_rows);
+
+As some of these rows are not used, and thus return immediately,
+you may want to read about interlacing in the PNG specification,
+and only update the rows that are actually used.
+
+Finishing a sequential write
+
+After you are finished writing the image, you should finish writing
+the file.  If you are interested in writing comments or time, you should
+pass an appropriately filled png_info pointer.  If you are not interested,
+you can pass NULL.
+
+    png_write_end(png_ptr, info_ptr);
+
+When you are done, you can free all memory used by libpng like this:
+
+    png_destroy_write_struct(&png_ptr, &info_ptr);
+
+It is also possible to individually free the info_ptr members that
+point to libpng-allocated storage with the following functions:
+
+    png_free_data(png_ptr, info_ptr, mask, n)
+    mask         - identifies data to be freed, a mask
+                   made up by the OR one or more of
+                   PNG_FREE_PLTE, PNG_FREE_TRNS,
+                   PNG_FREE_HIST, PNG_FREE_ICCP,
+                   PNG_FREE_SPLT, PNG_FREE_ROWS,
+                   PNG_FREE_PCAL, PNG_FREE_SCAL,
+                   PNG_FREE_TEXT, PNG_FREE_UNKN,
+                   or simply PNG_FREE_ALL
+    n            - sequence number of item to be freed
+                   (-1 for all items)
+
+These functions may be safely called when the relevant storage has
+already been freed, or has not yet been allocated, and will in that
+case do nothing.  The "n" parameter is ignored if only one item
+of the selected data type, such as PLTE, is allowed.  If "n" is not
+-1, and multiple items are allowed for the data type identified in
+the mask, such as text or splt, only the n'th item is freed.
+
+If you allocated data such as a palette that you passed in to libpng with
+png_set_*, you must not free it until just before the call to
+png_destroy_write_struct().
+
+For a more compact example of writing a PNG image, see the file example.c.
+
+V. Modifying/Customizing libpng:
+
+There are two issues here.  The first is changing how libpng does
+standard things like memory allocation, input/output, and error handling.
+The second deals with more complicated things like adding new chunks,
+adding new transformations, and generally changing how libpng works.
+
+All of the memory allocation, input/output, and error handling in libpng
+goes through callbacks that are user settable.  The default routines are
+in pngmem.c, pngrio.c, pngwio.c, and pngerror.c respectively.  To change
+these functions, call the appropriate png_set_*_fn() function.
+
+Memory allocation is done through the functions png_large_malloc(),
+png_malloc(), png_realloc(), png_large_free(), and png_free().  These
+currently just call the standard C functions.  The large functions must
+handle exactly 64K, but they don't have to handle more than that.  If
+your pointers can't access more then 64K at a time, you will want to set
+MAXSEG_64K in zlib.h.  Since it is unlikely that the method of handling
+memory allocation on a platform will change between applications, these
+functions must be modified in the library at compile time.
+
+Input/Output in libpng is done through png_read() and png_write(),
+which currently just call fread() and fwrite().  The FILE * is stored in
+png_struct and is initialized via png_init_io().  If you wish to change
+the method of I/O, the library supplies callbacks that you can set
+through the function png_set_read_fn() and png_set_write_fn() at run
+time, instead of calling the png_init_io() function.
+These functions
+also provide a void pointer that can be retrieved via the function
+png_get_io_ptr().  For example:
+
+    png_set_read_fn(png_structp read_ptr,
+        voidp read_io_ptr, png_rw_ptr read_data_fn)
+
+    png_set_write_fn(png_structp write_ptr,
+        voidp write_io_ptr, png_rw_ptr write_data_fn,
+        png_flush_ptr output_flush_fn);
+
+    voidp read_io_ptr = png_get_io_ptr(read_ptr);
+    voidp write_io_ptr = png_get_io_ptr(write_ptr);
+
+The replacement I/O functions must have prototypes as follows:
+
+    void user_read_data(png_structp png_ptr,
+        png_bytep data, png_uint_32 length);
+    void user_write_data(png_structp png_ptr,
+        png_bytep data, png_uint_32 length);
+    void user_flush_data(png_structp png_ptr);
+
+Supplying NULL for the read, write, or flush functions sets them back
+to using the default C stream functions.  It is an error to read from
+a write stream, and vice versa.
+
+Error handling in libpng is done through png_error() and png_warning().
+Errors handled through png_error() are fatal, meaning that png_error()
+should never return to its caller.  Currently, this is handled via
+setjmp() and longjmp() (unless you have compiled libpng with
+PNG_SETJMP_NOT_SUPPORTED, in which case it is handled via PNG_ABORT()),
+but you could change this to do things like exit() if you should wish.
+
+On non-fatal errors, png_warning() is called
+to print a warning message, and then control returns to the calling code.
+By default png_error() and png_warning() print a message on stderr via
+fprintf() unless the library is compiled with PNG_NO_CONSOLE_IO defined
+(because you don't want the messages) or PNG_NO_STDIO defined (because
+fprintf() isn't available).  If you wish to change the behavior of the error
+functions, you will need to set up your own message callbacks.  These
+functions are normally supplied at the time that the png_struct is created.
+It is also possible to redirect errors and warnings to your own replacement
+functions after png_create_*_struct() has been called by calling:
+
+    png_set_error_fn(png_structp png_ptr,
+        png_voidp error_ptr, png_error_ptr error_fn,
+        png_error_ptr warning_fn);
+
+    png_voidp error_ptr = png_get_error_ptr(png_ptr);
+
+If NULL is supplied for either error_fn or warning_fn, then the libpng
+default function will be used, calling fprintf() and/or longjmp() if a
+problem is encountered.  The replacement error functions should have
+parameters as follows:
+
+    void user_error_fn(png_structp png_ptr,
+        png_const_charp error_msg);
+    void user_warning_fn(png_structp png_ptr,
+        png_const_charp warning_msg);
+
+The motivation behind using setjmp() and longjmp() is the C++ throw and
+catch exception handling methods.  This makes the code much easier to write,
+as there is no need to check every return code of every function call.
+However, there are some uncertainties about the status of local variables
+after a longjmp, so the user may want to be careful about doing anything after
+setjmp returns non-zero besides returning itself.  Consult your compiler
+documentation for more details.  For an alternative approach, you may wish
+to use the "cexcept" facility (see http://cexcept.sourceforge.net).
+
+Custom chunks
+
+If you need to read or write custom chunks, you may need to get deeper
+into the libpng code.  The library now has mechanisms for storing
+and writing chunks of unknown type; you can even declare callbacks
+for custom chunks.  Hoewver, this may not be good enough if the
+library code itself needs to know about interactions between your
+chunk and existing `intrinsic' chunks.
+
+If you need to write a new intrinsic chunk, first read the PNG
+specification. Acquire a first level of
+understanding of how it works.  Pay particular attention to the
+sections that describe chunk names, and look at how other chunks were
+designed, so you can do things similarly.  Second, check out the
+sections of libpng that read and write chunks.  Try to find a chunk
+that is similar to yours and use it as a template.  More details can
+be found in the comments inside the code.  It is best to handle unknown
+chunks in a generic method, via callback functions, instead of by
+modifying libpng functions.
+
+If you wish to write your own transformation for the data, look through
+the part of the code that does the transformations, and check out some of
+the simpler ones to get an idea of how they work.  Try to find a similar
+transformation to the one you want to add and copy off of it.  More details
+can be found in the comments inside the code itself.
+
+Configuring for 16 bit platforms
+
+You will want to look into zconf.h to tell zlib (and thus libpng) that
+it cannot allocate more then 64K at a time.  Even if you can, the memory
+won't be accessible.  So limit zlib and libpng to 64K by defining MAXSEG_64K.
+
+Configuring for DOS
+
+For DOS users who only have access to the lower 640K, you will
+have to limit zlib's memory usage via a png_set_compression_mem_level()
+call.  See zlib.h or zconf.h in the zlib library for more information.
+
+Configuring for Medium Model
+
+Libpng's support for medium model has been tested on most of the popular
+compilers.  Make sure MAXSEG_64K gets defined, USE_FAR_KEYWORD gets
+defined, and FAR gets defined to far in pngconf.h, and you should be
+all set.  Everything in the library (except for zlib's structure) is
+expecting far data.  You must use the typedefs with the p or pp on
+the end for pointers (or at least look at them and be careful).  Make
+note that the rows of data are defined as png_bytepp, which is an
+unsigned char far * far *.
+
+Configuring for gui/windowing platforms:
+
+You will need to write new error and warning functions that use the GUI
+interface, as described previously, and set them to be the error and
+warning functions at the time that png_create_*_struct() is called,
+in order to have them available during the structure initialization.
+They can be changed later via png_set_error_fn().  On some compilers,
+you may also have to change the memory allocators (png_malloc, etc.).
+
+Configuring for compiler xxx:
+
+All includes for libpng are in pngconf.h.  If you need to add/change/delete
+an include, this is the place to do it.  The includes that are not
+needed outside libpng are protected by the PNG_INTERNAL definition,
+which is only defined for those routines inside libpng itself.  The
+files in libpng proper only include png.h, which includes pngconf.h.
+
+Configuring zlib:
+
+There are special functions to configure the compression.  Perhaps the
+most useful one changes the compression level, which currently uses
+input compression values in the range 0 - 9.  The library normally
+uses the default compression level (Z_DEFAULT_COMPRESSION = 6).  Tests
+have shown that for a large majority of images, compression values in
+the range 3-6 compress nearly as well as higher levels, and do so much
+faster.  For online applications it may be desirable to have maximum speed
+(Z_BEST_SPEED = 1).  With versions of zlib after v0.99, you can also
+specify no compression (Z_NO_COMPRESSION = 0), but this would create
+files larger than just storing the raw bitmap.  You can specify the
+compression level by calling:
+
+    png_set_compression_level(png_ptr, level);
+
+Another useful one is to reduce the memory level used by the library.
+The memory level defaults to 8, but it can be lowered if you are
+short on memory (running DOS, for example, where you only have 640K).
+
+    png_set_compression_mem_level(png_ptr, level);
+
+The other functions are for configuring zlib.  They are not recommended
+for normal use and may result in writing an invalid PNG file.  See
+zlib.h for more information on what these mean.
+
+    png_set_compression_strategy(png_ptr,
+        strategy);
+    png_set_compression_window_bits(png_ptr,
+        window_bits);
+    png_set_compression_method(png_ptr, method);
+    png_set_compression_buffer_size(png_ptr, size);
+
+Controlling row filtering
+
+If you want to control whether libpng uses filtering or not, which
+filters are used, and how it goes about picking row filters, you
+can call one of these functions.  The selection and configuration
+of row filters can have a significant impact on the size and
+encoding speed and a somewhat lesser impact on the decoding speed
+of an image.  Filtering is enabled by default for RGB and grayscale
+images (with and without alpha), but not for paletted images nor
+for any images with bit depths less than 8 bits/pixel.
+
+The 'method' parameter sets the main filtering method, which is
+currently only '0' in the PNG 1.2 specification.  The 'filters'
+parameter sets which filter(s), if any, should be used for each
+scanline.  Possible values are PNG_ALL_FILTERS and PNG_NO_FILTERS
+to turn filtering on and off, respectively.
+
+Individual filter types are PNG_FILTER_NONE, PNG_FILTER_SUB,
+PNG_FILTER_UP, PNG_FILTER_AVG, PNG_FILTER_PAETH, which can be bitwise
+ORed together with '|' to specify one or more filters to use.
+These filters are described in more detail in the PNG specification.  If
+you intend to change the filter type during the course of writing
+the image, you should start with flags set for all of the filters
+you intend to use so that libpng can initialize its internal
+structures appropriately for all of the filter types.
+
+    filters = PNG_FILTER_NONE | PNG_FILTER_SUB
+              PNG_FILTER_UP | PNG_FILTER_AVE |
+              PNG_FILTER_PAETH | PNG_ALL_FILTERS;
+    or
+    filters = one of PNG_FILTER_VALUE_NONE,
+              PNG_FILTER_VALUE_SUB, PNG_FILTER_VALUE_UP,
+              PNG_FILTER_VALUE_AVE, PNG_FILTER_VALUE_PAETH
+
+    png_set_filter(png_ptr, PNG_FILTER_TYPE_BASE,
+       filters);
+              The second parameter can also be
+              PNG_INTRAPIXEL_DIFFERENCING if you are
+              writing a PNG to be embedded in a MNG
+              datastream.  This parameter must be the
+              same as the value of filter_method used
+              in png_set_IHDR().
+
+It is also possible to influence how libpng chooses from among the
+available filters.  This is done in two ways - by telling it how
+important it is to keep the same filter for successive rows, and
+by telling it the relative computational costs of the filters.
+
+    double weights[3] = {1.5, 1.3, 1.1},
+       costs[PNG_FILTER_VALUE_LAST] =
+       {1.0, 1.3, 1.3, 1.5, 1.7};
+
+    png_set_filter_selection(png_ptr,
+       PNG_FILTER_SELECTION_WEIGHTED, 3,
+       weights, costs);
+
+The weights are multiplying factors that indicate to libpng that the
+row filter should be the same for successive rows unless another row filter
+is that many times better than the previous filter.  In the above example,
+if the previous 3 filters were SUB, SUB, NONE, the SUB filter could have a
+"sum of absolute differences" 1.5 x 1.3 times higher than other filters
+and still be chosen, while the NONE filter could have a sum 1.1 times
+higher than other filters and still be chosen.  Unspecified weights are
+taken to be 1.0, and the specified weights should probably be declining
+like those above in order to emphasize recent filters over older filters.
+
+The filter costs specify for each filter type a relative decoding cost
+to be considered when selecting row filters.  This means that filters
+with higher costs are less likely to be chosen over filters with lower
+costs, unless their "sum of absolute differences" is that much smaller.
+The costs do not necessarily reflect the exact computational speeds of
+the various filters, since this would unduly influence the final image
+size.
+
+Note that the numbers above were invented purely for this example and
+are given only to help explain the function usage.  Little testing has
+been done to find optimum values for either the costs or the weights.
+
+Removing unwanted object code
+
+There are a bunch of #define's in pngconf.h that control what parts of
+libpng are compiled.  All the defines end in _SUPPORTED.  If you are
+never going to use a capability, you can change the #define to #undef
+before recompiling libpng and save yourself code and data space, or
+you can turn off individual capabilities with defines that begin with
+PNG_NO_.
+
+You can also turn all of the transforms and ancillary chunk capabilities
+off en masse with compiler directives that define
+PNG_NO_READ[or WRITE]_TRANSFORMS, or PNG_NO_READ[or WRITE]_ANCILLARY_CHUNKS,
+or all four,
+along with directives to turn on any of the capabilities that you do
+want.  The PNG_NO_READ[or WRITE]_TRANSFORMS directives disable
+the extra transformations but still leave the library fully capable of reading
+and writing PNG files with all known public chunks
+Use of the PNG_NO_READ[or WRITE]_ANCILLARY_CHUNKS directive
+produces a library that is incapable of reading or writing ancillary chunks.
+If you are not using the progressive reading capability, you can
+turn that off with PNG_NO_PROGRESSIVE_READ (don't confuse
+this with the INTERLACING capability, which you'll still have).
+
+All the reading and writing specific code are in separate files, so the
+linker should only grab the files it needs.  However, if you want to
+make sure, or if you are building a stand alone library, all the
+reading files start with pngr and all the writing files start with
+pngw.  The files that don't match either (like png.c, pngtrans.c, etc.)
+are used for both reading and writing, and always need to be included.
+The progressive reader is in pngpread.c
+
+If you are creating or distributing a dynamically linked library (a .so
+or DLL file), you should not remove or disable any parts of the library,
+as this will cause applications linked with different versions of the
+library to fail if they call functions not available in your library.
+The size of the library itself should not be an issue, because only
+those sections that are actually used will be loaded into memory.
+
+Requesting debug printout
+
+The macro definition PNG_DEBUG can be used to request debugging
+printout.  Set it to an integer value in the range 0 to 3.  Higher
+numbers result in increasing amounts of debugging information.  The
+information is printed to the "stderr" file, unless another file
+name is specified in the PNG_DEBUG_FILE macro definition.
+
+When PNG_DEBUG > 0, the following functions (macros) become available:
+
+   png_debug(level, message)
+   png_debug1(level, message, p1)
+   png_debug2(level, message, p1, p2)
+
+in which "level" is compared to PNG_DEBUG to decide whether to print
+the message, "message" is the formatted string to be printed,
+and p1 and p2 are parameters that are to be embedded in the string
+according to printf-style formatting directives.  For example,
+
+   png_debug1(2, "foo=%d\n", foo);
+
+is expanded to
+
+   if(PNG_DEBUG > 2)
+     fprintf(PNG_DEBUG_FILE, "foo=%d\n", foo);
+
+When PNG_DEBUG is defined but is zero, the macros aren't defined, but you
+can still use PNG_DEBUG to control your own debugging:
+
+   #ifdef PNG_DEBUG
+       fprintf(stderr, ...
+   #endif
+
+When PNG_DEBUG = 1, the macros are defined, but only png_debug statements
+having level = 0 will be printed.  There aren't any such statements in
+this version of libpng, but if you insert some they will be printed.
+
+VI.  Runtime optimization
+
+A new feature in libpng 1.2.0 is the ability to dynamically switch between
+standard and optimized versions of some routines.  Currently these are
+limited to three computationally intensive tasks when reading PNG files:
+decoding row filters, expanding interlacing, and combining interlaced or
+transparent row data with previous row data.  Currently the optimized
+versions are available only for x86 (Intel, AMD, etc.) platforms with
+MMX support, though this may change in future versions.  (For example,
+the non-MMX assembler optimizations for zlib might become similarly
+runtime-selectable in future releases, in which case libpng could be
+extended to support them.  Alternatively, the compile-time choice of
+floating-point versus integer routines for gamma correction might become
+runtime-selectable.)
+
+Because such optimizations tend to be very platform- and compiler-dependent,
+both in how they are written and in how they perform, the new runtime code
+in libpng has been written to allow programs to query, enable, and disable
+either specific optimizations or all such optimizations.  For example, to
+enable all possible optimizations (bearing in mind that some "optimizations"
+may actually run more slowly in rare cases):
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       png_uint_32 mask, flags;
+
+       flags = png_get_asm_flags(png_ptr);
+       mask = png_get_asm_flagmask(PNG_SELECT_READ | PNG_SELECT_WRITE);
+       png_set_asm_flags(png_ptr, flags | mask);
+    #endif
+
+To enable only optimizations relevant to reading PNGs, use PNG_SELECT_READ
+by itself when calling png_get_asm_flagmask(); similarly for optimizing
+only writing.  To disable all optimizations:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       flags = png_get_asm_flags(png_ptr);
+       mask = png_get_asm_flagmask(PNG_SELECT_READ | PNG_SELECT_WRITE);
+       png_set_asm_flags(png_ptr, flags & ~mask);
+    #endif
+
+To enable or disable only MMX-related features, use png_get_mmx_flagmask()
+in place of png_get_asm_flagmask().  The mmx version takes one additional
+parameter:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       int selection = PNG_SELECT_READ | PNG_SELECT_WRITE;
+       int compilerID;
+
+       mask = png_get_mmx_flagmask(selection, &compilerID);
+    #endif
+
+On return, compilerID will indicate which version of the MMX assembler
+optimizations was compiled.  Currently two flavors exist:  Microsoft
+Visual C++ (compilerID == 1) and GNU C (a.k.a. gcc/gas, compilerID == 2).
+On non-x86 platforms or on systems compiled without MMX optimizations, a
+value of -1 is used.
+
+Note that both png_get_asm_flagmask() and png_get_mmx_flagmask() return
+all valid, settable optimization bits for the version of the library that's
+currently in use.  In the case of shared (dynamically linked) libraries,
+this may include optimizations that did not exist at the time the code was
+written and compiled.  It is also possible, of course, to enable only known,
+specific optimizations; for example:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       flags = PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  \
+             | PNG_ASM_FLAG_MMX_READ_INTERLACE    \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_SUB   \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_UP    \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_AVG   \
+             | PNG_ASM_FLAG_MMX_READ_FILTER_PAETH ;
+       png_set_asm_flags(png_ptr, flags);
+    #endif
+
+This method would enable only the MMX read-optimizations available at the
+time of libpng 1.2.0's release, regardless of whether a later version of
+the DLL were actually being used.  (Also note that these functions did not
+exist in versions older than 1.2.0, so any attempt to run a dynamically
+linked app on such an older version would fail.)
+
+To determine whether the processor supports MMX instructions at all, use
+the png_mmx_support() function:
+
+    #if defined(PNG_LIBPNG_VER) && (PNG_LIBPNG_VER >= 10200)
+       mmxsupport = png_mmx_support();
+    #endif
+
+It returns -1 if MMX support is not compiled into libpng, 0 if MMX code
+is compiled but MMX is not supported by the processor, or 1 if MMX support
+is fully available.  Note that png_mmx_support(), png_get_mmx_flagmask(),
+and png_get_asm_flagmask() all may be called without allocating and ini-
+tializing any PNG structures (for example, as part of a usage screen or
+"about" box).
+
+The following code can be used to prevent an application from using the
+thread_unsafe features, even if libpng was built with PNG_THREAD_UNSAFE_OK
+defined:
+
+#if defined(PNG_USE_PNGGCCRD) && defined(PNG_ASSEMBLER_CODE_SUPPORTED) \
+  && defined(PNG_THREAD_UNSAFE_OK)
+    /* Disable thread-unsafe features of pnggccrd */
+    if (png_access_version() >= 10200)
+    {
+      png_uint_32 mmx_disable_mask = 0;
+      png_uint_32 asm_flags;
+
+      mmx_disable_mask |= ( PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  \
+                          | PNG_ASM_FLAG_MMX_READ_FILTER_SUB   \
+                          | PNG_ASM_FLAG_MMX_READ_FILTER_AVG   \
+                          | PNG_ASM_FLAG_MMX_READ_FILTER_PAETH );
+      asm_flags = png_get_asm_flags(png_ptr);
+      png_set_asm_flags(png_ptr, asm_flags & ~mmx_disable_mask);
+    }
+#endif
+
+For more extensive examples of runtime querying, enabling and disabling
+of optimized features, see contrib/gregbook/readpng2.c in the libpng
+source-code distribution.
+
+
+VII.  MNG support
+
+The MNG specification (available at http://www.libpng.org/pub/mng) allows
+certain extensions to PNG for PNG images that are embedded in MNG datastreams.
+Libpng can support some of these extensions.  To enable them, use the
+png_permit_mng_features() function:
+
+   feature_set = png_permit_mng_features(png_ptr, mask)
+   mask is a png_uint_32 containing the logical OR of the
+        features you want to enable.  These include
+        PNG_FLAG_MNG_EMPTY_PLTE
+        PNG_FLAG_MNG_FILTER_64
+        PNG_ALL_MNG_FEATURES
+   feature_set is a png_32_uint that is the logical AND of
+      your mask with the set of MNG features that is
+      supported by the version of libpng that you are using.
+
+It is an error to use this function when reading or writing a standalone
+PNG file with the PNG 8-byte signature.  The PNG datastream must be wrapped
+in a MNG datastream.  As a minimum, it must have the MNG 8-byte signature
+and the MHDR and MEND chunks.  Libpng does not provide support for these
+or any other MNG chunks; your application must provide its own support for
+them.  You may wish to consider using libmng (available at
+http://www.libmng.com) instead.
+
+VIII.  Changes to Libpng from version 0.88
+
+It should be noted that versions of libpng later than 0.96 are not
+distributed by the original libpng author, Guy Schalnat, nor by
+Andreas Dilger, who had taken over from Guy during 1996 and 1997, and
+distributed versions 0.89 through 0.96, but rather by another member
+of the original PNG Group, Glenn Randers-Pehrson.  Guy and Andreas are
+still alive and well, but they have moved on to other things.
+
+The old libpng functions png_read_init(), png_write_init(),
+png_info_init(), png_read_destroy(), and png_write_destroy() have been
+moved to PNG_INTERNAL in version 0.95 to discourage their use.  These
+functions will be removed from libpng version 2.0.0.
+
+The preferred method of creating and initializing the libpng structures is
+via the png_create_read_struct(), png_create_write_struct(), and
+png_create_info_struct() because they isolate the size of the structures
+from the application, allow version error checking, and also allow the
+use of custom error handling routines during the initialization, which
+the old functions do not.  The functions png_read_destroy() and
+png_write_destroy() do not actually free the memory that libpng
+allocated for these structs, but just reset the data structures, so they
+can be used instead of png_destroy_read_struct() and
+png_destroy_write_struct() if you feel there is too much system overhead
+allocating and freeing the png_struct for each image read.
+
+Setting the error callbacks via png_set_message_fn() before
+png_read_init() as was suggested in libpng-0.88 is no longer supported
+because this caused applications that do not use custom error functions
+to fail if the png_ptr was not initialized to zero.  It is still possible
+to set the error callbacks AFTER png_read_init(), or to change them with
+png_set_error_fn(), which is essentially the same function, but with a new
+name to force compilation errors with applications that try to use the old
+method.
+
+Starting with version 1.0.7, you can find out which version of the library
+you are using at run-time:
+
+   png_uint_32 libpng_vn = png_access_version_number();
+
+The number libpng_vn is constructed from the major version, minor
+version with leading zero, and release number with leading zero,
+(e.g., libpng_vn for version 1.0.7 is 10007).
+
+You can also check which version of png.h you used when compiling your
+application:
+
+   png_uint_32 application_vn = PNG_LIBPNG_VER;
+
+IX. Y2K Compliance in libpng
+
+December 12, 2001
+
+Since the PNG Development group is an ad-hoc body, we can't make
+an official declaration.
+
+This is your unofficial assurance that libpng from version 0.71 and
+upward through 1.2.1 are Y2K compliant.  It is my belief that earlier
+versions were also Y2K compliant.
+
+Libpng only has three year fields.  One is a 2-byte unsigned integer that
+will hold years up to 65535.  The other two hold the date in text
+format, and will hold years up to 9999.
+
+The integer is
+    "png_uint_16 year" in png_time_struct.
+
+The strings are
+    "png_charp time_buffer" in png_struct and
+    "near_time_buffer", which is a local character string in png.c.
+
+There are seven time-related functions:
+
+    png_convert_to_rfc_1123() in png.c
+      (formerly png_convert_to_rfc_1152() in error)
+    png_convert_from_struct_tm() in pngwrite.c, called
+      in pngwrite.c
+    png_convert_from_time_t() in pngwrite.c
+    png_get_tIME() in pngget.c
+    png_handle_tIME() in pngrutil.c, called in pngread.c
+    png_set_tIME() in pngset.c
+    png_write_tIME() in pngwutil.c, called in pngwrite.c
+
+All appear to handle dates properly in a Y2K environment.  The
+png_convert_from_time_t() function calls gmtime() to convert from system
+clock time, which returns (year - 1900), which we properly convert to
+the full 4-digit year.  There is a possibility that applications using
+libpng are not passing 4-digit years into the png_convert_to_rfc_1123()
+function, or that they are incorrectly passing only a 2-digit year
+instead of "year - 1900" into the png_convert_from_struct_tm() function,
+but this is not under our control.  The libpng documentation has always
+stated that it works with 4-digit years, and the APIs have been
+documented as such.
+
+The tIME chunk itself is also Y2K compliant.  It uses a 2-byte unsigned
+integer to hold the year, and can hold years as large as 65535.
+
+zlib, upon which libpng depends, is also Y2K compliant.  It contains
+no date-related code.
+
+
+   Glenn Randers-Pehrson
+   libpng maintainer
+   PNG Development Group
diff --git a/png/libpngpf.3 b/png/libpngpf.3
new file mode 100644
index 000000000..ea1fae7a6
--- /dev/null
+++ b/png/libpngpf.3
@@ -0,0 +1,558 @@
+.TH LIBPNGPF 3 "December 12, 2001"
+.SH NAME
+libpng \- Portable Network Graphics (PNG) Reference Library 1.2.1
+(private functions)
+.SH SYNOPSIS
+\fB#include <png.h>\fP
+
+\fI\fB
+
+\fBvoid png_build_gamma_table (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_build_grayscale_palette (int \fP\fIbit_depth\fP\fB, png_colorp \fIpalette\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_calculate_crc (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIptr\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_check_chunk_name (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fIchunk_name\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_size_t png_check_keyword (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIkey\fP\fB, png_charpp \fInew_key\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_combine_row (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIrow\fP\fB, int \fImask\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_correct_palette (png_structp \fP\fIpng_ptr\fP\fB, png_colorp \fP\fIpalette\fP\fB, int \fInum_palette\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_crc_error (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_crc_finish (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIskip\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_crc_read (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIbuf\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_create_struct (int \fItype\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_voidp png_create_struct_2 (int \fP\fItype\fP\fB, png_malloc_ptr \fP\fImalloc_fn\fP\fB, png_voidp \fImem_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_charp png_decompress_chunk (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fIcomp_type\fP\fB, png_charp \fP\fIchunkdata\fP\fB, png_size_t \fP\fIchunklength\fP\fB, png_size_t \fP\fIprefix_length\fP\fB, png_size_t \fI*data_length\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_destroy_struct (png_voidp \fIstruct_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_destroy_struct_2 (png_voidp \fP\fIstruct_ptr\fP\fB, png_free_ptr \fP\fIfree_fn\fP\fB, png_voidp \fImem_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_background (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_color_16p \fP\fItrans_values\fP\fB, png_color_16p \fP\fIbackground\fP\fB, png_color_16p \fP\fIbackground_1\fP\fB, png_bytep \fP\fIgamma_table\fP\fB, png_bytep \fP\fIgamma_from_1\fP\fB, png_bytep \fP\fIgamma_to_1\fP\fB, png_uint_16pp \fP\fIgamma_16\fP\fB, png_uint_16pp \fP\fIgamma_16_from_1\fP\fB, png_uint_16pp \fP\fIgamma_16_to_1\fP\fB, int \fIgamma_shift\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_bgr (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_chop (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_dither (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_bytep \fP\fIpalette_lookup\fP\fB, png_bytep \fIdither_lookup\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_expand (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_color_16p \fItrans_value\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_expand_palette (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_colorp \fP\fIpalette\fP\fB, png_bytep \fP\fItrans\fP\fB, int \fInum_trans\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_gamma (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_bytep \fP\fIgamma_table\fP\fB, png_uint_16pp \fP\fIgamma_16_table\fP\fB, int \fIgamma_shift\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_gray_to_rgb (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_invert (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_pack (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_uint_32 \fIbit_depth\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_packswap (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_read_filler (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_uint_32 \fP\fIfiller\fP\fB, png_uint_32 \fIflags\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_read_interlace (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, int \fP\fIpass\fP\fB, png_uint_32 \fItransformations\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_read_invert_alpha (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_read_swap_alpha (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_read_transformations (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_do_rgb_to_gray (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_shift (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_color_8p \fIbit_depth\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_strip_filler (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_uint_32 \fIflags\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_swap (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_unpack (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_unshift (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_color_8p \fIsig_bits\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_write_interlace (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, int \fIpass\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_write_invert_alpha (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_write_swap_alpha (png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_do_write_transformations (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid *png_far_to_near (png_structp png_ptr,png_voidp \fP\fIptr\fP\fB, int \fIcheck\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_flush (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_int_32 png_get_int_32 (png_bytep \fIbuf\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_16 png_get_uint_16 (png_bytep \fIbuf\fP\fB);\fP
+
+\fI\fB
+
+\fBpng_uint_32 png_get_uint_32 (png_bytep \fIbuf\fP\fB);\fP
+
+\fI\fB
+
+\fBint png_handle_as_unknown (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fIchunk_name\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_bKGD (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_cHRM (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_gAMA (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_hIST (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_IEND (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_IHDR (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_iCCP (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_iTXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_oFFs (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_pCAL (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_pHYs (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_PLTE (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_sBIT (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_sCAL (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_sPLT (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_sRGB (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_tEXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_tIME (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_tRNS (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_unknown (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_handle_zTXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_info_destroy (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_init_mmx_flags (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_init_read_transformations (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_process_IDAT_data (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIbuffer\fP\fB, png_size_t \fIbuffer_length\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_process_some_data (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_check_crc (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_crc_finish (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_crc_skip (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_fill_buffer (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIbuffer\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_handle_tEXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_handle_unknown (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_handle_zTXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fP\fIinfo_ptr\fP\fB, png_uint_32 \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_have_end (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_have_info (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_have_row (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fIrow\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_process_row (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_read_chunk (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_read_end (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_read_IDAT (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_read_sig (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_read_tEXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_read_zTXt (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_restore_buffer (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIbuffer\fP\fB, png_size_t \fIbuffer_length\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_push_save_buffer (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_data (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIdata\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_filter_row (png_structp \fP\fIpng_ptr\fP\fB, png_row_infop \fP\fIrow_info\fP\fB, png_bytep \fP\fIrow\fP\fB, png_bytep \fP\fIprev_row\fP\fB, int \fIfilter\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_finish_row (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_push_finish_row (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_start_row (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_read_transform_info (png_structp \fP\fIpng_ptr\fP\fB, png_infop \fIinfo_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_reset_crc (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_save_int_32 (png_bytep \fP\fIbuf\fP\fB, png_int_32 \fIi\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_save_uint_16 (png_bytep \fP\fIbuf\fP\fB, unsigned int \fIi\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_save_uint_32 (png_bytep \fP\fIbuf\fP\fB, png_uint_32 \fIi\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_bKGD (png_structp \fP\fIpng_ptr\fP\fB, png_color_16p \fP\fIvalues\fP\fB, int \fIcolor_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_cHRM (png_structp \fP\fIpng_ptr\fP\fB, double \fP\fIwhite_x\fP\fB, double \fP\fIwhite_y\fP\fB, double \fP\fIred_x\fP\fB, double \fP\fIred_y\fP\fB, double \fP\fIgreen_x\fP\fB, double \fP\fIgreen_y\fP\fB, double \fP\fIblue_x\fP\fB, double \fIblue_y\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_cHRM_fixed (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fP\fIwhite_x\fP\fB, png_uint_32 \fP\fIwhite_y\fP\fB, png_uint_32 \fP\fIred_x\fP\fB, png_uint_32 \fP\fIred_y\fP\fB, png_uint_32 \fP\fIgreen_x\fP\fB, png_uint_32 \fP\fIgreen_y\fP\fB, png_uint_32 \fP\fIblue_x\fP\fB, png_uint_32 \fIblue_y\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_data (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIdata\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_filtered_row (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fIfiltered_row\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_find_filter (png_structp \fP\fIpng_ptr\fP\fB, png_row_infop \fIrow_info\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_finish_row (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_gAMA (png_structp \fP\fIpng_ptr\fP\fB, double \fIfile_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_gAMA_fixed (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fIint_file_gamma\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_hIST (png_structp \fP\fIpng_ptr\fP\fB, png_uint_16p \fP\fIhist\fP\fB, int \fInum_hist\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_iCCP (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIname\fP\fB, int \fP\fIcompression_type\fP\fB, png_charp \fP\fIprofile\fP\fB, int \fIproflen\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_IDAT (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fIdata\fP\fB, png_size_t \fIlength\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_IEND (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_IHDR (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fP\fIwidth\fP\fB, png_uint_32 \fP\fIheight\fP\fB, int \fP\fIbit_depth\fP\fB, int \fP\fIcolor_type\fP\fB, int \fP\fIcompression_type\fP\fB, int \fP\fIfilter_type\fP\fB, int \fIinterlace_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_iTXt (png_structp \fP\fIpng_ptr\fP\fB, int \fP\fIcompression\fP\fB, png_charp \fP\fIkey\fP\fB, png_charp \fP\fIlang\fP\fB, png_charp \fP\fItranslated_key\fP\fB, png_charp \fItext)\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_oFFs (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fP\fIx_offset\fP\fB, png_uint_32 \fP\fIy_offset\fP\fB, int \fIunit_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_pCAL (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIpurpose\fP\fB, png_int_32 \fP\fIX0\fP\fB, png_int_32 \fP\fIX1\fP\fB, int \fP\fItype\fP\fB, int \fP\fInparams\fP\fB, png_charp \fP\fIunits\fP\fB, png_charpp \fIparams\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_pHYs (png_structp \fP\fIpng_ptr\fP\fB, png_uint_32 \fP\fIx_pixels_per_unit\fP\fB, png_uint_32 \fP\fIy_pixels_per_unit\fP\fB, int \fIunit_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_PLTE (png_structp \fP\fIpng_ptr\fP\fB, png_colorp \fP\fIpalette\fP\fB, png_uint_32 \fInum_pal\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_sBIT (png_structp \fP\fIpng_ptr\fP\fB, png_color_8p \fP\fIsbit\fP\fB, int \fIcolor_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_sCAL (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIunit\fP\fB, double \fP\fIwidth\fP\fB, double \fIheight\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_sCAL_s (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIunit\fP\fB, png_charp \fP\fIwidth\fP\fB, png_charp \fIheight\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_sig (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_sRGB (png_structp \fP\fIpng_ptr\fP\fB, int \fIintent\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_sPLT (png_structp \fP\fIpng_ptr\fP\fB, png_spalette_p \fIpalette\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_start_row (png_structp \fIpng_ptr\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_tEXt (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIkey\fP\fB, png_charp \fP\fItext\fP\fB, png_size_t \fItext_len\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_tIME (png_structp \fP\fIpng_ptr\fP\fB, png_timep \fImod_time\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_tRNS (png_structp \fP\fIpng_ptr\fP\fB, png_bytep \fP\fItrans\fP\fB, png_color_16p \fP\fIvalues\fP\fB, int \fP\fInumber\fP\fB, int \fIcolor_type\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_write_zTXt (png_structp \fP\fIpng_ptr\fP\fB, png_charp \fP\fIkey\fP\fB, png_charp \fP\fItext\fP\fB, png_size_t \fP\fItext_len\fP\fB, int \fIcompression\fP\fB);\fP
+
+\fI\fB
+
+\fBvoidpf png_zalloc (voidpf \fP\fIpng_ptr\fP\fB, uInt \fP\fIitems\fP\fB, uInt \fIsize\fP\fB);\fP
+
+\fI\fB
+
+\fBvoid png_zfree (voidpf \fP\fIpng_ptr\fP\fB, voidpf \fIptr\fP\fB);\fP
+
+\fI\fB
+
+.SH DESCRIPTION
+The functions listed above are used privately by libpng
+and are not recommended for use by applications.  They are
+not "exported" to applications using shared libraries.  They
+are listed alphabetically here as an aid to libpng maintainers.
+See png.h for more information on these functions.
+
+.SH SEE ALSO
+.BR libpng (3),
+.BR png (5).
+
+.SH AUTHOR
+Glenn Randers-Pehrson
diff --git a/png/makedepend b/png/makedepend
new file mode 100644
index 000000000..bbb12e487
--- /dev/null
+++ b/png/makedepend
@@ -0,0 +1,17 @@
+# DO NOT DELETE
+
+png.o: png.h pngconf.h
+pngset.o: png.h pngconf.h
+pngget.o: png.h pngconf.h
+pngrutil.o: png.h pngconf.h
+pngtrans.o: png.h pngconf.h
+pngwutil.o: png.h pngconf.h
+pngread.o: png.h pngconf.h
+pngrio.o: png.h pngconf.h
+pngwio.o: png.h pngconf.h
+pngwrite.o: png.h pngconf.h
+pngrtran.o: png.h pngconf.h
+pngwtran.o: png.h pngconf.h
+pngmem.o: png.h pngconf.h
+pngerror.o: png.h pngconf.h
+pngpread.o: png.h pngconf.h
diff --git a/png/png.5 b/png/png.5
new file mode 100644
index 000000000..26cea335c
--- /dev/null
+++ b/png/png.5
@@ -0,0 +1,64 @@
+.TH PNG 5 "December 12, 2001"
+.SH NAME
+png \- Portable Network Graphics (PNG) format
+.SH DESCRIPTION
+PNG (Portable Network Graphics) is an extensible file format for the
+lossless, portable, well-compressed storage of raster images. PNG provides
+a patent-free replacement for GIF and can also replace many
+common uses of TIFF. Indexed-color, grayscale, and truecolor images are
+supported, plus an optional alpha channel. Sample depths range from
+1 to 16 bits.
+.br
+
+PNG is designed to work well in online viewing applications, such as the
+World Wide Web, so it is fully streamable with a progressive display
+option. PNG is robust, providing both full file integrity checking and
+fast, simple detection of common transmission errors. Also, PNG can store
+gamma and chromaticity data for improved color matching on heterogeneous
+platforms.
+
+.SH "SEE ALSO"
+.BR libpng (3),
+.BR zlib (3),
+.BR deflate (5),
+and
+.BR zlib (5).
+.LP
+PNG 1.2 specification, July 1999:
+.IP
+.br
+http://www.libpng.org/pub/png
+.br
+or ftp://ftp.uu.net/graphics/png/documents
+.LP
+PNG 1.0 specification, October 1996:
+.IP
+.br
+RFC 2083
+.IP
+.br
+ftp://ds.internic.net/rfc/rfc2083.txt
+.br
+or (as a W3C Recommendation) at
+.br
+http://www.w3.org/TR/REC-png.html
+.SH AUTHORS
+This man page: Glenn Randers-Pehrson
+.LP
+Portable Network Graphics (PNG) Specification Version 1.2 (July 8, 1999):
+Glenn Randers-Pehrson and others (png-list@ccrc.wustl.edu).
+.LP
+Portable Network Graphics (PNG) Specification Version 1.0 (October 1, 1996):
+Thomas Boutell and others (png-list@ccrc.wustl.edu).
+.LP
+
+
+.SH COPYRIGHT NOTICE
+The PNG-1.2 specification is copyright (c) 1999 Glenn Randers-Pehrson.
+See the specification for conditions of use and distribution.
+.LP
+The PNG-1.0 specification is copyright (c) 1996 Massachussets Institute of
+Technology.  See the specification for conditions of use and distribution.
+.LP
+.\" end of man page
+
diff --git a/png/png.c b/png/png.c
new file mode 100644
index 000000000..72bd00eba
--- /dev/null
+++ b/png/png.c
@@ -0,0 +1,777 @@
+
+/* png.c - location for general purpose libpng functions
+ *
+ * libpng version 1.2.1 - December 12, 2001
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ */
+
+#define PNG_INTERNAL
+#define PNG_NO_EXTERN
+#include "png.h"
+
+/* Generate a compiler error if there is an old png.h in the search path. */
+typedef version_1_2_1 Your_png_h_is_not_version_1_2_1;
+
+/* Version information for C files.  This had better match the version
+ * string defined in png.h.  */
+
+#ifdef PNG_USE_GLOBAL_ARRAYS
+/* png_libpng_ver was changed to a function in version 1.0.5c */
+const char png_libpng_ver[18] = "1.2.1";
+
+/* png_sig was changed to a function in version 1.0.5c */
+/* Place to hold the signature string for a PNG file. */
+const png_byte FARDATA png_sig[8] = {137, 80, 78, 71, 13, 10, 26, 10};
+
+/* Invoke global declarations for constant strings for known chunk types */
+PNG_IHDR;
+PNG_IDAT;
+PNG_IEND;
+PNG_PLTE;
+PNG_bKGD;
+PNG_cHRM;
+PNG_gAMA;
+PNG_hIST;
+PNG_iCCP;
+PNG_iTXt;
+PNG_oFFs;
+PNG_pCAL;
+PNG_sCAL;
+PNG_pHYs;
+PNG_sBIT;
+PNG_sPLT;
+PNG_sRGB;
+PNG_tEXt;
+PNG_tIME;
+PNG_tRNS;
+PNG_zTXt;
+
+/* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+/* start of interlace block */
+const int FARDATA png_pass_start[] = {0, 4, 0, 2, 0, 1, 0};
+
+/* offset to next interlace block */
+const int FARDATA png_pass_inc[] = {8, 8, 4, 4, 2, 2, 1};
+
+/* start of interlace block in the y direction */
+const int FARDATA png_pass_ystart[] = {0, 0, 4, 0, 2, 0, 1};
+
+/* offset to next interlace block in the y direction */
+const int FARDATA png_pass_yinc[] = {8, 8, 8, 4, 4, 2, 2};
+
+/* width of interlace block (used in assembler routines only) */
+#ifdef PNG_HAVE_ASSEMBLER_COMBINE_ROW
+const int FARDATA png_pass_width[] = {8, 4, 4, 2, 2, 1, 1};
+#endif
+
+/* Height of interlace block.  This is not currently used - if you need
+ * it, uncomment it here and in png.h
+const int FARDATA png_pass_height[] = {8, 8, 4, 4, 2, 2, 1};
+*/
+
+/* Mask to determine which pixels are valid in a pass */
+const int FARDATA png_pass_mask[] = {0x80, 0x08, 0x88, 0x22, 0xaa, 0x55, 0xff};
+
+/* Mask to determine which pixels to overwrite while displaying */
+const int FARDATA png_pass_dsp_mask[]
+   = {0xff, 0x0f, 0xff, 0x33, 0xff, 0x55, 0xff};
+
+#endif
+
+/* Tells libpng that we have already handled the first "num_bytes" bytes
+ * of the PNG file signature.  If the PNG data is embedded into another
+ * stream we can set num_bytes = 8 so that libpng will not attempt to read
+ * or write any of the magic bytes before it starts on the IHDR.
+ */
+
+void PNGAPI
+png_set_sig_bytes(png_structp png_ptr, int num_bytes)
+{
+   png_debug(1, "in png_set_sig_bytes\n");
+   if (num_bytes > 8)
+      png_error(png_ptr, "Too many bytes for PNG signature.");
+
+   png_ptr->sig_bytes = (png_byte)(num_bytes < 0 ? 0 : num_bytes);
+}
+
+/* Checks whether the supplied bytes match the PNG signature.  We allow
+ * checking less than the full 8-byte signature so that those apps that
+ * already read the first few bytes of a file to determine the file type
+ * can simply check the remaining bytes for extra assurance.  Returns
+ * an integer less than, equal to, or greater than zero if sig is found,
+ * respectively, to be less than, to match, or be greater than the correct
+ * PNG signature (this is the same behaviour as strcmp, memcmp, etc).
+ */
+int PNGAPI
+png_sig_cmp(png_bytep sig, png_size_t start, png_size_t num_to_check)
+{
+   png_byte png_signature[8] = {137, 80, 78, 71, 13, 10, 26, 10};
+   if (num_to_check > 8)
+      num_to_check = 8;
+   else if (num_to_check < 1)
+      return (0);
+
+   if (start > 7)
+      return (0);
+
+   if (start + num_to_check > 8)
+      num_to_check = 8 - start;
+
+   return ((int)(png_memcmp(&sig[start], &png_signature[start], num_to_check)));
+}
+
+/* (Obsolete) function to check signature bytes.  It does not allow one
+ * to check a partial signature.  This function might be removed in the
+ * future - use png_sig_cmp().  Returns true (nonzero) if the file is a PNG.
+ */
+int PNGAPI
+png_check_sig(png_bytep sig, int num)
+{
+  return ((int)!png_sig_cmp(sig, (png_size_t)0, (png_size_t)num));
+}
+
+/* Function to allocate memory for zlib and clear it to 0. */
+voidpf /* PRIVATE */
+png_zalloc(voidpf png_ptr, uInt items, uInt size)
+{
+   png_uint_32 num_bytes = (png_uint_32)items * size;
+   png_voidp ptr = (png_voidp)png_malloc((png_structp)png_ptr, num_bytes);
+
+#ifndef PNG_NO_ZALLOC_ZERO
+   if (num_bytes > (png_uint_32)0x8000L)
+   {
+      png_memset(ptr, 0, (png_size_t)0x8000L);
+      png_memset((png_bytep)ptr + (png_size_t)0x8000L, 0,
+         (png_size_t)(num_bytes - (png_uint_32)0x8000L));
+   }
+   else
+   {
+      png_memset(ptr, 0, (png_size_t)num_bytes);
+   }
+#endif
+   return ((voidpf)ptr);
+}
+
+/* function to free memory for zlib */
+void /* PRIVATE */
+png_zfree(voidpf png_ptr, voidpf ptr)
+{
+   png_free((png_structp)png_ptr, (png_voidp)ptr);
+}
+
+/* Reset the CRC variable to 32 bits of 1's.  Care must be taken
+ * in case CRC is > 32 bits to leave the top bits 0.
+ */
+void /* PRIVATE */
+png_reset_crc(png_structp png_ptr)
+{
+   png_ptr->crc = crc32(0, Z_NULL, 0);
+}
+
+/* Calculate the CRC over a section of data.  We can only pass as
+ * much data to this routine as the largest single buffer size.  We
+ * also check that this data will actually be used before going to the
+ * trouble of calculating it.
+ */
+void /* PRIVATE */
+png_calculate_crc(png_structp png_ptr, png_bytep ptr, png_size_t length)
+{
+   int need_crc = 1;
+
+   if (png_ptr->chunk_name[0] & 0x20)                     /* ancillary */
+   {
+      if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_MASK) ==
+          (PNG_FLAG_CRC_ANCILLARY_USE | PNG_FLAG_CRC_ANCILLARY_NOWARN))
+         need_crc = 0;
+   }
+   else                                                    /* critical */
+   {
+      if (png_ptr->flags & PNG_FLAG_CRC_CRITICAL_IGNORE)
+         need_crc = 0;
+   }
+
+   if (need_crc)
+      png_ptr->crc = crc32(png_ptr->crc, ptr, (uInt)length);
+}
+
+/* Allocate the memory for an info_struct for the application.  We don't
+ * really need the png_ptr, but it could potentially be useful in the
+ * future.  This should be used in favour of malloc(sizeof(png_info))
+ * and png_info_init() so that applications that want to use a shared
+ * libpng don't have to be recompiled if png_info changes size.
+ */
+png_infop PNGAPI
+png_create_info_struct(png_structp png_ptr)
+{
+   png_infop info_ptr;
+
+   png_debug(1, "in png_create_info_struct\n");
+   if(png_ptr == NULL) return (NULL);
+#ifdef PNG_USER_MEM_SUPPORTED
+   info_ptr = (png_infop)png_create_struct_2(PNG_STRUCT_INFO,
+      png_ptr->malloc_fn, png_ptr->mem_ptr);
+#else
+   info_ptr = (png_infop)png_create_struct(PNG_STRUCT_INFO);
+#endif
+   if (info_ptr != NULL)
+      png_info_init_3(&info_ptr, sizeof(png_info));
+
+   return (info_ptr);
+}
+
+/* This function frees the memory associated with a single info struct.
+ * Normally, one would use either png_destroy_read_struct() or
+ * png_destroy_write_struct() to free an info struct, but this may be
+ * useful for some applications.
+ */
+void PNGAPI
+png_destroy_info_struct(png_structp png_ptr, png_infopp info_ptr_ptr)
+{
+   png_infop info_ptr = NULL;
+
+   png_debug(1, "in png_destroy_info_struct\n");
+   if (info_ptr_ptr != NULL)
+      info_ptr = *info_ptr_ptr;
+
+   if (info_ptr != NULL)
+   {
+      png_info_destroy(png_ptr, info_ptr);
+
+#ifdef PNG_USER_MEM_SUPPORTED
+      png_destroy_struct_2((png_voidp)info_ptr, png_ptr->free_fn,
+          png_ptr->mem_ptr);
+#else
+      png_destroy_struct((png_voidp)info_ptr);
+#endif
+      *info_ptr_ptr = NULL;
+   }
+}
+
+/* Initialize the info structure.  This is now an internal function (0.89)
+ * and applications using it are urged to use png_create_info_struct()
+ * instead.
+ */
+#undef png_info_init
+void PNGAPI
+png_info_init(png_infop info_ptr)
+{
+   /* We only come here via pre-1.0.12-compiled applications */
+   png_info_init_3(&info_ptr, 0);
+}
+
+void PNGAPI
+png_info_init_3(png_infopp ptr_ptr, png_size_t png_info_struct_size)
+{
+   png_infop info_ptr = *ptr_ptr;
+
+   png_debug(1, "in png_info_init_3\n");
+
+   if(sizeof(png_info) > png_info_struct_size)
+     {
+       png_destroy_struct(info_ptr);
+       info_ptr = (png_infop)png_create_struct(PNG_STRUCT_INFO);
+       *ptr_ptr = info_ptr;
+     }
+
+   /* set everything to 0 */
+   png_memset(info_ptr, 0, sizeof (png_info));
+}
+
+#ifdef PNG_FREE_ME_SUPPORTED
+void PNGAPI
+png_data_freer(png_structp png_ptr, png_infop info_ptr,
+   int freer, png_uint_32 mask)
+{
+   png_debug(1, "in png_data_freer\n");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+   if(freer == PNG_DESTROY_WILL_FREE_DATA)
+      info_ptr->free_me |= mask;
+   else if(freer == PNG_USER_WILL_FREE_DATA)
+      info_ptr->free_me &= ~mask;
+   else
+      png_warning(png_ptr,
+         "Unknown freer parameter in png_data_freer.");
+}
+#endif
+
+void PNGAPI
+png_free_data(png_structp png_ptr, png_infop info_ptr, png_uint_32 mask,
+   int num)
+{
+   png_debug(1, "in png_free_data\n");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+#if defined(PNG_TEXT_SUPPORTED)
+/* free text item num or (if num == -1) all text items */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_TEXT) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_TEXT)
+#endif
+{
+   if (num != -1)
+   {
+     if (info_ptr->text && info_ptr->text[num].key)
+     {
+         png_free(png_ptr, info_ptr->text[num].key);
+         info_ptr->text[num].key = NULL;
+     }
+   }
+   else
+   {
+       int i;
+       for (i = 0; i < info_ptr->num_text; i++)
+           png_free_data(png_ptr, info_ptr, PNG_FREE_TEXT, i);
+       png_free(png_ptr, info_ptr->text);
+       info_ptr->text = NULL;
+       info_ptr->num_text=0;
+   }
+}
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED)
+/* free any tRNS entry */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_TRNS) & info_ptr->free_me)
+#else
+if ((mask & PNG_FREE_TRNS) && (png_ptr->flags & PNG_FLAG_FREE_TRNS))
+#endif
+{
+    png_free(png_ptr, info_ptr->trans);
+    info_ptr->valid &= ~PNG_INFO_tRNS;
+    info_ptr->trans = NULL;
+}
+#endif
+
+#if defined(PNG_sCAL_SUPPORTED)
+/* free any sCAL entry */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_SCAL) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_SCAL)
+#endif
+{
+#if defined(PNG_FIXED_POINT_SUPPORTED) && !defined(PNG_FLOATING_POINT_SUPPORTED)
+    png_free(png_ptr, info_ptr->scal_s_width);
+    png_free(png_ptr, info_ptr->scal_s_height);
+    info_ptr->scal_s_width = NULL;
+    info_ptr->scal_s_height = NULL;
+#endif
+    info_ptr->valid &= ~PNG_INFO_sCAL;
+}
+#endif
+
+#if defined(PNG_pCAL_SUPPORTED)
+/* free any pCAL entry */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_PCAL) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_PCAL)
+#endif
+{
+    png_free(png_ptr, info_ptr->pcal_purpose);
+    png_free(png_ptr, info_ptr->pcal_units);
+    info_ptr->pcal_purpose = NULL;
+    info_ptr->pcal_units = NULL;
+    if (info_ptr->pcal_params != NULL)
+    {
+        int i;
+        for (i = 0; i < (int)info_ptr->pcal_nparams; i++)
+        {
+          png_free(png_ptr, info_ptr->pcal_params[i]);
+          info_ptr->pcal_params[i]=NULL;
+        }
+        png_free(png_ptr, info_ptr->pcal_params);
+        info_ptr->pcal_params = NULL;
+    }
+    info_ptr->valid &= ~PNG_INFO_pCAL;
+}
+#endif
+
+#if defined(PNG_iCCP_SUPPORTED)
+/* free any iCCP entry */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_ICCP) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_ICCP)
+#endif
+{
+    png_free(png_ptr, info_ptr->iccp_name);
+    png_free(png_ptr, info_ptr->iccp_profile);
+    info_ptr->iccp_name = NULL;
+    info_ptr->iccp_profile = NULL;
+    info_ptr->valid &= ~PNG_INFO_iCCP;
+}
+#endif
+
+#if defined(PNG_sPLT_SUPPORTED)
+/* free a given sPLT entry, or (if num == -1) all sPLT entries */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_SPLT) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_SPLT)
+#endif
+{
+   if (num != -1)
+   {
+      if(info_ptr->splt_palettes)
+      {
+          png_free(png_ptr, info_ptr->splt_palettes[num].name);
+          png_free(png_ptr, info_ptr->splt_palettes[num].entries);
+          info_ptr->splt_palettes[num].name = NULL;
+          info_ptr->splt_palettes[num].entries = NULL;
+      }
+   }
+   else
+   {
+       if(info_ptr->splt_palettes_num)
+       {
+         int i;
+         for (i = 0; i < (int)info_ptr->splt_palettes_num; i++)
+            png_free_data(png_ptr, info_ptr, PNG_FREE_SPLT, i);
+
+         png_free(png_ptr, info_ptr->splt_palettes);
+         info_ptr->splt_palettes = NULL;
+         info_ptr->splt_palettes_num = 0;
+       }
+       info_ptr->valid &= ~PNG_INFO_sPLT;
+   }
+}
+#endif
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_UNKN) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_UNKN)
+#endif
+{
+   if (num != -1)
+   {
+       if(info_ptr->unknown_chunks)
+       {
+          png_free(png_ptr, info_ptr->unknown_chunks[num].data);
+          info_ptr->unknown_chunks[num].data = NULL;
+       }
+   }
+   else
+   {
+       int i;
+
+       if(info_ptr->unknown_chunks_num)
+       {
+         for (i = 0; i < (int)info_ptr->unknown_chunks_num; i++)
+            png_free_data(png_ptr, info_ptr, PNG_FREE_UNKN, i);
+
+         png_free(png_ptr, info_ptr->unknown_chunks);
+         info_ptr->unknown_chunks = NULL;
+         info_ptr->unknown_chunks_num = 0;
+       }
+   }
+}
+#endif
+
+#if defined(PNG_hIST_SUPPORTED)
+/* free any hIST entry */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_HIST)  & info_ptr->free_me)
+#else
+if ((mask & PNG_FREE_HIST) && (png_ptr->flags & PNG_FLAG_FREE_HIST))
+#endif
+{
+    png_free(png_ptr, info_ptr->hist);
+    info_ptr->hist = NULL;
+    info_ptr->valid &= ~PNG_INFO_hIST;
+}
+#endif
+
+/* free any PLTE entry that was internally allocated */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_PLTE) & info_ptr->free_me)
+#else
+if ((mask & PNG_FREE_PLTE) && (png_ptr->flags & PNG_FLAG_FREE_PLTE))
+#endif
+{
+    png_zfree(png_ptr, info_ptr->palette);
+    info_ptr->palette = NULL;
+    info_ptr->valid &= ~PNG_INFO_PLTE;
+    info_ptr->num_palette = 0;
+}
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+/* free any image bits attached to the info structure */
+#ifdef PNG_FREE_ME_SUPPORTED
+if ((mask & PNG_FREE_ROWS) & info_ptr->free_me)
+#else
+if (mask & PNG_FREE_ROWS)
+#endif
+{
+    if(info_ptr->row_pointers)
+    {
+       int row;
+       for (row = 0; row < (int)info_ptr->height; row++)
+       {
+          png_free(png_ptr, info_ptr->row_pointers[row]);
+          info_ptr->row_pointers[row]=NULL;
+       }
+       png_free(png_ptr, info_ptr->row_pointers);
+       info_ptr->row_pointers=NULL;
+    }
+    info_ptr->valid &= ~PNG_INFO_IDAT;
+}
+#endif
+
+#ifdef PNG_FREE_ME_SUPPORTED
+   if(num == -1)
+     info_ptr->free_me &= ~mask;
+   else
+     info_ptr->free_me &= ~(mask & ~PNG_FREE_MUL);
+#endif
+}
+
+/* This is an internal routine to free any memory that the info struct is
+ * pointing to before re-using it or freeing the struct itself.  Recall
+ * that png_free() checks for NULL pointers for us.
+ */
+void /* PRIVATE */
+png_info_destroy(png_structp png_ptr, png_infop info_ptr)
+{
+   png_debug(1, "in png_info_destroy\n");
+
+   png_free_data(png_ptr, info_ptr, PNG_FREE_ALL, -1);
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+   if (png_ptr->num_chunk_list)
+   {
+       png_free(png_ptr, png_ptr->chunk_list);
+       png_ptr->chunk_list=NULL;
+       png_ptr->num_chunk_list=0;
+   }
+#endif
+
+   png_info_init_3(&info_ptr, sizeof(png_info));
+}
+
+/* This function returns a pointer to the io_ptr associated with the user
+ * functions.  The application should free any memory associated with this
+ * pointer before png_write_destroy() or png_read_destroy() are called.
+ */
+png_voidp PNGAPI
+png_get_io_ptr(png_structp png_ptr)
+{
+   return (png_ptr->io_ptr);
+}
+
+#if !defined(PNG_NO_STDIO)
+/* Initialize the default input/output functions for the PNG file.  If you
+ * use your own read or write routines, you can call either png_set_read_fn()
+ * or png_set_write_fn() instead of png_init_io().  If you have defined
+ * PNG_NO_STDIO, you must use a function of your own because "FILE *" isn't
+ * necessarily available.
+ */
+void PNGAPI
+png_init_io(png_structp png_ptr, png_FILE_p fp)
+{
+   png_debug(1, "in png_init_io\n");
+   png_ptr->io_ptr = (png_voidp)fp;
+}
+#endif
+
+#if defined(PNG_TIME_RFC1123_SUPPORTED)
+/* Convert the supplied time into an RFC 1123 string suitable for use in
+ * a "Creation Time" or other text-based time string.
+ */
+png_charp PNGAPI
+png_convert_to_rfc1123(png_structp png_ptr, png_timep ptime)
+{
+   static PNG_CONST char short_months[12][4] =
+        {"Jan", "Feb", "Mar", "Apr", "May", "Jun",
+         "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"};
+
+   if (png_ptr->time_buffer == NULL)
+   {
+      png_ptr->time_buffer = (png_charp)png_malloc(png_ptr, (png_uint_32)(29*
+         sizeof(char)));
+   }
+
+#if defined(_WIN32_WCE)
+   {
+      wchar_t time_buf[29];
+      wsprintf(time_buf, TEXT("%d %S %d %02d:%02d:%02d +0000"),
+          ptime->day % 32, short_months[(ptime->month - 1) % 12],
+        ptime->year, ptime->hour % 24, ptime->minute % 60,
+          ptime->second % 61);
+      WideCharToMultiByte(CP_ACP, 0, time_buf, -1, png_ptr->time_buffer, 29,
+          NULL, NULL);
+   }
+#else
+#ifdef USE_FAR_KEYWORD
+   {
+      char near_time_buf[29];
+      sprintf(near_time_buf, "%d %s %d %02d:%02d:%02d +0000",
+          ptime->day % 32, short_months[(ptime->month - 1) % 12],
+          ptime->year, ptime->hour % 24, ptime->minute % 60,
+          ptime->second % 61);
+      png_memcpy(png_ptr->time_buffer, near_time_buf,
+          29*sizeof(char));
+   }
+#else
+   sprintf(png_ptr->time_buffer, "%d %s %d %02d:%02d:%02d +0000",
+       ptime->day % 32, short_months[(ptime->month - 1) % 12],
+       ptime->year, ptime->hour % 24, ptime->minute % 60,
+       ptime->second % 61);
+#endif
+#endif /* _WIN32_WCE */
+   return ((png_charp)png_ptr->time_buffer);
+}
+#endif /* PNG_TIME_RFC1123_SUPPORTED */
+
+#if 0
+/* Signature string for a PNG file. */
+png_bytep PNGAPI
+png_sig_bytes(void)
+{
+   return ((png_bytep)"\211\120\116\107\015\012\032\012");
+}
+#endif
+
+png_charp PNGAPI
+png_get_copyright(png_structp png_ptr)
+{
+   if (png_ptr != NULL || png_ptr == NULL)  /* silence compiler warning */
+   return ((png_charp) "\n libpng version 1.2.1 - December 12, 2001\n\
+   Copyright (c) 1998-2001 Glenn Randers-Pehrson\n\
+   Copyright (c) 1996, 1997 Andreas Dilger\n\
+   Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.\n");
+   return ((png_charp) "");
+}
+
+/* The following return the library version as a short string in the
+ * format 1.0.0 through 99.99.99zz.  To get the version of *.h files used
+ * with your application, print out PNG_LIBPNG_VER_STRING, which is defined
+ * in png.h.
+ */
+
+png_charp PNGAPI
+png_get_libpng_ver(png_structp png_ptr)
+{
+   /* Version of *.c files used when building libpng */
+   if(png_ptr != NULL) /* silence compiler warning about unused png_ptr */
+      return((png_charp) "1.2.1");
+   return((png_charp) "1.2.1");
+}
+
+png_charp PNGAPI
+png_get_header_ver(png_structp png_ptr)
+{
+   /* Version of *.h files used when building libpng */
+   if(png_ptr != NULL) /* silence compiler warning about unused png_ptr */
+      return((png_charp) PNG_LIBPNG_VER_STRING);
+   return((png_charp) PNG_LIBPNG_VER_STRING);
+}
+
+png_charp PNGAPI
+png_get_header_version(png_structp png_ptr)
+{
+   /* Returns longer string containing both version and date */
+   if(png_ptr != NULL) /* silence compiler warning about unused png_ptr */
+      return((png_charp) PNG_HEADER_VERSION_STRING);
+   return((png_charp) PNG_HEADER_VERSION_STRING);
+}
+
+#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+int /* PRIVATE */
+png_handle_as_unknown(png_structp png_ptr, png_bytep chunk_name)
+{
+   /* check chunk_name and return "keep" value if it's on the list, else 0 */
+   int i;
+   png_bytep p;
+   if((png_ptr == NULL && chunk_name == NULL) || png_ptr->num_chunk_list<=0)
+      return 0;
+   p=png_ptr->chunk_list+png_ptr->num_chunk_list*5-5;
+   for (i = png_ptr->num_chunk_list; i; i--, p-=5)
+      if (!png_memcmp(chunk_name, p, 4))
+        return ((int)*(p+4));
+   return 0;
+}
+#endif
+
+/* This function, added to libpng-1.0.6g, is untested. */
+int PNGAPI
+png_reset_zstream(png_structp png_ptr)
+{
+   return (inflateReset(&png_ptr->zstream));
+}
+
+/* This function was added to libpng-1.0.7 */
+png_uint_32 PNGAPI
+png_access_version_number(void)
+{
+   /* Version of *.c files used when building libpng */
+   return((png_uint_32) 10201L);
+}
+
+
+#if defined(PNG_ASSEMBLER_CODE_SUPPORTED)
+    /* GRR:  could add this:   && defined(PNG_MMX_CODE_SUPPORTED) */
+/* this INTERNAL function was added to libpng 1.2.0 */
+void /* PRIVATE */
+png_init_mmx_flags (png_structp png_ptr)
+{
+    png_ptr->mmx_rowbytes_threshold = 0;
+    png_ptr->mmx_bitdepth_threshold = 0;
+
+#  if (defined(PNG_USE_PNGVCRD) || defined(PNG_USE_PNGGCCRD))
+
+    png_ptr->asm_flags |= PNG_ASM_FLAG_MMX_SUPPORT_COMPILED;
+
+    if (png_mmx_support()) {
+        png_ptr->asm_flags |= PNG_ASM_FLAG_MMX_SUPPORT_IN_CPU
+#    ifdef PNG_HAVE_ASSEMBLER_COMBINE_ROW
+                              | PNG_ASM_FLAG_MMX_READ_COMBINE_ROW
+#    endif
+#    ifdef PNG_HAVE_ASSEMBLER_READ_INTERLACE
+                              | PNG_ASM_FLAG_MMX_READ_INTERLACE
+#    endif
+#    ifndef PNG_HAVE_ASSEMBLER_READ_FILTER_ROW
+                              ;
+#    else
+                              | PNG_ASM_FLAG_MMX_READ_FILTER_SUB
+                              | PNG_ASM_FLAG_MMX_READ_FILTER_UP
+                              | PNG_ASM_FLAG_MMX_READ_FILTER_AVG
+                              | PNG_ASM_FLAG_MMX_READ_FILTER_PAETH ;
+
+        png_ptr->mmx_rowbytes_threshold = PNG_MMX_ROWBYTES_THRESHOLD_DEFAULT;
+        png_ptr->mmx_bitdepth_threshold = PNG_MMX_BITDEPTH_THRESHOLD_DEFAULT;
+#    endif
+    } else {
+        png_ptr->asm_flags &= ~( PNG_ASM_FLAG_MMX_SUPPORT_IN_CPU
+                               | PNG_MMX_READ_FLAGS
+                               | PNG_MMX_WRITE_FLAGS );
+    }
+
+#  else /* !((PNGVCRD || PNGGCCRD) && PNG_ASSEMBLER_CODE_SUPPORTED)) */
+
+    /* clear all MMX flags; no support is compiled in */
+    png_ptr->asm_flags &= ~( PNG_MMX_FLAGS );
+
+#  endif /* ?(PNGVCRD || PNGGCCRD) */
+}
+
+#endif /* !(PNG_ASSEMBLER_CODE_SUPPORTED) */
+
+/* this function was added to libpng 1.2.0 */
+#if !defined(PNG_USE_PNGGCCRD) && \
+    !(defined(PNG_ASSEMBLER_CODE_SUPPORTED) && defined(PNG_USE_PNGVCRD))
+int PNGAPI
+png_mmx_support(void)
+{
+    return -1;
+}
+#endif
diff --git a/png/png.h b/png/png.h
new file mode 100644
index 000000000..77fc42078
--- /dev/null
+++ b/png/png.h
@@ -0,0 +1,3200 @@
+
+/* png.h - header file for PNG reference library
+ *
+ * libpng version 1.2.1 - December 12, 2001
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * Authors and maintainers:
+ *  libpng versions 0.71, May 1995, through 0.88, January 1996: Guy Schalnat
+ *  libpng versions 0.89c, June 1996, through 0.96, May 1997: Andreas Dilger
+ *  libpng versions 0.97, January 1998, through 1.2.1 - December 12, 2001: Glenn
+ *  See also "Contributing Authors", below.
+ *
+ * Note about libpng version numbers:
+ *
+ *    Due to various miscommunications, unforeseen code incompatibilities
+ *    and occasional factors outside the authors' control, version numbering
+ *    on the library has not always been consistent and straightforward.
+ *    The following table summarizes matters since version 0.89c, which was
+ *    the first widely used release:
+ *
+ *    source                 png.h  png.h  shared-lib
+ *    version                string   int  version
+ *    -------                ------ -----  ----------
+ *    0.89c "1.0 beta 3"     0.89      89  1.0.89
+ *    0.90  "1.0 beta 4"     0.90      90  0.90  [should have been 2.0.90]
+ *    0.95  "1.0 beta 5"     0.95      95  0.95  [should have been 2.0.95]
+ *    0.96  "1.0 beta 6"     0.96      96  0.96  [should have been 2.0.96]
+ *    0.97b "1.00.97 beta 7" 1.00.97   97  1.0.1 [should have been 2.0.97]
+ *    0.97c                  0.97      97  2.0.97
+ *    0.98                   0.98      98  2.0.98
+ *    0.99                   0.99      98  2.0.99
+ *    0.99a-m                0.99      99  2.0.99
+ *    1.00                   1.00     100  2.1.0 [100 should be 10000]
+ *    1.0.0      (from here on, the   100  2.1.0 [100 should be 10000]
+ *    1.0.1       png.h string is   10001  2.1.0
+ *    1.0.1a-e    identical to the  10002  from here on, the shared library
+ *    1.0.2       source version)   10002  is 2.V where V is the source code
+ *    1.0.2a-b                      10003  version, except as noted.
+ *    1.0.3                         10003
+ *    1.0.3a-d                      10004
+ *    1.0.4                         10004
+ *    1.0.4a-f                      10005
+ *    1.0.5 (+ 2 patches)           10005
+ *    1.0.5a-d                      10006
+ *    1.0.5e-r                      10100 (not source compatible)
+ *    1.0.5s-v                      10006 (not binary compatible)
+ *    1.0.6 (+ 3 patches)           10006 (still binary incompatible)
+ *    1.0.6d-f                      10007 (still binary incompatible)
+ *    1.0.6g                        10007
+ *    1.0.6h                        10007  10.6h (testing xy.z so-numbering)
+ *    1.0.6i                        10007  10.6i
+ *    1.0.6j                        10007  2.1.0.6j (incompatible with 1.0.0)
+ *    1.0.7beta11-14        DLLNUM  10007  2.1.0.7beta11-14 (binary compatible)
+ *    1.0.7beta15-18           1    10007  2.1.0.7beta15-18 (binary compatible)
+ *    1.0.7rc1-2               1    10007  2.1.0.7rc1-2 (binary compatible)
+ *    1.0.7                    1    10007  (still compatible)
+ *    1.0.8beta1-4             1    10008  2.1.0.8beta1-4
+ *    1.0.8rc1                 1    10008  2.1.0.8rc1
+ *    1.0.8                    1    10008  2.1.0.8
+ *    1.0.9beta1-6             1    10009  2.1.0.9beta1-6
+ *    1.0.9rc1                 1    10009  2.1.0.9rc1
+ *    1.0.9beta7-10            1    10009  2.1.0.9beta7-10
+ *    1.0.9rc2                 1    10009  2.1.0.9rc2
+ *    1.0.9                    1    10009  2.1.0.9
+ *    1.0.10beta1              1    10010  2.1.0.10beta1
+ *    1.0.10rc1                1    10010  2.1.0.10rc1
+ *    1.0.10                   1    10010  2.1.0.10
+ *    1.0.11beta1-3            1    10011  2.1.0.11beta1-3
+ *    1.0.11rc1                1    10011  2.1.0.11rc1
+ *    1.0.11                   1    10011  2.1.0.11
+ *    1.0.12beta1-2            2    10012  2.1.0.12beta1-2
+ *    1.0.12rc1                2    10012  2.1.0.12rc1
+ *    1.0.12                   2    10012  2.1.0.12
+ *    1.1.0a-f                 -    10100  2.1.1.0a-f (branch abandoned)
+ *    1.2.0beta1-2             2    10200  2.1.2.0beta1-2
+ *    1.2.0beta3-5             3    10200  3.1.2.0beta3-5
+ *    1.2.0rc1                 3    10200  3.1.2.0rc1
+ *    1.2.0                    3    10200  3.1.2.0
+ *    1.2.1beta1-4             3    10201  3.1.2.1beta1-4
+ *    1.2.1rc1-2               3    10201  3.1.2.1rc1-2
+ *    1.2.1                    3    10201  3.1.2.1
+ *
+ *    Henceforth the source version will match the shared-library major
+ *    and minor numbers; the shared-library major version number will be
+ *    used for changes in backward compatibility, as it is intended.  The
+ *    PNG_LIBPNG_VER macro, which is not used within libpng but is available
+ *    for applications, is an unsigned integer of the form xyyzz corresponding
+ *    to the source version x.y.z (leading zeros in y and z).  Beta versions
+ *    were given the previous public release number plus a letter, until
+ *    version 1.0.6j; from then on they were given the upcoming public
+ *    release number plus "betaNN" or "rcN".
+ *
+ *    Binary incompatibility exists only when applications make direct access
+ *    to the info_ptr or png_ptr members through png.h, and the compiled
+ *    application is loaded with a different version of the library.
+ *
+ *    DLLNUM will change each time there are forward or backward changes
+ *    in binary compatibility (e.g., when a new feature is added).
+ *
+ * See libpng.txt or libpng.3 for more information.  The PNG specification
+ * is available as RFC 2083 <ftp://ftp.uu.net/graphics/png/documents/>
+ * and as a W3C Recommendation <http://www.w3.org/TR/REC.png.html>
+ */
+
+/*
+ * COPYRIGHT NOTICE, DISCLAIMER, and LICENSE:
+ *
+ * If you modify libpng you may insert additional notices immediately following
+ * this sentence.
+ *
+ * libpng versions 1.0.7, July 1, 2000, through  1.2.1, December 12, 2001, are
+ * Copyright (c) 2000, 2001 Glenn Randers-Pehrson, and are
+ * distributed according to the same disclaimer and license as libpng-1.0.6
+ * with the following individuals added to the list of Contributing Authors
+ *
+ *    Simon-Pierre Cadieux
+ *    Eric S. Raymond
+ *    Gilles Vollant
+ *
+ * and with the following additions to the disclaimer:
+ *
+ *    There is no warranty against interference with your enjoyment of the
+ *    library or against infringement.  There is no warranty that our
+ *    efforts or the library will fulfill any of your particular purposes
+ *    or needs.  This library is provided with all faults, and the entire
+ *    risk of satisfactory quality, performance, accuracy, and effort is with
+ *    the user.
+ *
+ * libpng versions 0.97, January 1998, through 1.0.6, March 20, 2000, are
+ * Copyright (c) 1998, 1999, 2000 Glenn Randers-Pehrson
+ * Distributed according to the same disclaimer and license as libpng-0.96,
+ * with the following individuals added to the list of Contributing Authors:
+ *
+ *    Tom Lane
+ *    Glenn Randers-Pehrson
+ *    Willem van Schaik
+ *
+ * libpng versions 0.89, June 1996, through 0.96, May 1997, are
+ * Copyright (c) 1996, 1997 Andreas Dilger
+ * Distributed according to the same disclaimer and license as libpng-0.88,
+ * with the following individuals added to the list of Contributing Authors:
+ *
+ *    John Bowler
+ *    Kevin Bracey
+ *    Sam Bushell
+ *    Magnus Holmgren
+ *    Greg Roelofs
+ *    Tom Tanner
+ *
+ * libpng versions 0.5, May 1995, through 0.88, January 1996, are
+ * Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.
+ *
+ * For the purposes of this copyright and license, "Contributing Authors"
+ * is defined as the following set of individuals:
+ *
+ *    Andreas Dilger
+ *    Dave Martindale
+ *    Guy Eric Schalnat
+ *    Paul Schmidt
+ *    Tim Wegner
+ *
+ * The PNG Reference Library is supplied "AS IS".  The Contributing Authors
+ * and Group 42, Inc. disclaim all warranties, expressed or implied,
+ * including, without limitation, the warranties of merchantability and of
+ * fitness for any purpose.  The Contributing Authors and Group 42, Inc.
+ * assume no liability for direct, indirect, incidental, special, exemplary,
+ * or consequential damages, which may result from the use of the PNG
+ * Reference Library, even if advised of the possibility of such damage.
+ *
+ * Permission is hereby granted to use, copy, modify, and distribute this
+ * source code, or portions hereof, for any purpose, without fee, subject
+ * to the following restrictions:
+ *
+ * 1. The origin of this source code must not be misrepresented.
+ *
+ * 2. Altered versions must be plainly marked as such and
+ * must not be misrepresented as being the original source.
+ *
+ * 3. This Copyright notice may not be removed or altered from
+ *    any source or altered source distribution.
+ *
+ * The Contributing Authors and Group 42, Inc. specifically permit, without
+ * fee, and encourage the use of this source code as a component to
+ * supporting the PNG file format in commercial products.  If you use this
+ * source code in a product, acknowledgment is not required but would be
+ * appreciated.
+ */
+
+/*
+ * A "png_get_copyright" function is available, for convenient use in "about"
+ * boxes and the like:
+ *
+ * printf("%s",png_get_copyright(NULL));
+ *
+ * Also, the PNG logo (in PNG format, of course) is supplied in the
+ * files "pngbar.png" and "pngbar.jpg (88x31) and "pngnow.png" (98x31).
+ */
+
+/*
+ * Libpng is OSI Certified Open Source Software.  OSI Certified is a
+ * certification mark of the Open Source Initiative.
+ */
+
+/*
+ * The contributing authors would like to thank all those who helped
+ * with testing, bug fixes, and patience.  This wouldn't have been
+ * possible without all of you.
+ *
+ * Thanks to Frank J. T. Wojcik for helping with the documentation.
+ */
+
+/*
+ * Y2K compliance in libpng:
+ * =========================
+ *
+ *    December 12, 2001
+ *
+ *    Since the PNG Development group is an ad-hoc body, we can't make
+ *    an official declaration.
+ *
+ *    This is your unofficial assurance that libpng from version 0.71 and
+ *    upward through 1.2.1 are Y2K compliant.  It is my belief that earlier
+ *    versions were also Y2K compliant.
+ *
+ *    Libpng only has three year fields.  One is a 2-byte unsigned integer
+ *    that will hold years up to 65535.  The other two hold the date in text
+ *    format, and will hold years up to 9999.
+ *
+ *    The integer is
+ *        "png_uint_16 year" in png_time_struct.
+ *
+ *    The strings are
+ *        "png_charp time_buffer" in png_struct and
+ *        "near_time_buffer", which is a local character string in png.c.
+ *
+ *    There are seven time-related functions:
+ *        png.c: png_convert_to_rfc_1123() in png.c
+ *          (formerly png_convert_to_rfc_1152() in error)
+ *        png_convert_from_struct_tm() in pngwrite.c, called in pngwrite.c
+ *        png_convert_from_time_t() in pngwrite.c
+ *        png_get_tIME() in pngget.c
+ *        png_handle_tIME() in pngrutil.c, called in pngread.c
+ *        png_set_tIME() in pngset.c
+ *        png_write_tIME() in pngwutil.c, called in pngwrite.c
+ *
+ *    All handle dates properly in a Y2K environment.  The
+ *    png_convert_from_time_t() function calls gmtime() to convert from system
+ *    clock time, which returns (year - 1900), which we properly convert to
+ *    the full 4-digit year.  There is a possibility that applications using
+ *    libpng are not passing 4-digit years into the png_convert_to_rfc_1123()
+ *    function, or that they are incorrectly passing only a 2-digit year
+ *    instead of "year - 1900" into the png_convert_from_struct_tm() function,
+ *    but this is not under our control.  The libpng documentation has always
+ *    stated that it works with 4-digit years, and the APIs have been
+ *    documented as such.
+ *
+ *    The tIME chunk itself is also Y2K compliant.  It uses a 2-byte unsigned
+ *    integer to hold the year, and can hold years as large as 65535.
+ *
+ *    zlib, upon which libpng depends, is also Y2K compliant.  It contains
+ *    no date-related code.
+ *
+ *       Glenn Randers-Pehrson
+ *       libpng maintainer
+ *       PNG Development Group
+ */
+
+#ifndef PNG_H
+#define PNG_H
+
+/* This is not the place to learn how to use libpng.  The file libpng.txt
+ * describes how to use libpng, and the file example.c summarizes it
+ * with some code on which to build.  This file is useful for looking
+ * at the actual function definitions and structure components.
+ */
+
+/* Version information for png.h - this should match the version in png.c */
+#define PNG_LIBPNG_VER_STRING "1.2.1"
+
+#define PNG_LIBPNG_VER_SONUM   3
+#define PNG_LIBPNG_VER_DLLNUM  %DLLNUM%
+
+/* These should match the first 3 components of PNG_LIBPNG_VER_STRING: */
+#define PNG_LIBPNG_VER_MAJOR   1
+#define PNG_LIBPNG_VER_MINOR   2
+#define PNG_LIBPNG_VER_RELEASE 1
+/* This should match the numeric part of the final component of
+ * PNG_LIBPNG_VER_STRING, omitting any leading zero: */
+
+#define PNG_LIBPNG_VER_BUILD  0
+
+#define PNG_LIBPNG_BUILD_ALPHA    1
+#define PNG_LIBPNG_BUILD_BETA     2
+#define PNG_LIBPNG_BUILD_RC       3
+#define PNG_LIBPNG_BUILD_STABLE   4
+#define PNG_LIBPNG_BUILD_TYPEMASK 7
+#define PNG_LIBPNG_BUILD_PATCH    8 /* Can be OR'ed with STABLE only */
+#define PNG_LIBPNG_BUILD_TYPE 4
+
+/* Careful here.  At one time, Guy wanted to use 082, but that would be octal.
+ * We must not include leading zeros.
+ * Versions 0.7 through 1.0.0 were in the range 0 to 100 here (only
+ * version 1.0.0 was mis-numbered 100 instead of 10000).  From
+ * version 1.0.1 it's    xxyyzz, where x=major, y=minor, z=release */
+#define PNG_LIBPNG_VER 10201 /* 1.2.1 */
+
+#ifndef PNG_VERSION_INFO_ONLY
+
+/* include the compression library's header */
+#include "zlib.h"
+
+/* include all user configurable info, including optional assembler routines */
+#include "pngconf.h"
+
+/* Inhibit C++ name-mangling for libpng functions but not for system calls. */
+#ifdef __cplusplus
+extern "C" {
+#endif /* __cplusplus */
+
+/* This file is arranged in several sections.  The first section contains
+ * structure and type definitions.  The second section contains the external
+ * library functions, while the third has the internal library functions,
+ * which applications aren't expected to use directly.
+ */
+
+#ifndef PNG_NO_TYPECAST_NULL
+#define int_p_NULL                (int *)NULL
+#define png_bytep_NULL            (png_bytep)NULL
+#define png_bytepp_NULL           (png_bytepp)NULL
+#define png_doublep_NULL          (png_doublep)NULL
+#define png_error_ptr_NULL        (png_error_ptr)NULL
+#define png_flush_ptr_NULL        (png_flush_ptr)NULL
+#define png_free_ptr_NULL         (png_free_ptr)NULL
+#define png_infopp_NULL           (png_infopp)NULL
+#define png_malloc_ptr_NULL       (png_malloc_ptr)NULL
+#define png_read_status_ptr_NULL  (png_read_status_ptr)NULL
+#define png_rw_ptr_NULL           (png_rw_ptr)NULL
+#define png_structp_NULL          (png_structp)NULL
+#define png_uint_16p_NULL         (png_uint_16p)NULL
+#define png_voidp_NULL            (png_voidp)NULL
+#define png_write_status_ptr_NULL (png_write_status_ptr)NULL
+#else
+#define int_p_NULL                NULL
+#define png_bytep_NULL            NULL
+#define png_bytepp_NULL           NULL
+#define png_doublep_NULL          NULL
+#define png_error_ptr_NULL        NULL
+#define png_flush_ptr_NULL        NULL
+#define png_free_ptr_NULL         NULL
+#define png_infopp_NULL           NULL
+#define png_malloc_ptr_NULL       NULL
+#define png_read_status_ptr_NULL  NULL
+#define png_rw_ptr_NULL           NULL
+#define png_structp_NULL          NULL
+#define png_uint_16p_NULL         NULL
+#define png_voidp_NULL            NULL
+#define png_write_status_ptr_NULL NULL
+#endif
+
+/* variables declared in png.c - only it needs to define PNG_NO_EXTERN */
+#if !defined(PNG_NO_EXTERN) || defined(PNG_ALWAYS_EXTERN)
+/* Version information for C files, stored in png.c.  This had better match
+ * the version above.
+ */
+#ifdef PNG_USE_GLOBAL_ARRAYS
+PNG_EXPORT_VAR (const char) png_libpng_ver[18];
+  /* need room for 99.99.99beta99z*/
+#else
+#define png_libpng_ver png_get_header_ver(NULL)
+#endif
+
+#ifdef PNG_USE_GLOBAL_ARRAYS
+/* This was removed in version 1.0.5c */
+/* Structures to facilitate easy interlacing.  See png.c for more details */
+PNG_EXPORT_VAR (const int FARDATA) png_pass_start[7];
+PNG_EXPORT_VAR (const int FARDATA) png_pass_inc[7];
+PNG_EXPORT_VAR (const int FARDATA) png_pass_ystart[7];
+PNG_EXPORT_VAR (const int FARDATA) png_pass_yinc[7];
+PNG_EXPORT_VAR (const int FARDATA) png_pass_mask[7];
+PNG_EXPORT_VAR (const int FARDATA) png_pass_dsp_mask[7];
+#ifdef PNG_HAVE_ASSEMBLER_COMBINE_ROW
+PNG_EXPORT_VAR (const int FARDATA) png_pass_width[7];
+#endif
+/* This isn't currently used.  If you need it, see png.c for more details.
+PNG_EXPORT_VAR (const int FARDATA) png_pass_height[7];
+*/
+#endif
+
+#endif /* PNG_NO_EXTERN */
+
+/* Three color definitions.  The order of the red, green, and blue, (and the
+ * exact size) is not important, although the size of the fields need to
+ * be png_byte or png_uint_16 (as defined below).
+ */
+typedef struct png_color_struct
+{
+   png_byte red;
+   png_byte green;
+   png_byte blue;
+} png_color;
+typedef png_color FAR * png_colorp;
+typedef png_color FAR * FAR * png_colorpp;
+
+typedef struct png_color_16_struct
+{
+   png_byte index;    /* used for palette files */
+   png_uint_16 red;   /* for use in red green blue files */
+   png_uint_16 green;
+   png_uint_16 blue;
+   png_uint_16 gray;  /* for use in grayscale files */
+} png_color_16;
+typedef png_color_16 FAR * png_color_16p;
+typedef png_color_16 FAR * FAR * png_color_16pp;
+
+typedef struct png_color_8_struct
+{
+   png_byte red;   /* for use in red green blue files */
+   png_byte green;
+   png_byte blue;
+   png_byte gray;  /* for use in grayscale files */
+   png_byte alpha; /* for alpha channel files */
+} png_color_8;
+typedef png_color_8 FAR * png_color_8p;
+typedef png_color_8 FAR * FAR * png_color_8pp;
+
+/*
+ * The following two structures are used for the in-core representation
+ * of sPLT chunks.
+ */
+typedef struct png_sPLT_entry_struct
+{
+   png_uint_16 red;
+   png_uint_16 green;
+   png_uint_16 blue;
+   png_uint_16 alpha;
+   png_uint_16 frequency;
+} png_sPLT_entry;
+typedef png_sPLT_entry FAR * png_sPLT_entryp;
+typedef png_sPLT_entry FAR * FAR * png_sPLT_entrypp;
+
+/*  When the depth of the sPLT palette is 8 bits, the color and alpha samples
+ *  occupy the LSB of their respective members, and the MSB of each member
+ *  is zero-filled.  The frequency member always occupies the full 16 bits.
+ */
+
+typedef struct png_sPLT_struct
+{
+   png_charp name;           /* palette name */
+   png_byte depth;           /* depth of palette samples */
+   png_sPLT_entryp entries;  /* palette entries */
+   png_int_32 nentries;      /* number of palette entries */
+} png_sPLT_t;
+typedef png_sPLT_t FAR * png_sPLT_tp;
+typedef png_sPLT_t FAR * FAR * png_sPLT_tpp;
+
+#ifdef PNG_TEXT_SUPPORTED
+/* png_text holds the contents of a text/ztxt/itxt chunk in a PNG file,
+ * and whether that contents is compressed or not.  The "key" field
+ * points to a regular zero-terminated C string.  The "text", "lang", and
+ * "lang_key" fields can be regular C strings, empty strings, or NULL pointers.
+ * However, the * structure returned by png_get_text() will always contain
+ * regular zero-terminated C strings (possibly empty), never NULL pointers,
+ * so they can be safely used in printf() and other string-handling functions.
+ */
+typedef struct png_text_struct
+{
+   int  compression;       /* compression value:
+                             -1: tEXt, none
+                              0: zTXt, deflate
+                              1: iTXt, none
+                              2: iTXt, deflate  */
+   png_charp key;          /* keyword, 1-79 character description of "text" */
+   png_charp text;         /* comment, may be an empty string (ie "")
+                              or a NULL pointer */
+   png_size_t text_length; /* length of the text string */
+#ifdef PNG_iTXt_SUPPORTED
+   png_size_t itxt_length; /* length of the itxt string */
+   png_charp lang;         /* language code, 0-79 characters
+                              or a NULL pointer */
+   png_charp lang_key;     /* keyword translated UTF-8 string, 0 or more
+                              chars or a NULL pointer */
+#endif
+} png_text;
+typedef png_text FAR * png_textp;
+typedef png_text FAR * FAR * png_textpp;
+#endif
+
+/* Supported compression types for text in PNG files (tEXt, and zTXt).
+ * The values of the PNG_TEXT_COMPRESSION_ defines should NOT be changed. */
+#define PNG_TEXT_COMPRESSION_NONE_WR -3
+#define PNG_TEXT_COMPRESSION_zTXt_WR -2
+#define PNG_TEXT_COMPRESSION_NONE    -1
+#define PNG_TEXT_COMPRESSION_zTXt     0
+#define PNG_ITXT_COMPRESSION_NONE     1
+#define PNG_ITXT_COMPRESSION_zTXt     2
+#define PNG_TEXT_COMPRESSION_LAST     3  /* Not a valid value */
+
+/* png_time is a way to hold the time in an machine independent way.
+ * Two conversions are provided, both from time_t and struct tm.  There
+ * is no portable way to convert to either of these structures, as far
+ * as I know.  If you know of a portable way, send it to me.  As a side
+ * note - PNG has always been Year 2000 compliant!
+ */
+typedef struct png_time_struct
+{
+   png_uint_16 year; /* full year, as in, 1995 */
+   png_byte month;   /* month of year, 1 - 12 */
+   png_byte day;     /* day of month, 1 - 31 */
+   png_byte hour;    /* hour of day, 0 - 23 */
+   png_byte minute;  /* minute of hour, 0 - 59 */
+   png_byte second;  /* second of minute, 0 - 60 (for leap seconds) */
+} png_time;
+typedef png_time FAR * png_timep;
+typedef png_time FAR * FAR * png_timepp;
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+/* png_unknown_chunk is a structure to hold queued chunks for which there is
+ * no specific support.  The idea is that we can use this to queue
+ * up private chunks for output even though the library doesn't actually
+ * know about their semantics.
+ */
+typedef struct png_unknown_chunk_t
+{
+    png_byte name[5];
+    png_byte *data;
+    png_size_t size;
+
+    /* libpng-using applications should NOT directly modify this byte. */
+    png_byte location; /* mode of operation at read time */
+}
+png_unknown_chunk;
+typedef png_unknown_chunk FAR * png_unknown_chunkp;
+typedef png_unknown_chunk FAR * FAR * png_unknown_chunkpp;
+#endif
+
+/* png_info is a structure that holds the information in a PNG file so
+ * that the application can find out the characteristics of the image.
+ * If you are reading the file, this structure will tell you what is
+ * in the PNG file.  If you are writing the file, fill in the information
+ * you want to put into the PNG file, then call png_write_info().
+ * The names chosen should be very close to the PNG specification, so
+ * consult that document for information about the meaning of each field.
+ *
+ * With libpng < 0.95, it was only possible to directly set and read the
+ * the values in the png_info_struct, which meant that the contents and
+ * order of the values had to remain fixed.  With libpng 0.95 and later,
+ * however, there are now functions that abstract the contents of
+ * png_info_struct from the application, so this makes it easier to use
+ * libpng with dynamic libraries, and even makes it possible to use
+ * libraries that don't have all of the libpng ancillary chunk-handing
+ * functionality.
+ *
+ * In any case, the order of the parameters in png_info_struct should NOT
+ * be changed for as long as possible to keep compatibility with applications
+ * that use the old direct-access method with png_info_struct.
+ *
+ * The following members may have allocated storage attached that should be
+ * cleaned up before the structure is discarded: palette, trans, text,
+ * pcal_purpose, pcal_units, pcal_params, hist, iccp_name, iccp_profile,
+ * splt_palettes, scal_unit, row_pointers, and unknowns.   By default, these
+ * are automatically freed when the info structure is deallocated, if they were
+ * allocated internally by libpng.  This behavior can be changed by means
+ * of the png_data_freer() function.
+ *
+ * More allocation details: all the chunk-reading functions that
+ * change these members go through the corresponding png_set_*
+ * functions.  A function to clear these members is available: see
+ * png_free_data().  The png_set_* functions do not depend on being
+ * able to point info structure members to any of the storage they are
+ * passed (they make their own copies), EXCEPT that the png_set_text
+ * functions use the same storage passed to them in the text_ptr or
+ * itxt_ptr structure argument, and the png_set_rows and png_set_unknowns
+ * functions do not make their own copies.
+ */
+typedef struct png_info_struct
+{
+   /* the following are necessary for every PNG file */
+   png_uint_32 width;       /* width of image in pixels (from IHDR) */
+   png_uint_32 height;      /* height of image in pixels (from IHDR) */
+   png_uint_32 valid;       /* valid chunk data (see PNG_INFO_ below) */
+   png_uint_32 rowbytes;    /* bytes needed to hold an untransformed row */
+   png_colorp palette;      /* array of color values (valid & PNG_INFO_PLTE) */
+   png_uint_16 num_palette; /* number of color entries in "palette" (PLTE) */
+   png_uint_16 num_trans;   /* number of transparent palette color (tRNS) */
+   png_byte bit_depth;      /* 1, 2, 4, 8, or 16 bits/channel (from IHDR) */
+   png_byte color_type;     /* see PNG_COLOR_TYPE_ below (from IHDR) */
+   /* The following three should have been named *_method not *_type */
+   png_byte compression_type; /* must be PNG_COMPRESSION_TYPE_BASE (IHDR) */
+   png_byte filter_type;    /* must be PNG_FILTER_TYPE_BASE (from IHDR) */
+   png_byte interlace_type; /* One of PNG_INTERLACE_NONE, PNG_INTERLACE_ADAM7 */
+
+   /* The following is informational only on read, and not used on writes. */
+   png_byte channels;       /* number of data channels per pixel (1, 2, 3, 4)*/
+   png_byte pixel_depth;    /* number of bits per pixel */
+   png_byte spare_byte;     /* to align the data, and for future use */
+   png_byte signature[8];   /* magic bytes read by libpng from start of file */
+
+   /* The rest of the data is optional.  If you are reading, check the
+    * valid field to see if the information in these are valid.  If you
+    * are writing, set the valid field to those chunks you want written,
+    * and initialize the appropriate fields below.
+    */
+
+#if defined(PNG_gAMA_SUPPORTED) && defined(PNG_FLOATING_POINT_SUPPORTED)
+   /* The gAMA chunk describes the gamma characteristics of the system
+    * on which the image was created, normally in the range [1.0, 2.5].
+    * Data is valid if (valid & PNG_INFO_gAMA) is non-zero.
+    */
+   float gamma; /* gamma value of image, if (valid & PNG_INFO_gAMA) */
+#endif
+
+#if defined(PNG_sRGB_SUPPORTED)
+    /* GR-P, 0.96a */
+    /* Data valid if (valid & PNG_INFO_sRGB) non-zero. */
+   png_byte srgb_intent; /* sRGB rendering intent [0, 1, 2, or 3] */
+#endif
+
+#if defined(PNG_TEXT_SUPPORTED)
+   /* The tEXt, and zTXt chunks contain human-readable textual data in
+    * uncompressed, compressed, and optionally compressed forms, respectively.
+    * The data in "text" is an array of pointers to uncompressed,
+    * null-terminated C strings. Each chunk has a keyword that describes the
+    * textual data contained in that chunk.  Keywords are not required to be
+    * unique, and the text string may be empty.  Any number of text chunks may
+    * be in an image.
+    */
+   int num_text; /* number of comments read/to write */
+   int max_text; /* current size of text array */
+   png_textp text; /* array of comments read/to write */
+#endif /* PNG_TEXT_SUPPORTED */
+
+#if defined(PNG_tIME_SUPPORTED)
+   /* The tIME chunk holds the last time the displayed image data was
+    * modified.  See the png_time struct for the contents of this struct.
+    */
+   png_time mod_time;
+#endif
+
+#if defined(PNG_sBIT_SUPPORTED)
+   /* The sBIT chunk specifies the number of significant high-order bits
+    * in the pixel data.  Values are in the range [1, bit_depth], and are
+    * only specified for the channels in the pixel data.  The contents of
+    * the low-order bits is not specified.  Data is valid if
+    * (valid & PNG_INFO_sBIT) is non-zero.
+    */
+   png_color_8 sig_bit; /* significant bits in color channels */
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED) || defined(PNG_READ_EXPAND_SUPPORTED) || \
+defined(PNG_READ_BACKGROUND_SUPPORTED)
+   /* The tRNS chunk supplies transparency data for paletted images and
+    * other image types that don't need a full alpha channel.  There are
+    * "num_trans" transparency values for a paletted image, stored in the
+    * same order as the palette colors, starting from index 0.  Values
+    * for the data are in the range [0, 255], ranging from fully transparent
+    * to fully opaque, respectively.  For non-paletted images, there is a
+    * single color specified that should be treated as fully transparent.
+    * Data is valid if (valid & PNG_INFO_tRNS) is non-zero.
+    */
+   png_bytep trans; /* transparent values for paletted image */
+   png_color_16 trans_values; /* transparent color for non-palette image */
+#endif
+
+#if defined(PNG_bKGD_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+   /* The bKGD chunk gives the suggested image background color if the
+    * display program does not have its own background color and the image
+    * is needs to composited onto a background before display.  The colors
+    * in "background" are normally in the same color space/depth as the
+    * pixel data.  Data is valid if (valid & PNG_INFO_bKGD) is non-zero.
+    */
+   png_color_16 background;
+#endif
+
+#if defined(PNG_oFFs_SUPPORTED)
+   /* The oFFs chunk gives the offset in "offset_unit_type" units rightwards
+    * and downwards from the top-left corner of the display, page, or other
+    * application-specific co-ordinate space.  See the PNG_OFFSET_ defines
+    * below for the unit types.  Valid if (valid & PNG_INFO_oFFs) non-zero.
+    */
+   png_int_32 x_offset; /* x offset on page */
+   png_int_32 y_offset; /* y offset on page */
+   png_byte offset_unit_type; /* offset units type */
+#endif
+
+#if defined(PNG_pHYs_SUPPORTED)
+   /* The pHYs chunk gives the physical pixel density of the image for
+    * display or printing in "phys_unit_type" units (see PNG_RESOLUTION_
+    * defines below).  Data is valid if (valid & PNG_INFO_pHYs) is non-zero.
+    */
+   png_uint_32 x_pixels_per_unit; /* horizontal pixel density */
+   png_uint_32 y_pixels_per_unit; /* vertical pixel density */
+   png_byte phys_unit_type; /* resolution type (see PNG_RESOLUTION_ below) */
+#endif
+
+#if defined(PNG_hIST_SUPPORTED)
+   /* The hIST chunk contains the relative frequency or importance of the
+    * various palette entries, so that a viewer can intelligently select a
+    * reduced-color palette, if required.  Data is an array of "num_palette"
+    * values in the range [0,65535]. Data valid if (valid & PNG_INFO_hIST)
+    * is non-zero.
+    */
+   png_uint_16p hist;
+#endif
+
+#ifdef PNG_cHRM_SUPPORTED
+   /* The cHRM chunk describes the CIE color characteristics of the monitor
+    * on which the PNG was created.  This data allows the viewer to do gamut
+    * mapping of the input image to ensure that the viewer sees the same
+    * colors in the image as the creator.  Values are in the range
+    * [0.0, 0.8].  Data valid if (valid & PNG_INFO_cHRM) non-zero.
+    */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   float x_white;
+   float y_white;
+   float x_red;
+   float y_red;
+   float x_green;
+   float y_green;
+   float x_blue;
+   float y_blue;
+#endif
+#endif
+
+#if defined(PNG_pCAL_SUPPORTED)
+   /* The pCAL chunk describes a transformation between the stored pixel
+    * values and original physical data values used to create the image.
+    * The integer range [0, 2^bit_depth - 1] maps to the floating-point
+    * range given by [pcal_X0, pcal_X1], and are further transformed by a
+    * (possibly non-linear) transformation function given by "pcal_type"
+    * and "pcal_params" into "pcal_units".  Please see the PNG_EQUATION_
+    * defines below, and the PNG-Group's PNG extensions document for a
+    * complete description of the transformations and how they should be
+    * implemented, and for a description of the ASCII parameter strings.
+    * Data values are valid if (valid & PNG_INFO_pCAL) non-zero.
+    */
+   png_charp pcal_purpose;  /* pCAL chunk description string */
+   png_int_32 pcal_X0;      /* minimum value */
+   png_int_32 pcal_X1;      /* maximum value */
+   png_charp pcal_units;    /* Latin-1 string giving physical units */
+   png_charpp pcal_params;  /* ASCII strings containing parameter values */
+   png_byte pcal_type;      /* equation type (see PNG_EQUATION_ below) */
+   png_byte pcal_nparams;   /* number of parameters given in pcal_params */
+#endif
+
+/* New members added in libpng-1.0.6 */
+#ifdef PNG_FREE_ME_SUPPORTED
+   png_uint_32 free_me;     /* flags items libpng is responsible for freeing */
+#endif
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+   /* storage for unknown chunks that the library doesn't recognize. */
+   png_unknown_chunkp unknown_chunks;
+   png_size_t unknown_chunks_num;
+#endif
+
+#if defined(PNG_iCCP_SUPPORTED)
+   /* iCCP chunk data. */
+   png_charp iccp_name;     /* profile name */
+   png_charp iccp_profile;  /* International Color Consortium profile data */
+                            /* Note to maintainer: should be png_bytep */
+   png_uint_32 iccp_proflen;  /* ICC profile data length */
+   png_byte iccp_compression; /* Always zero */
+#endif
+
+#if defined(PNG_sPLT_SUPPORTED)
+   /* data on sPLT chunks (there may be more than one). */
+   png_sPLT_tp splt_palettes;
+   png_uint_32 splt_palettes_num;
+#endif
+
+#if defined(PNG_sCAL_SUPPORTED)
+   /* The sCAL chunk describes the actual physical dimensions of the
+    * subject matter of the graphic.  The chunk contains a unit specification
+    * a byte value, and two ASCII strings representing floating-point
+    * values.  The values are width and height corresponsing to one pixel
+    * in the image.  This external representation is converted to double
+    * here.  Data values are valid if (valid & PNG_INFO_sCAL) is non-zero.
+    */
+   png_byte scal_unit;         /* unit of physical scale */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   double scal_pixel_width;    /* width of one pixel */
+   double scal_pixel_height;   /* height of one pixel */
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_charp scal_s_width;     /* string containing height */
+   png_charp scal_s_height;    /* string containing width */
+#endif
+#endif
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+   /* Memory has been allocated if (valid & PNG_ALLOCATED_INFO_ROWS) non-zero */
+   /* Data valid if (valid & PNG_INFO_IDAT) non-zero */
+   png_bytepp row_pointers;        /* the image bits */
+#endif
+
+#if defined(PNG_FIXED_POINT_SUPPORTED) && defined(PNG_gAMA_SUPPORTED)
+   png_fixed_point int_gamma; /* gamma of image, if (valid & PNG_INFO_gAMA) */
+#endif
+
+#if defined(PNG_cHRM_SUPPORTED) && defined(PNG_FIXED_POINT_SUPPORTED)
+   png_fixed_point int_x_white;
+   png_fixed_point int_y_white;
+   png_fixed_point int_x_red;
+   png_fixed_point int_y_red;
+   png_fixed_point int_x_green;
+   png_fixed_point int_y_green;
+   png_fixed_point int_x_blue;
+   png_fixed_point int_y_blue;
+#endif
+
+} png_info;
+
+typedef png_info FAR * png_infop;
+typedef png_info FAR * FAR * png_infopp;
+
+/* Maximum positive integer used in PNG is (2^31)-1 */
+#define PNG_MAX_UINT ((png_uint_32)0x7fffffffL)
+
+/* These describe the color_type field in png_info. */
+/* color type masks */
+#define PNG_COLOR_MASK_PALETTE    1
+#define PNG_COLOR_MASK_COLOR      2
+#define PNG_COLOR_MASK_ALPHA      4
+
+/* color types.  Note that not all combinations are legal */
+#define PNG_COLOR_TYPE_GRAY 0
+#define PNG_COLOR_TYPE_PALETTE  (PNG_COLOR_MASK_COLOR | PNG_COLOR_MASK_PALETTE)
+#define PNG_COLOR_TYPE_RGB        (PNG_COLOR_MASK_COLOR)
+#define PNG_COLOR_TYPE_RGB_ALPHA  (PNG_COLOR_MASK_COLOR | PNG_COLOR_MASK_ALPHA)
+#define PNG_COLOR_TYPE_GRAY_ALPHA (PNG_COLOR_MASK_ALPHA)
+/* aliases */
+#define PNG_COLOR_TYPE_RGBA  PNG_COLOR_TYPE_RGB_ALPHA
+#define PNG_COLOR_TYPE_GA  PNG_COLOR_TYPE_GRAY_ALPHA
+
+/* This is for compression type. PNG 1.0-1.2 only define the single type. */
+#define PNG_COMPRESSION_TYPE_BASE 0 /* Deflate method 8, 32K window */
+#define PNG_COMPRESSION_TYPE_DEFAULT PNG_COMPRESSION_TYPE_BASE
+
+/* This is for filter type. PNG 1.0-1.2 only define the single type. */
+#define PNG_FILTER_TYPE_BASE      0 /* Single row per-byte filtering */
+#define PNG_INTRAPIXEL_DIFFERENCING 64 /* Used only in MNG datastreams */
+#define PNG_FILTER_TYPE_DEFAULT   PNG_FILTER_TYPE_BASE
+
+/* These are for the interlacing type.  These values should NOT be changed. */
+#define PNG_INTERLACE_NONE        0 /* Non-interlaced image */
+#define PNG_INTERLACE_ADAM7       1 /* Adam7 interlacing */
+#define PNG_INTERLACE_LAST        2 /* Not a valid value */
+
+/* These are for the oFFs chunk.  These values should NOT be changed. */
+#define PNG_OFFSET_PIXEL          0 /* Offset in pixels */
+#define PNG_OFFSET_MICROMETER     1 /* Offset in micrometers (1/10^6 meter) */
+#define PNG_OFFSET_LAST           2 /* Not a valid value */
+
+/* These are for the pCAL chunk.  These values should NOT be changed. */
+#define PNG_EQUATION_LINEAR       0 /* Linear transformation */
+#define PNG_EQUATION_BASE_E       1 /* Exponential base e transform */
+#define PNG_EQUATION_ARBITRARY    2 /* Arbitrary base exponential transform */
+#define PNG_EQUATION_HYPERBOLIC   3 /* Hyperbolic sine transformation */
+#define PNG_EQUATION_LAST         4 /* Not a valid value */
+
+/* These are for the sCAL chunk.  These values should NOT be changed. */
+#define PNG_SCALE_UNKNOWN         0 /* unknown unit (image scale) */
+#define PNG_SCALE_METER           1 /* meters per pixel */
+#define PNG_SCALE_RADIAN          2 /* radians per pixel */
+#define PNG_SCALE_LAST            3 /* Not a valid value */
+
+/* These are for the pHYs chunk.  These values should NOT be changed. */
+#define PNG_RESOLUTION_UNKNOWN    0 /* pixels/unknown unit (aspect ratio) */
+#define PNG_RESOLUTION_METER      1 /* pixels/meter */
+#define PNG_RESOLUTION_LAST       2 /* Not a valid value */
+
+/* These are for the sRGB chunk.  These values should NOT be changed. */
+#define PNG_sRGB_INTENT_PERCEPTUAL 0
+#define PNG_sRGB_INTENT_RELATIVE   1
+#define PNG_sRGB_INTENT_SATURATION 2
+#define PNG_sRGB_INTENT_ABSOLUTE   3
+#define PNG_sRGB_INTENT_LAST       4 /* Not a valid value */
+
+/* This is for text chunks */
+#define PNG_KEYWORD_MAX_LENGTH     79
+
+/* Maximum number of entries in PLTE/sPLT/tRNS arrays */
+#define PNG_MAX_PALETTE_LENGTH	   256
+
+/* These determine if an ancillary chunk's data has been successfully read
+ * from the PNG header, or if the application has filled in the corresponding
+ * data in the info_struct to be written into the output file.  The values
+ * of the PNG_INFO_<chunk> defines should NOT be changed.
+ */
+#define PNG_INFO_gAMA 0x0001
+#define PNG_INFO_sBIT 0x0002
+#define PNG_INFO_cHRM 0x0004
+#define PNG_INFO_PLTE 0x0008
+#define PNG_INFO_tRNS 0x0010
+#define PNG_INFO_bKGD 0x0020
+#define PNG_INFO_hIST 0x0040
+#define PNG_INFO_pHYs 0x0080
+#define PNG_INFO_oFFs 0x0100
+#define PNG_INFO_tIME 0x0200
+#define PNG_INFO_pCAL 0x0400
+#define PNG_INFO_sRGB 0x0800   /* GR-P, 0.96a */
+#define PNG_INFO_iCCP 0x1000   /* ESR, 1.0.6 */
+#define PNG_INFO_sPLT 0x2000   /* ESR, 1.0.6 */
+#define PNG_INFO_sCAL 0x4000   /* ESR, 1.0.6 */
+#define PNG_INFO_IDAT 0x8000L  /* ESR, 1.0.6 */
+
+/* This is used for the transformation routines, as some of them
+ * change these values for the row.  It also should enable using
+ * the routines for other purposes.
+ */
+typedef struct png_row_info_struct
+{
+   png_uint_32 width; /* width of row */
+   png_uint_32 rowbytes; /* number of bytes in row */
+   png_byte color_type; /* color type of row */
+   png_byte bit_depth; /* bit depth of row */
+   png_byte channels; /* number of channels (1, 2, 3, or 4) */
+   png_byte pixel_depth; /* bits per pixel (depth * channels) */
+} png_row_info;
+
+typedef png_row_info FAR * png_row_infop;
+typedef png_row_info FAR * FAR * png_row_infopp;
+
+/* These are the function types for the I/O functions and for the functions
+ * that allow the user to override the default I/O functions with his or her
+ * own.  The png_error_ptr type should match that of user-supplied warning
+ * and error functions, while the png_rw_ptr type should match that of the
+ * user read/write data functions.
+ */
+typedef struct png_struct_def png_struct;
+typedef png_struct FAR * png_structp;
+
+typedef void (PNGAPI *png_error_ptr) PNGARG((png_structp, png_const_charp));
+typedef void (PNGAPI *png_rw_ptr) PNGARG((png_structp, png_bytep, png_size_t));
+typedef void (PNGAPI *png_flush_ptr) PNGARG((png_structp));
+typedef void (PNGAPI *png_read_status_ptr) PNGARG((png_structp, png_uint_32,
+   int));
+typedef void (PNGAPI *png_write_status_ptr) PNGARG((png_structp, png_uint_32,
+   int));
+
+#ifdef PNG_PROGRESSIVE_READ_SUPPORTED
+typedef void (PNGAPI *png_progressive_info_ptr) PNGARG((png_structp, png_infop));
+typedef void (PNGAPI *png_progressive_end_ptr) PNGARG((png_structp, png_infop));
+typedef void (PNGAPI *png_progressive_row_ptr) PNGARG((png_structp, png_bytep,
+   png_uint_32, int));
+#endif
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_LEGACY_SUPPORTED)
+typedef void (PNGAPI *png_user_transform_ptr) PNGARG((png_structp,
+    png_row_infop, png_bytep));
+#endif
+
+#if defined(PNG_USER_CHUNKS_SUPPORTED)
+typedef int (PNGAPI *png_user_chunk_ptr) PNGARG((png_structp, png_unknown_chunkp));
+#endif
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+typedef void (PNGAPI *png_unknown_chunk_ptr) PNGARG((png_structp));
+#endif
+
+/* Transform masks for the high-level interface */
+#define PNG_TRANSFORM_IDENTITY       0x0000    /* read and write */
+#define PNG_TRANSFORM_STRIP_16       0x0001    /* read only */
+#define PNG_TRANSFORM_STRIP_ALPHA    0x0002    /* read only */
+#define PNG_TRANSFORM_PACKING        0x0004    /* read and write */
+#define PNG_TRANSFORM_PACKSWAP       0x0008    /* read and write */
+#define PNG_TRANSFORM_EXPAND         0x0010    /* read only */
+#define PNG_TRANSFORM_INVERT_MONO    0x0020    /* read and write */
+#define PNG_TRANSFORM_SHIFT          0x0040    /* read and write */
+#define PNG_TRANSFORM_BGR            0x0080    /* read and write */
+#define PNG_TRANSFORM_SWAP_ALPHA     0x0100    /* read and write */
+#define PNG_TRANSFORM_SWAP_ENDIAN    0x0200    /* read and write */
+#define PNG_TRANSFORM_INVERT_ALPHA   0x0400    /* read and write */
+#define PNG_TRANSFORM_STRIP_FILLER   0x0800    /* WRITE only */
+
+/* Flags for MNG supported features */
+#define PNG_FLAG_MNG_EMPTY_PLTE     0x01
+#define PNG_FLAG_MNG_FILTER_64      0x04
+#define PNG_ALL_MNG_FEATURES        0x05
+
+typedef png_voidp (*png_malloc_ptr) PNGARG((png_structp, png_size_t));
+typedef void (*png_free_ptr) PNGARG((png_structp, png_voidp));
+
+/* The structure that holds the information to read and write PNG files.
+ * The only people who need to care about what is inside of this are the
+ * people who will be modifying the library for their own special needs.
+ * It should NOT be accessed directly by an application, except to store
+ * the jmp_buf.
+ */
+
+struct png_struct_def
+{
+#ifdef PNG_SETJMP_SUPPORTED
+   jmp_buf jmpbuf;            /* used in png_error */
+#endif
+   png_error_ptr error_fn;    /* function for printing errors and aborting */
+   png_error_ptr warning_fn;  /* function for printing warnings */
+   png_voidp error_ptr;       /* user supplied struct for error functions */
+   png_rw_ptr write_data_fn;  /* function for writing output data */
+   png_rw_ptr read_data_fn;   /* function for reading input data */
+   png_voidp io_ptr;          /* ptr to application struct for I/O functions*/
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED)
+   png_user_transform_ptr read_user_transform_fn; /* user read transform */
+#endif
+
+#if defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED)
+   png_user_transform_ptr write_user_transform_fn; /* user write transform */
+#endif
+
+/* These were added in libpng-1.0.2 */
+#if defined(PNG_USER_TRANSFORM_PTR_SUPPORTED)
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED)
+   png_voidp user_transform_ptr; /* user supplied struct for user transform */
+   png_byte user_transform_depth;    /* bit depth of user transformed pixels */
+   png_byte user_transform_channels; /* channels in user transformed pixels */
+#endif
+#endif
+
+   png_uint_32 mode;          /* tells us where we are in the PNG file */
+   png_uint_32 flags;         /* flags indicating various things to libpng */
+   png_uint_32 transformations; /* which transformations to perform */
+
+   z_stream zstream;          /* pointer to decompression structure (below) */
+   png_bytep zbuf;            /* buffer for zlib */
+   png_size_t zbuf_size;      /* size of zbuf */
+   int zlib_level;            /* holds zlib compression level */
+   int zlib_method;           /* holds zlib compression method */
+   int zlib_window_bits;      /* holds zlib compression window bits */
+   int zlib_mem_level;        /* holds zlib compression memory level */
+   int zlib_strategy;         /* holds zlib compression strategy */
+
+   png_uint_32 width;         /* width of image in pixels */
+   png_uint_32 height;        /* height of image in pixels */
+   png_uint_32 num_rows;      /* number of rows in current pass */
+   png_uint_32 usr_width;     /* width of row at start of write */
+   png_uint_32 rowbytes;      /* size of row in bytes */
+   png_uint_32 irowbytes;     /* size of current interlaced row in bytes */
+   png_uint_32 iwidth;        /* width of current interlaced row in pixels */
+   png_uint_32 row_number;    /* current row in interlace pass */
+   png_bytep prev_row;        /* buffer to save previous (unfiltered) row */
+   png_bytep row_buf;         /* buffer to save current (unfiltered) row */
+   png_bytep sub_row;         /* buffer to save "sub" row when filtering */
+   png_bytep up_row;          /* buffer to save "up" row when filtering */
+   png_bytep avg_row;         /* buffer to save "avg" row when filtering */
+   png_bytep paeth_row;       /* buffer to save "Paeth" row when filtering */
+   png_row_info row_info;     /* used for transformation routines */
+
+   png_uint_32 idat_size;     /* current IDAT size for read */
+   png_uint_32 crc;           /* current chunk CRC value */
+   png_colorp palette;        /* palette from the input file */
+   png_uint_16 num_palette;   /* number of color entries in palette */
+   png_uint_16 num_trans;     /* number of transparency values */
+   png_byte chunk_name[5];    /* null-terminated name of current chunk */
+   png_byte compression;      /* file compression type (always 0) */
+   png_byte filter;           /* file filter type (always 0) */
+   png_byte interlaced;       /* PNG_INTERLACE_NONE, PNG_INTERLACE_ADAM7 */
+   png_byte pass;             /* current interlace pass (0 - 6) */
+   png_byte do_filter;        /* row filter flags (see PNG_FILTER_ below ) */
+   png_byte color_type;       /* color type of file */
+   png_byte bit_depth;        /* bit depth of file */
+   png_byte usr_bit_depth;    /* bit depth of users row */
+   png_byte pixel_depth;      /* number of bits per pixel */
+   png_byte channels;         /* number of channels in file */
+   png_byte usr_channels;     /* channels at start of write */
+   png_byte sig_bytes;        /* magic bytes read/written from start of file */
+
+#if defined(PNG_READ_FILLER_SUPPORTED) || defined(PNG_WRITE_FILLER_SUPPORTED)
+#ifdef PNG_LEGACY_SUPPORTED
+   png_byte filler;           /* filler byte for pixel expansion */
+#else
+   png_uint_16 filler;           /* filler bytes for pixel expansion */
+#endif
+#endif
+
+#if defined(PNG_bKGD_SUPPORTED)
+   png_byte background_gamma_type;
+#  ifdef PNG_FLOATING_POINT_SUPPORTED
+   float background_gamma;
+#  endif
+   png_color_16 background;   /* background color in screen gamma space */
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   png_color_16 background_1; /* background normalized to gamma 1.0 */
+#endif
+#endif /* PNG_bKGD_SUPPORTED */
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+   png_flush_ptr output_flush_fn;/* Function for flushing output */
+   png_uint_32 flush_dist;    /* how many rows apart to flush, 0 - no flush */
+   png_uint_32 flush_rows;    /* number of rows written since last flush */
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+   int gamma_shift;      /* number of "insignificant" bits 16-bit gamma */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   float gamma;          /* file gamma value */
+   float screen_gamma;   /* screen gamma value (display_exponent) */
+#endif
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+   png_bytep gamma_table;     /* gamma table for 8-bit depth files */
+   png_bytep gamma_from_1;    /* converts from 1.0 to screen */
+   png_bytep gamma_to_1;      /* converts from file to 1.0 */
+   png_uint_16pp gamma_16_table; /* gamma table for 16-bit depth files */
+   png_uint_16pp gamma_16_from_1; /* converts from 1.0 to screen */
+   png_uint_16pp gamma_16_to_1; /* converts from file to 1.0 */
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_sBIT_SUPPORTED)
+   png_color_8 sig_bit;       /* significant bits in each available channel */
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED) || defined(PNG_WRITE_SHIFT_SUPPORTED)
+   png_color_8 shift;         /* shift for significant bit tranformation */
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED) \
+ || defined(PNG_READ_EXPAND_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+   png_bytep trans;           /* transparency values for paletted files */
+   png_color_16 trans_values; /* transparency values for non-paletted files */
+#endif
+
+   png_read_status_ptr read_row_fn;   /* called after each row is decoded */
+   png_write_status_ptr write_row_fn; /* called after each row is encoded */
+#ifdef PNG_PROGRESSIVE_READ_SUPPORTED
+   png_progressive_info_ptr info_fn; /* called after header data fully read */
+   png_progressive_row_ptr row_fn;   /* called after each prog. row is decoded */
+   png_progressive_end_ptr end_fn;   /* called after image is complete */
+   png_bytep save_buffer_ptr;        /* current location in save_buffer */
+   png_bytep save_buffer;            /* buffer for previously read data */
+   png_bytep current_buffer_ptr;     /* current location in current_buffer */
+   png_bytep current_buffer;         /* buffer for recently used data */
+   png_uint_32 push_length;          /* size of current input chunk */
+   png_uint_32 skip_length;          /* bytes to skip in input data */
+   png_size_t save_buffer_size;      /* amount of data now in save_buffer */
+   png_size_t save_buffer_max;       /* total size of save_buffer */
+   png_size_t buffer_size;           /* total amount of available input data */
+   png_size_t current_buffer_size;   /* amount of data now in current_buffer */
+   int process_mode;                 /* what push library is currently doing */
+   int cur_palette;                  /* current push library palette index */
+
+#  if defined(PNG_TEXT_SUPPORTED)
+     png_size_t current_text_size;   /* current size of text input data */
+     png_size_t current_text_left;   /* how much text left to read in input */
+     png_charp current_text;         /* current text chunk buffer */
+     png_charp current_text_ptr;     /* current location in current_text */
+#  endif /* PNG_PROGRESSIVE_READ_SUPPORTED && PNG_TEXT_SUPPORTED */
+
+#endif /* PNG_PROGRESSIVE_READ_SUPPORTED */
+
+#if defined(__TURBOC__) && !defined(_Windows) && !defined(__FLAT__)
+/* for the Borland special 64K segment handler */
+   png_bytepp offset_table_ptr;
+   png_bytep offset_table;
+   png_uint_16 offset_table_number;
+   png_uint_16 offset_table_count;
+   png_uint_16 offset_table_count_free;
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+   png_bytep palette_lookup;         /* lookup table for dithering */
+   png_bytep dither_index;           /* index translation for palette files */
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED) || defined(PNG_hIST_SUPPORTED)
+   png_uint_16p hist;                /* histogram */
+#endif
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+   png_byte heuristic_method;        /* heuristic for row filter selection */
+   png_byte num_prev_filters;        /* number of weights for previous rows */
+   png_bytep prev_filters;           /* filter type(s) of previous row(s) */
+   png_uint_16p filter_weights;      /* weight(s) for previous line(s) */
+   png_uint_16p inv_filter_weights;  /* 1/weight(s) for previous line(s) */
+   png_uint_16p filter_costs;        /* relative filter calculation cost */
+   png_uint_16p inv_filter_costs;    /* 1/relative filter calculation cost */
+#endif
+
+#if defined(PNG_TIME_RFC1123_SUPPORTED)
+   png_charp time_buffer;            /* String to hold RFC 1123 time text */
+#endif
+
+/* New members added in libpng-1.0.6 */
+
+#ifdef PNG_FREE_ME_SUPPORTED
+   png_uint_32 free_me;       /* flags items libpng is responsible for freeing */
+#endif
+
+#if defined(PNG_USER_CHUNKS_SUPPORTED)
+   png_voidp user_chunk_ptr;
+   png_user_chunk_ptr read_user_chunk_fn; /* user read chunk handler */
+#endif
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+   int num_chunk_list;
+   png_bytep chunk_list;
+#endif
+
+/* New members added in libpng-1.0.3 */
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+   png_byte rgb_to_gray_status;
+   /* These were changed from png_byte in libpng-1.0.6 */
+   png_uint_16 rgb_to_gray_red_coeff;
+   png_uint_16 rgb_to_gray_green_coeff;
+   png_uint_16 rgb_to_gray_blue_coeff;
+#endif
+
+/* New member added in libpng-1.0.4 (renamed in 1.0.9) */
+#if defined(PNG_MNG_FEATURES_SUPPORTED) || \
+    defined(PNG_READ_EMPTY_PLTE_SUPPORTED) || \
+    defined(PNG_WRITE_EMPTY_PLTE_SUPPORTED)
+/* changed from png_byte to png_uint_32 at version 1.2.0 */
+   png_uint_32 mng_features_permitted;
+#endif
+
+/* New member added in libpng-1.0.7 */
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+   png_fixed_point int_gamma;
+#endif
+
+/* New member added in libpng-1.0.9, ifdef'ed out in 1.0.12, enabled in 1.2.0 */
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   png_byte filter_type;
+#endif
+
+#if defined(PNG_DEBUG) && defined(PNG_USE_PNGGCCRD)
+/* New member added in libpng-1.0.10, ifdef'ed out in 1.2.0 */
+   png_uint_32 row_buf_size;
+#endif
+
+/* New members added in libpng-1.2.0 */
+#if defined(PNG_ASSEMBLER_CODE_SUPPORTED)
+   png_byte     mmx_bitdepth_threshold;
+   png_uint_32  mmx_rowbytes_threshold;
+   png_uint_32  asm_flags;
+#endif
+
+/* New members added in libpng-1.0.2 but first enabled by default in 1.2.0 */
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_voidp mem_ptr;                /* user supplied struct for mem functions */
+   png_malloc_ptr malloc_fn;         /* function for allocating memory */
+   png_free_ptr free_fn;             /* function for freeing memory */
+#endif
+
+   png_bytep big_row_buf;         /* buffer to save current (unfiltered) row */
+
+};
+
+
+/* This prevents a compiler error in png_get_copyright() in png.c if png.c
+   and png.h are both at version 1.2.1
+ */
+typedef png_structp version_1_2_1;
+
+typedef png_struct FAR * FAR * png_structpp;
+
+/* Here are the function definitions most commonly used.  This is not
+ * the place to find out how to use libpng.  See libpng.txt for the
+ * full explanation, see example.c for the summary.  This just provides
+ * a simple one line description of the use of each function.
+ */
+
+/* Returns the version number of the library */
+extern PNG_EXPORT(png_uint_32,png_access_version_number) PNGARG((void));
+
+/* Tell lib we have already handled the first <num_bytes> magic bytes.
+ * Handling more than 8 bytes from the beginning of the file is an error.
+ */
+extern PNG_EXPORT(void,png_set_sig_bytes) PNGARG((png_structp png_ptr,
+   int num_bytes));
+
+/* Check sig[start] through sig[start + num_to_check - 1] to see if it's a
+ * PNG file.  Returns zero if the supplied bytes match the 8-byte PNG
+ * signature, and non-zero otherwise.  Having num_to_check == 0 or
+ * start > 7 will always fail (ie return non-zero).
+ */
+extern PNG_EXPORT(int,png_sig_cmp) PNGARG((png_bytep sig, png_size_t start,
+   png_size_t num_to_check));
+
+/* Simple signature checking function.  This is the same as calling
+ * png_check_sig(sig, n) := !png_sig_cmp(sig, 0, n).
+ */
+extern PNG_EXPORT(int,png_check_sig) PNGARG((png_bytep sig, int num));
+
+/* Allocate and initialize png_ptr struct for reading, and any other memory. */
+extern PNG_EXPORT(png_structp,png_create_read_struct)
+   PNGARG((png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn));
+
+/* Allocate and initialize png_ptr struct for writing, and any other memory */
+extern PNG_EXPORT(png_structp,png_create_write_struct)
+   PNGARG((png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn));
+
+extern PNG_EXPORT(png_uint_32,png_get_compression_buffer_size)
+   PNGARG((png_structp png_ptr));
+
+extern PNG_EXPORT(void,png_set_compression_buffer_size)
+   PNGARG((png_structp png_ptr, png_uint_32 size));
+
+/* Reset the compression stream */
+extern PNG_EXPORT(int,png_reset_zstream) PNGARG((png_structp png_ptr));
+
+/* New functions added in libpng-1.0.2 (not enabled by default until 1.2.0) */
+#ifdef PNG_USER_MEM_SUPPORTED
+extern PNG_EXPORT(png_structp,png_create_read_struct_2)
+   PNGARG((png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn, png_voidp mem_ptr,
+   png_malloc_ptr malloc_fn, png_free_ptr free_fn));
+extern PNG_EXPORT(png_structp,png_create_write_struct_2)
+   PNGARG((png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn, png_voidp mem_ptr,
+   png_malloc_ptr malloc_fn, png_free_ptr free_fn));
+#endif
+
+/* Write a PNG chunk - size, type, (optional) data, CRC. */
+extern PNG_EXPORT(void,png_write_chunk) PNGARG((png_structp png_ptr,
+   png_bytep chunk_name, png_bytep data, png_size_t length));
+
+/* Write the start of a PNG chunk - length and chunk name. */
+extern PNG_EXPORT(void,png_write_chunk_start) PNGARG((png_structp png_ptr,
+   png_bytep chunk_name, png_uint_32 length));
+
+/* Write the data of a PNG chunk started with png_write_chunk_start(). */
+extern PNG_EXPORT(void,png_write_chunk_data) PNGARG((png_structp png_ptr,
+   png_bytep data, png_size_t length));
+
+/* Finish a chunk started with png_write_chunk_start() (includes CRC). */
+extern PNG_EXPORT(void,png_write_chunk_end) PNGARG((png_structp png_ptr));
+
+/* Allocate and initialize the info structure */
+extern PNG_EXPORT(png_infop,png_create_info_struct)
+   PNGARG((png_structp png_ptr));
+
+/* Initialize the info structure (old interface - DEPRECATED) */
+extern PNG_EXPORT(void,png_info_init) PNGARG((png_infop info_ptr));
+#define png_info_init(info_ptr) png_info_init_3(&info_ptr, sizeof(png_info));
+extern PNG_EXPORT(void,png_info_init_3) PNGARG((png_infopp info_ptr,
+    png_size_t png_info_struct_size));
+
+/* Writes all the PNG information before the image. */
+extern PNG_EXPORT(void,png_write_info_before_PLTE) PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+extern PNG_EXPORT(void,png_write_info) PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+/* read the information before the actual image data. */
+extern PNG_EXPORT(void,png_read_info) PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+#if defined(PNG_TIME_RFC1123_SUPPORTED)
+extern PNG_EXPORT(png_charp,png_convert_to_rfc1123)
+   PNGARG((png_structp png_ptr, png_timep ptime));
+#endif
+
+#if !defined(_WIN32_WCE)
+/* "time.h" functions are not supported on WindowsCE */
+#if defined(PNG_WRITE_tIME_SUPPORTED)
+/* convert from a struct tm to png_time */
+extern PNG_EXPORT(void,png_convert_from_struct_tm) PNGARG((png_timep ptime,
+   struct tm FAR * ttime));
+
+/* convert from time_t to png_time.  Uses gmtime() */
+extern PNG_EXPORT(void,png_convert_from_time_t) PNGARG((png_timep ptime,
+   time_t ttime));
+#endif /* PNG_WRITE_tIME_SUPPORTED */
+#endif /* _WIN32_WCE */
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+/* Expand data to 24-bit RGB, or 8-bit grayscale, with alpha if available. */
+extern PNG_EXPORT(void,png_set_expand) PNGARG((png_structp png_ptr));
+extern PNG_EXPORT(void,png_set_gray_1_2_4_to_8) PNGARG((png_structp png_ptr));
+extern PNG_EXPORT(void,png_set_palette_to_rgb) PNGARG((png_structp png_ptr));
+extern PNG_EXPORT(void,png_set_tRNS_to_alpha) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED)
+/* Use blue, green, red order for pixels. */
+extern PNG_EXPORT(void,png_set_bgr) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+/* Expand the grayscale to 24-bit RGB if necessary. */
+extern PNG_EXPORT(void,png_set_gray_to_rgb) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+/* Reduce RGB to grayscale. */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_rgb_to_gray) PNGARG((png_structp png_ptr,
+   int error_action, double red, double green ));
+#endif
+extern PNG_EXPORT(void,png_set_rgb_to_gray_fixed) PNGARG((png_structp png_ptr,
+   int error_action, png_fixed_point red, png_fixed_point green ));
+extern PNG_EXPORT(png_byte,png_get_rgb_to_gray_status) PNGARG((png_structp
+   png_ptr));
+#endif
+
+extern PNG_EXPORT(void,png_build_grayscale_palette) PNGARG((int bit_depth,
+   png_colorp palette));
+
+#if defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+extern PNG_EXPORT(void,png_set_strip_alpha) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED) || \
+    defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED)
+extern PNG_EXPORT(void,png_set_swap_alpha) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED) || \
+    defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+extern PNG_EXPORT(void,png_set_invert_alpha) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_FILLER_SUPPORTED) || defined(PNG_WRITE_FILLER_SUPPORTED)
+/* Add a filler byte to 24-bit RGB images. */
+extern PNG_EXPORT(void,png_set_filler) PNGARG((png_structp png_ptr,
+   png_uint_32 filler, int flags));
+/* The values of the PNG_FILLER_ defines should NOT be changed */
+#define PNG_FILLER_BEFORE 0
+#define PNG_FILLER_AFTER 1
+#endif /* PNG_READ_FILLER_SUPPORTED || PNG_WRITE_FILLER_SUPPORTED */
+
+#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED)
+/* Swap bytes in 16-bit depth files. */
+extern PNG_EXPORT(void,png_set_swap) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_PACK_SUPPORTED) || defined(PNG_WRITE_PACK_SUPPORTED)
+/* Use 1 byte per pixel in 1, 2, or 4-bit depth files. */
+extern PNG_EXPORT(void,png_set_packing) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED) || defined(PNG_WRITE_PACKSWAP_SUPPORTED)
+/* Swap packing order of pixels in bytes. */
+extern PNG_EXPORT(void,png_set_packswap) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED) || defined(PNG_WRITE_SHIFT_SUPPORTED)
+/* Converts files to legal bit depths. */
+extern PNG_EXPORT(void,png_set_shift) PNGARG((png_structp png_ptr,
+   png_color_8p true_bits));
+#endif
+
+#if defined(PNG_READ_INTERLACING_SUPPORTED) || \
+    defined(PNG_WRITE_INTERLACING_SUPPORTED)
+/* Have the code handle the interlacing.  Returns the number of passes. */
+extern PNG_EXPORT(int,png_set_interlace_handling) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_INVERT_SUPPORTED) || defined(PNG_WRITE_INVERT_SUPPORTED)
+/* Invert monochrome files */
+extern PNG_EXPORT(void,png_set_invert_mono) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+/* Handle alpha and tRNS by replacing with a background color. */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_background) PNGARG((png_structp png_ptr,
+   png_color_16p background_color, int background_gamma_code,
+   int need_expand, double background_gamma));
+#endif
+#define PNG_BACKGROUND_GAMMA_UNKNOWN 0
+#define PNG_BACKGROUND_GAMMA_SCREEN  1
+#define PNG_BACKGROUND_GAMMA_FILE    2
+#define PNG_BACKGROUND_GAMMA_UNIQUE  3
+#endif
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+/* strip the second byte of information from a 16-bit depth file. */
+extern PNG_EXPORT(void,png_set_strip_16) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+/* Turn on dithering, and reduce the palette to the number of colors available. */
+extern PNG_EXPORT(void,png_set_dither) PNGARG((png_structp png_ptr,
+   png_colorp palette, int num_palette, int maximum_colors,
+   png_uint_16p histogram, int full_dither));
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+/* Handle gamma correction. Screen_gamma=(display_exponent) */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_gamma) PNGARG((png_structp png_ptr,
+   double screen_gamma, double default_file_gamma));
+#endif
+#endif
+
+#if defined(PNG_READ_EMPTY_PLTE_SUPPORTED) || \
+    defined(PNG_WRITE_EMPTY_PLTE_SUPPORTED)
+/* Permit or disallow empty PLTE (0: not permitted, 1: permitted) */
+/* Deprecated and will be removed.  Use png_permit_mng_features() instead. */
+extern PNG_EXPORT(void,png_permit_empty_plte) PNGARG((png_structp png_ptr,
+   int empty_plte_permitted));
+#endif
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+/* Set how many lines between output flushes - 0 for no flushing */
+extern PNG_EXPORT(void,png_set_flush) PNGARG((png_structp png_ptr, int nrows));
+/* Flush the current PNG output buffer */
+extern PNG_EXPORT(void,png_write_flush) PNGARG((png_structp png_ptr));
+#endif
+
+/* optional update palette with requested transformations */
+extern PNG_EXPORT(void,png_start_read_image) PNGARG((png_structp png_ptr));
+
+/* optional call to update the users info structure */
+extern PNG_EXPORT(void,png_read_update_info) PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+/* read a one or more rows of image data.*/
+extern PNG_EXPORT(void,png_read_rows) PNGARG((png_structp png_ptr,
+   png_bytepp row, png_bytepp display_row, png_uint_32 num_rows));
+
+/* read a row of data.*/
+extern PNG_EXPORT(void,png_read_row) PNGARG((png_structp png_ptr,
+   png_bytep row,
+   png_bytep display_row));
+
+/* read the whole image into memory at once. */
+extern PNG_EXPORT(void,png_read_image) PNGARG((png_structp png_ptr,
+   png_bytepp image));
+
+/* write a row of image data */
+extern PNG_EXPORT(void,png_write_row) PNGARG((png_structp png_ptr,
+   png_bytep row));
+
+/* write a few rows of image data */
+extern PNG_EXPORT(void,png_write_rows) PNGARG((png_structp png_ptr,
+   png_bytepp row, png_uint_32 num_rows));
+
+/* write the image data */
+extern PNG_EXPORT(void,png_write_image) PNGARG((png_structp png_ptr,
+   png_bytepp image));
+
+/* writes the end of the PNG file. */
+extern PNG_EXPORT(void,png_write_end) PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+/* read the end of the PNG file. */
+extern PNG_EXPORT(void,png_read_end) PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+/* free any memory associated with the png_info_struct */
+extern PNG_EXPORT(void,png_destroy_info_struct) PNGARG((png_structp png_ptr,
+   png_infopp info_ptr_ptr));
+
+/* free any memory associated with the png_struct and the png_info_structs */
+extern PNG_EXPORT(void,png_destroy_read_struct) PNGARG((png_structpp
+   png_ptr_ptr, png_infopp info_ptr_ptr, png_infopp end_info_ptr_ptr));
+
+/* free all memory used by the read (old method - NOT DLL EXPORTED) */
+extern void png_read_destroy PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_infop end_info_ptr));
+
+/* free any memory associated with the png_struct and the png_info_structs */
+extern PNG_EXPORT(void,png_destroy_write_struct)
+   PNGARG((png_structpp png_ptr_ptr, png_infopp info_ptr_ptr));
+
+/* free any memory used in info_ptr struct (old method - NOT DLL EXPORTED) */
+extern void png_write_destroy_info PNGARG((png_infop info_ptr));
+
+/* free any memory used in png_ptr struct (old method - NOT DLL EXPORTED) */
+extern void png_write_destroy PNGARG((png_structp png_ptr));
+
+/* set the libpng method of handling chunk CRC errors */
+extern PNG_EXPORT(void,png_set_crc_action) PNGARG((png_structp png_ptr,
+   int crit_action, int ancil_action));
+
+/* Values for png_set_crc_action() to say how to handle CRC errors in
+ * ancillary and critical chunks, and whether to use the data contained
+ * therein.  Note that it is impossible to "discard" data in a critical
+ * chunk.  For versions prior to 0.90, the action was always error/quit,
+ * whereas in version 0.90 and later, the action for CRC errors in ancillary
+ * chunks is warn/discard.  These values should NOT be changed.
+ *
+ *      value                       action:critical     action:ancillary
+ */
+#define PNG_CRC_DEFAULT       0  /* error/quit          warn/discard data */
+#define PNG_CRC_ERROR_QUIT    1  /* error/quit          error/quit        */
+#define PNG_CRC_WARN_DISCARD  2  /* (INVALID)           warn/discard data */
+#define PNG_CRC_WARN_USE      3  /* warn/use data       warn/use data     */
+#define PNG_CRC_QUIET_USE     4  /* quiet/use data      quiet/use data    */
+#define PNG_CRC_NO_CHANGE     5  /* use current value   use current value */
+
+/* These functions give the user control over the scan-line filtering in
+ * libpng and the compression methods used by zlib.  These functions are
+ * mainly useful for testing, as the defaults should work with most users.
+ * Those users who are tight on memory or want faster performance at the
+ * expense of compression can modify them.  See the compression library
+ * header file (zlib.h) for an explination of the compression functions.
+ */
+
+/* set the filtering method(s) used by libpng.  Currently, the only valid
+ * value for "method" is 0.
+ */
+extern PNG_EXPORT(void,png_set_filter) PNGARG((png_structp png_ptr, int method,
+   int filters));
+
+/* Flags for png_set_filter() to say which filters to use.  The flags
+ * are chosen so that they don't conflict with real filter types
+ * below, in case they are supplied instead of the #defined constants.
+ * These values should NOT be changed.
+ */
+#define PNG_NO_FILTERS     0x00
+#define PNG_FILTER_NONE    0x08
+#define PNG_FILTER_SUB     0x10
+#define PNG_FILTER_UP      0x20
+#define PNG_FILTER_AVG     0x40
+#define PNG_FILTER_PAETH   0x80
+#define PNG_ALL_FILTERS (PNG_FILTER_NONE | PNG_FILTER_SUB | PNG_FILTER_UP | \
+                         PNG_FILTER_AVG | PNG_FILTER_PAETH)
+
+/* Filter values (not flags) - used in pngwrite.c, pngwutil.c for now.
+ * These defines should NOT be changed.
+ */
+#define PNG_FILTER_VALUE_NONE  0
+#define PNG_FILTER_VALUE_SUB   1
+#define PNG_FILTER_VALUE_UP    2
+#define PNG_FILTER_VALUE_AVG   3
+#define PNG_FILTER_VALUE_PAETH 4
+#define PNG_FILTER_VALUE_LAST  5
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED) /* EXPERIMENTAL */
+/* The "heuristic_method" is given by one of the PNG_FILTER_HEURISTIC_
+ * defines, either the default (minimum-sum-of-absolute-differences), or
+ * the experimental method (weighted-minimum-sum-of-absolute-differences).
+ *
+ * Weights are factors >= 1.0, indicating how important it is to keep the
+ * filter type consistent between rows.  Larger numbers mean the current
+ * filter is that many times as likely to be the same as the "num_weights"
+ * previous filters.  This is cumulative for each previous row with a weight.
+ * There needs to be "num_weights" values in "filter_weights", or it can be
+ * NULL if the weights aren't being specified.  Weights have no influence on
+ * the selection of the first row filter.  Well chosen weights can (in theory)
+ * improve the compression for a given image.
+ *
+ * Costs are factors >= 1.0 indicating the relative decoding costs of a
+ * filter type.  Higher costs indicate more decoding expense, and are
+ * therefore less likely to be selected over a filter with lower computational
+ * costs.  There needs to be a value in "filter_costs" for each valid filter
+ * type (given by PNG_FILTER_VALUE_LAST), or it can be NULL if you aren't
+ * setting the costs.  Costs try to improve the speed of decompression without
+ * unduly increasing the compressed image size.
+ *
+ * A negative weight or cost indicates the default value is to be used, and
+ * values in the range [0.0, 1.0) indicate the value is to remain unchanged.
+ * The default values for both weights and costs are currently 1.0, but may
+ * change if good general weighting/cost heuristics can be found.  If both
+ * the weights and costs are set to 1.0, this degenerates the WEIGHTED method
+ * to the UNWEIGHTED method, but with added encoding time/computation.
+ */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_filter_heuristics) PNGARG((png_structp png_ptr,
+   int heuristic_method, int num_weights, png_doublep filter_weights,
+   png_doublep filter_costs));
+#endif
+#endif /*  PNG_WRITE_WEIGHTED_FILTER_SUPPORTED */
+
+/* Heuristic used for row filter selection.  These defines should NOT be
+ * changed.
+ */
+#define PNG_FILTER_HEURISTIC_DEFAULT    0  /* Currently "UNWEIGHTED" */
+#define PNG_FILTER_HEURISTIC_UNWEIGHTED 1  /* Used by libpng < 0.95 */
+#define PNG_FILTER_HEURISTIC_WEIGHTED   2  /* Experimental feature */
+#define PNG_FILTER_HEURISTIC_LAST       3  /* Not a valid value */
+
+/* Set the library compression level.  Currently, valid values range from
+ * 0 - 9, corresponding directly to the zlib compression levels 0 - 9
+ * (0 - no compression, 9 - "maximal" compression).  Note that tests have
+ * shown that zlib compression levels 3-6 usually perform as well as level 9
+ * for PNG images, and do considerably fewer caclulations.  In the future,
+ * these values may not correspond directly to the zlib compression levels.
+ */
+extern PNG_EXPORT(void,png_set_compression_level) PNGARG((png_structp png_ptr,
+   int level));
+
+extern PNG_EXPORT(void,png_set_compression_mem_level)
+   PNGARG((png_structp png_ptr, int mem_level));
+
+extern PNG_EXPORT(void,png_set_compression_strategy)
+   PNGARG((png_structp png_ptr, int strategy));
+
+extern PNG_EXPORT(void,png_set_compression_window_bits)
+   PNGARG((png_structp png_ptr, int window_bits));
+
+extern PNG_EXPORT(void,png_set_compression_method) PNGARG((png_structp png_ptr,
+   int method));
+
+/* These next functions are called for input/output, memory, and error
+ * handling.  They are in the file pngrio.c, pngwio.c, and pngerror.c,
+ * and call standard C I/O routines such as fread(), fwrite(), and
+ * fprintf().  These functions can be made to use other I/O routines
+ * at run time for those applications that need to handle I/O in a
+ * different manner by calling png_set_???_fn().  See libpng.txt for
+ * more information.
+ */
+
+#if !defined(PNG_NO_STDIO)
+/* Initialize the input/output for the PNG file to the default functions. */
+extern PNG_EXPORT(void,png_init_io) PNGARG((png_structp png_ptr, png_FILE_p fp));
+#endif
+
+/* Replace the (error and abort), and warning functions with user
+ * supplied functions.  If no messages are to be printed you must still
+ * write and use replacement functions. The replacement error_fn should
+ * still do a longjmp to the last setjmp location if you are using this
+ * method of error handling.  If error_fn or warning_fn is NULL, the
+ * default function will be used.
+ */
+
+extern PNG_EXPORT(void,png_set_error_fn) PNGARG((png_structp png_ptr,
+   png_voidp error_ptr, png_error_ptr error_fn, png_error_ptr warning_fn));
+
+/* Return the user pointer associated with the error functions */
+extern PNG_EXPORT(png_voidp,png_get_error_ptr) PNGARG((png_structp png_ptr));
+
+/* Replace the default data output functions with a user supplied one(s).
+ * If buffered output is not used, then output_flush_fn can be set to NULL.
+ * If PNG_WRITE_FLUSH_SUPPORTED is not defined at libpng compile time
+ * output_flush_fn will be ignored (and thus can be NULL).
+ */
+extern PNG_EXPORT(void,png_set_write_fn) PNGARG((png_structp png_ptr,
+   png_voidp io_ptr, png_rw_ptr write_data_fn, png_flush_ptr output_flush_fn));
+
+/* Replace the default data input function with a user supplied one. */
+extern PNG_EXPORT(void,png_set_read_fn) PNGARG((png_structp png_ptr,
+   png_voidp io_ptr, png_rw_ptr read_data_fn));
+
+/* Return the user pointer associated with the I/O functions */
+extern PNG_EXPORT(png_voidp,png_get_io_ptr) PNGARG((png_structp png_ptr));
+
+extern PNG_EXPORT(void,png_set_read_status_fn) PNGARG((png_structp png_ptr,
+   png_read_status_ptr read_row_fn));
+
+extern PNG_EXPORT(void,png_set_write_status_fn) PNGARG((png_structp png_ptr,
+   png_write_status_ptr write_row_fn));
+
+#ifdef PNG_USER_MEM_SUPPORTED
+/* Replace the default memory allocation functions with user supplied one(s). */
+extern PNG_EXPORT(void,png_set_mem_fn) PNGARG((png_structp png_ptr,
+   png_voidp mem_ptr, png_malloc_ptr malloc_fn, png_free_ptr free_fn));
+/* Return the user pointer associated with the memory functions */
+extern PNG_EXPORT(png_voidp,png_get_mem_ptr) PNGARG((png_structp png_ptr));
+#endif
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_LEGACY_SUPPORTED)
+extern PNG_EXPORT(void,png_set_read_user_transform_fn) PNGARG((png_structp
+   png_ptr, png_user_transform_ptr read_user_transform_fn));
+#endif
+
+#if defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_LEGACY_SUPPORTED)
+extern PNG_EXPORT(void,png_set_write_user_transform_fn) PNGARG((png_structp
+   png_ptr, png_user_transform_ptr write_user_transform_fn));
+#endif
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_LEGACY_SUPPORTED)
+extern PNG_EXPORT(void,png_set_user_transform_info) PNGARG((png_structp
+   png_ptr, png_voidp user_transform_ptr, int user_transform_depth,
+   int user_transform_channels));
+/* Return the user pointer associated with the user transform functions */
+extern PNG_EXPORT(png_voidp,png_get_user_transform_ptr)
+   PNGARG((png_structp png_ptr));
+#endif
+
+#ifdef PNG_USER_CHUNKS_SUPPORTED
+extern PNG_EXPORT(void,png_set_read_user_chunk_fn) PNGARG((png_structp png_ptr,
+   png_voidp user_chunk_ptr, png_user_chunk_ptr read_user_chunk_fn));
+extern PNG_EXPORT(png_voidp,png_get_user_chunk_ptr) PNGARG((png_structp
+   png_ptr));
+#endif
+
+#ifdef PNG_PROGRESSIVE_READ_SUPPORTED
+/* Sets the function callbacks for the push reader, and a pointer to a
+ * user-defined structure available to the callback functions.
+ */
+extern PNG_EXPORT(void,png_set_progressive_read_fn) PNGARG((png_structp png_ptr,
+   png_voidp progressive_ptr,
+   png_progressive_info_ptr info_fn, png_progressive_row_ptr row_fn,
+   png_progressive_end_ptr end_fn));
+
+/* returns the user pointer associated with the push read functions */
+extern PNG_EXPORT(png_voidp,png_get_progressive_ptr)
+   PNGARG((png_structp png_ptr));
+
+/* function to be called when data becomes available */
+extern PNG_EXPORT(void,png_process_data) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_bytep buffer, png_size_t buffer_size));
+
+/* function that combines rows.  Not very much different than the
+ * png_combine_row() call.  Is this even used?????
+ */
+extern PNG_EXPORT(void,png_progressive_combine_row) PNGARG((png_structp png_ptr,
+   png_bytep old_row, png_bytep new_row));
+#endif /* PNG_PROGRESSIVE_READ_SUPPORTED */
+
+extern PNG_EXPORT(png_voidp,png_malloc) PNGARG((png_structp png_ptr,
+   png_uint_32 size));
+
+/* frees a pointer allocated by png_malloc() */
+extern PNG_EXPORT(void,png_free) PNGARG((png_structp png_ptr, png_voidp ptr));
+
+/* Free data that was allocated internally */
+extern PNG_EXPORT(void,png_free_data) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 free_me, int num));
+#ifdef PNG_FREE_ME_SUPPORTED
+/* Reassign responsibility for freeing existing data, whether allocated
+ * by libpng or by the application */
+extern PNG_EXPORT(void,png_data_freer) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int freer, png_uint_32 mask));
+#endif
+/* assignments for png_data_freer */
+#define PNG_DESTROY_WILL_FREE_DATA 1
+#define PNG_SET_WILL_FREE_DATA 1
+#define PNG_USER_WILL_FREE_DATA 2
+/* Flags for png_ptr->free_me and info_ptr->free_me */
+#define PNG_FREE_HIST 0x0008
+#define PNG_FREE_ICCP 0x0010
+#define PNG_FREE_SPLT 0x0020
+#define PNG_FREE_ROWS 0x0040
+#define PNG_FREE_PCAL 0x0080
+#define PNG_FREE_SCAL 0x0100
+#define PNG_FREE_UNKN 0x0200
+#define PNG_FREE_LIST 0x0400
+#define PNG_FREE_PLTE 0x1000
+#define PNG_FREE_TRNS 0x2000
+#define PNG_FREE_TEXT 0x4000
+#define PNG_FREE_ALL  0x7fff
+#define PNG_FREE_MUL  0x4220 /* PNG_FREE_SPLT|PNG_FREE_TEXT|PNG_FREE_UNKN */
+
+#ifdef PNG_USER_MEM_SUPPORTED
+extern PNG_EXPORT(png_voidp,png_malloc_default) PNGARG((png_structp png_ptr,
+   png_uint_32 size));
+extern PNG_EXPORT(void,png_free_default) PNGARG((png_structp png_ptr,
+   png_voidp ptr));
+#endif
+
+extern PNG_EXPORT(png_voidp,png_memcpy_check) PNGARG((png_structp png_ptr,
+   png_voidp s1, png_voidp s2, png_uint_32 size));
+
+extern PNG_EXPORT(png_voidp,png_memset_check) PNGARG((png_structp png_ptr,
+   png_voidp s1, int value, png_uint_32 size));
+
+#if defined(USE_FAR_KEYWORD)  /* memory model conversion function */
+extern void *png_far_to_near PNGARG((png_structp png_ptr,png_voidp ptr,
+   int check));
+#endif /* USE_FAR_KEYWORD */
+
+/* Fatal error in PNG image of libpng - can't continue */
+extern PNG_EXPORT(void,png_error) PNGARG((png_structp png_ptr,
+   png_const_charp error));
+
+/* The same, but the chunk name is prepended to the error string. */
+extern PNG_EXPORT(void,png_chunk_error) PNGARG((png_structp png_ptr,
+   png_const_charp error));
+
+/* Non-fatal error in libpng.  Can continue, but may have a problem. */
+extern PNG_EXPORT(void,png_warning) PNGARG((png_structp png_ptr,
+   png_const_charp message));
+
+/* Non-fatal error in libpng, chunk name is prepended to message. */
+extern PNG_EXPORT(void,png_chunk_warning) PNGARG((png_structp png_ptr,
+   png_const_charp message));
+
+/* The png_set_<chunk> functions are for storing values in the png_info_struct.
+ * Similarly, the png_get_<chunk> calls are used to read values from the
+ * png_info_struct, either storing the parameters in the passed variables, or
+ * setting pointers into the png_info_struct where the data is stored.  The
+ * png_get_<chunk> functions return a non-zero value if the data was available
+ * in info_ptr, or return zero and do not change any of the parameters if the
+ * data was not available.
+ *
+ * These functions should be used instead of directly accessing png_info
+ * to avoid problems with future changes in the size and internal layout of
+ * png_info_struct.
+ */
+/* Returns "flag" if chunk data is valid in info_ptr. */
+extern PNG_EXPORT(png_uint_32,png_get_valid) PNGARG((png_structp png_ptr,
+png_infop info_ptr, png_uint_32 flag));
+
+/* Returns number of bytes needed to hold a transformed row. */
+extern PNG_EXPORT(png_uint_32,png_get_rowbytes) PNGARG((png_structp png_ptr,
+png_infop info_ptr));
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+/* Returns row_pointers, which is an array of pointers to scanlines that was
+returned from png_read_png(). */
+extern PNG_EXPORT(png_bytepp,png_get_rows) PNGARG((png_structp png_ptr,
+png_infop info_ptr));
+/* Set row_pointers, which is an array of pointers to scanlines for use
+by png_write_png(). */
+extern PNG_EXPORT(void,png_set_rows) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_bytepp row_pointers));
+#endif
+
+/* Returns number of color channels in image. */
+extern PNG_EXPORT(png_byte,png_get_channels) PNGARG((png_structp png_ptr,
+png_infop info_ptr));
+
+#ifdef PNG_EASY_ACCESS_SUPPORTED
+/* Returns image width in pixels. */
+extern PNG_EXPORT(png_uint_32, png_get_image_width) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image height in pixels. */
+extern PNG_EXPORT(png_uint_32, png_get_image_height) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image bit_depth. */
+extern PNG_EXPORT(png_byte, png_get_bit_depth) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image color_type. */
+extern PNG_EXPORT(png_byte, png_get_color_type) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image filter_type. */
+extern PNG_EXPORT(png_byte, png_get_filter_type) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image interlace_type. */
+extern PNG_EXPORT(png_byte, png_get_interlace_type) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image compression_type. */
+extern PNG_EXPORT(png_byte, png_get_compression_type) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns image resolution in pixels per meter, from pHYs chunk data. */
+extern PNG_EXPORT(png_uint_32, png_get_pixels_per_meter) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+extern PNG_EXPORT(png_uint_32, png_get_x_pixels_per_meter) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+extern PNG_EXPORT(png_uint_32, png_get_y_pixels_per_meter) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+/* Returns pixel aspect ratio, computed from pHYs chunk data.  */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(float, png_get_pixel_aspect_ratio) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+#endif
+
+/* Returns image x, y offset in pixels or microns, from oFFs chunk data. */
+extern PNG_EXPORT(png_int_32, png_get_x_offset_pixels) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+extern PNG_EXPORT(png_int_32, png_get_y_offset_pixels) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+extern PNG_EXPORT(png_int_32, png_get_x_offset_microns) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+extern PNG_EXPORT(png_int_32, png_get_y_offset_microns) PNGARG((png_structp
+png_ptr, png_infop info_ptr));
+
+#endif /* PNG_EASY_ACCESS_SUPPORTED */
+
+/* Returns pointer to signature string read from PNG header */
+extern PNG_EXPORT(png_bytep,png_get_signature) PNGARG((png_structp png_ptr,
+png_infop info_ptr));
+
+#if defined(PNG_bKGD_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_bKGD) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_color_16p *background));
+#endif
+
+#if defined(PNG_bKGD_SUPPORTED)
+extern PNG_EXPORT(void,png_set_bKGD) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_color_16p background));
+#endif
+
+#if defined(PNG_cHRM_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(png_uint_32,png_get_cHRM) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, double *white_x, double *white_y, double *red_x,
+   double *red_y, double *green_x, double *green_y, double *blue_x,
+   double *blue_y));
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+extern PNG_EXPORT(png_uint_32,png_get_cHRM_fixed) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_fixed_point *int_white_x, png_fixed_point
+   *int_white_y, png_fixed_point *int_red_x, png_fixed_point *int_red_y,
+   png_fixed_point *int_green_x, png_fixed_point *int_green_y, png_fixed_point
+   *int_blue_x, png_fixed_point *int_blue_y));
+#endif
+#endif
+
+#if defined(PNG_cHRM_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_cHRM) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, double white_x, double white_y, double red_x,
+   double red_y, double green_x, double green_y, double blue_x, double blue_y));
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_cHRM_fixed) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_fixed_point int_white_x, png_fixed_point int_white_y,
+   png_fixed_point int_red_x, png_fixed_point int_red_y, png_fixed_point
+   int_green_x, png_fixed_point int_green_y, png_fixed_point int_blue_x,
+   png_fixed_point int_blue_y));
+#endif
+#endif
+
+#if defined(PNG_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(png_uint_32,png_get_gAMA) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, double *file_gamma));
+#endif
+extern PNG_EXPORT(png_uint_32,png_get_gAMA_fixed) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_fixed_point *int_file_gamma));
+#endif
+
+#if defined(PNG_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_gAMA) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, double file_gamma));
+#endif
+extern PNG_EXPORT(void,png_set_gAMA_fixed) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_fixed_point int_file_gamma));
+#endif
+
+#if defined(PNG_hIST_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_hIST) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_16p *hist));
+#endif
+
+#if defined(PNG_hIST_SUPPORTED)
+extern PNG_EXPORT(void,png_set_hIST) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_16p hist));
+#endif
+
+extern PNG_EXPORT(png_uint_32,png_get_IHDR) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 *width, png_uint_32 *height,
+   int *bit_depth, int *color_type, int *interlace_method,
+   int *compression_method, int *filter_method));
+
+extern PNG_EXPORT(void,png_set_IHDR) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 width, png_uint_32 height, int bit_depth,
+   int color_type, int interlace_method, int compression_method,
+   int filter_method));
+
+#if defined(PNG_oFFs_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_oFFs) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_int_32 *offset_x, png_int_32 *offset_y,
+   int *unit_type));
+#endif
+
+#if defined(PNG_oFFs_SUPPORTED)
+extern PNG_EXPORT(void,png_set_oFFs) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_int_32 offset_x, png_int_32 offset_y,
+   int unit_type));
+#endif
+
+#if defined(PNG_pCAL_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_pCAL) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_charp *purpose, png_int_32 *X0, png_int_32 *X1,
+   int *type, int *nparams, png_charp *units, png_charpp *params));
+#endif
+
+#if defined(PNG_pCAL_SUPPORTED)
+extern PNG_EXPORT(void,png_set_pCAL) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_charp purpose, png_int_32 X0, png_int_32 X1,
+   int type, int nparams, png_charp units, png_charpp params));
+#endif
+
+#if defined(PNG_pHYs_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_pHYs) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 *res_x, png_uint_32 *res_y, int *unit_type));
+#endif
+
+#if defined(PNG_pHYs_SUPPORTED)
+extern PNG_EXPORT(void,png_set_pHYs) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 res_x, png_uint_32 res_y, int unit_type));
+#endif
+
+extern PNG_EXPORT(png_uint_32,png_get_PLTE) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_colorp *palette, int *num_palette));
+
+extern PNG_EXPORT(void,png_set_PLTE) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_colorp palette, int num_palette));
+
+#if defined(PNG_sBIT_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_sBIT) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_color_8p *sig_bit));
+#endif
+
+#if defined(PNG_sBIT_SUPPORTED)
+extern PNG_EXPORT(void,png_set_sBIT) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_color_8p sig_bit));
+#endif
+
+#if defined(PNG_sRGB_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_sRGB) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int *intent));
+#endif
+
+#if defined(PNG_sRGB_SUPPORTED)
+extern PNG_EXPORT(void,png_set_sRGB) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int intent));
+extern PNG_EXPORT(void,png_set_sRGB_gAMA_and_cHRM) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int intent));
+#endif
+
+#if defined(PNG_iCCP_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_iCCP) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_charpp name, int *compression_type,
+   png_charpp profile, png_uint_32 *proflen));
+   /* Note to maintainer: profile should be png_bytepp */
+#endif
+
+#if defined(PNG_iCCP_SUPPORTED)
+extern PNG_EXPORT(void,png_set_iCCP) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_charp name, int compression_type,
+   png_charp profile, png_uint_32 proflen));
+   /* Note to maintainer: profile should be png_bytep */
+#endif
+
+#if defined(PNG_sPLT_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_sPLT) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_sPLT_tpp entries));
+#endif
+
+#if defined(PNG_sPLT_SUPPORTED)
+extern PNG_EXPORT(void,png_set_sPLT) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_sPLT_tp entries, int nentries));
+#endif
+
+#if defined(PNG_TEXT_SUPPORTED)
+/* png_get_text also returns the number of text chunks in *num_text */
+extern PNG_EXPORT(png_uint_32,png_get_text) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_textp *text_ptr, int *num_text));
+#endif
+
+/*
+ *  Note while png_set_text() will accept a structure whose text,
+ *  language, and  translated keywords are NULL pointers, the structure
+ *  returned by png_get_text will always contain regular
+ *  zero-terminated C strings.  They might be empty strings but
+ *  they will never be NULL pointers.
+ */
+
+#if defined(PNG_TEXT_SUPPORTED)
+extern PNG_EXPORT(void,png_set_text) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_textp text_ptr, int num_text));
+#endif
+
+#if defined(PNG_tIME_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_tIME) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_timep *mod_time));
+#endif
+
+#if defined(PNG_tIME_SUPPORTED)
+extern PNG_EXPORT(void,png_set_tIME) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_timep mod_time));
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED)
+extern PNG_EXPORT(png_uint_32,png_get_tRNS) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_bytep *trans, int *num_trans,
+   png_color_16p *trans_values));
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED)
+extern PNG_EXPORT(void,png_set_tRNS) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_bytep trans, int num_trans,
+   png_color_16p trans_values));
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED)
+#endif
+
+#if defined(PNG_sCAL_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(png_uint_32,png_get_sCAL) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int *unit, double *width, double *height));
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+extern PNG_EXPORT(png_uint_32,png_get_sCAL_s) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int *unit, png_charpp swidth, png_charpp sheight));
+#endif
+#endif
+#endif /* PNG_sCAL_SUPPORTED */
+
+#if defined(PNG_sCAL_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_sCAL) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int unit, double width, double height));
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+extern PNG_EXPORT(void,png_set_sCAL_s) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int unit, png_charp swidth, png_charp sheight));
+#endif
+#endif /* PNG_sCAL_SUPPORTED || PNG_WRITE_sCAL_SUPPORTED */
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+/* provide a list of chunks and how they are to be handled, if the built-in
+   handling or default unknown chunk handling is not desired.  Any chunks not
+   listed will be handled in the default manner.  The IHDR and IEND chunks
+   must not be listed.
+      keep = 0: follow default behavour
+           = 1: do not keep
+           = 2: keep only if safe-to-copy
+           = 3: keep even if unsafe-to-copy
+*/
+extern PNG_EXPORT(void, png_set_keep_unknown_chunks) PNGARG((png_structp
+   png_ptr, int keep, png_bytep chunk_list, int num_chunks));
+extern PNG_EXPORT(void, png_set_unknown_chunks) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_unknown_chunkp unknowns, int num_unknowns));
+extern PNG_EXPORT(void, png_set_unknown_chunk_location)
+   PNGARG((png_structp png_ptr, png_infop info_ptr, int chunk, int location));
+extern PNG_EXPORT(png_uint_32,png_get_unknown_chunks) PNGARG((png_structp
+   png_ptr, png_infop info_ptr, png_unknown_chunkpp entries));
+#endif
+
+/* Png_free_data() will turn off the "valid" flag for anything it frees.
+   If you need to turn it off for a chunk that your application has freed,
+   you can use png_set_invalid(png_ptr, info_ptr, PNG_INFO_CHNK); */
+extern PNG_EXPORT(void, png_set_invalid) PNGARG((png_structp png_ptr,
+   png_infop info_ptr, int mask));
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+/* The "params" pointer is currently not used and is for future expansion. */
+extern PNG_EXPORT(void, png_read_png) PNGARG((png_structp png_ptr,
+                        png_infop info_ptr,
+                        int transforms,
+                        png_voidp params));
+extern PNG_EXPORT(void, png_write_png) PNGARG((png_structp png_ptr,
+                        png_infop info_ptr,
+                        int transforms,
+                        png_voidp params));
+#endif
+
+/* Define PNG_DEBUG at compile time for debugging information.  Higher
+ * numbers for PNG_DEBUG mean more debugging information.  This has
+ * only been added since version 0.95 so it is not implemented throughout
+ * libpng yet, but more support will be added as needed.
+ */
+#ifdef PNG_DEBUG
+#if (PNG_DEBUG > 0)
+#if !defined(PNG_DEBUG_FILE) && defined(_MSC_VER)
+#include <crtdbg.h>
+#if (PNG_DEBUG > 1)
+#define png_debug(l,m)  _RPT0(_CRT_WARN,m)
+#define png_debug1(l,m,p1)  _RPT1(_CRT_WARN,m,p1)
+#define png_debug2(l,m,p1,p2) _RPT2(_CRT_WARN,m,p1,p2)
+#endif
+#else /* PNG_DEBUG_FILE || !_MSC_VER */
+#ifndef PNG_DEBUG_FILE
+#define PNG_DEBUG_FILE stderr
+#endif /* PNG_DEBUG_FILE */
+#if (PNG_DEBUG > 1)
+#define png_debug(l,m) \
+{ \
+     int num_tabs=l; \
+     fprintf(PNG_DEBUG_FILE,"%s"m,(num_tabs==1 ? "\t" : \
+       (num_tabs==2 ? "\t\t":(num_tabs>2 ? "\t\t\t":"")))); \
+}
+#define png_debug1(l,m,p1) \
+{ \
+     int num_tabs=l; \
+     fprintf(PNG_DEBUG_FILE,"%s"m,(num_tabs==1 ? "\t" : \
+       (num_tabs==2 ? "\t\t":(num_tabs>2 ? "\t\t\t":""))),p1); \
+}
+#define png_debug2(l,m,p1,p2) \
+{ \
+     int num_tabs=l; \
+     fprintf(PNG_DEBUG_FILE,"%s"m,(num_tabs==1 ? "\t" : \
+       (num_tabs==2 ? "\t\t":(num_tabs>2 ? "\t\t\t":""))),p1,p2); \
+}
+#endif /* (PNG_DEBUG > 1) */
+#endif /* _MSC_VER */
+#endif /* (PNG_DEBUG > 0) */
+#endif /* PNG_DEBUG */
+#ifndef png_debug
+#define png_debug(l, m)
+#endif
+#ifndef png_debug1
+#define png_debug1(l, m, p1)
+#endif
+#ifndef png_debug2
+#define png_debug2(l, m, p1, p2)
+#endif
+
+extern PNG_EXPORT(png_bytep,png_sig_bytes) PNGARG((void));
+
+extern PNG_EXPORT(png_charp,png_get_copyright) PNGARG((png_structp png_ptr));
+extern PNG_EXPORT(png_charp,png_get_header_ver) PNGARG((png_structp png_ptr));
+extern PNG_EXPORT(png_charp,png_get_header_version) PNGARG((png_structp png_ptr));
+extern PNG_EXPORT(png_charp,png_get_libpng_ver) PNGARG((png_structp png_ptr));
+
+#ifdef PNG_MNG_FEATURES_SUPPORTED
+extern PNG_EXPORT(png_uint_32,png_permit_mng_features) PNGARG((png_structp
+   png_ptr, png_uint_32 mng_features_permitted));
+#endif
+
+/* Added to version 1.2.0 */
+#if defined(PNG_ASSEMBLER_CODE_SUPPORTED)
+#define PNG_ASM_FLAG_MMX_SUPPORT_COMPILED  0x01		/* not user-settable */
+#define PNG_ASM_FLAG_MMX_SUPPORT_IN_CPU    0x02		/* not user-settable */
+#define PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  0x04
+#define PNG_ASM_FLAG_MMX_READ_INTERLACE    0x08
+#define PNG_ASM_FLAG_MMX_READ_FILTER_SUB   0x10
+#define PNG_ASM_FLAG_MMX_READ_FILTER_UP    0x20
+#define PNG_ASM_FLAG_MMX_READ_FILTER_AVG   0x40
+#define PNG_ASM_FLAG_MMX_READ_FILTER_PAETH 0x80
+#define PNG_ASM_FLAGS_INITIALIZED          0x80000000	/* not user-settable */
+
+#define PNG_MMX_READ_FLAGS ( PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  \
+                           | PNG_ASM_FLAG_MMX_READ_INTERLACE    \
+                           | PNG_ASM_FLAG_MMX_READ_FILTER_SUB   \
+                           | PNG_ASM_FLAG_MMX_READ_FILTER_UP    \
+                           | PNG_ASM_FLAG_MMX_READ_FILTER_AVG   \
+                           | PNG_ASM_FLAG_MMX_READ_FILTER_PAETH )
+#define PNG_MMX_WRITE_FLAGS ( 0 )
+
+#define PNG_MMX_FLAGS ( PNG_ASM_FLAG_MMX_SUPPORT_COMPILED \
+                      | PNG_ASM_FLAG_MMX_SUPPORT_IN_CPU   \
+                      | PNG_MMX_READ_FLAGS                \
+                      | PNG_MMX_WRITE_FLAGS )
+
+#define PNG_SELECT_READ   1
+#define PNG_SELECT_WRITE  2
+
+
+/* pngget.c */
+extern PNG_EXPORT(png_uint_32,png_get_mmx_flagmask)
+   PNGARG((int flag_select, int *compilerID));
+
+/* pngget.c */
+extern PNG_EXPORT(png_uint_32,png_get_asm_flagmask)
+   PNGARG((int flag_select));
+
+/* pngget.c */
+extern PNG_EXPORT(png_uint_32,png_get_asm_flags)
+   PNGARG((png_structp png_ptr));
+
+/* pngget.c */
+extern PNG_EXPORT(png_byte,png_get_mmx_bitdepth_threshold)
+   PNGARG((png_structp png_ptr));
+
+/* pngget.c */
+extern PNG_EXPORT(png_uint_32,png_get_mmx_rowbytes_threshold)
+   PNGARG((png_structp png_ptr));
+
+/* pngset.c */
+extern PNG_EXPORT(void,png_set_asm_flags)
+   PNGARG((png_structp png_ptr, png_uint_32 asm_flags));
+
+/* pngset.c */
+extern PNG_EXPORT(void,png_set_mmx_thresholds)
+   PNGARG((png_structp png_ptr, png_byte mmx_bitdepth_threshold,
+   png_uint_32 mmx_rowbytes_threshold));
+
+#endif /* PNG_ASSEMBLER_CODE_SUPPORTED */
+
+/* png.c, pnggccrd.c, or pngvcrd.c */
+extern PNG_EXPORT(int,png_mmx_support) PNGARG((void));
+
+/* Strip the prepended error numbers ("#nnn ") from error and warning
+ * messages before passing them to the error or warning handler. */
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+extern PNG_EXPORT(void,png_set_strip_error_numbers) PNGARG((png_structp
+   png_ptr, png_uint_32 strip_mode));
+#endif
+
+/* Maintainer: Put new public prototypes here ^, in libpng.3, and project defs */
+
+#define PNG_HEADER_VERSION_STRING \
+   " libpng version 1.2.1 - December 12, 2001 (header)\n"
+
+#ifdef PNG_READ_COMPOSITE_NODIV_SUPPORTED
+/* With these routines we avoid an integer divide, which will be slower on
+ * most machines.  However, it does take more operations than the corresponding
+ * divide method, so it may be slower on a few RISC systems.  There are two
+ * shifts (by 8 or 16 bits) and an addition, versus a single integer divide.
+ *
+ * Note that the rounding factors are NOT supposed to be the same!  128 and
+ * 32768 are correct for the NODIV code; 127 and 32767 are correct for the
+ * standard method.
+ *
+ * [Optimized code by Greg Roelofs and Mark Adler...blame us for bugs. :-) ]
+ */
+
+ /* fg and bg should be in `gamma 1.0' space; alpha is the opacity          */
+
+#  define png_composite(composite, fg, alpha, bg)                            \
+     { png_uint_16 temp = (png_uint_16)((png_uint_16)(fg) * (png_uint_16)(alpha) \
+                        +        (png_uint_16)(bg)*(png_uint_16)(255 -       \
+                        (png_uint_16)(alpha)) + (png_uint_16)128);           \
+       (composite) = (png_byte)((temp + (temp >> 8)) >> 8); }
+
+#  define png_composite_16(composite, fg, alpha, bg)                         \
+     { png_uint_32 temp = (png_uint_32)((png_uint_32)(fg) * (png_uint_32)(alpha) \
+                        + (png_uint_32)(bg)*(png_uint_32)(65535L -           \
+                        (png_uint_32)(alpha)) + (png_uint_32)32768L);        \
+       (composite) = (png_uint_16)((temp + (temp >> 16)) >> 16); }
+
+#else  /* standard method using integer division */
+
+#  define png_composite(composite, fg, alpha, bg)                            \
+     (composite) = (png_byte)(((png_uint_16)(fg) * (png_uint_16)(alpha) +    \
+       (png_uint_16)(bg) * (png_uint_16)(255 - (png_uint_16)(alpha)) +       \
+       (png_uint_16)127) / 255)
+
+#  define png_composite_16(composite, fg, alpha, bg)                         \
+     (composite) = (png_uint_16)(((png_uint_32)(fg) * (png_uint_32)(alpha) + \
+       (png_uint_32)(bg)*(png_uint_32)(65535L - (png_uint_32)(alpha)) +      \
+       (png_uint_32)32767) / (png_uint_32)65535L)
+
+#endif /* PNG_READ_COMPOSITE_NODIV_SUPPORTED */
+
+/* These next functions are used internally in the code.  They generally
+ * shouldn't be used unless you are writing code to add or replace some
+ * functionality in libpng.  More information about most functions can
+ * be found in the files where the functions are located.
+ */
+
+#if defined(PNG_INTERNAL)
+
+/* Various modes of operation.  Note that after an init, mode is set to
+ * zero automatically when the structure is created.
+ */
+#define PNG_HAVE_IHDR               0x01
+#define PNG_HAVE_PLTE               0x02
+#define PNG_HAVE_IDAT               0x04
+#define PNG_AFTER_IDAT              0x08
+#define PNG_HAVE_IEND               0x10
+#define PNG_HAVE_gAMA               0x20
+#define PNG_HAVE_cHRM               0x40
+#define PNG_HAVE_sRGB               0x80
+#define PNG_HAVE_CHUNK_HEADER      0x100
+#define PNG_WROTE_tIME             0x200
+#define PNG_WROTE_INFO_BEFORE_PLTE 0x400
+#define PNG_BACKGROUND_IS_GRAY     0x800
+#define PNG_HAVE_PNG_SIGNATURE    0x1000
+
+/* flags for the transformations the PNG library does on the image data */
+#define PNG_BGR                0x0001
+#define PNG_INTERLACE          0x0002
+#define PNG_PACK               0x0004
+#define PNG_SHIFT              0x0008
+#define PNG_SWAP_BYTES         0x0010
+#define PNG_INVERT_MONO        0x0020
+#define PNG_DITHER             0x0040
+#define PNG_BACKGROUND         0x0080
+#define PNG_BACKGROUND_EXPAND  0x0100
+                          /*   0x0200 unused */
+#define PNG_16_TO_8            0x0400
+#define PNG_RGBA               0x0800
+#define PNG_EXPAND             0x1000
+#define PNG_GAMMA              0x2000
+#define PNG_GRAY_TO_RGB        0x4000
+#define PNG_FILLER             0x8000L
+#define PNG_PACKSWAP          0x10000L
+#define PNG_SWAP_ALPHA        0x20000L
+#define PNG_STRIP_ALPHA       0x40000L
+#define PNG_INVERT_ALPHA      0x80000L
+#define PNG_USER_TRANSFORM   0x100000L
+#define PNG_RGB_TO_GRAY_ERR  0x200000L
+#define PNG_RGB_TO_GRAY_WARN 0x400000L
+#define PNG_RGB_TO_GRAY      0x600000L  /* two bits, RGB_TO_GRAY_ERR|WARN */
+
+/* flags for png_create_struct */
+#define PNG_STRUCT_PNG   0x0001
+#define PNG_STRUCT_INFO  0x0002
+
+/* Scaling factor for filter heuristic weighting calculations */
+#define PNG_WEIGHT_SHIFT 8
+#define PNG_WEIGHT_FACTOR (1<<(PNG_WEIGHT_SHIFT))
+#define PNG_COST_SHIFT 3
+#define PNG_COST_FACTOR (1<<(PNG_COST_SHIFT))
+
+/* flags for the png_ptr->flags rather than declaring a byte for each one */
+#define PNG_FLAG_ZLIB_CUSTOM_STRATEGY     0x0001
+#define PNG_FLAG_ZLIB_CUSTOM_LEVEL        0x0002
+#define PNG_FLAG_ZLIB_CUSTOM_MEM_LEVEL    0x0004
+#define PNG_FLAG_ZLIB_CUSTOM_WINDOW_BITS  0x0008
+#define PNG_FLAG_ZLIB_CUSTOM_METHOD       0x0010
+#define PNG_FLAG_ZLIB_FINISHED            0x0020
+#define PNG_FLAG_ROW_INIT                 0x0040
+#define PNG_FLAG_FILLER_AFTER             0x0080
+#define PNG_FLAG_CRC_ANCILLARY_USE        0x0100
+#define PNG_FLAG_CRC_ANCILLARY_NOWARN     0x0200
+#define PNG_FLAG_CRC_CRITICAL_USE         0x0400
+#define PNG_FLAG_CRC_CRITICAL_IGNORE      0x0800
+#define PNG_FLAG_FREE_PLTE                0x1000
+#define PNG_FLAG_FREE_TRNS                0x2000
+#define PNG_FLAG_FREE_HIST                0x4000
+#define PNG_FLAG_KEEP_UNKNOWN_CHUNKS      0x8000L
+#define PNG_FLAG_KEEP_UNSAFE_CHUNKS       0x10000L
+#define PNG_FLAG_LIBRARY_MISMATCH         0x20000L
+#define PNG_FLAG_STRIP_ERROR_NUMBERS      0x40000L
+#define PNG_FLAG_STRIP_ERROR_TEXT         0x80000L
+
+/* For use in png_set_keep_unknown, png_handle_as_unknown */
+#define HANDLE_CHUNK_AS_DEFAULT   0
+#define HANDLE_CHUNK_NEVER        1
+#define HANDLE_CHUNK_IF_SAFE      2
+#define HANDLE_CHUNK_ALWAYS       3
+
+#define PNG_FLAG_CRC_ANCILLARY_MASK (PNG_FLAG_CRC_ANCILLARY_USE | \
+                                     PNG_FLAG_CRC_ANCILLARY_NOWARN)
+
+#define PNG_FLAG_CRC_CRITICAL_MASK  (PNG_FLAG_CRC_CRITICAL_USE | \
+                                     PNG_FLAG_CRC_CRITICAL_IGNORE)
+
+#define PNG_FLAG_CRC_MASK           (PNG_FLAG_CRC_ANCILLARY_MASK | \
+                                     PNG_FLAG_CRC_CRITICAL_MASK)
+
+/* save typing and make code easier to understand */
+#define PNG_COLOR_DIST(c1, c2) (abs((int)((c1).red) - (int)((c2).red)) + \
+   abs((int)((c1).green) - (int)((c2).green)) + \
+   abs((int)((c1).blue) - (int)((c2).blue)))
+
+/* variables declared in png.c - only it needs to define PNG_NO_EXTERN */
+#if !defined(PNG_NO_EXTERN) || defined(PNG_ALWAYS_EXTERN)
+/* place to hold the signature string for a PNG file. */
+#ifdef PNG_USE_GLOBAL_ARRAYS
+   PNG_EXPORT_VAR (const png_byte FARDATA) png_sig[8];
+#else
+#define png_sig png_sig_bytes(NULL)
+#endif
+#endif /* PNG_NO_EXTERN */
+
+/* Constant strings for known chunk types.  If you need to add a chunk,
+ * define the name here, and add an invocation of the macro in png.c and
+ * wherever it's needed.
+ */
+#define PNG_IHDR const png_byte png_IHDR[5] = { 73,  72,  68,  82, '\0'}
+#define PNG_IDAT const png_byte png_IDAT[5] = { 73,  68,  65,  84, '\0'}
+#define PNG_IEND const png_byte png_IEND[5] = { 73,  69,  78,  68, '\0'}
+#define PNG_PLTE const png_byte png_PLTE[5] = { 80,  76,  84,  69, '\0'}
+#define PNG_bKGD const png_byte png_bKGD[5] = { 98,  75,  71,  68, '\0'}
+#define PNG_cHRM const png_byte png_cHRM[5] = { 99,  72,  82,  77, '\0'}
+#define PNG_gAMA const png_byte png_gAMA[5] = {103,  65,  77,  65, '\0'}
+#define PNG_hIST const png_byte png_hIST[5] = {104,  73,  83,  84, '\0'}
+#define PNG_iCCP const png_byte png_iCCP[5] = {105,  67,  67,  80, '\0'}
+#define PNG_iTXt const png_byte png_iTXt[5] = {105,  84,  88, 116, '\0'}
+#define PNG_oFFs const png_byte png_oFFs[5] = {111,  70,  70, 115, '\0'}
+#define PNG_pCAL const png_byte png_pCAL[5] = {112,  67,  65,  76, '\0'}
+#define PNG_sCAL const png_byte png_sCAL[5] = {115,  67,  65,  76, '\0'}
+#define PNG_pHYs const png_byte png_pHYs[5] = {112,  72,  89, 115, '\0'}
+#define PNG_sBIT const png_byte png_sBIT[5] = {115,  66,  73,  84, '\0'}
+#define PNG_sPLT const png_byte png_sPLT[5] = {115,  80,  76,  84, '\0'}
+#define PNG_sRGB const png_byte png_sRGB[5] = {115,  82,  71,  66, '\0'}
+#define PNG_tEXt const png_byte png_tEXt[5] = {116,  69,  88, 116, '\0'}
+#define PNG_tIME const png_byte png_tIME[5] = {116,  73,  77,  69, '\0'}
+#define PNG_tRNS const png_byte png_tRNS[5] = {116,  82,  78,  83, '\0'}
+#define PNG_zTXt const png_byte png_zTXt[5] = {122,  84,  88, 116, '\0'}
+
+#ifdef PNG_USE_GLOBAL_ARRAYS
+PNG_EXPORT_VAR (const png_byte FARDATA) png_IHDR[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_IDAT[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_IEND[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_PLTE[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_bKGD[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_cHRM[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_gAMA[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_hIST[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_iCCP[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_iTXt[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_oFFs[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_pCAL[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_sCAL[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_pHYs[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_sBIT[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_sPLT[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_sRGB[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_tEXt[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_tIME[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_tRNS[5];
+PNG_EXPORT_VAR (const png_byte FARDATA) png_zTXt[5];
+#endif /* PNG_USE_GLOBAL_ARRAYS */
+
+
+/* Inline macros to do direct reads of bytes from the input buffer.  These
+ * require that you are using an architecture that uses PNG byte ordering
+ * (MSB first) and supports unaligned data storage.  I think that PowerPC
+ * in big-endian mode and 680x0 are the only ones that will support this.
+ * The x86 line of processors definitely do not.  The png_get_int_32()
+ * routine also assumes we are using two's complement format for negative
+ * values, which is almost certainly true.
+ */
+#if defined(PNG_READ_BIG_ENDIAN_SUPPORTED)
+#  if defined(PNG_pCAL_SUPPORTED) || defined(PNG_oFFs_SUPPORTED)
+#    define png_get_int_32(buf) ( *((png_int_32p) (buf)))
+#  endif
+#  define png_get_uint_32(buf) ( *((png_uint_32p) (buf)))
+#  define png_get_uint_16(buf) ( *((png_uint_16p) (buf)))
+#else
+#  if defined(PNG_pCAL_SUPPORTED) || defined(PNG_oFFs_SUPPORTED)
+PNG_EXTERN png_int_32 png_get_int_32 PNGARG((png_bytep buf));
+#  endif
+PNG_EXTERN png_uint_32 png_get_uint_32 PNGARG((png_bytep buf));
+PNG_EXTERN png_uint_16 png_get_uint_16 PNGARG((png_bytep buf));
+#endif /* !PNG_READ_BIG_ENDIAN_SUPPORTED */
+
+/* Initialize png_ptr struct for reading, and allocate any other memory.
+ * (old interface - DEPRECATED - use png_create_read_struct instead).
+ */
+extern PNG_EXPORT(void,png_read_init) PNGARG((png_structp png_ptr));
+#define png_read_init(png_ptr) png_read_init_3(&png_ptr, \
+    PNG_LIBPNG_VER_STRING,  sizeof(png_struct));
+extern PNG_EXPORT(void,png_read_init_3) PNGARG((png_structpp ptr_ptr,
+    png_const_charp user_png_ver, png_size_t png_struct_size));
+extern PNG_EXPORT(void,png_read_init_2) PNGARG((png_structp png_ptr,
+    png_const_charp user_png_ver, png_size_t png_struct_size, png_size_t
+    png_info_size));
+
+/* Initialize png_ptr struct for writing, and allocate any other memory.
+ * (old interface - DEPRECATED - use png_create_write_struct instead).
+ */
+extern PNG_EXPORT(void,png_write_init) PNGARG((png_structp png_ptr));
+#define png_write_init(png_ptr) png_write_init_3(&png_ptr, \
+    PNG_LIBPNG_VER_STRING, sizeof(png_struct));
+extern PNG_EXPORT(void,png_write_init_3) PNGARG((png_structpp ptr_ptr,
+    png_const_charp user_png_ver, png_size_t png_struct_size));
+extern PNG_EXPORT(void,png_write_init_2) PNGARG((png_structp png_ptr,
+    png_const_charp user_png_ver, png_size_t png_struct_size, png_size_t
+    png_info_size));
+
+/* Allocate memory for an internal libpng struct */
+PNG_EXTERN png_voidp png_create_struct PNGARG((int type));
+
+/* Free memory from internal libpng struct */
+PNG_EXTERN void png_destroy_struct PNGARG((png_voidp struct_ptr));
+
+PNG_EXTERN png_voidp png_create_struct_2 PNGARG((int type, png_malloc_ptr
+  malloc_fn, png_voidp mem_ptr));
+PNG_EXTERN void png_destroy_struct_2 PNGARG((png_voidp struct_ptr,
+   png_free_ptr free_fn, png_voidp mem_ptr));
+
+/* Free any memory that info_ptr points to and reset struct. */
+PNG_EXTERN void png_info_destroy PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+/* Function to allocate memory for zlib. */
+PNG_EXTERN voidpf png_zalloc PNGARG((voidpf png_ptr, uInt items, uInt size));
+
+/* Function to free memory for zlib */
+PNG_EXTERN void png_zfree PNGARG((voidpf png_ptr, voidpf ptr));
+
+/* Reset the CRC variable */
+PNG_EXTERN void png_reset_crc PNGARG((png_structp png_ptr));
+
+/* Write the "data" buffer to whatever output you are using. */
+PNG_EXTERN void png_write_data PNGARG((png_structp png_ptr, png_bytep data,
+   png_size_t length));
+
+/* Read data from whatever input you are using into the "data" buffer */
+PNG_EXTERN void png_read_data PNGARG((png_structp png_ptr, png_bytep data,
+   png_size_t length));
+
+/* Read bytes into buf, and update png_ptr->crc */
+PNG_EXTERN void png_crc_read PNGARG((png_structp png_ptr, png_bytep buf,
+   png_size_t length));
+
+/* Decompress data in a chunk that uses compression */
+#if defined(PNG_zTXt_SUPPORTED) || defined(PNG_iTXt_SUPPORTED) || \
+    defined(PNG_iCCP_SUPPORTED) || defined(PNG_sPLT_SUPPORTED)
+PNG_EXTERN png_charp png_decompress_chunk PNGARG((png_structp png_ptr,
+   int comp_type, png_charp chunkdata, png_size_t chunklength,
+   png_size_t prefix_length, png_size_t *data_length));
+#endif
+
+/* Read "skip" bytes, read the file crc, and (optionally) verify png_ptr->crc */
+PNG_EXTERN int png_crc_finish PNGARG((png_structp png_ptr, png_uint_32 skip));
+
+/* Read the CRC from the file and compare it to the libpng calculated CRC */
+PNG_EXTERN int png_crc_error PNGARG((png_structp png_ptr));
+
+/* Calculate the CRC over a section of data.  Note that we are only
+ * passing a maximum of 64K on systems that have this as a memory limit,
+ * since this is the maximum buffer size we can specify.
+ */
+PNG_EXTERN void png_calculate_crc PNGARG((png_structp png_ptr, png_bytep ptr,
+   png_size_t length));
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+PNG_EXTERN void png_flush PNGARG((png_structp png_ptr));
+#endif
+
+/* Place a 32-bit number into a buffer in PNG byte order (big-endian).
+ * The only currently known PNG chunks that use signed numbers are
+ * the ancillary extension chunks, oFFs and pCAL.
+ */
+PNG_EXTERN void png_save_uint_32 PNGARG((png_bytep buf, png_uint_32 i));
+
+#if defined(PNG_WRITE_pCAL_SUPPORTED) || defined(PNG_WRITE_oFFs_SUPPORTED)
+PNG_EXTERN void png_save_int_32 PNGARG((png_bytep buf, png_int_32 i));
+#endif
+
+/* Place a 16-bit number into a buffer in PNG byte order.
+ * The parameter is declared unsigned int, not png_uint_16,
+ * just to avoid potential problems on pre-ANSI C compilers.
+ */
+PNG_EXTERN void png_save_uint_16 PNGARG((png_bytep buf, unsigned int i));
+
+/* simple function to write the signature */
+PNG_EXTERN void png_write_sig PNGARG((png_structp png_ptr));
+
+/* write various chunks */
+
+/* Write the IHDR chunk, and update the png_struct with the necessary
+ * information.
+ */
+PNG_EXTERN void png_write_IHDR PNGARG((png_structp png_ptr, png_uint_32 width,
+   png_uint_32 height,
+   int bit_depth, int color_type, int compression_method, int filter_method,
+   int interlace_method));
+
+PNG_EXTERN void png_write_PLTE PNGARG((png_structp png_ptr, png_colorp palette,
+   png_uint_32 num_pal));
+
+PNG_EXTERN void png_write_IDAT PNGARG((png_structp png_ptr, png_bytep data,
+   png_size_t length));
+
+PNG_EXTERN void png_write_IEND PNGARG((png_structp png_ptr));
+
+#if defined(PNG_WRITE_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+PNG_EXTERN void png_write_gAMA PNGARG((png_structp png_ptr, double file_gamma));
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+PNG_EXTERN void png_write_gAMA_fixed PNGARG((png_structp png_ptr, png_fixed_point
+    file_gamma));
+#endif
+#endif
+
+#if defined(PNG_WRITE_sBIT_SUPPORTED)
+PNG_EXTERN void png_write_sBIT PNGARG((png_structp png_ptr, png_color_8p sbit,
+   int color_type));
+#endif
+
+#if defined(PNG_WRITE_cHRM_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+PNG_EXTERN void png_write_cHRM PNGARG((png_structp png_ptr,
+   double white_x, double white_y,
+   double red_x, double red_y, double green_x, double green_y,
+   double blue_x, double blue_y));
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+PNG_EXTERN void png_write_cHRM_fixed PNGARG((png_structp png_ptr,
+   png_fixed_point int_white_x, png_fixed_point int_white_y,
+   png_fixed_point int_red_x, png_fixed_point int_red_y, png_fixed_point
+   int_green_x, png_fixed_point int_green_y, png_fixed_point int_blue_x,
+   png_fixed_point int_blue_y));
+#endif
+#endif
+
+#if defined(PNG_WRITE_sRGB_SUPPORTED)
+PNG_EXTERN void png_write_sRGB PNGARG((png_structp png_ptr,
+   int intent));
+#endif
+
+#if defined(PNG_WRITE_iCCP_SUPPORTED)
+PNG_EXTERN void png_write_iCCP PNGARG((png_structp png_ptr,
+   png_charp name, int compression_type,
+   png_charp profile, int proflen));
+   /* Note to maintainer: profile should be png_bytep */
+#endif
+
+#if defined(PNG_WRITE_sPLT_SUPPORTED)
+PNG_EXTERN void png_write_sPLT PNGARG((png_structp png_ptr,
+   png_sPLT_tp palette));
+#endif
+
+#if defined(PNG_WRITE_tRNS_SUPPORTED)
+PNG_EXTERN void png_write_tRNS PNGARG((png_structp png_ptr, png_bytep trans,
+   png_color_16p values, int number, int color_type));
+#endif
+
+#if defined(PNG_WRITE_bKGD_SUPPORTED)
+PNG_EXTERN void png_write_bKGD PNGARG((png_structp png_ptr,
+   png_color_16p values, int color_type));
+#endif
+
+#if defined(PNG_WRITE_hIST_SUPPORTED)
+PNG_EXTERN void png_write_hIST PNGARG((png_structp png_ptr, png_uint_16p hist,
+   int num_hist));
+#endif
+
+#if defined(PNG_WRITE_TEXT_SUPPORTED) || defined(PNG_WRITE_pCAL_SUPPORTED) || \
+    defined(PNG_WRITE_iCCP_SUPPORTED) || defined(PNG_WRITE_sPLT_SUPPORTED)
+PNG_EXTERN png_size_t png_check_keyword PNGARG((png_structp png_ptr,
+   png_charp key, png_charpp new_key));
+#endif
+
+#if defined(PNG_WRITE_tEXt_SUPPORTED)
+PNG_EXTERN void png_write_tEXt PNGARG((png_structp png_ptr, png_charp key,
+   png_charp text, png_size_t text_len));
+#endif
+
+#if defined(PNG_WRITE_zTXt_SUPPORTED)
+PNG_EXTERN void png_write_zTXt PNGARG((png_structp png_ptr, png_charp key,
+   png_charp text, png_size_t text_len, int compression));
+#endif
+
+#if defined(PNG_WRITE_iTXt_SUPPORTED)
+PNG_EXTERN void png_write_iTXt PNGARG((png_structp png_ptr,
+   int compression, png_charp key, png_charp lang, png_charp lang_key,
+   png_charp text));
+#endif
+
+#if defined(PNG_WRITE_oFFs_SUPPORTED)
+PNG_EXTERN void png_write_oFFs PNGARG((png_structp png_ptr,
+   png_int_32 x_offset, png_int_32 y_offset, int unit_type));
+#endif
+
+#if defined(PNG_WRITE_pCAL_SUPPORTED)
+PNG_EXTERN void png_write_pCAL PNGARG((png_structp png_ptr, png_charp purpose,
+   png_int_32 X0, png_int_32 X1, int type, int nparams,
+   png_charp units, png_charpp params));
+#endif
+
+#if defined(PNG_WRITE_pHYs_SUPPORTED)
+PNG_EXTERN void png_write_pHYs PNGARG((png_structp png_ptr,
+   png_uint_32 x_pixels_per_unit, png_uint_32 y_pixels_per_unit,
+   int unit_type));
+#endif
+
+#if defined(PNG_WRITE_tIME_SUPPORTED)
+PNG_EXTERN void png_write_tIME PNGARG((png_structp png_ptr,
+   png_timep mod_time));
+#endif
+
+#if defined(PNG_WRITE_sCAL_SUPPORTED)
+#if defined(PNG_FLOATING_POINT_SUPPORTED) && !defined(PNG_NO_STDIO)
+PNG_EXTERN void png_write_sCAL PNGARG((png_structp png_ptr,
+   int unit, double width, double height));
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+PNG_EXTERN void png_write_sCAL_s PNGARG((png_structp png_ptr,
+   int unit, png_charp width, png_charp height));
+#endif
+#endif
+#endif
+
+/* Called when finished processing a row of data */
+PNG_EXTERN void png_write_finish_row PNGARG((png_structp png_ptr));
+
+/* Internal use only.   Called before first row of data */
+PNG_EXTERN void png_write_start_row PNGARG((png_structp png_ptr));
+
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+PNG_EXTERN void png_build_gamma_table PNGARG((png_structp png_ptr));
+#endif
+
+/* combine a row of data, dealing with alpha, etc. if requested */
+PNG_EXTERN void png_combine_row PNGARG((png_structp png_ptr, png_bytep row,
+   int mask));
+
+#if defined(PNG_READ_INTERLACING_SUPPORTED)
+/* expand an interlaced row */
+/* OLD pre-1.0.9 interface:
+PNG_EXTERN void png_do_read_interlace PNGARG((png_row_infop row_info,
+   png_bytep row, int pass, png_uint_32 transformations));
+ */
+PNG_EXTERN void png_do_read_interlace PNGARG((png_structp png_ptr));
+#endif
+
+/* GRR TO DO (2.0 or whenever):  simplify other internal calling interfaces */
+
+#if defined(PNG_WRITE_INTERLACING_SUPPORTED)
+/* grab pixels out of a row for an interlaced pass */
+PNG_EXTERN void png_do_write_interlace PNGARG((png_row_infop row_info,
+   png_bytep row, int pass));
+#endif
+
+/* unfilter a row */
+PNG_EXTERN void png_read_filter_row PNGARG((png_structp png_ptr,
+   png_row_infop row_info, png_bytep row, png_bytep prev_row, int filter));
+
+/* Choose the best filter to use and filter the row data */
+PNG_EXTERN void png_write_find_filter PNGARG((png_structp png_ptr,
+   png_row_infop row_info));
+
+/* Write out the filtered row. */
+PNG_EXTERN void png_write_filtered_row PNGARG((png_structp png_ptr,
+   png_bytep filtered_row));
+/* finish a row while reading, dealing with interlacing passes, etc. */
+PNG_EXTERN void png_read_finish_row PNGARG((png_structp png_ptr));
+
+/* initialize the row buffers, etc. */
+PNG_EXTERN void png_read_start_row PNGARG((png_structp png_ptr));
+/* optional call to update the users info structure */
+PNG_EXTERN void png_read_transform_info PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+
+/* these are the functions that do the transformations */
+#if defined(PNG_READ_FILLER_SUPPORTED)
+PNG_EXTERN void png_do_read_filler PNGARG((png_row_infop row_info,
+   png_bytep row, png_uint_32 filler, png_uint_32 flags));
+#endif
+
+#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED)
+PNG_EXTERN void png_do_read_swap_alpha PNGARG((png_row_infop row_info,
+   png_bytep row));
+#endif
+
+#if defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED)
+PNG_EXTERN void png_do_write_swap_alpha PNGARG((png_row_infop row_info,
+   png_bytep row));
+#endif
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED)
+PNG_EXTERN void png_do_read_invert_alpha PNGARG((png_row_infop row_info,
+   png_bytep row));
+#endif
+
+#if defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+PNG_EXTERN void png_do_write_invert_alpha PNGARG((png_row_infop row_info,
+   png_bytep row));
+#endif
+
+#if defined(PNG_WRITE_FILLER_SUPPORTED) || \
+    defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+PNG_EXTERN void png_do_strip_filler PNGARG((png_row_infop row_info,
+   png_bytep row, png_uint_32 flags));
+#endif
+
+#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED)
+PNG_EXTERN void png_do_swap PNGARG((png_row_infop row_info, png_bytep row));
+#endif
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED) || defined(PNG_WRITE_PACKSWAP_SUPPORTED)
+PNG_EXTERN void png_do_packswap PNGARG((png_row_infop row_info, png_bytep row));
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+PNG_EXTERN int png_do_rgb_to_gray PNGARG((png_structp png_ptr, png_row_infop
+   row_info, png_bytep row));
+#endif
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+PNG_EXTERN void png_do_gray_to_rgb PNGARG((png_row_infop row_info,
+   png_bytep row));
+#endif
+
+#if defined(PNG_READ_PACK_SUPPORTED)
+PNG_EXTERN void png_do_unpack PNGARG((png_row_infop row_info, png_bytep row));
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED)
+PNG_EXTERN void png_do_unshift PNGARG((png_row_infop row_info, png_bytep row,
+   png_color_8p sig_bits));
+#endif
+
+#if defined(PNG_READ_INVERT_SUPPORTED) || defined(PNG_WRITE_INVERT_SUPPORTED)
+PNG_EXTERN void png_do_invert PNGARG((png_row_infop row_info, png_bytep row));
+#endif
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+PNG_EXTERN void png_do_chop PNGARG((png_row_infop row_info, png_bytep row));
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+PNG_EXTERN void png_do_dither PNGARG((png_row_infop row_info,
+   png_bytep row, png_bytep palette_lookup, png_bytep dither_lookup));
+
+#  if defined(PNG_CORRECT_PALETTE_SUPPORTED)
+PNG_EXTERN void png_correct_palette PNGARG((png_structp png_ptr,
+   png_colorp palette, int num_palette));
+#  endif
+#endif
+
+#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED)
+PNG_EXTERN void png_do_bgr PNGARG((png_row_infop row_info, png_bytep row));
+#endif
+
+#if defined(PNG_WRITE_PACK_SUPPORTED)
+PNG_EXTERN void png_do_pack PNGARG((png_row_infop row_info,
+   png_bytep row, png_uint_32 bit_depth));
+#endif
+
+#if defined(PNG_WRITE_SHIFT_SUPPORTED)
+PNG_EXTERN void png_do_shift PNGARG((png_row_infop row_info, png_bytep row,
+   png_color_8p bit_depth));
+#endif
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+PNG_EXTERN void png_do_background PNGARG((png_row_infop row_info, png_bytep row,
+   png_color_16p trans_values, png_color_16p background,
+   png_color_16p background_1,
+   png_bytep gamma_table, png_bytep gamma_from_1, png_bytep gamma_to_1,
+   png_uint_16pp gamma_16, png_uint_16pp gamma_16_from_1,
+   png_uint_16pp gamma_16_to_1, int gamma_shift));
+#else
+PNG_EXTERN void png_do_background PNGARG((png_row_infop row_info, png_bytep row,
+   png_color_16p trans_values, png_color_16p background));
+#endif
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+PNG_EXTERN void png_do_gamma PNGARG((png_row_infop row_info, png_bytep row,
+   png_bytep gamma_table, png_uint_16pp gamma_16_table,
+   int gamma_shift));
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+PNG_EXTERN void png_do_expand_palette PNGARG((png_row_infop row_info,
+   png_bytep row, png_colorp palette, png_bytep trans, int num_trans));
+PNG_EXTERN void png_do_expand PNGARG((png_row_infop row_info,
+   png_bytep row, png_color_16p trans_value));
+#endif
+
+/* The following decodes the appropriate chunks, and does error correction,
+ * then calls the appropriate callback for the chunk if it is valid.
+ */
+
+/* decode the IHDR chunk */
+PNG_EXTERN void png_handle_IHDR PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+PNG_EXTERN void png_handle_PLTE PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+PNG_EXTERN void png_handle_IEND PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+
+#if defined(PNG_READ_bKGD_SUPPORTED)
+PNG_EXTERN void png_handle_bKGD PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_cHRM_SUPPORTED)
+PNG_EXTERN void png_handle_cHRM PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_gAMA_SUPPORTED)
+PNG_EXTERN void png_handle_gAMA PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_hIST_SUPPORTED)
+PNG_EXTERN void png_handle_hIST PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_iCCP_SUPPORTED)
+extern void png_handle_iCCP PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif /* PNG_READ_iCCP_SUPPORTED */
+
+#if defined(PNG_READ_iTXt_SUPPORTED)
+PNG_EXTERN void png_handle_iTXt PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_oFFs_SUPPORTED)
+PNG_EXTERN void png_handle_oFFs PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_pCAL_SUPPORTED)
+PNG_EXTERN void png_handle_pCAL PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_pHYs_SUPPORTED)
+PNG_EXTERN void png_handle_pHYs PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_sBIT_SUPPORTED)
+PNG_EXTERN void png_handle_sBIT PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_sCAL_SUPPORTED)
+PNG_EXTERN void png_handle_sCAL PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_sPLT_SUPPORTED)
+extern void png_handle_sPLT PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif /* PNG_READ_sPLT_SUPPORTED */
+
+#if defined(PNG_READ_sRGB_SUPPORTED)
+PNG_EXTERN void png_handle_sRGB PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_tEXt_SUPPORTED)
+PNG_EXTERN void png_handle_tEXt PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_tIME_SUPPORTED)
+PNG_EXTERN void png_handle_tIME PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_tRNS_SUPPORTED)
+PNG_EXTERN void png_handle_tRNS PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#if defined(PNG_READ_zTXt_SUPPORTED)
+PNG_EXTERN void png_handle_zTXt PNGARG((png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 length));
+#endif
+
+#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+PNG_EXTERN int png_handle_as_unknown PNGARG((png_structp png_ptr, png_bytep
+   chunk_name));
+#endif
+
+PNG_EXTERN void png_handle_unknown PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 length));
+
+PNG_EXTERN void png_check_chunk_name PNGARG((png_structp png_ptr,
+   png_bytep chunk_name));
+
+/* handle the transformations for reading and writing */
+PNG_EXTERN void png_do_read_transformations PNGARG((png_structp png_ptr));
+PNG_EXTERN void png_do_write_transformations PNGARG((png_structp png_ptr));
+
+PNG_EXTERN void png_init_read_transformations PNGARG((png_structp png_ptr));
+
+#ifdef PNG_PROGRESSIVE_READ_SUPPORTED
+PNG_EXTERN void png_push_read_chunk PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+PNG_EXTERN void png_push_read_sig PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+PNG_EXTERN void png_push_check_crc PNGARG((png_structp png_ptr));
+PNG_EXTERN void png_push_crc_skip PNGARG((png_structp png_ptr,
+   png_uint_32 length));
+PNG_EXTERN void png_push_crc_finish PNGARG((png_structp png_ptr));
+PNG_EXTERN void png_push_fill_buffer PNGARG((png_structp png_ptr,
+   png_bytep buffer, png_size_t length));
+PNG_EXTERN void png_push_save_buffer PNGARG((png_structp png_ptr));
+PNG_EXTERN void png_push_restore_buffer PNGARG((png_structp png_ptr,
+   png_bytep buffer, png_size_t buffer_length));
+PNG_EXTERN void png_push_read_IDAT PNGARG((png_structp png_ptr));
+PNG_EXTERN void png_process_IDAT_data PNGARG((png_structp png_ptr,
+   png_bytep buffer, png_size_t buffer_length));
+PNG_EXTERN void png_push_process_row PNGARG((png_structp png_ptr));
+PNG_EXTERN void png_push_handle_unknown PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 length));
+PNG_EXTERN void png_push_have_info PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+PNG_EXTERN void png_push_have_end PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+PNG_EXTERN void png_push_have_row PNGARG((png_structp png_ptr, png_bytep row));
+PNG_EXTERN void png_push_read_end PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+PNG_EXTERN void png_process_some_data PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+PNG_EXTERN void png_read_push_finish_row PNGARG((png_structp png_ptr));
+#if defined(PNG_READ_tEXt_SUPPORTED)
+PNG_EXTERN void png_push_handle_tEXt PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 length));
+PNG_EXTERN void png_push_read_tEXt PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+PNG_EXTERN void png_push_handle_zTXt PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 length));
+PNG_EXTERN void png_push_read_zTXt PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+PNG_EXTERN void png_push_handle_iTXt PNGARG((png_structp png_ptr,
+   png_infop info_ptr, png_uint_32 length));
+PNG_EXTERN void png_push_read_iTXt PNGARG((png_structp png_ptr,
+   png_infop info_ptr));
+#endif
+
+#endif /* PNG_PROGRESSIVE_READ_SUPPORTED */
+
+#ifdef PNG_MNG_FEATURES_SUPPORTED
+PNG_EXTERN void png_do_read_intrapixel PNGARG((png_row_infop row_info,
+   png_bytep row));
+PNG_EXTERN void png_do_write_intrapixel PNGARG((png_row_infop row_info,
+   png_bytep row));
+#endif
+
+#if defined(PNG_ASSEMBLER_CODE_SUPPORTED)
+/* png.c */ /* PRIVATE */
+PNG_EXTERN void png_init_mmx_flags PNGARG((png_structp png_ptr));
+#endif
+/* Maintainer: Put new private prototypes here ^ and in libpngpf.3 */
+
+#endif /* PNG_INTERNAL */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* PNG_VERSION_INFO_ONLY */
+/* do not put anything past this line */
+#endif /* PNG_H */
diff --git a/png/pngasmrd.h b/png/pngasmrd.h
new file mode 100644
index 000000000..b96a46417
--- /dev/null
+++ b/png/pngasmrd.h
@@ -0,0 +1,11 @@
+/* pngasmrd.h - assembler version of utilities to read a PNG file
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 2001 Glenn Randers-Pehrson
+ *
+ */
+
+/* This file is obsolete in libpng-1.0.9 and later; its contents now appear
+ * at the end of pngconf.h.
+ */
diff --git a/png/pngconf.h b/png/pngconf.h
new file mode 100644
index 000000000..6167301d5
--- /dev/null
+++ b/png/pngconf.h
@@ -0,0 +1,1336 @@
+/* pngconf.h - machine configurable file for libpng
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+/* Any machine specific code is near the front of this file, so if you
+ * are configuring libpng for a machine, you may want to read the section
+ * starting here down to where it starts to typedef png_color, png_text,
+ * and png_info.
+ */
+
+#ifndef PNGCONF_H
+#define PNGCONF_H
+
+/* This is the size of the compression buffer, and thus the size of
+ * an IDAT chunk.  Make this whatever size you feel is best for your
+ * machine.  One of these will be allocated per png_struct.  When this
+ * is full, it writes the data to the disk, and does some other
+ * calculations.  Making this an extremely small size will slow
+ * the library down, but you may want to experiment to determine
+ * where it becomes significant, if you are concerned with memory
+ * usage.  Note that zlib allocates at least 32Kb also.  For readers,
+ * this describes the size of the buffer available to read the data in.
+ * Unless this gets smaller than the size of a row (compressed),
+ * it should not make much difference how big this is.
+ */
+
+#ifndef PNG_ZBUF_SIZE
+#  define PNG_ZBUF_SIZE 8192
+#endif
+
+/* Enable if you want a write-only libpng */
+
+#ifndef PNG_NO_READ_SUPPORTED
+#  define PNG_READ_SUPPORTED
+#endif
+
+/* Enable if you want a read-only libpng */
+
+#ifndef PNG_NO_WRITE_SUPPORTED
+#  define PNG_WRITE_SUPPORTED
+#endif
+
+/* Enabled by default in 1.2.0.  You can disable this if you don't need to
+   support PNGs that are embedded in MNG datastreams */
+#ifndef PNG_NO_MNG_FEATURES
+#  ifndef PNG_MNG_FEATURES_SUPPORTED
+#    define PNG_MNG_FEATURES_SUPPORTED
+#  endif
+#endif
+
+#ifndef PNG_NO_FLOATING_POINT_SUPPORTED
+#  ifndef PNG_FLOATING_POINT_SUPPORTED
+#    define PNG_FLOATING_POINT_SUPPORTED
+#  endif
+#endif
+
+/* If you are running on a machine where you cannot allocate more
+ * than 64K of memory at once, uncomment this.  While libpng will not
+ * normally need that much memory in a chunk (unless you load up a very
+ * large file), zlib needs to know how big of a chunk it can use, and
+ * libpng thus makes sure to check any memory allocation to verify it
+ * will fit into memory.
+#define PNG_MAX_MALLOC_64K
+ */
+#if defined(MAXSEG_64K) && !defined(PNG_MAX_MALLOC_64K)
+#  define PNG_MAX_MALLOC_64K
+#endif
+
+/* Special munging to support doing things the 'cygwin' way:
+ * 'Normal' png-on-win32 defines/defaults:
+ *   PNG_BUILD_DLL -- building dll
+ *   PNG_USE_DLL   -- building an application, linking to dll
+ *   (no define)   -- building static library, or building an
+ *                    application and linking to the static lib
+ * 'Cygwin' defines/defaults:
+ *   PNG_BUILD_DLL -- building the dll
+ *   (no define)   -- building an application, linking to the dll
+ *   PNG_STATIC    -- building the static lib, or building an application
+ *                    that links to the static lib.
+ *   ALL_STATIC    -- building various static libs, or building an application
+ *                    that links to the static libs.
+ * Thus,
+ * a cygwin user should define either PNG_BUILD_DLL or PNG_STATIC, and
+ * this bit of #ifdefs will define the 'correct' config variables based on
+ * that. If a cygwin user *wants* to define 'PNG_USE_DLL' that's okay, but
+ * unnecessary.
+ *
+ * Also, the precedence order is:
+ *   ALL_STATIC (since we can't #undef something outside our namespace)
+ *   PNG_BUILD_DLL
+ *   PNG_STATIC
+ *   (nothing) == PNG_USE_DLL 
+ */
+#if defined(__CYGWIN__)
+#  if defined(ALL_STATIC)
+#    if defined(PNG_BUILD_DLL)
+#      undef PNG_BUILD_DLL
+#    endif
+#    if defined(PNG_USE_DLL)
+#      undef PNG_USE_DLL
+#    endif
+#    if defined(PNG_DLL)
+#      undef PNG_DLL
+#    endif
+#    if !defined(PNG_STATIC)
+#      define PNG_STATIC
+#    endif
+#  else
+#    if defined (PNG_BUILD_DLL)
+#      if defined(PNG_STATIC)
+#        undef PNG_STATIC
+#      endif
+#      if defined(PNG_USE_DLL)
+#        undef PNG_USE_DLL
+#      endif
+#      if !defined(PNG_DLL)
+#        define PNG_DLL
+#      endif
+#    else
+#      if defined(PNG_STATIC)
+#        if defined(PNG_USE_DLL)
+#          undef PNG_USE_DLL
+#        endif
+#        if defined(PNG_DLL)
+#          undef PNG_DLL
+#        endif
+#      else
+#        if !defined(PNG_USE_DLL)
+#          define PNG_USE_DLL
+#        endif
+#        if !defined(PNG_DLL)
+#          define PNG_DLL
+#        endif
+#      endif  
+#    endif  
+#  endif
+#endif
+
+/* This protects us against compilers that run on a windowing system
+ * and thus don't have or would rather us not use the stdio types:
+ * stdin, stdout, and stderr.  The only one currently used is stderr
+ * in png_error() and png_warning().  #defining PNG_NO_CONSOLE_IO will
+ * prevent these from being compiled and used. #defining PNG_NO_STDIO
+ * will also prevent these, plus will prevent the entire set of stdio
+ * macros and functions (FILE *, printf, etc.) from being compiled and used,
+ * unless (PNG_DEBUG > 0) has been #defined.
+ *
+ * #define PNG_NO_CONSOLE_IO
+ * #define PNG_NO_STDIO
+ */
+
+#if defined(_WIN32_WCE)
+#  include <windows.h>
+   /* Console I/O functions are not supported on WindowsCE */
+#  define PNG_NO_CONSOLE_IO
+#  ifdef PNG_DEBUG
+#    undef PNG_DEBUG
+#  endif
+#endif
+
+#ifdef PNG_BUILD_DLL
+#  ifndef PNG_CONSOLE_IO_SUPPORTED
+#    ifndef PNG_NO_CONSOLE_IO
+#      define PNG_NO_CONSOLE_IO
+#    endif
+#  endif
+#endif
+
+#  ifdef PNG_NO_STDIO
+#    ifndef PNG_NO_CONSOLE_IO
+#      define PNG_NO_CONSOLE_IO
+#    endif
+#    ifdef PNG_DEBUG
+#      if (PNG_DEBUG > 0)
+#        include <stdio.h>
+#      endif
+#    endif
+#  else
+#    if !defined(_WIN32_WCE)
+/* "stdio.h" functions are not supported on WindowsCE */
+#      include <stdio.h>
+#    endif
+#  endif
+
+/* This macro protects us against machines that don't have function
+ * prototypes (ie K&R style headers).  If your compiler does not handle
+ * function prototypes, define this macro and use the included ansi2knr.
+ * I've always been able to use _NO_PROTO as the indicator, but you may
+ * need to drag the empty declaration out in front of here, or change the
+ * ifdef to suit your own needs.
+ */
+#ifndef PNGARG
+
+#ifdef OF /* zlib prototype munger */
+#  define PNGARG(arglist) OF(arglist)
+#else
+
+#ifdef _NO_PROTO
+#  define PNGARG(arglist) ()
+#  ifndef PNG_TYPECAST_NULL
+#     define PNG_TYPECAST_NULL
+#  endif
+#else
+#  define PNGARG(arglist) arglist
+#endif /* _NO_PROTO */
+
+#endif /* OF */
+
+#endif /* PNGARG */
+
+/* Try to determine if we are compiling on a Mac.  Note that testing for
+ * just __MWERKS__ is not good enough, because the Codewarrior is now used
+ * on non-Mac platforms.
+ */
+#ifndef MACOS
+#  if (defined(__MWERKS__) && defined(macintosh)) || defined(applec) || \
+      defined(THINK_C) || defined(__SC__) || defined(TARGET_OS_MAC)
+#    define MACOS
+#  endif
+#endif
+
+/* enough people need this for various reasons to include it here */
+#if !defined(MACOS) && !defined(RISCOS) && !defined(_WIN32_WCE)
+#  include <sys/types.h>
+#endif
+
+#if !defined(PNG_SETJMP_NOT_SUPPORTED) && !defined(PNG_NO_SETJMP_SUPPORTED)
+#  define PNG_SETJMP_SUPPORTED
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+/* This is an attempt to force a single setjmp behaviour on Linux.  If
+ * the X config stuff didn't define _BSD_SOURCE we wouldn't need this.
+ */
+
+#  ifdef __linux__
+#    ifdef _BSD_SOURCE
+#      define PNG_SAVE_BSD_SOURCE
+#      undef _BSD_SOURCE
+#    endif
+#    ifdef _SETJMP_H
+/* Explanation added by debian maintainer Philippe Troin <phil@fifi.org>.
+   There are two versions of setjmp, depending wether or not we compile for
+   BSD. They are incompatible and can cause crashes. The PNG people force
+   here a unique behavior for setjmp. It you get the error below,
+   then include <png.h> before <setjmp.h>.
+*/
+#error png.h already includes setjmp.h with some additional fixup.
+#    endif
+#  endif /* __linux__ */
+
+   /* include setjmp.h for error handling */
+#  include <setjmp.h>
+
+#  ifdef __linux__
+#    ifdef PNG_SAVE_BSD_SOURCE
+#      define _BSD_SOURCE
+#      undef PNG_SAVE_BSD_SOURCE
+#    endif
+#  endif /* __linux__ */
+#endif /* PNG_SETJMP_SUPPORTED */
+
+#ifdef BSD
+#  include <strings.h>
+#else
+#  include <string.h>
+#endif
+
+/* Other defines for things like memory and the like can go here.  */
+#ifdef PNG_INTERNAL
+
+#include <stdlib.h>
+
+/* The functions exported by PNG_EXTERN are PNG_INTERNAL functions, which
+ * aren't usually used outside the library (as far as I know), so it is
+ * debatable if they should be exported at all.  In the future, when it is
+ * possible to have run-time registry of chunk-handling functions, some of
+ * these will be made available again.
+#define PNG_EXTERN extern
+ */
+#define PNG_EXTERN
+
+/* Other defines specific to compilers can go here.  Try to keep
+ * them inside an appropriate ifdef/endif pair for portability.
+ */
+
+#if defined(PNG_FLOATING_POINT_SUPPORTED)
+#  if defined(MACOS)
+     /* We need to check that <math.h> hasn't already been included earlier
+      * as it seems it doesn't agree with <fp.h>, yet we should really use
+      * <fp.h> if possible.
+      */
+#    if !defined(__MATH_H__) && !defined(__MATH_H) && !defined(__cmath__)
+#      include <fp.h>
+#    endif
+#  else
+#    include <math.h>
+#  endif
+#  if defined(_AMIGA) && defined(__SASC) && defined(_M68881)
+     /* Amiga SAS/C: We must include builtin FPU functions when compiling using
+      * MATH=68881
+      */
+#    include <m68881.h>
+#  endif
+#endif
+
+/* Codewarrior on NT has linking problems without this. */
+#if (defined(__MWERKS__) && defined(WIN32)) || defined(__STDC__)
+#  define PNG_ALWAYS_EXTERN
+#endif
+
+/* For some reason, Borland C++ defines memcmp, etc. in mem.h, not
+ * stdlib.h like it should (I think).  Or perhaps this is a C++
+ * "feature"?
+ */
+#ifdef __TURBOC__
+#  include <mem.h>
+#  include "alloc.h"
+#endif
+
+#if defined(_MSC_VER) && (defined(WIN32) || defined(_Windows) || \
+    defined(_WINDOWS) || defined(_WIN32) || defined(__WIN32__))
+#  include <malloc.h>
+#endif
+
+/* This controls how fine the dithering gets.  As this allocates
+ * a largish chunk of memory (32K), those who are not as concerned
+ * with dithering quality can decrease some or all of these.
+ */
+#ifndef PNG_DITHER_RED_BITS
+#  define PNG_DITHER_RED_BITS 5
+#endif
+#ifndef PNG_DITHER_GREEN_BITS
+#  define PNG_DITHER_GREEN_BITS 5
+#endif
+#ifndef PNG_DITHER_BLUE_BITS
+#  define PNG_DITHER_BLUE_BITS 5
+#endif
+
+/* This controls how fine the gamma correction becomes when you
+ * are only interested in 8 bits anyway.  Increasing this value
+ * results in more memory being used, and more pow() functions
+ * being called to fill in the gamma tables.  Don't set this value
+ * less then 8, and even that may not work (I haven't tested it).
+ */
+
+#ifndef PNG_MAX_GAMMA_8
+#  define PNG_MAX_GAMMA_8 11
+#endif
+
+/* This controls how much a difference in gamma we can tolerate before
+ * we actually start doing gamma conversion.
+ */
+#ifndef PNG_GAMMA_THRESHOLD
+#  define PNG_GAMMA_THRESHOLD 0.05
+#endif
+
+#endif /* PNG_INTERNAL */
+
+/* The following uses const char * instead of char * for error
+ * and warning message functions, so some compilers won't complain.
+ * If you do not want to use const, define PNG_NO_CONST here.
+ */
+
+#ifndef PNG_NO_CONST
+#  define PNG_CONST const
+#else
+#  define PNG_CONST
+#endif
+
+/* The following defines give you the ability to remove code from the
+ * library that you will not be using.  I wish I could figure out how to
+ * automate this, but I can't do that without making it seriously hard
+ * on the users.  So if you are not using an ability, change the #define
+ * to and #undef, and that part of the library will not be compiled.  If
+ * your linker can't find a function, you may want to make sure the
+ * ability is defined here.  Some of these depend upon some others being
+ * defined.  I haven't figured out all the interactions here, so you may
+ * have to experiment awhile to get everything to compile.  If you are
+ * creating or using a shared library, you probably shouldn't touch this,
+ * as it will affect the size of the structures, and this will cause bad
+ * things to happen if the library and/or application ever change.
+ */
+
+/* Any features you will not be using can be undef'ed here */
+
+/* GR-P, 0.96a: Set "*TRANSFORMS_SUPPORTED as default but allow user
+ * to turn it off with "*TRANSFORMS_NOT_SUPPORTED" or *PNG_NO_*_TRANSFORMS
+ * on the compile line, then pick and choose which ones to define without
+ * having to edit this file. It is safe to use the *TRANSFORMS_NOT_SUPPORTED
+ * if you only want to have a png-compliant reader/writer but don't need
+ * any of the extra transformations.  This saves about 80 kbytes in a
+ * typical installation of the library. (PNG_NO_* form added in version
+ * 1.0.1c, for consistency)
+ */
+
+/* The size of the png_text structure changed in libpng-1.0.6 when
+ * iTXt is supported.  It is turned off by default, to support old apps
+ * that malloc the png_text structure instead of calling png_set_text()
+ * and letting libpng malloc it.  It will be turned on by default in
+ * libpng-1.3.0.
+ */
+
+#ifndef PNG_iTXt_SUPPORTED
+#  if !defined(PNG_READ_iTXt_SUPPORTED) && !defined(PNG_NO_READ_iTXt)
+#    define PNG_NO_READ_iTXt
+#  endif
+#  if !defined(PNG_WRITE_iTXt_SUPPORTED) && !defined(PNG_NO_WRITE_iTXt)
+#    define PNG_NO_WRITE_iTXt
+#  endif
+#endif
+
+/* The following support, added after version 1.0.0, can be turned off here en
+ * masse by defining PNG_LEGACY_SUPPORTED in case you need binary compatibility
+ * with old applications that require the length of png_struct and png_info
+ * to remain unchanged.
+ */
+
+#ifdef PNG_LEGACY_SUPPORTED
+#  define PNG_NO_FREE_ME
+#  define PNG_NO_READ_UNKNOWN_CHUNKS
+#  define PNG_NO_WRITE_UNKNOWN_CHUNKS
+#  define PNG_NO_READ_USER_CHUNKS
+#  define PNG_NO_READ_iCCP
+#  define PNG_NO_WRITE_iCCP
+#  define PNG_NO_READ_iTXt
+#  define PNG_NO_WRITE_iTXt
+#  define PNG_NO_READ_sCAL
+#  define PNG_NO_WRITE_sCAL
+#  define PNG_NO_READ_sPLT
+#  define PNG_NO_WRITE_sPLT
+#  define PNG_NO_INFO_IMAGE
+#  define PNG_NO_READ_RGB_TO_GRAY
+#  define PNG_NO_READ_USER_TRANSFORM
+#  define PNG_NO_WRITE_USER_TRANSFORM
+#  define PNG_NO_USER_MEM
+#  define PNG_NO_READ_EMPTY_PLTE
+#  define PNG_NO_MNG_FEATURES
+#  define PNG_NO_FIXED_POINT_SUPPORTED
+#endif
+
+/* Ignore attempt to turn off both floating and fixed point support */
+#if !defined(PNG_FLOATING_POINT_SUPPORTED) || \
+    !defined(PNG_NO_FIXED_POINT_SUPPORTED)
+#  define PNG_FIXED_POINT_SUPPORTED
+#endif
+
+#ifndef PNG_NO_FREE_ME
+#  define PNG_FREE_ME_SUPPORTED
+#endif
+
+#if defined(PNG_READ_SUPPORTED)
+
+#if !defined(PNG_READ_TRANSFORMS_NOT_SUPPORTED) && \
+      !defined(PNG_NO_READ_TRANSFORMS)
+#  define PNG_READ_TRANSFORMS_SUPPORTED
+#endif
+
+#ifdef PNG_READ_TRANSFORMS_SUPPORTED
+#  ifndef PNG_NO_READ_EXPAND
+#    define PNG_READ_EXPAND_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_SHIFT
+#    define PNG_READ_SHIFT_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_PACK
+#    define PNG_READ_PACK_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_BGR
+#    define PNG_READ_BGR_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_SWAP
+#    define PNG_READ_SWAP_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_PACKSWAP
+#    define PNG_READ_PACKSWAP_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_INVERT
+#    define PNG_READ_INVERT_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_DITHER
+#    define PNG_READ_DITHER_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_BACKGROUND
+#    define PNG_READ_BACKGROUND_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_16_TO_8
+#    define PNG_READ_16_TO_8_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_FILLER
+#    define PNG_READ_FILLER_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_GAMMA
+#    define PNG_READ_GAMMA_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_GRAY_TO_RGB
+#    define PNG_READ_GRAY_TO_RGB_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_SWAP_ALPHA
+#    define PNG_READ_SWAP_ALPHA_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_INVERT_ALPHA
+#    define PNG_READ_INVERT_ALPHA_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_STRIP_ALPHA
+#    define PNG_READ_STRIP_ALPHA_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_USER_TRANSFORM
+#    define PNG_READ_USER_TRANSFORM_SUPPORTED
+#  endif
+#  ifndef PNG_NO_READ_RGB_TO_GRAY
+#    define PNG_READ_RGB_TO_GRAY_SUPPORTED
+#  endif
+#endif /* PNG_READ_TRANSFORMS_SUPPORTED */
+
+#if !defined(PNG_NO_PROGRESSIVE_READ) && \
+ !defined(PNG_PROGRESSIVE_READ_NOT_SUPPORTED)  /* if you don't do progressive */
+#  define PNG_PROGRESSIVE_READ_SUPPORTED     /* reading.  This is not talking */
+#endif                               /* about interlacing capability!  You'll */
+              /* still have interlacing unless you change the following line: */
+
+#define PNG_READ_INTERLACING_SUPPORTED /* required for PNG-compliant decoders */
+
+#ifndef PNG_NO_READ_COMPOSITE_NODIV
+#  ifndef PNG_NO_READ_COMPOSITED_NODIV  /* libpng-1.0.x misspelling */
+#    define PNG_READ_COMPOSITE_NODIV_SUPPORTED   /* well tested on Intel, SGI */
+#  endif
+#endif
+
+/* Deprecated, will be removed from version 2.0.0.
+   Use PNG_MNG_FEATURES_SUPPORTED instead. */
+#ifndef PNG_NO_READ_EMPTY_PLTE
+#  define PNG_READ_EMPTY_PLTE_SUPPORTED
+#endif
+
+#endif /* PNG_READ_SUPPORTED */
+
+#if defined(PNG_WRITE_SUPPORTED)
+
+# if !defined(PNG_WRITE_TRANSFORMS_NOT_SUPPORTED) && \
+    !defined(PNG_NO_WRITE_TRANSFORMS)
+#  define PNG_WRITE_TRANSFORMS_SUPPORTED
+#endif
+
+#ifdef PNG_WRITE_TRANSFORMS_SUPPORTED
+#  ifndef PNG_NO_WRITE_SHIFT
+#    define PNG_WRITE_SHIFT_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_PACK
+#    define PNG_WRITE_PACK_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_BGR
+#    define PNG_WRITE_BGR_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_SWAP
+#    define PNG_WRITE_SWAP_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_PACKSWAP
+#    define PNG_WRITE_PACKSWAP_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_INVERT
+#    define PNG_WRITE_INVERT_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_FILLER
+#    define PNG_WRITE_FILLER_SUPPORTED   /* same as WRITE_STRIP_ALPHA */
+#  endif
+#  ifndef PNG_NO_WRITE_SWAP_ALPHA
+#    define PNG_WRITE_SWAP_ALPHA_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_INVERT_ALPHA
+#    define PNG_WRITE_INVERT_ALPHA_SUPPORTED
+#  endif
+#  ifndef PNG_NO_WRITE_USER_TRANSFORM
+#    define PNG_WRITE_USER_TRANSFORM_SUPPORTED
+#  endif
+#endif /* PNG_WRITE_TRANSFORMS_SUPPORTED */
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED)
+#  ifndef PNG_NO_USER_TRANSFORM_PTR
+#    define PNG_USER_TRANSFORM_PTR_SUPPORTED
+#  endif
+#endif
+
+#define PNG_WRITE_INTERLACING_SUPPORTED  /* not required for PNG-compliant
+                                            encoders, but can cause trouble
+                                            if left undefined */
+
+#if !defined(PNG_NO_WRITE_WEIGHTED_FILTER) && \
+     defined(PNG_FLOATING_POINT_SUPPORTED)
+#  define PNG_WRITE_WEIGHTED_FILTER_SUPPORTED
+#endif
+
+#ifndef PNG_NO_ERROR_NUMBERS
+#define PNG_ERROR_NUMBERS_SUPPORTED
+#endif
+
+#ifndef PNG_NO_WRITE_FLUSH
+#  define PNG_WRITE_FLUSH_SUPPORTED
+#endif
+
+/* Deprecated, see PNG_MNG_FEATURES_SUPPORTED, above */
+#ifndef PNG_NO_WRITE_EMPTY_PLTE
+#  define PNG_WRITE_EMPTY_PLTE_SUPPORTED
+#endif
+
+#endif /* PNG_WRITE_SUPPORTED */
+
+#ifndef PNG_NO_STDIO
+#  define PNG_TIME_RFC1123_SUPPORTED
+#endif
+
+/* This adds extra functions in pngget.c for accessing data from the
+ * info pointer (added in version 0.99)
+ * png_get_image_width()
+ * png_get_image_height()
+ * png_get_bit_depth()
+ * png_get_color_type()
+ * png_get_compression_type()
+ * png_get_filter_type()
+ * png_get_interlace_type()
+ * png_get_pixel_aspect_ratio()
+ * png_get_pixels_per_meter()
+ * png_get_x_offset_pixels()
+ * png_get_y_offset_pixels()
+ * png_get_x_offset_microns()
+ * png_get_y_offset_microns()
+ */
+#if !defined(PNG_NO_EASY_ACCESS) && !defined(PNG_EASY_ACCESS_SUPPORTED)
+#  define PNG_EASY_ACCESS_SUPPORTED
+#endif
+
+/* PNG_ASSEMBLER_CODE was enabled by default in version 1.2.0 
+   even when PNG_USE_PNGVCRD or PNG_USE_PNGGCCRD is not defined */
+#if defined(PNG_READ_SUPPORTED) && !defined(PNG_NO_ASSEMBLER_CODE)
+#  ifndef PNG_ASSEMBLER_CODE_SUPPORTED
+#    define PNG_ASSEMBLER_CODE_SUPPORTED
+#  endif
+#  if !defined(PNG_MMX_CODE_SUPPORTED) && !defined(PNG_NO_MMX_CODE)
+#    define PNG_MMX_CODE_SUPPORTED
+#  endif
+#endif
+
+/* If you are sure that you don't need thread safety and you are compiling
+   with PNG_USE_PNGCCRD for an MMX application, you can define this for
+   faster execution.  See pnggccrd.c.
+#define PNG_THREAD_UNSAFE_OK
+*/
+
+#if !defined(PNG_NO_USER_MEM) && !defined(PNG_USER_MEM_SUPPORTED)
+#  define PNG_USER_MEM_SUPPORTED
+#endif
+
+/* These are currently experimental features, define them if you want */
+
+/* very little testing */
+/*
+#ifdef PNG_READ_SUPPORTED
+#  ifndef PNG_READ_16_TO_8_ACCURATE_SCALE_SUPPORTED
+#    define PNG_READ_16_TO_8_ACCURATE_SCALE_SUPPORTED
+#  endif
+#endif
+*/
+
+/* This is only for PowerPC big-endian and 680x0 systems */
+/* some testing */
+/*
+#ifdef PNG_READ_SUPPORTED
+#  ifndef PNG_PNG_READ_BIG_ENDIAN_SUPPORTED
+#    define PNG_READ_BIG_ENDIAN_SUPPORTED
+#  endif
+#endif
+*/
+
+/* Buggy compilers (e.g., gcc 2.7.2.2) need this */
+/*
+#define PNG_NO_POINTER_INDEXING
+*/
+
+/* These functions are turned off by default, as they will be phased out. */
+/*
+#define  PNG_USELESS_TESTS_SUPPORTED
+#define  PNG_CORRECT_PALETTE_SUPPORTED
+*/
+
+/* Any chunks you are not interested in, you can undef here.  The
+ * ones that allocate memory may be expecially important (hIST,
+ * tEXt, zTXt, tRNS, pCAL).  Others will just save time and make png_info
+ * a bit smaller.
+ */
+
+#if defined(PNG_READ_SUPPORTED) && \
+    !defined(PNG_READ_ANCILLARY_CHUNKS_NOT_SUPPORTED) && \
+    !defined(PNG_NO_READ_ANCILLARY_CHUNKS)
+#  define PNG_READ_ANCILLARY_CHUNKS_SUPPORTED
+#endif
+
+#if defined(PNG_WRITE_SUPPORTED) && \
+    !defined(PNG_WRITE_ANCILLARY_CHUNKS_NOT_SUPPORTED) && \
+    !defined(PNG_NO_WRITE_ANCILLARY_CHUNKS)
+#  define PNG_WRITE_ANCILLARY_CHUNKS_SUPPORTED
+#endif
+
+#ifdef PNG_READ_ANCILLARY_CHUNKS_SUPPORTED
+
+#ifdef PNG_NO_READ_TEXT
+#  define PNG_NO_READ_iTXt
+#  define PNG_NO_READ_tEXt
+#  define PNG_NO_READ_zTXt
+#endif
+#ifndef PNG_NO_READ_bKGD
+#  define PNG_READ_bKGD_SUPPORTED
+#  define PNG_bKGD_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_cHRM
+#  define PNG_READ_cHRM_SUPPORTED
+#  define PNG_cHRM_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_gAMA
+#  define PNG_READ_gAMA_SUPPORTED
+#  define PNG_gAMA_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_hIST
+#  define PNG_READ_hIST_SUPPORTED
+#  define PNG_hIST_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_iCCP
+#  define PNG_READ_iCCP_SUPPORTED
+#  define PNG_iCCP_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_iTXt
+#  ifndef PNG_READ_iTXt_SUPPORTED
+#    define PNG_READ_iTXt_SUPPORTED
+#  endif
+#  ifndef PNG_iTXt_SUPPORTED
+#    define PNG_iTXt_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_READ_oFFs
+#  define PNG_READ_oFFs_SUPPORTED
+#  define PNG_oFFs_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_pCAL
+#  define PNG_READ_pCAL_SUPPORTED
+#  define PNG_pCAL_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_sCAL
+#  define PNG_READ_sCAL_SUPPORTED
+#  define PNG_sCAL_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_pHYs
+#  define PNG_READ_pHYs_SUPPORTED
+#  define PNG_pHYs_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_sBIT
+#  define PNG_READ_sBIT_SUPPORTED
+#  define PNG_sBIT_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_sPLT
+#  define PNG_READ_sPLT_SUPPORTED
+#  define PNG_sPLT_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_sRGB
+#  define PNG_READ_sRGB_SUPPORTED
+#  define PNG_sRGB_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_tEXt
+#  define PNG_READ_tEXt_SUPPORTED
+#  define PNG_tEXt_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_tIME
+#  define PNG_READ_tIME_SUPPORTED
+#  define PNG_tIME_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_tRNS
+#  define PNG_READ_tRNS_SUPPORTED
+#  define PNG_tRNS_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_zTXt
+#  define PNG_READ_zTXt_SUPPORTED
+#  define PNG_zTXt_SUPPORTED
+#endif
+#ifndef PNG_NO_READ_UNKNOWN_CHUNKS
+#  define PNG_READ_UNKNOWN_CHUNKS_SUPPORTED
+#  ifndef PNG_UNKNOWN_CHUNKS_SUPPORTED
+#    define PNG_UNKNOWN_CHUNKS_SUPPORTED
+#  endif
+#  ifndef PNG_NO_HANDLE_AS_UNKNOWN
+#    define PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+#  endif
+#endif
+#if !defined(PNG_NO_READ_USER_CHUNKS) && \
+     defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED)
+#  define PNG_READ_USER_CHUNKS_SUPPORTED
+#  define PNG_USER_CHUNKS_SUPPORTED
+#  ifdef PNG_NO_READ_UNKNOWN_CHUNKS
+#    undef PNG_NO_READ_UNKNOWN_CHUNKS
+#  endif
+#  ifdef PNG_NO_HANDLE_AS_UNKNOWN
+#    undef PNG_NO_HANDLE_AS_UNKNOWN
+#  endif
+#endif
+#ifndef PNG_NO_READ_OPT_PLTE
+#  define PNG_READ_OPT_PLTE_SUPPORTED /* only affects support of the */
+#endif                      /* optional PLTE chunk in RGB and RGBA images */
+#if defined(PNG_READ_iTXt_SUPPORTED) || defined(PNG_READ_tEXt_SUPPORTED) || \
+    defined(PNG_READ_zTXt_SUPPORTED)
+#  define PNG_READ_TEXT_SUPPORTED
+#  define PNG_TEXT_SUPPORTED
+#endif
+
+#endif /* PNG_READ_ANCILLARY_CHUNKS_SUPPORTED */
+
+#ifdef PNG_WRITE_ANCILLARY_CHUNKS_SUPPORTED
+
+#ifdef PNG_NO_WRITE_TEXT
+#  define PNG_NO_WRITE_iTXt
+#  define PNG_NO_WRITE_tEXt
+#  define PNG_NO_WRITE_zTXt
+#endif
+#ifndef PNG_NO_WRITE_bKGD
+#  define PNG_WRITE_bKGD_SUPPORTED
+#  ifndef PNG_bKGD_SUPPORTED
+#    define PNG_bKGD_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_cHRM
+#  define PNG_WRITE_cHRM_SUPPORTED
+#  ifndef PNG_cHRM_SUPPORTED
+#    define PNG_cHRM_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_gAMA
+#  define PNG_WRITE_gAMA_SUPPORTED
+#  ifndef PNG_gAMA_SUPPORTED
+#    define PNG_gAMA_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_hIST
+#  define PNG_WRITE_hIST_SUPPORTED
+#  ifndef PNG_hIST_SUPPORTED
+#    define PNG_hIST_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_iCCP
+#  define PNG_WRITE_iCCP_SUPPORTED
+#  ifndef PNG_iCCP_SUPPORTED
+#    define PNG_iCCP_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_iTXt
+#  ifndef PNG_WRITE_iTXt_SUPPORTED
+#    define PNG_WRITE_iTXt_SUPPORTED
+#  endif
+#  ifndef PNG_iTXt_SUPPORTED
+#    define PNG_iTXt_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_oFFs
+#  define PNG_WRITE_oFFs_SUPPORTED
+#  ifndef PNG_oFFs_SUPPORTED
+#    define PNG_oFFs_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_pCAL
+#  define PNG_WRITE_pCAL_SUPPORTED
+#  ifndef PNG_pCAL_SUPPORTED
+#    define PNG_pCAL_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_sCAL
+#  define PNG_WRITE_sCAL_SUPPORTED
+#  ifndef PNG_sCAL_SUPPORTED
+#    define PNG_sCAL_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_pHYs
+#  define PNG_WRITE_pHYs_SUPPORTED
+#  ifndef PNG_pHYs_SUPPORTED
+#    define PNG_pHYs_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_sBIT
+#  define PNG_WRITE_sBIT_SUPPORTED
+#  ifndef PNG_sBIT_SUPPORTED
+#    define PNG_sBIT_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_sPLT
+#  define PNG_WRITE_sPLT_SUPPORTED
+#  ifndef PNG_sPLT_SUPPORTED
+#    define PNG_sPLT_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_sRGB
+#  define PNG_WRITE_sRGB_SUPPORTED
+#  ifndef PNG_sRGB_SUPPORTED
+#    define PNG_sRGB_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_tEXt
+#  define PNG_WRITE_tEXt_SUPPORTED
+#  ifndef PNG_tEXt_SUPPORTED
+#    define PNG_tEXt_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_tIME
+#  define PNG_WRITE_tIME_SUPPORTED
+#  ifndef PNG_tIME_SUPPORTED
+#    define PNG_tIME_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_tRNS
+#  define PNG_WRITE_tRNS_SUPPORTED
+#  ifndef PNG_tRNS_SUPPORTED
+#    define PNG_tRNS_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_zTXt
+#  define PNG_WRITE_zTXt_SUPPORTED
+#  ifndef PNG_zTXt_SUPPORTED
+#    define PNG_zTXt_SUPPORTED
+#  endif
+#endif
+#ifndef PNG_NO_WRITE_UNKNOWN_CHUNKS
+#  define PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED
+#  ifndef PNG_UNKNOWN_CHUNKS_SUPPORTED
+#    define PNG_UNKNOWN_CHUNKS_SUPPORTED
+#  endif
+#  ifndef PNG_NO_HANDLE_AS_UNKNOWN
+#     ifndef PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+#       define PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+#     endif
+#  endif
+#endif
+#if defined(PNG_WRITE_iTXt_SUPPORTED) || defined(PNG_WRITE_tEXt_SUPPORTED) || \
+    defined(PNG_WRITE_zTXt_SUPPORTED)
+#  define PNG_WRITE_TEXT_SUPPORTED
+#  ifndef PNG_TEXT_SUPPORTED
+#    define PNG_TEXT_SUPPORTED
+#  endif
+#endif
+
+#endif /* PNG_WRITE_ANCILLARY_CHUNKS_SUPPORTED */
+
+/* Turn this off to disable png_read_png() and
+ * png_write_png() and leave the row_pointers member
+ * out of the info structure.
+ */
+#ifndef PNG_NO_INFO_IMAGE
+#  define PNG_INFO_IMAGE_SUPPORTED
+#endif
+
+/* need the time information for reading tIME chunks */
+#if defined(PNG_tIME_SUPPORTED)
+#  if !defined(_WIN32_WCE)
+     /* "time.h" functions are not supported on WindowsCE */
+#    include <time.h>
+#  endif
+#endif
+
+/* Some typedefs to get us started.  These should be safe on most of the
+ * common platforms.  The typedefs should be at least as large as the
+ * numbers suggest (a png_uint_32 must be at least 32 bits long), but they
+ * don't have to be exactly that size.  Some compilers dislike passing
+ * unsigned shorts as function parameters, so you may be better off using
+ * unsigned int for png_uint_16.  Likewise, for 64-bit systems, you may
+ * want to have unsigned int for png_uint_32 instead of unsigned long.
+ */
+
+typedef unsigned long png_uint_32;
+typedef long png_int_32;
+typedef unsigned short png_uint_16;
+typedef short png_int_16;
+typedef unsigned char png_byte;
+
+/* This is usually size_t.  It is typedef'ed just in case you need it to
+   change (I'm not sure if you will or not, so I thought I'd be safe) */
+typedef size_t png_size_t;
+
+/* The following is needed for medium model support.  It cannot be in the
+ * PNG_INTERNAL section.  Needs modification for other compilers besides
+ * MSC.  Model independent support declares all arrays and pointers to be
+ * large using the far keyword.  The zlib version used must also support
+ * model independent data.  As of version zlib 1.0.4, the necessary changes
+ * have been made in zlib.  The USE_FAR_KEYWORD define triggers other
+ * changes that are needed. (Tim Wegner)
+ */
+
+/* Separate compiler dependencies (problem here is that zlib.h always
+   defines FAR. (SJT) */
+#ifdef __BORLANDC__
+#  if defined(__LARGE__) || defined(__HUGE__) || defined(__COMPACT__)
+#    define LDATA 1
+#  else
+#    define LDATA 0
+#  endif
+   /* GRR:  why is Cygwin in here?  Cygwin is not Borland C... */
+#  if !defined(__WIN32__) && !defined(__FLAT__) && !defined(__CYGWIN__)
+#    define PNG_MAX_MALLOC_64K
+#    if (LDATA != 1)
+#      ifndef FAR
+#        define FAR __far
+#      endif
+#      define USE_FAR_KEYWORD
+#    endif   /* LDATA != 1 */
+     /* Possibly useful for moving data out of default segment.
+      * Uncomment it if you want. Could also define FARDATA as
+      * const if your compiler supports it. (SJT)
+#    define FARDATA FAR
+      */
+#  endif  /* __WIN32__, __FLAT__, __CYGWIN__ */
+#endif   /* __BORLANDC__ */
+
+
+/* Suggest testing for specific compiler first before testing for
+ * FAR.  The Watcom compiler defines both __MEDIUM__ and M_I86MM,
+ * making reliance oncertain keywords suspect. (SJT)
+ */
+
+/* MSC Medium model */
+#if defined(FAR)
+#  if defined(M_I86MM)
+#    define USE_FAR_KEYWORD
+#    define FARDATA FAR
+#    include <dos.h>
+#  endif
+#endif
+
+/* SJT: default case */
+#ifndef FAR
+#  define FAR
+#endif
+
+/* At this point FAR is always defined */
+#ifndef FARDATA
+#  define FARDATA
+#endif
+
+/* Typedef for floating-point numbers that are converted
+   to fixed-point with a multiple of 100,000, e.g., int_gamma */
+typedef png_int_32 png_fixed_point;
+
+/* Add typedefs for pointers */
+typedef void            FAR * png_voidp;
+typedef png_byte        FAR * png_bytep;
+typedef png_uint_32     FAR * png_uint_32p;
+typedef png_int_32      FAR * png_int_32p;
+typedef png_uint_16     FAR * png_uint_16p;
+typedef png_int_16      FAR * png_int_16p;
+typedef PNG_CONST char  FAR * png_const_charp;
+typedef char            FAR * png_charp;
+typedef png_fixed_point FAR * png_fixed_point_p;
+
+#ifndef PNG_NO_STDIO
+#if defined(_WIN32_WCE)
+typedef HANDLE                png_FILE_p;
+#else
+typedef FILE                * png_FILE_p;
+#endif
+#endif
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+typedef double          FAR * png_doublep;
+#endif
+
+/* Pointers to pointers; i.e. arrays */
+typedef png_byte        FAR * FAR * png_bytepp;
+typedef png_uint_32     FAR * FAR * png_uint_32pp;
+typedef png_int_32      FAR * FAR * png_int_32pp;
+typedef png_uint_16     FAR * FAR * png_uint_16pp;
+typedef png_int_16      FAR * FAR * png_int_16pp;
+typedef PNG_CONST char  FAR * FAR * png_const_charpp;
+typedef char            FAR * FAR * png_charpp;
+typedef png_fixed_point FAR * FAR * png_fixed_point_pp;
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+typedef double          FAR * FAR * png_doublepp;
+#endif
+
+/* Pointers to pointers to pointers; i.e., pointer to array */
+typedef char            FAR * FAR * FAR * png_charppp;
+
+/* libpng typedefs for types in zlib. If zlib changes
+ * or another compression library is used, then change these.
+ * Eliminates need to change all the source files.
+ */
+typedef charf *         png_zcharp;
+typedef charf * FAR *   png_zcharpp;
+typedef z_stream FAR *  png_zstreamp;
+
+/*
+ * Define PNG_BUILD_DLL if the module being built is a Windows
+ * LIBPNG DLL.
+ *
+ * Define PNG_USE_DLL if you want to *link* to the Windows LIBPNG DLL.
+ * It is equivalent to Microsoft predefined macro _DLL that is
+ * automatically defined when you compile using the share
+ * version of the CRT (C Run-Time library)
+ *
+ * The cygwin mods make this behavior a little different:
+ * Define PNG_BUILD_DLL if you are building a dll for use with cygwin
+ * Define PNG_STATIC if you are building a static library for use with cygwin,
+ *   -or- if you are building an application that you want to link to the
+ *   static library.
+ * PNG_USE_DLL is defined by default (no user action needed) unless one of
+ *   the other flags is defined.
+ */
+
+#if !defined(PNG_DLL) && (defined(PNG_BUILD_DLL) || defined(PNG_USE_DLL))
+#  define PNG_DLL
+#endif
+/* If CYGWIN, then disallow GLOBAL ARRAYS unless building a static lib.
+ * When building a static lib, default to no GLOBAL ARRAYS, but allow
+ * command-line override
+ */
+#if defined(__CYGWIN__)
+#  if !defined(PNG_STATIC)
+#    if defined(PNG_USE_GLOBAL_ARRAYS)
+#      undef PNG_USE_GLOBAL_ARRAYS
+#    endif
+#    if !defined(PNG_USE_LOCAL_ARRAYS)
+#      define PNG_USE_LOCAL_ARRAYS
+#    endif
+#  else
+#    if defined(PNG_USE_LOCAL_ARRAYS) || defined(PNG_NO_GLOBAL_ARRAYS)
+#      if defined(PNG_USE_GLOBAL_ARRAYS)
+#        undef PNG_USE_GLOBAL_ARRAYS
+#      endif
+#    endif
+#  endif
+#  if !defined(PNG_USE_LOCAL_ARRAYS) && !defined(PNG_USE_GLOBAL_ARRAYS)
+#    define PNG_USE_LOCAL_ARRAYS
+#  endif
+#endif
+
+/* Do not use global arrays (helps with building DLL's)
+ * They are no longer used in libpng itself, since version 1.0.5c,
+ * but might be required for some pre-1.0.5c applications.
+ */
+#if !defined(PNG_USE_LOCAL_ARRAYS) && !defined(PNG_USE_GLOBAL_ARRAYS)
+#  if defined(PNG_NO_GLOBAL_ARRAYS) || (defined(__GNUC__) && defined(PNG_DLL))
+#    define PNG_USE_LOCAL_ARRAYS
+#  else
+#    define PNG_USE_GLOBAL_ARRAYS
+#  endif
+#endif
+
+
+#ifndef PNGAPI
+
+#if defined(__MINGW32__) || defined(__CYGWIN__) && !defined(PNG_MODULEDEF)
+#  ifndef PNG_NO_MODULEDEF
+#    define PNG_NO_MODULEDEF
+#  endif
+#endif
+
+#if !defined(PNG_IMPEXP) && defined(PNG_BUILD_DLL) && !defined(PNG_NO_MODULEDEF)
+#  define PNG_IMPEXP
+#endif
+
+#if defined(PNG_DLL) || defined(_DLL) || defined(__DLL__ ) || \
+    (( defined(_Windows) || defined(_WINDOWS) || \
+       defined(WIN32) || defined(_WIN32) || defined(__WIN32__) \
+	  ) && !defined(__CYGWIN__))
+
+#  if defined(__GNUC__) || (defined (_MSC_VER) && (_MSC_VER >= 800))
+#    define PNGAPI __cdecl
+#  else
+#    define PNGAPI _cdecl
+#  endif
+
+#  if !defined(PNG_IMPEXP) && (!defined(PNG_DLL) || \
+       0 /* WINCOMPILER_WITH_NO_SUPPORT_FOR_DECLIMPEXP */)
+#     define PNG_IMPEXP
+#  endif
+
+#  if !defined(PNG_IMPEXP)
+
+#     define PNG_EXPORT_TYPE1(type,symbol)  PNG_IMPEXP type PNGAPI symbol
+#     define PNG_EXPORT_TYPE2(type,symbol)  type PNG_IMPEXP PNGAPI symbol
+
+      /* Borland/Microsoft */
+#     if defined(_MSC_VER) || defined(__BORLANDC__)
+#        if (_MSC_VER >= 800) || (__BORLANDC__ >= 0x500)
+#           define PNG_EXPORT PNG_EXPORT_TYPE1
+#        else
+#           define PNG_EXPORT PNG_EXPORT_TYPE2
+#           if defined(PNG_BUILD_DLL)
+#              define PNG_IMPEXP __export
+#           else
+#              define PNG_IMPEXP /*__import*/ /* doesn't exist AFAIK in
+                                                 VC++*/
+#           endif                             /* Exists in Borland C++ for
+                                                 C++ classes (== huge) */
+#        endif
+#     endif
+
+#     if !defined(PNG_IMPEXP)
+#        if defined(PNG_BUILD_DLL)
+#           define PNG_IMPEXP __declspec(dllexport)
+#        else
+#           define PNG_IMPEXP __declspec(dllimport)
+#        endif
+#     endif
+#  endif  /* PNG_IMPEXP */
+#else /* !(DLL || non-cygwin WINDOWS) */
+#  if defined(__CYGWIN__) && !defined(PNG_DLL)
+#    if !defined(PNG_IMPEXP)
+#      define PNG_IMPEXP
+#    endif
+#    define PNGAPI __cdecl
+#  else
+#    if (defined(__IBMC__) || defined(IBMCPP__)) && defined(__OS2__)
+#      define PNGAPI _System
+#      define PNG_IMPEXP
+#    else
+#      if 0 /* ... other platforms, with other meanings */
+#      else
+#        define PNGAPI
+#        define PNG_IMPEXP
+#      endif
+#    endif
+#  endif
+#endif
+#endif
+
+#ifndef PNGAPI
+#  define PNGAPI
+#endif
+#ifndef PNG_IMPEXP
+#  define PNG_IMPEXP
+#endif
+
+#ifndef PNG_EXPORT
+#  define PNG_EXPORT(type,symbol) PNG_IMPEXP type PNGAPI symbol
+#endif
+
+#ifdef PNG_USE_GLOBAL_ARRAYS
+#  ifndef PNG_EXPORT_VAR
+#    define PNG_EXPORT_VAR(type) extern PNG_IMPEXP type
+#  endif
+#endif
+
+/* User may want to use these so they are not in PNG_INTERNAL. Any library
+ * functions that are passed far data must be model independent.
+ */
+
+#ifndef PNG_ABORT
+#  define PNG_ABORT() abort()
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+#  define png_jmpbuf(png_ptr) ((png_ptr)->jmpbuf)
+#else
+#  define png_jmpbuf(png_ptr) \
+   (LIBPNG_WAS_COMPILED_WITH__PNG_SETJMP_NOT_SUPPORTED)
+#endif
+
+#if defined(USE_FAR_KEYWORD)  /* memory model independent fns */
+/* use this to make far-to-near assignments */
+#  define CHECK   1
+#  define NOCHECK 0
+#  define CVT_PTR(ptr) (png_far_to_near(png_ptr,ptr,CHECK))
+#  define CVT_PTR_NOCHECK(ptr) (png_far_to_near(png_ptr,ptr,NOCHECK))
+#  define png_strcpy _fstrcpy
+#  define png_strlen _fstrlen
+#  define png_memcmp _fmemcmp      /* SJT: added */
+#  define png_memcpy _fmemcpy
+#  define png_memset _fmemset
+#else /* use the usual functions */
+#  define CVT_PTR(ptr)         (ptr)
+#  define CVT_PTR_NOCHECK(ptr) (ptr)
+#  define png_strcpy strcpy
+#  define png_strlen strlen
+#  define png_memcmp memcmp     /* SJT: added */
+#  define png_memcpy memcpy
+#  define png_memset memset
+#endif
+/* End of memory model independent support */
+
+/* Just a little check that someone hasn't tried to define something
+ * contradictory.
+ */
+#if (PNG_ZBUF_SIZE > 65536) && defined(PNG_MAX_MALLOC_64K)
+#  undef PNG_ZBUF_SIZE
+#  define PNG_ZBUF_SIZE 65536
+#endif
+
+#ifdef PNG_READ_SUPPORTED
+/* Prior to libpng-1.0.9, this block was in pngasmrd.h */
+#if defined(PNG_INTERNAL)
+
+/* These are the default thresholds before the MMX code kicks in; if either
+ * rowbytes or bitdepth is below the threshold, plain C code is used.  These
+ * can be overridden at runtime via the png_set_mmx_thresholds() call in
+ * libpng 1.2.0 and later.  The values below were chosen by Intel.
+ */
+
+#ifndef PNG_MMX_ROWBYTES_THRESHOLD_DEFAULT
+#  define PNG_MMX_ROWBYTES_THRESHOLD_DEFAULT  128  /*  >=  */
+#endif
+#ifndef PNG_MMX_BITDEPTH_THRESHOLD_DEFAULT
+#  define PNG_MMX_BITDEPTH_THRESHOLD_DEFAULT  9    /*  >=  */   
+#endif
+
+/* Set this in the makefile for VC++ on Pentium, not here. */
+/* Platform must be Pentium.  Makefile must assemble and load pngvcrd.c .
+ * MMX will be detected at run time and used if present.
+ */
+#ifdef PNG_USE_PNGVCRD
+#  define PNG_HAVE_ASSEMBLER_COMBINE_ROW
+#  define PNG_HAVE_ASSEMBLER_READ_INTERLACE
+#  define PNG_HAVE_ASSEMBLER_READ_FILTER_ROW
+#endif
+
+/* Set this in the makefile for gcc/as on Pentium, not here. */
+/* Platform must be Pentium.  Makefile must assemble and load pnggccrd.c .
+ * MMX will be detected at run time and used if present.
+ */
+#ifdef PNG_USE_PNGGCCRD
+#  define PNG_HAVE_ASSEMBLER_COMBINE_ROW
+#  define PNG_HAVE_ASSEMBLER_READ_INTERLACE
+#  define PNG_HAVE_ASSEMBLER_READ_FILTER_ROW
+#endif
+/* - see pnggccrd.c for info about what is currently enabled */
+
+#endif /* PNG_INTERNAL */
+#endif /* PNG_READ_SUPPORTED */
+
+#endif /* PNGCONF_H */
+
diff --git a/png/pngerror.c b/png/pngerror.c
new file mode 100644
index 000000000..b6ad8d866
--- /dev/null
+++ b/png/pngerror.c
@@ -0,0 +1,289 @@
+
+/* pngerror.c - stub functions for i/o and memory allocation
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file provides a location for all error handling.  Users who
+ * need special error handling are expected to write replacement functions
+ * and use png_set_error_fn() to use those functions.  See the instructions
+ * at each function.
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+static void /* PRIVATE */
+png_default_error PNGARG((png_structp png_ptr,
+                                      png_const_charp message));
+static void /* PRIVATE */
+png_default_warning PNGARG((png_structp png_ptr,
+                                        png_const_charp message));
+
+/* This function is called whenever there is a fatal error.  This function
+ * should not be changed.  If there is a need to handle errors differently,
+ * you should supply a replacement error function and use png_set_error_fn()
+ * to replace the error function at run-time.
+ */
+void PNGAPI
+png_error(png_structp png_ptr, png_const_charp message)
+{
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+   char msg[16];
+   if (png_ptr->flags&(PNG_FLAG_STRIP_ERROR_NUMBERS|PNG_FLAG_STRIP_ERROR_TEXT))
+   {
+     int offset = 0;
+     if (*message == '#')
+     {
+         for (offset=1; offset<15; offset++)
+            if (*(message+offset) == ' ')
+                break;
+         if (png_ptr->flags&PNG_FLAG_STRIP_ERROR_TEXT)
+         {
+            int i;
+            for (i=0; i<offset-1; i++)
+               msg[i]=message[i+1];
+            msg[i]='\0';
+            message=msg;
+         }
+         else
+            message+=offset;
+     }
+     else
+     {
+         if (png_ptr->flags&PNG_FLAG_STRIP_ERROR_TEXT)
+         {
+            msg[0]='0';        
+            msg[1]='\0';
+            message=msg;
+         }
+     }
+   }
+#endif
+   if (png_ptr->error_fn != NULL)
+      (*(png_ptr->error_fn))(png_ptr, message);
+
+   /* if the following returns or doesn't exist, use the default function,
+      which will not return */
+   png_default_error(png_ptr, message);
+}
+
+/* This function is called whenever there is a non-fatal error.  This function
+ * should not be changed.  If there is a need to handle warnings differently,
+ * you should supply a replacement warning function and use
+ * png_set_error_fn() to replace the warning function at run-time.
+ */
+void PNGAPI
+png_warning(png_structp png_ptr, png_const_charp message)
+{
+     int offset = 0;
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+   if (png_ptr->flags&(PNG_FLAG_STRIP_ERROR_NUMBERS|PNG_FLAG_STRIP_ERROR_TEXT))
+#endif
+   {
+     if (*message == '#')
+     {
+         for (offset=1; offset<15; offset++)
+            if (*(message+offset) == ' ')
+                break;
+     }
+   }
+   if (png_ptr->warning_fn != NULL)
+      (*(png_ptr->warning_fn))(png_ptr, (png_const_charp)(message+offset));
+   else
+      png_default_warning(png_ptr, (png_const_charp)(message+offset));
+}
+
+/* These utilities are used internally to build an error message that relates
+ * to the current chunk.  The chunk name comes from png_ptr->chunk_name,
+ * this is used to prefix the message.  The message is limited in length
+ * to 63 bytes, the name characters are output as hex digits wrapped in []
+ * if the character is invalid.
+ */
+#define isnonalpha(c) ((c) < 41 || (c) > 122 || ((c) > 90 && (c) < 97))
+static PNG_CONST char png_digit[16] = {
+   '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E',
+   'F' };
+
+static void /* PRIVATE */
+png_format_buffer(png_structp png_ptr, png_charp buffer, png_const_charp
+   message)
+{
+   int iout = 0, iin = 0;
+
+   while (iin < 4)
+   {
+      int c = png_ptr->chunk_name[iin++];
+      if (isnonalpha(c))
+      {
+         buffer[iout++] = '[';
+         buffer[iout++] = png_digit[(c & 0xf0) >> 4];
+         buffer[iout++] = png_digit[c & 0x0f];
+         buffer[iout++] = ']';
+      }
+      else
+      {
+         buffer[iout++] = (png_byte)c;
+      }
+   }
+
+   if (message == NULL)
+      buffer[iout] = 0;
+   else
+   {
+      buffer[iout++] = ':';
+      buffer[iout++] = ' ';
+      png_memcpy(buffer+iout, message, 64);
+      buffer[iout+63] = 0;
+   }
+}
+
+void PNGAPI
+png_chunk_error(png_structp png_ptr, png_const_charp message)
+{
+   char msg[18+64];
+   png_format_buffer(png_ptr, msg, message);
+   png_error(png_ptr, msg);
+}
+
+void PNGAPI
+png_chunk_warning(png_structp png_ptr, png_const_charp message)
+{
+   char msg[18+64];
+   png_format_buffer(png_ptr, msg, message);
+   png_warning(png_ptr, msg);
+}
+
+/* This is the default error handling function.  Note that replacements for
+ * this function MUST NOT RETURN, or the program will likely crash.  This
+ * function is used by default, or if the program supplies NULL for the
+ * error function pointer in png_set_error_fn().
+ */
+static void /* PRIVATE */
+png_default_error(png_structp png_ptr, png_const_charp message)
+{
+#ifndef PNG_NO_CONSOLE_IO
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+   if (*message == '#')
+   {
+     int offset;
+     char error_number[16];
+     for (offset=0; offset<15; offset++)
+     {
+         error_number[offset] = *(message+offset+1);
+         if (*(message+offset) == ' ')
+             break;
+     }
+     if((offset > 1) && (offset < 15))
+     {
+       error_number[offset-1]='\0';
+       fprintf(stderr, "libpng error no. %s: %s\n", error_number, message+offset);
+     }
+     else
+       fprintf(stderr, "libpng error: %s, offset=%d\n", message,offset);
+   }
+   else
+#endif
+   fprintf(stderr, "libpng error: %s\n", message);
+#else
+   if (message)
+     /* make compiler happy */ ;
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+#  ifdef USE_FAR_KEYWORD
+   {
+      jmp_buf jmpbuf;
+      png_memcpy(jmpbuf,png_ptr->jmpbuf,sizeof(jmp_buf));
+      longjmp(jmpbuf, 1);
+   }
+#  else
+   longjmp(png_ptr->jmpbuf, 1);
+# endif
+#else
+   if (png_ptr)
+     /* make compiler happy */ ;
+   PNG_ABORT();
+#endif
+}
+
+/* This function is called when there is a warning, but the library thinks
+ * it can continue anyway.  Replacement functions don't have to do anything
+ * here if you don't want them to.  In the default configuration, png_ptr is
+ * not used, but it is passed in case it may be useful.
+ */
+static void /* PRIVATE */
+png_default_warning(png_structp png_ptr, png_const_charp message)
+{
+#ifndef PNG_NO_CONSOLE_IO
+#  ifdef PNG_ERROR_NUMBERS_SUPPORTED
+   if (*message == '#')
+   {
+     int offset;
+     char warning_number[16];
+     for (offset=0; offset<15; offset++)
+     {
+        warning_number[offset]=*(message+offset+1);
+        if (*(message+offset) == ' ')
+            break;
+     }
+     if((offset > 1) && (offset < 15))
+     {
+       warning_number[offset-1]='\0';
+       fprintf(stderr, "libpng warning no. %s: %s\n", warning_number,
+          message+offset);
+     }
+     else
+       fprintf(stderr, "libpng warning: %s\n", message);
+   }
+   else
+#  endif
+     fprintf(stderr, "libpng warning: %s\n", message);
+#else
+   if (message)
+     /* appease compiler */ ;
+#endif
+   if (png_ptr)
+      return;
+}
+
+/* This function is called when the application wants to use another method
+ * of handling errors and warnings.  Note that the error function MUST NOT
+ * return to the calling routine or serious problems will occur.  The return
+ * method used in the default routine calls longjmp(png_ptr->jmpbuf, 1)
+ */
+void PNGAPI
+png_set_error_fn(png_structp png_ptr, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warning_fn)
+{
+   png_ptr->error_ptr = error_ptr;
+   png_ptr->error_fn = error_fn;
+   png_ptr->warning_fn = warning_fn;
+}
+
+
+/* This function returns a pointer to the error_ptr associated with the user
+ * functions.  The application should free any memory associated with this
+ * pointer before png_write_destroy and png_read_destroy are called.
+ */
+png_voidp PNGAPI
+png_get_error_ptr(png_structp png_ptr)
+{
+   return ((png_voidp)png_ptr->error_ptr);
+}
+
+
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+void
+png_set_strip_error_numbers(png_structp png_ptr, png_uint_32 strip_mode)
+{
+   if(png_ptr != NULL)
+   {
+     png_ptr->flags &=
+       ((~(PNG_FLAG_STRIP_ERROR_NUMBERS|PNG_FLAG_STRIP_ERROR_TEXT))&strip_mode);
+   }
+}
+#endif
diff --git a/png/pngget.c b/png/pngget.c
new file mode 100644
index 000000000..386fe55ea
--- /dev/null
+++ b/png/pngget.c
@@ -0,0 +1,917 @@
+
+/* pngget.c - retrieval of values from info struct
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+png_uint_32 PNGAPI
+png_get_valid(png_structp png_ptr, png_infop info_ptr, png_uint_32 flag)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+      return(info_ptr->valid & flag);
+   else
+      return(0);
+}
+
+png_uint_32 PNGAPI
+png_get_rowbytes(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+      return(info_ptr->rowbytes);
+   else
+      return(0);
+}
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+png_bytepp PNGAPI
+png_get_rows(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+      return(info_ptr->row_pointers);
+   else
+      return(0);
+}
+#endif
+
+#ifdef PNG_EASY_ACCESS_SUPPORTED
+/* easy access to info, added in libpng-0.99 */
+png_uint_32 PNGAPI
+png_get_image_width(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->width;
+   }
+   return (0);
+}
+
+png_uint_32 PNGAPI
+png_get_image_height(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->height;
+   }
+   return (0);
+}
+
+png_byte PNGAPI
+png_get_bit_depth(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->bit_depth;
+   }
+   return (0);
+}
+
+png_byte PNGAPI
+png_get_color_type(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->color_type;
+   }
+   return (0);
+}
+
+png_byte PNGAPI
+png_get_filter_type(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->filter_type;
+   }
+   return (0);
+}
+
+png_byte PNGAPI
+png_get_interlace_type(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->interlace_type;
+   }
+   return (0);
+}
+
+png_byte PNGAPI
+png_get_compression_type(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+   {
+      return info_ptr->compression_type;
+   }
+   return (0);
+}
+
+png_uint_32 PNGAPI
+png_get_x_pixels_per_meter(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_pHYs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_pHYs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_x_pixels_per_meter");
+      if(info_ptr->phys_unit_type != PNG_RESOLUTION_METER)
+          return (0);
+      else return (info_ptr->x_pixels_per_unit);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+png_uint_32 PNGAPI
+png_get_y_pixels_per_meter(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_pHYs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_pHYs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_y_pixels_per_meter");
+      if(info_ptr->phys_unit_type != PNG_RESOLUTION_METER)
+          return (0);
+      else return (info_ptr->y_pixels_per_unit);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+png_uint_32 PNGAPI
+png_get_pixels_per_meter(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_pHYs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_pHYs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_pixels_per_meter");
+      if(info_ptr->phys_unit_type != PNG_RESOLUTION_METER ||
+         info_ptr->x_pixels_per_unit != info_ptr->y_pixels_per_unit)
+          return (0);
+      else return (info_ptr->x_pixels_per_unit);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+float PNGAPI
+png_get_pixel_aspect_ratio(png_structp png_ptr, png_infop info_ptr)
+   {
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_pHYs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_pHYs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_aspect_ratio");
+      if (info_ptr->x_pixels_per_unit == 0)
+         return ((float)0.0);
+      else
+         return ((float)((float)info_ptr->y_pixels_per_unit
+            /(float)info_ptr->x_pixels_per_unit));
+   }
+#else
+   return (0.0);
+#endif
+   return ((float)0.0);
+}
+#endif
+
+png_int_32 PNGAPI
+png_get_x_offset_microns(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_oFFs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_oFFs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_x_offset_microns");
+      if(info_ptr->offset_unit_type != PNG_OFFSET_MICROMETER)
+          return (0);
+      else return (info_ptr->x_offset);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+png_int_32 PNGAPI
+png_get_y_offset_microns(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_oFFs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_oFFs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_y_offset_microns");
+      if(info_ptr->offset_unit_type != PNG_OFFSET_MICROMETER)
+          return (0);
+      else return (info_ptr->y_offset);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+png_int_32 PNGAPI
+png_get_x_offset_pixels(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_oFFs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_oFFs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_x_offset_microns");
+      if(info_ptr->offset_unit_type != PNG_OFFSET_PIXEL)
+          return (0);
+      else return (info_ptr->x_offset);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+png_int_32 PNGAPI
+png_get_y_offset_pixels(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+#if defined(PNG_oFFs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_oFFs)
+   {
+      png_debug1(1, "in %s retrieval function\n", "png_get_y_offset_microns");
+      if(info_ptr->offset_unit_type != PNG_OFFSET_PIXEL)
+          return (0);
+      else return (info_ptr->y_offset);
+   }
+#else
+   return (0);
+#endif
+   return (0);
+}
+
+#if defined(PNG_INCH_CONVERSIONS) && defined(PNG_FLOATING_POINT_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_pixels_per_inch(png_structp png_ptr, png_infop info_ptr)
+{
+   return ((png_uint_32)((float)png_get_pixels_per_meter(png_ptr, info_ptr)
+     *.0254 +.5));
+}
+
+png_uint_32 PNGAPI
+png_get_x_pixels_per_inch(png_structp png_ptr, png_infop info_ptr)
+{
+   return ((png_uint_32)((float)png_get_x_pixels_per_meter(png_ptr, info_ptr)
+     *.0254 +.5));
+}
+
+png_uint_32 PNGAPI
+png_get_y_pixels_per_inch(png_structp png_ptr, png_infop info_ptr)
+{
+   return ((png_uint_32)((float)png_get_y_pixels_per_meter(png_ptr, info_ptr)
+     *.0254 +.5));
+}
+
+float PNGAPI
+png_get_x_offset_inches(png_structp png_ptr, png_infop info_ptr)
+{
+   return ((float)png_get_x_offset_microns(png_ptr, info_ptr)
+     *.00003937);
+}
+
+float PNGAPI
+png_get_y_offset_inches(png_structp png_ptr, png_infop info_ptr)
+{
+   return ((float)png_get_y_offset_microns(png_ptr, info_ptr)
+     *.00003937);
+}
+
+#if defined(PNG_pHYs_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_pHYs_dpi(png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 *res_x, png_uint_32 *res_y, int *unit_type)
+{
+   png_uint_32 retval = 0;
+
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_pHYs))
+   {
+      png_debug1(1, "in %s retrieval function\n", "pHYs");
+      if (res_x != NULL)
+      {
+         *res_x = info_ptr->x_pixels_per_unit;
+         retval |= PNG_INFO_pHYs;
+      }
+      if (res_y != NULL)
+      {
+         *res_y = info_ptr->y_pixels_per_unit;
+         retval |= PNG_INFO_pHYs;
+      }
+      if (unit_type != NULL)
+      {
+         *unit_type = (int)info_ptr->phys_unit_type;
+         retval |= PNG_INFO_pHYs;
+         if(*unit_type == 1)
+         {
+            if (res_x != NULL) *res_x = (png_uint_32)(*res_x * .0254 + .50);
+            if (res_y != NULL) *res_y = (png_uint_32)(*res_y * .0254 + .50);
+         }
+      }
+   }
+   return (retval);
+}
+#endif /* PNG_pHYs_SUPPORTED */
+#endif  /* PNG_INCH_CONVERSIONS && PNG_FLOATING_POINT_SUPPORTED */
+
+/* png_get_channels really belongs in here, too, but it's been around longer */
+
+#endif  /* PNG_EASY_ACCESS_SUPPORTED */
+
+png_byte PNGAPI
+png_get_channels(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+      return(info_ptr->channels);
+   else
+      return (0);
+}
+
+png_bytep PNGAPI
+png_get_signature(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr != NULL && info_ptr != NULL)
+      return(info_ptr->signature);
+   else
+      return (NULL);
+}
+
+#if defined(PNG_bKGD_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_bKGD(png_structp png_ptr, png_infop info_ptr,
+   png_color_16p *background)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_bKGD)
+      && background != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "bKGD");
+      *background = &(info_ptr->background);
+      return (PNG_INFO_bKGD);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_cHRM_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+png_uint_32 PNGAPI
+png_get_cHRM(png_structp png_ptr, png_infop info_ptr,
+   double *white_x, double *white_y, double *red_x, double *red_y,
+   double *green_x, double *green_y, double *blue_x, double *blue_y)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_cHRM))
+   {
+      png_debug1(1, "in %s retrieval function\n", "cHRM");
+      if (white_x != NULL)
+         *white_x = (double)info_ptr->x_white;
+      if (white_y != NULL)
+         *white_y = (double)info_ptr->y_white;
+      if (red_x != NULL)
+         *red_x = (double)info_ptr->x_red;
+      if (red_y != NULL)
+         *red_y = (double)info_ptr->y_red;
+      if (green_x != NULL)
+         *green_x = (double)info_ptr->x_green;
+      if (green_y != NULL)
+         *green_y = (double)info_ptr->y_green;
+      if (blue_x != NULL)
+         *blue_x = (double)info_ptr->x_blue;
+      if (blue_y != NULL)
+         *blue_y = (double)info_ptr->y_blue;
+      return (PNG_INFO_cHRM);
+   }
+   return (0);
+}
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+png_uint_32 PNGAPI
+png_get_cHRM_fixed(png_structp png_ptr, png_infop info_ptr,
+   png_fixed_point *white_x, png_fixed_point *white_y, png_fixed_point *red_x,
+   png_fixed_point *red_y, png_fixed_point *green_x, png_fixed_point *green_y,
+   png_fixed_point *blue_x, png_fixed_point *blue_y)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_cHRM))
+   {
+      png_debug1(1, "in %s retrieval function\n", "cHRM");
+      if (white_x != NULL)
+         *white_x = info_ptr->int_x_white;
+      if (white_y != NULL)
+         *white_y = info_ptr->int_y_white;
+      if (red_x != NULL)
+         *red_x = info_ptr->int_x_red;
+      if (red_y != NULL)
+         *red_y = info_ptr->int_y_red;
+      if (green_x != NULL)
+         *green_x = info_ptr->int_x_green;
+      if (green_y != NULL)
+         *green_y = info_ptr->int_y_green;
+      if (blue_x != NULL)
+         *blue_x = info_ptr->int_x_blue;
+      if (blue_y != NULL)
+         *blue_y = info_ptr->int_y_blue;
+      return (PNG_INFO_cHRM);
+   }
+   return (0);
+}
+#endif
+#endif
+
+#if defined(PNG_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+png_uint_32 PNGAPI
+png_get_gAMA(png_structp png_ptr, png_infop info_ptr, double *file_gamma)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_gAMA)
+      && file_gamma != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "gAMA");
+      *file_gamma = (double)info_ptr->gamma;
+      return (PNG_INFO_gAMA);
+   }
+   return (0);
+}
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+png_uint_32 PNGAPI
+png_get_gAMA_fixed(png_structp png_ptr, png_infop info_ptr,
+    png_fixed_point *int_file_gamma)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_gAMA)
+      && int_file_gamma != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "gAMA");
+      *int_file_gamma = info_ptr->int_gamma;
+      return (PNG_INFO_gAMA);
+   }
+   return (0);
+}
+#endif
+#endif
+
+#if defined(PNG_sRGB_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_sRGB(png_structp png_ptr, png_infop info_ptr, int *file_srgb_intent)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_sRGB)
+      && file_srgb_intent != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "sRGB");
+      *file_srgb_intent = (int)info_ptr->srgb_intent;
+      return (PNG_INFO_sRGB);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_iCCP_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_iCCP(png_structp png_ptr, png_infop info_ptr,
+             png_charpp name, int *compression_type,
+             png_charpp profile, png_uint_32 *proflen)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_iCCP)
+      && name != NULL && profile != NULL && proflen != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "iCCP");
+      *name = info_ptr->iccp_name;
+      *profile = info_ptr->iccp_profile;
+      /* compression_type is a dummy so the API won't have to change
+         if we introduce multiple compression types later. */
+      *proflen = (int)info_ptr->iccp_proflen;
+      *compression_type = (int)info_ptr->iccp_compression;
+      return (PNG_INFO_iCCP);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_sPLT_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_sPLT(png_structp png_ptr, png_infop info_ptr,
+             png_sPLT_tpp spalettes)
+{
+   if (png_ptr != NULL && info_ptr != NULL && spalettes != NULL)
+     *spalettes = info_ptr->splt_palettes;
+   return ((png_uint_32)info_ptr->splt_palettes_num);
+}
+#endif
+
+#if defined(PNG_hIST_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_hIST(png_structp png_ptr, png_infop info_ptr, png_uint_16p *hist)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_hIST)
+      && hist != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "hIST");
+      *hist = info_ptr->hist;
+      return (PNG_INFO_hIST);
+   }
+   return (0);
+}
+#endif
+
+png_uint_32 PNGAPI
+png_get_IHDR(png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 *width, png_uint_32 *height, int *bit_depth,
+   int *color_type, int *interlace_type, int *compression_type,
+   int *filter_type)
+
+{
+   if (png_ptr != NULL && info_ptr != NULL && width != NULL && height != NULL &&
+      bit_depth != NULL && color_type != NULL)
+   {
+      int pixel_depth, channels;
+      png_uint_32 rowbytes_per_pixel;
+
+      png_debug1(1, "in %s retrieval function\n", "IHDR");
+      *width = info_ptr->width;
+      *height = info_ptr->height;
+      *bit_depth = info_ptr->bit_depth;
+      *color_type = info_ptr->color_type;
+      if (compression_type != NULL)
+         *compression_type = info_ptr->compression_type;
+      if (filter_type != NULL)
+         *filter_type = info_ptr->filter_type;
+      if (interlace_type != NULL)
+         *interlace_type = info_ptr->interlace_type;
+
+      /* check for potential overflow of rowbytes */
+      if (*color_type == PNG_COLOR_TYPE_PALETTE)
+         channels = 1;
+      else if (*color_type & PNG_COLOR_MASK_COLOR)
+         channels = 3;
+      else
+         channels = 1;
+      if (*color_type & PNG_COLOR_MASK_ALPHA)
+         channels++;
+      pixel_depth = *bit_depth * channels;
+      rowbytes_per_pixel = (pixel_depth + 7) >> 3;
+      if ((*width > (PNG_MAX_UINT/rowbytes_per_pixel) - 64))
+      {
+         png_error(png_ptr,
+            "Width too large for libpng to process image data.");
+      }
+      return (1);
+   }
+   return (0);
+}
+
+#if defined(PNG_oFFs_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_oFFs(png_structp png_ptr, png_infop info_ptr,
+   png_int_32 *offset_x, png_int_32 *offset_y, int *unit_type)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_oFFs)
+      && offset_x != NULL && offset_y != NULL && unit_type != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "oFFs");
+      *offset_x = info_ptr->x_offset;
+      *offset_y = info_ptr->y_offset;
+      *unit_type = (int)info_ptr->offset_unit_type;
+      return (PNG_INFO_oFFs);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_pCAL_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_pCAL(png_structp png_ptr, png_infop info_ptr,
+   png_charp *purpose, png_int_32 *X0, png_int_32 *X1, int *type, int *nparams,
+   png_charp *units, png_charpp *params)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_pCAL)
+      && purpose != NULL && X0 != NULL && X1 != NULL && type != NULL &&
+      nparams != NULL && units != NULL && params != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "pCAL");
+      *purpose = info_ptr->pcal_purpose;
+      *X0 = info_ptr->pcal_X0;
+      *X1 = info_ptr->pcal_X1;
+      *type = (int)info_ptr->pcal_type;
+      *nparams = (int)info_ptr->pcal_nparams;
+      *units = info_ptr->pcal_units;
+      *params = info_ptr->pcal_params;
+      return (PNG_INFO_pCAL);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_sCAL_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+png_uint_32 PNGAPI
+png_get_sCAL(png_structp png_ptr, png_infop info_ptr,
+             int *unit, double *width, double *height)
+{
+    if (png_ptr != NULL && info_ptr != NULL &&
+       (info_ptr->valid & PNG_INFO_sCAL))
+    {
+        *unit = info_ptr->scal_unit;
+        *width = info_ptr->scal_pixel_width;
+        *height = info_ptr->scal_pixel_height;
+        return (PNG_INFO_sCAL);
+    }
+    return(0);
+}
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+png_uint_32 PNGAPI
+png_get_sCAL_s(png_structp png_ptr, png_infop info_ptr,
+             int *unit, png_charpp width, png_charpp height)
+{
+    if (png_ptr != NULL && info_ptr != NULL &&
+       (info_ptr->valid & PNG_INFO_sCAL))
+    {
+        *unit = info_ptr->scal_unit;
+        *width = info_ptr->scal_s_width;
+        *height = info_ptr->scal_s_height;
+        return (PNG_INFO_sCAL);
+    }
+    return(0);
+}
+#endif
+#endif
+#endif
+
+#if defined(PNG_pHYs_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_pHYs(png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 *res_x, png_uint_32 *res_y, int *unit_type)
+{
+   png_uint_32 retval = 0;
+
+   if (png_ptr != NULL && info_ptr != NULL &&
+      (info_ptr->valid & PNG_INFO_pHYs))
+   {
+      png_debug1(1, "in %s retrieval function\n", "pHYs");
+      if (res_x != NULL)
+      {
+         *res_x = info_ptr->x_pixels_per_unit;
+         retval |= PNG_INFO_pHYs;
+      }
+      if (res_y != NULL)
+      {
+         *res_y = info_ptr->y_pixels_per_unit;
+         retval |= PNG_INFO_pHYs;
+      }
+      if (unit_type != NULL)
+      {
+         *unit_type = (int)info_ptr->phys_unit_type;
+         retval |= PNG_INFO_pHYs;
+      }
+   }
+   return (retval);
+}
+#endif
+
+png_uint_32 PNGAPI
+png_get_PLTE(png_structp png_ptr, png_infop info_ptr, png_colorp *palette,
+   int *num_palette)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_PLTE)
+       && palette != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "PLTE");
+      *palette = info_ptr->palette;
+      *num_palette = info_ptr->num_palette;
+      png_debug1(3, "num_palette = %d\n", *num_palette);
+      return (PNG_INFO_PLTE);
+   }
+   return (0);
+}
+
+#if defined(PNG_sBIT_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_sBIT(png_structp png_ptr, png_infop info_ptr, png_color_8p *sig_bit)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_sBIT)
+      && sig_bit != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "sBIT");
+      *sig_bit = &(info_ptr->sig_bit);
+      return (PNG_INFO_sBIT);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_TEXT_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_text(png_structp png_ptr, png_infop info_ptr, png_textp *text_ptr,
+   int *num_text)
+{
+   if (png_ptr != NULL && info_ptr != NULL && info_ptr->num_text > 0)
+   {
+      png_debug1(1, "in %s retrieval function\n",
+         (png_ptr->chunk_name[0] == '\0' ? "text"
+             : (png_const_charp)png_ptr->chunk_name));
+      if (text_ptr != NULL)
+         *text_ptr = info_ptr->text;
+      if (num_text != NULL)
+         *num_text = info_ptr->num_text;
+      return ((png_uint_32)info_ptr->num_text);
+   }
+   if (num_text != NULL)
+     *num_text = 0;
+   return(0);
+}
+#endif
+
+#if defined(PNG_tIME_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_tIME(png_structp png_ptr, png_infop info_ptr, png_timep *mod_time)
+{
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_tIME)
+       && mod_time != NULL)
+   {
+      png_debug1(1, "in %s retrieval function\n", "tIME");
+      *mod_time = &(info_ptr->mod_time);
+      return (PNG_INFO_tIME);
+   }
+   return (0);
+}
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_tRNS(png_structp png_ptr, png_infop info_ptr,
+   png_bytep *trans, int *num_trans, png_color_16p *trans_values)
+{
+   png_uint_32 retval = 0;
+   if (png_ptr != NULL && info_ptr != NULL && (info_ptr->valid & PNG_INFO_tRNS))
+   {
+      png_debug1(1, "in %s retrieval function\n", "tRNS");
+      if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+          if (trans != NULL)
+          {
+             *trans = info_ptr->trans;
+             retval |= PNG_INFO_tRNS;
+          }
+          if (trans_values != NULL)
+             *trans_values = &(info_ptr->trans_values);
+      }
+      else /* if (info_ptr->color_type != PNG_COLOR_TYPE_PALETTE) */
+      {
+          if (trans_values != NULL)
+          {
+             *trans_values = &(info_ptr->trans_values);
+             retval |= PNG_INFO_tRNS;
+          }
+          if(trans != NULL)
+             *trans = NULL;
+      }
+      if(num_trans != NULL)
+      {
+         *num_trans = info_ptr->num_trans;
+         retval |= PNG_INFO_tRNS;
+      }
+   }
+   return (retval);
+}
+#endif
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+png_uint_32 PNGAPI
+png_get_unknown_chunks(png_structp png_ptr, png_infop info_ptr,
+             png_unknown_chunkpp unknowns)
+{
+   if (png_ptr != NULL && info_ptr != NULL && unknowns != NULL)
+     *unknowns = info_ptr->unknown_chunks;
+   return ((png_uint_32)info_ptr->unknown_chunks_num);
+}
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+png_byte PNGAPI
+png_get_rgb_to_gray_status (png_structp png_ptr)
+{
+   return (png_byte)(png_ptr? png_ptr->rgb_to_gray_status : 0);
+}
+#endif
+
+#if defined(PNG_USER_CHUNKS_SUPPORTED)
+png_voidp PNGAPI
+png_get_user_chunk_ptr(png_structp png_ptr)
+{
+   return (png_ptr? png_ptr->user_chunk_ptr : NULL);
+}
+#endif
+
+
+png_uint_32 PNGAPI
+png_get_compression_buffer_size(png_structp png_ptr)
+{
+   return (png_uint_32)(png_ptr? png_ptr->zbuf_size : 0L);
+}
+
+
+#ifdef PNG_ASSEMBLER_CODE_SUPPORTED
+/* this function was added to libpng 1.2.0 and should exist by default*/
+png_uint_32 PNGAPI
+png_get_asm_flags (png_structp png_ptr)
+{
+    return (png_uint_32)(png_ptr? png_ptr->asm_flags : 0L);
+}
+
+/* this function was added to libpng 1.2.0 and should exist by default */
+png_uint_32 PNGAPI
+png_get_asm_flagmask (int flag_select)
+{
+    png_uint_32 settable_asm_flags = 0;
+
+    if (flag_select & PNG_SELECT_READ)
+        settable_asm_flags |=
+          PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  |
+          PNG_ASM_FLAG_MMX_READ_INTERLACE    |
+          PNG_ASM_FLAG_MMX_READ_FILTER_SUB   |
+          PNG_ASM_FLAG_MMX_READ_FILTER_UP    |
+          PNG_ASM_FLAG_MMX_READ_FILTER_AVG   |
+          PNG_ASM_FLAG_MMX_READ_FILTER_PAETH ;
+          /* no non-MMX flags yet */
+
+#if 0
+    /* GRR:  no write-flags yet, either, but someday... */
+    if (flag_select & PNG_SELECT_WRITE)
+        settable_asm_flags |=
+          PNG_ASM_FLAG_MMX_WRITE_ [whatever] ;
+#endif /* 0 */
+
+    return settable_asm_flags;  /* _theoretically_ settable capabilities only */
+}
+#endif /* PNG_ASSEMBLER_CODE_SUPPORTED */
+
+
+#if defined(PNG_ASSEMBLER_CODE_SUPPORTED)
+    /* GRR:  could add this:   && defined(PNG_MMX_CODE_SUPPORTED) */
+/* this function was added to libpng 1.2.0 */
+png_uint_32 PNGAPI
+png_get_mmx_flagmask (int flag_select, int *compilerID)
+{
+    png_uint_32 settable_mmx_flags = 0;
+
+    if (flag_select & PNG_SELECT_READ)
+        settable_mmx_flags |=
+          PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  |
+          PNG_ASM_FLAG_MMX_READ_INTERLACE    |
+          PNG_ASM_FLAG_MMX_READ_FILTER_SUB   |
+          PNG_ASM_FLAG_MMX_READ_FILTER_UP    |
+          PNG_ASM_FLAG_MMX_READ_FILTER_AVG   |
+          PNG_ASM_FLAG_MMX_READ_FILTER_PAETH ;
+#if 0
+    /* GRR:  no MMX write support yet, but someday... */
+    if (flag_select & PNG_SELECT_WRITE)
+        settable_mmx_flags |=
+          PNG_ASM_FLAG_MMX_WRITE_ [whatever] ;
+#endif /* 0 */
+
+    if (compilerID != NULL) {
+#ifdef PNG_USE_PNGVCRD
+        *compilerID = 1;    /* MSVC */
+#else
+#ifdef PNG_USE_PNGGCCRD
+        *compilerID = 2;    /* gcc/gas */
+#else
+        *compilerID = -1;   /* unknown (i.e., no asm/MMX code compiled) */
+#endif
+#endif
+    }
+
+    return settable_mmx_flags;  /* _theoretically_ settable capabilities only */
+}
+
+/* this function was added to libpng 1.2.0 */
+png_byte PNGAPI
+png_get_mmx_bitdepth_threshold (png_structp png_ptr)
+{
+    return (png_byte)(png_ptr? png_ptr->mmx_bitdepth_threshold : 0);
+}
+
+/* this function was added to libpng 1.2.0 */
+png_uint_32 PNGAPI
+png_get_mmx_rowbytes_threshold (png_structp png_ptr)
+{
+    return (png_uint_32)(png_ptr? png_ptr->mmx_rowbytes_threshold : 0L);
+}
+#endif /* PNG_ASSEMBLER_CODE_SUPPORTED */
diff --git a/png/pngmem.c b/png/pngmem.c
new file mode 100644
index 000000000..b1741a407
--- /dev/null
+++ b/png/pngmem.c
@@ -0,0 +1,517 @@
+
+/* pngmem.c - stub functions for memory allocation
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file provides a location for all memory allocation.  Users who
+ * need special memory handling are expected to supply replacement
+ * functions for png_malloc() and png_free(), and to use
+ * png_create_read_struct_2() and png_create_write_struct_2() to
+ * identify the replacement functions.
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+/* Borland DOS special memory handler */
+#if defined(__TURBOC__) && !defined(_Windows) && !defined(__FLAT__)
+/* if you change this, be sure to change the one in png.h also */
+
+/* Allocate memory for a png_struct.  The malloc and memset can be replaced
+   by a single call to calloc() if this is thought to improve performance. */
+png_voidp /* PRIVATE */
+png_create_struct(int type)
+{
+#ifdef PNG_USER_MEM_SUPPORTED
+   return (png_create_struct_2(type, png_malloc_ptr_NULL, png_voidp_NULL));
+}
+
+/* Alternate version of png_create_struct, for use with user-defined malloc. */
+png_voidp /* PRIVATE */
+png_create_struct_2(int type, png_malloc_ptr malloc_fn, png_voidp mem_ptr)
+{
+#endif /* PNG_USER_MEM_SUPPORTED */
+   png_size_t size;
+   png_voidp struct_ptr;
+
+   if (type == PNG_STRUCT_INFO)
+     size = sizeof(png_info);
+   else if (type == PNG_STRUCT_PNG)
+     size = sizeof(png_struct);
+   else
+     return (NULL);
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   if(malloc_fn != NULL)
+   {
+      png_struct dummy_struct;
+      png_structp png_ptr = &dummy_struct;
+      png_ptr->mem_ptr=mem_ptr;
+      struct_ptr = (*(malloc_fn))(png_ptr, (png_uint_32)size);
+   }
+   else
+#endif /* PNG_USER_MEM_SUPPORTED */
+      struct_ptr = (png_voidp)farmalloc(size));
+   if (struct_ptr != NULL)
+      png_memset(struct_ptr, 0, size);
+   return (struct_ptr);
+}
+
+/* Free memory allocated by a png_create_struct() call */
+void /* PRIVATE */
+png_destroy_struct(png_voidp struct_ptr)
+{
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_destroy_struct_2(struct_ptr, png_free_ptr_NULL, png_voidp_NULL);
+}
+
+/* Free memory allocated by a png_create_struct() call */
+void /* PRIVATE */
+png_destroy_struct_2(png_voidp struct_ptr, png_free_ptr free_fn,
+    png_voidp mem_ptr)
+{
+#endif
+   if (struct_ptr != NULL)
+   {
+#ifdef PNG_USER_MEM_SUPPORTED
+      if(free_fn != NULL)
+      {
+         png_struct dummy_struct;
+         png_structp png_ptr = &dummy_struct;
+         png_ptr->mem_ptr=mem_ptr;
+         (*(free_fn))(png_ptr, struct_ptr);
+         return;
+      }
+#endif /* PNG_USER_MEM_SUPPORTED */
+      farfree (struct_ptr);
+   }
+}
+
+/* Allocate memory.  For reasonable files, size should never exceed
+ * 64K.  However, zlib may allocate more then 64K if you don't tell
+ * it not to.  See zconf.h and png.h for more information. zlib does
+ * need to allocate exactly 64K, so whatever you call here must
+ * have the ability to do that.
+ *
+ * Borland seems to have a problem in DOS mode for exactly 64K.
+ * It gives you a segment with an offset of 8 (perhaps to store its
+ * memory stuff).  zlib doesn't like this at all, so we have to
+ * detect and deal with it.  This code should not be needed in
+ * Windows or OS/2 modes, and only in 16 bit mode.  This code has
+ * been updated by Alexander Lehmann for version 0.89 to waste less
+ * memory.
+ *
+ * Note that we can't use png_size_t for the "size" declaration,
+ * since on some systems a png_size_t is a 16-bit quantity, and as a
+ * result, we would be truncating potentially larger memory requests
+ * (which should cause a fatal error) and introducing major problems.
+ */
+png_voidp PNGAPI
+png_malloc(png_structp png_ptr, png_uint_32 size)
+{
+#ifndef PNG_USER_MEM_SUPPORTED
+   png_voidp ret;
+#endif
+   if (png_ptr == NULL || size == 0)
+      return (NULL);
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   if(png_ptr->malloc_fn != NULL)
+   {
+       ret = ((png_voidp)(*(png_ptr->malloc_fn))(png_ptr, (png_size_t)size));
+       if (ret == NULL)
+          png_error(png_ptr, "Out of memory!");
+       return (ret);
+   }
+   else
+       return png_malloc_default(png_ptr, size);
+}
+
+png_voidp PNGAPI
+png_malloc_default(png_structp png_ptr, png_uint_32 size)
+{
+   png_voidp ret;
+#endif /* PNG_USER_MEM_SUPPORTED */
+
+#ifdef PNG_MAX_MALLOC_64K
+   if (size > (png_uint_32)65536L)
+      png_error(png_ptr, "Cannot Allocate > 64K");
+#endif
+
+   if (size == (png_uint_32)65536L)
+   {
+      if (png_ptr->offset_table == NULL)
+      {
+         /* try to see if we need to do any of this fancy stuff */
+         ret = farmalloc(size);
+         if (ret == NULL || ((png_size_t)ret & 0xffff))
+         {
+            int num_blocks;
+            png_uint_32 total_size;
+            png_bytep table;
+            int i;
+            png_byte huge * hptr;
+
+            if (ret != NULL)
+            {
+               farfree(ret);
+               ret = NULL;
+            }
+
+            if(png_ptr->zlib_window_bits > 14)
+               num_blocks = (int)(1 << (png_ptr->zlib_window_bits - 14));
+            else
+               num_blocks = 1;
+            if (png_ptr->zlib_mem_level >= 7)
+               num_blocks += (int)(1 << (png_ptr->zlib_mem_level - 7));
+            else
+               num_blocks++;
+
+            total_size = ((png_uint_32)65536L) * (png_uint_32)num_blocks+16;
+
+            table = farmalloc(total_size);
+
+            if (table == NULL)
+            {
+               png_error(png_ptr, "Out Of Memory."); /* Note "O" and "M" */
+            }
+
+            if ((png_size_t)table & 0xfff0)
+            {
+               png_error(png_ptr, "Farmalloc didn't return normalized pointer");
+            }
+
+            png_ptr->offset_table = table;
+            png_ptr->offset_table_ptr = farmalloc(num_blocks *
+               sizeof (png_bytep));
+
+            if (png_ptr->offset_table_ptr == NULL)
+            {
+               png_error(png_ptr, "Out Of memory.");
+            }
+
+            hptr = (png_byte huge *)table;
+            if ((png_size_t)hptr & 0xf)
+            {
+               hptr = (png_byte huge *)((long)(hptr) & 0xfffffff0L);
+               hptr = hptr + 16L;  /* "hptr += 16L" fails on Turbo C++ 3.0 */
+            }
+            for (i = 0; i < num_blocks; i++)
+            {
+               png_ptr->offset_table_ptr[i] = (png_bytep)hptr;
+               hptr = hptr + (png_uint_32)65536L;  /* "+=" fails on TC++3.0 */
+            }
+
+            png_ptr->offset_table_number = num_blocks;
+            png_ptr->offset_table_count = 0;
+            png_ptr->offset_table_count_free = 0;
+         }
+      }
+
+      if (png_ptr->offset_table_count >= png_ptr->offset_table_number)
+         png_error(png_ptr, "Out of Memory.");
+
+      ret = png_ptr->offset_table_ptr[png_ptr->offset_table_count++];
+   }
+   else
+      ret = farmalloc(size);
+
+   if (ret == NULL)
+   {
+      png_error(png_ptr, "Out of memory."); /* Note "o" and "m" */
+   }
+
+   return (ret);
+}
+
+/* free a pointer allocated by png_malloc().  In the default
+   configuration, png_ptr is not used, but is passed in case it
+   is needed.  If ptr is NULL, return without taking any action. */
+void PNGAPI
+png_free(png_structp png_ptr, png_voidp ptr)
+{
+   if (png_ptr == NULL || ptr == NULL)
+      return;
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   if (png_ptr->free_fn != NULL)
+   {
+      (*(png_ptr->free_fn))(png_ptr, ptr);
+      return;
+   }
+   else png_free_default(png_ptr, ptr);
+}
+
+void PNGAPI
+png_free_default(png_structp png_ptr, png_voidp ptr)
+{
+#endif /* PNG_USER_MEM_SUPPORTED */
+
+   if (png_ptr->offset_table != NULL)
+   {
+      int i;
+
+      for (i = 0; i < png_ptr->offset_table_count; i++)
+      {
+         if (ptr == png_ptr->offset_table_ptr[i])
+         {
+            ptr = NULL;
+            png_ptr->offset_table_count_free++;
+            break;
+         }
+      }
+      if (png_ptr->offset_table_count_free == png_ptr->offset_table_count)
+      {
+         farfree(png_ptr->offset_table);
+         farfree(png_ptr->offset_table_ptr);
+         png_ptr->offset_table = NULL;
+         png_ptr->offset_table_ptr = NULL;
+      }
+   }
+
+   if (ptr != NULL)
+   {
+      farfree(ptr);
+   }
+}
+
+#else /* Not the Borland DOS special memory handler */
+
+/* Allocate memory for a png_struct or a png_info.  The malloc and
+   memset can be replaced by a single call to calloc() if this is thought
+   to improve performance noticably.*/
+png_voidp /* PRIVATE */
+png_create_struct(int type)
+{
+#ifdef PNG_USER_MEM_SUPPORTED
+   return (png_create_struct_2(type, png_malloc_ptr_NULL, png_voidp_NULL));
+}
+
+/* Allocate memory for a png_struct or a png_info.  The malloc and
+   memset can be replaced by a single call to calloc() if this is thought
+   to improve performance noticably.*/
+png_voidp /* PRIVATE */
+png_create_struct_2(int type, png_malloc_ptr malloc_fn, png_voidp mem_ptr)
+{
+#endif /* PNG_USER_MEM_SUPPORTED */
+   png_size_t size;
+   png_voidp struct_ptr;
+
+   if (type == PNG_STRUCT_INFO)
+      size = sizeof(png_info);
+   else if (type == PNG_STRUCT_PNG)
+      size = sizeof(png_struct);
+   else
+      return (NULL);
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   if(malloc_fn != NULL)
+   {
+      png_struct dummy_struct;
+      png_structp png_ptr = &dummy_struct;
+      png_ptr->mem_ptr=mem_ptr;
+      struct_ptr = (*(malloc_fn))(png_ptr, size);
+      if (struct_ptr != NULL)
+         png_memset(struct_ptr, 0, size);
+      return (struct_ptr);
+   }
+#endif /* PNG_USER_MEM_SUPPORTED */
+
+#if defined(__TURBOC__) && !defined(__FLAT__)
+   if ((struct_ptr = (png_voidp)farmalloc(size)) != NULL)
+#else
+# if defined(_MSC_VER) && defined(MAXSEG_64K)
+   if ((struct_ptr = (png_voidp)halloc(size,1)) != NULL)
+# else
+   if ((struct_ptr = (png_voidp)malloc(size)) != NULL)
+# endif
+#endif
+   {
+      png_memset(struct_ptr, 0, size);
+   }
+
+   return (struct_ptr);
+}
+
+
+/* Free memory allocated by a png_create_struct() call */
+void /* PRIVATE */
+png_destroy_struct(png_voidp struct_ptr)
+{
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_destroy_struct_2(struct_ptr, png_free_ptr_NULL, png_voidp_NULL);
+}
+
+/* Free memory allocated by a png_create_struct() call */
+void /* PRIVATE */
+png_destroy_struct_2(png_voidp struct_ptr, png_free_ptr free_fn,
+    png_voidp mem_ptr)
+{
+#endif /* PNG_USER_MEM_SUPPORTED */
+   if (struct_ptr != NULL)
+   {
+#ifdef PNG_USER_MEM_SUPPORTED
+      if(free_fn != NULL)
+      {
+         png_struct dummy_struct;
+         png_structp png_ptr = &dummy_struct;
+         png_ptr->mem_ptr=mem_ptr;
+         (*(free_fn))(png_ptr, struct_ptr);
+         return;
+      }
+#endif /* PNG_USER_MEM_SUPPORTED */
+#if defined(__TURBOC__) && !defined(__FLAT__)
+      farfree(struct_ptr);
+#else
+# if defined(_MSC_VER) && defined(MAXSEG_64K)
+      hfree(struct_ptr);
+# else
+      free(struct_ptr);
+# endif
+#endif
+   }
+}
+
+
+/* Allocate memory.  For reasonable files, size should never exceed
+   64K.  However, zlib may allocate more then 64K if you don't tell
+   it not to.  See zconf.h and png.h for more information.  zlib does
+   need to allocate exactly 64K, so whatever you call here must
+   have the ability to do that. */
+
+png_voidp PNGAPI
+png_malloc(png_structp png_ptr, png_uint_32 size)
+{
+   png_voidp ret;
+   if (png_ptr == NULL || size == 0)
+      return (NULL);
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   if(png_ptr->malloc_fn != NULL)
+   {
+       ret = ((png_voidp)(*(png_ptr->malloc_fn))(png_ptr, size));
+       if (ret == NULL)
+          png_error(png_ptr, "Out of Memory!");
+       return (ret);
+   }
+   else
+       return (png_malloc_default(png_ptr, size));
+}
+png_voidp /* PRIVATE */
+png_malloc_default(png_structp png_ptr, png_uint_32 size)
+{
+   png_voidp ret;
+#endif /* PNG_USER_MEM_SUPPORTED */
+
+#ifdef PNG_MAX_MALLOC_64K
+   if (size > (png_uint_32)65536L)
+      png_error(png_ptr, "Cannot Allocate > 64K");
+#endif
+
+#if defined(__TURBOC__) && !defined(__FLAT__)
+   ret = farmalloc(size);
+#else
+# if defined(_MSC_VER) && defined(MAXSEG_64K)
+   ret = halloc(size, 1);
+# else
+   ret = malloc((size_t)size);
+# endif
+#endif
+
+   if (ret == NULL)
+      png_error(png_ptr, "Out of Memory");
+
+   return (ret);
+}
+
+/* Free a pointer allocated by png_malloc().  If ptr is NULL, return
+   without taking any action. */
+void PNGAPI
+png_free(png_structp png_ptr, png_voidp ptr)
+{
+   if (png_ptr == NULL || ptr == NULL)
+      return;
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   if (png_ptr->free_fn != NULL)
+   {
+      (*(png_ptr->free_fn))(png_ptr, ptr);
+      return;
+   }
+   else png_free_default(png_ptr, ptr);
+}
+void /* PRIVATE */
+png_free_default(png_structp png_ptr, png_voidp ptr)
+{
+   if (png_ptr == NULL || ptr == NULL)
+      return;
+
+#endif /* PNG_USER_MEM_SUPPORTED */
+
+#if defined(__TURBOC__) && !defined(__FLAT__)
+   farfree(ptr);
+#else
+# if defined(_MSC_VER) && defined(MAXSEG_64K)
+   hfree(ptr);
+# else
+   free(ptr);
+# endif
+#endif
+}
+
+#endif /* Not Borland DOS special memory handler */
+
+png_voidp /* PRIVATE */
+png_memcpy_check (png_structp png_ptr, png_voidp s1, png_voidp s2,
+   png_uint_32 length)
+{
+   png_size_t size;
+
+   size = (png_size_t)length;
+   if ((png_uint_32)size != length)
+      png_error(png_ptr,"Overflow in png_memcpy_check.");
+
+   return(png_memcpy (s1, s2, size));
+}
+
+png_voidp /* PRIVATE */
+png_memset_check (png_structp png_ptr, png_voidp s1, int value,
+   png_uint_32 length)
+{
+   png_size_t size;
+
+   size = (png_size_t)length;
+   if ((png_uint_32)size != length)
+      png_error(png_ptr,"Overflow in png_memset_check.");
+
+   return (png_memset (s1, value, size));
+
+}
+
+#ifdef PNG_USER_MEM_SUPPORTED
+/* This function is called when the application wants to use another method
+ * of allocating and freeing memory.
+ */
+void PNGAPI
+png_set_mem_fn(png_structp png_ptr, png_voidp mem_ptr, png_malloc_ptr
+  malloc_fn, png_free_ptr free_fn)
+{
+   png_ptr->mem_ptr = mem_ptr;
+   png_ptr->malloc_fn = malloc_fn;
+   png_ptr->free_fn = free_fn;
+}
+
+/* This function returns a pointer to the mem_ptr associated with the user
+ * functions.  The application should free any memory associated with this
+ * pointer before png_write_destroy and png_read_destroy are called.
+ */
+png_voidp PNGAPI
+png_get_mem_ptr(png_structp png_ptr)
+{
+   return ((png_voidp)png_ptr->mem_ptr);
+}
+#endif /* PNG_USER_MEM_SUPPORTED */
diff --git a/png/pngpread.c b/png/pngpread.c
new file mode 100644
index 000000000..8e93ecf69
--- /dev/null
+++ b/png/pngpread.c
@@ -0,0 +1,1509 @@
+
+/* pngpread.c - read a png file in push mode
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+#ifdef PNG_PROGRESSIVE_READ_SUPPORTED
+
+/* push model modes */
+#define PNG_READ_SIG_MODE   0
+#define PNG_READ_CHUNK_MODE 1
+#define PNG_READ_IDAT_MODE  2
+#define PNG_SKIP_MODE       3
+#define PNG_READ_tEXt_MODE  4
+#define PNG_READ_zTXt_MODE  5
+#define PNG_READ_DONE_MODE  6
+#define PNG_READ_iTXt_MODE  7
+#define PNG_ERROR_MODE      8
+
+void PNGAPI
+png_process_data(png_structp png_ptr, png_infop info_ptr,
+   png_bytep buffer, png_size_t buffer_size)
+{
+   png_push_restore_buffer(png_ptr, buffer, buffer_size);
+
+   while (png_ptr->buffer_size)
+   {
+      png_process_some_data(png_ptr, info_ptr);
+   }
+}
+
+/* What we do with the incoming data depends on what we were previously
+ * doing before we ran out of data...
+ */
+void /* PRIVATE */
+png_process_some_data(png_structp png_ptr, png_infop info_ptr)
+{
+   switch (png_ptr->process_mode)
+   {
+      case PNG_READ_SIG_MODE:
+      {
+         png_push_read_sig(png_ptr, info_ptr);
+         break;
+      }
+      case PNG_READ_CHUNK_MODE:
+      {
+         png_push_read_chunk(png_ptr, info_ptr);
+         break;
+      }
+      case PNG_READ_IDAT_MODE:
+      {
+         png_push_read_IDAT(png_ptr);
+         break;
+      }
+#if defined(PNG_READ_tEXt_SUPPORTED)
+      case PNG_READ_tEXt_MODE:
+      {
+         png_push_read_tEXt(png_ptr, info_ptr);
+         break;
+      }
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+      case PNG_READ_zTXt_MODE:
+      {
+         png_push_read_zTXt(png_ptr, info_ptr);
+         break;
+      }
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+      case PNG_READ_iTXt_MODE:
+      {
+         png_push_read_iTXt(png_ptr, info_ptr);
+         break;
+      }
+#endif
+      case PNG_SKIP_MODE:
+      {
+         png_push_crc_finish(png_ptr);
+         break;
+      }
+      default:
+      {
+         png_ptr->buffer_size = 0;
+         break;
+      }
+   }
+}
+
+/* Read any remaining signature bytes from the stream and compare them with
+ * the correct PNG signature.  It is possible that this routine is called
+ * with bytes already read from the signature, either because they have been
+ * checked by the calling application, or because of multiple calls to this
+ * routine.
+ */
+void /* PRIVATE */
+png_push_read_sig(png_structp png_ptr, png_infop info_ptr)
+{
+   png_size_t num_checked = png_ptr->sig_bytes,
+             num_to_check = 8 - num_checked;
+
+   if (png_ptr->buffer_size < num_to_check)
+   {
+      num_to_check = png_ptr->buffer_size;
+   }
+
+   png_push_fill_buffer(png_ptr, &(info_ptr->signature[num_checked]),
+      num_to_check);
+   png_ptr->sig_bytes = (png_byte)(png_ptr->sig_bytes+num_to_check);
+
+   if (png_sig_cmp(info_ptr->signature, num_checked, num_to_check))
+   {
+      if (num_checked < 4 &&
+          png_sig_cmp(info_ptr->signature, num_checked, num_to_check - 4))
+         png_error(png_ptr, "Not a PNG file");
+      else
+         png_error(png_ptr, "PNG file corrupted by ASCII conversion");
+   }
+   else
+   {
+      if (png_ptr->sig_bytes >= 8)
+      {
+         png_ptr->process_mode = PNG_READ_CHUNK_MODE;
+      }
+   }
+}
+
+void /* PRIVATE */
+png_push_read_chunk(png_structp png_ptr, png_infop info_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+      PNG_IHDR;
+      PNG_IDAT;
+      PNG_IEND;
+      PNG_PLTE;
+#if defined(PNG_READ_bKGD_SUPPORTED)
+      PNG_bKGD;
+#endif
+#if defined(PNG_READ_cHRM_SUPPORTED)
+      PNG_cHRM;
+#endif
+#if defined(PNG_READ_gAMA_SUPPORTED)
+      PNG_gAMA;
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+      PNG_hIST;
+#endif
+#if defined(PNG_READ_iCCP_SUPPORTED)
+      PNG_iCCP;
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+      PNG_iTXt;
+#endif
+#if defined(PNG_READ_oFFs_SUPPORTED)
+      PNG_oFFs;
+#endif
+#if defined(PNG_READ_pCAL_SUPPORTED)
+      PNG_pCAL;
+#endif
+#if defined(PNG_READ_pHYs_SUPPORTED)
+      PNG_pHYs;
+#endif
+#if defined(PNG_READ_sBIT_SUPPORTED)
+      PNG_sBIT;
+#endif
+#if defined(PNG_READ_sCAL_SUPPORTED)
+      PNG_sCAL;
+#endif
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      PNG_sRGB;
+#endif
+#if defined(PNG_READ_sPLT_SUPPORTED)
+      PNG_sPLT;
+#endif
+#if defined(PNG_READ_tEXt_SUPPORTED)
+      PNG_tEXt;
+#endif
+#if defined(PNG_READ_tIME_SUPPORTED)
+      PNG_tIME;
+#endif
+#if defined(PNG_READ_tRNS_SUPPORTED)
+      PNG_tRNS;
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+      PNG_zTXt;
+#endif
+#endif /* PNG_USE_LOCAL_ARRAYS */
+   /* First we make sure we have enough data for the 4 byte chunk name
+    * and the 4 byte chunk length before proceeding with decoding the
+    * chunk data.  To fully decode each of these chunks, we also make
+    * sure we have enough data in the buffer for the 4 byte CRC at the
+    * end of every chunk (except IDAT, which is handled separately).
+    */
+   if (!(png_ptr->mode & PNG_HAVE_CHUNK_HEADER))
+   {
+      png_byte chunk_length[4];
+
+      if (png_ptr->buffer_size < 8)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_push_fill_buffer(png_ptr, chunk_length, 4);
+      png_ptr->push_length = png_get_uint_32(chunk_length);
+      png_reset_crc(png_ptr);
+      png_crc_read(png_ptr, png_ptr->chunk_name, 4);
+      png_ptr->mode |= PNG_HAVE_CHUNK_HEADER;
+   }
+
+   if (!png_memcmp(png_ptr->chunk_name, png_IHDR, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_IHDR(png_ptr, info_ptr, png_ptr->push_length);
+   }
+   else if (!png_memcmp(png_ptr->chunk_name, png_PLTE, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_PLTE(png_ptr, info_ptr, png_ptr->push_length);
+   }
+   else if (!png_memcmp(png_ptr->chunk_name, (png_bytep)png_IDAT, 4))
+   {
+      /* If we reach an IDAT chunk, this means we have read all of the
+       * header chunks, and we can start reading the image (or if this
+       * is called after the image has been read - we have an error).
+       */
+      if (png_ptr->mode & PNG_HAVE_IDAT)
+      {
+         if (png_ptr->push_length == 0)
+            return;
+
+         if (png_ptr->mode & PNG_AFTER_IDAT)
+            png_error(png_ptr, "Too many IDAT's found");
+      }
+
+      png_ptr->idat_size = png_ptr->push_length;
+      png_ptr->mode |= PNG_HAVE_IDAT;
+      png_ptr->process_mode = PNG_READ_IDAT_MODE;
+      png_push_have_info(png_ptr, info_ptr);
+      png_ptr->zstream.avail_out = (uInt)png_ptr->irowbytes;
+      png_ptr->zstream.next_out = png_ptr->row_buf;
+      return;
+   }
+   else if (!png_memcmp(png_ptr->chunk_name, png_IEND, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_IEND(png_ptr, info_ptr, png_ptr->push_length);
+      png_ptr->process_mode = PNG_READ_DONE_MODE;
+      png_push_have_end(png_ptr, info_ptr);
+   }
+#if defined(PNG_READ_gAMA_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_gAMA, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_gAMA(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_sBIT_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_sBIT, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_sBIT(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_cHRM_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_cHRM, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_cHRM(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_sRGB_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_sRGB, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_sRGB(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_iCCP_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_iCCP, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_iCCP(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_sPLT_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_sPLT, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_sPLT(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_tRNS_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_tRNS, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_tRNS(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_bKGD_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_bKGD, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_bKGD(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_hIST, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_hIST(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_pHYs_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_pHYs, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_pHYs(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_oFFs_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_oFFs, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_oFFs(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_pCAL_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_pCAL, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_pCAL(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_sCAL_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_sCAL, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_sCAL(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_tIME_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_tIME, 4))
+   {
+      if (png_ptr->push_length + 4 > png_ptr->buffer_size)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_handle_tIME(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_tEXt_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_tEXt, 4))
+   {
+      png_push_handle_tEXt(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_zTXt, 4))
+   {
+      png_push_handle_zTXt(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+   else if (!png_memcmp(png_ptr->chunk_name, png_iTXt, 4))
+   {
+      png_push_handle_iTXt(png_ptr, info_ptr, png_ptr->push_length);
+   }
+#endif
+   else
+   {
+      png_push_handle_unknown(png_ptr, info_ptr, png_ptr->push_length);
+   }
+
+   png_ptr->mode &= ~PNG_HAVE_CHUNK_HEADER;
+}
+
+void /* PRIVATE */
+png_push_crc_skip(png_structp png_ptr, png_uint_32 skip)
+{
+   png_ptr->process_mode = PNG_SKIP_MODE;
+   png_ptr->skip_length = skip;
+}
+
+void /* PRIVATE */
+png_push_crc_finish(png_structp png_ptr)
+{
+   if (png_ptr->skip_length && png_ptr->save_buffer_size)
+   {
+      png_size_t save_size;
+
+      if (png_ptr->skip_length < (png_uint_32)png_ptr->save_buffer_size)
+         save_size = (png_size_t)png_ptr->skip_length;
+      else
+         save_size = png_ptr->save_buffer_size;
+
+      png_calculate_crc(png_ptr, png_ptr->save_buffer_ptr, save_size);
+
+      png_ptr->skip_length -= save_size;
+      png_ptr->buffer_size -= save_size;
+      png_ptr->save_buffer_size -= save_size;
+      png_ptr->save_buffer_ptr += save_size;
+   }
+   if (png_ptr->skip_length && png_ptr->current_buffer_size)
+   {
+      png_size_t save_size;
+
+      if (png_ptr->skip_length < (png_uint_32)png_ptr->current_buffer_size)
+         save_size = (png_size_t)png_ptr->skip_length;
+      else
+         save_size = png_ptr->current_buffer_size;
+
+      png_calculate_crc(png_ptr, png_ptr->current_buffer_ptr, save_size);
+
+      png_ptr->skip_length -= save_size;
+      png_ptr->buffer_size -= save_size;
+      png_ptr->current_buffer_size -= save_size;
+      png_ptr->current_buffer_ptr += save_size;
+   }
+   if (!png_ptr->skip_length)
+   {
+      if (png_ptr->buffer_size < 4)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_crc_finish(png_ptr, 0);
+      png_ptr->process_mode = PNG_READ_CHUNK_MODE;
+   }
+}
+
+void /* PRIVATE */
+png_push_fill_buffer(png_structp png_ptr, png_bytep buffer, png_size_t length)
+{
+   png_bytep ptr;
+
+   ptr = buffer;
+   if (png_ptr->save_buffer_size)
+   {
+      png_size_t save_size;
+
+      if (length < png_ptr->save_buffer_size)
+         save_size = length;
+      else
+         save_size = png_ptr->save_buffer_size;
+
+      png_memcpy(ptr, png_ptr->save_buffer_ptr, save_size);
+      length -= save_size;
+      ptr += save_size;
+      png_ptr->buffer_size -= save_size;
+      png_ptr->save_buffer_size -= save_size;
+      png_ptr->save_buffer_ptr += save_size;
+   }
+   if (length && png_ptr->current_buffer_size)
+   {
+      png_size_t save_size;
+
+      if (length < png_ptr->current_buffer_size)
+         save_size = length;
+      else
+         save_size = png_ptr->current_buffer_size;
+
+      png_memcpy(ptr, png_ptr->current_buffer_ptr, save_size);
+      png_ptr->buffer_size -= save_size;
+      png_ptr->current_buffer_size -= save_size;
+      png_ptr->current_buffer_ptr += save_size;
+   }
+}
+
+void /* PRIVATE */
+png_push_save_buffer(png_structp png_ptr)
+{
+   if (png_ptr->save_buffer_size)
+   {
+      if (png_ptr->save_buffer_ptr != png_ptr->save_buffer)
+      {
+         png_size_t i,istop;
+         png_bytep sp;
+         png_bytep dp;
+
+         istop = png_ptr->save_buffer_size;
+         for (i = 0, sp = png_ptr->save_buffer_ptr, dp = png_ptr->save_buffer;
+            i < istop; i++, sp++, dp++)
+         {
+            *dp = *sp;
+         }
+      }
+   }
+   if (png_ptr->save_buffer_size + png_ptr->current_buffer_size >
+      png_ptr->save_buffer_max)
+   {
+      png_size_t new_max;
+      png_bytep old_buffer;
+
+      new_max = png_ptr->save_buffer_size + png_ptr->current_buffer_size + 256;
+      old_buffer = png_ptr->save_buffer;
+      png_ptr->save_buffer = (png_bytep)png_malloc(png_ptr,
+         (png_uint_32)new_max);
+      png_memcpy(png_ptr->save_buffer, old_buffer, png_ptr->save_buffer_size);
+      png_free(png_ptr, old_buffer);
+      png_ptr->save_buffer_max = new_max;
+   }
+   if (png_ptr->current_buffer_size)
+   {
+      png_memcpy(png_ptr->save_buffer + png_ptr->save_buffer_size,
+         png_ptr->current_buffer_ptr, png_ptr->current_buffer_size);
+      png_ptr->save_buffer_size += png_ptr->current_buffer_size;
+      png_ptr->current_buffer_size = 0;
+   }
+   png_ptr->save_buffer_ptr = png_ptr->save_buffer;
+   png_ptr->buffer_size = 0;
+}
+
+void /* PRIVATE */
+png_push_restore_buffer(png_structp png_ptr, png_bytep buffer,
+   png_size_t buffer_length)
+{
+   png_ptr->current_buffer = buffer;
+   png_ptr->current_buffer_size = buffer_length;
+   png_ptr->buffer_size = buffer_length + png_ptr->save_buffer_size;
+   png_ptr->current_buffer_ptr = png_ptr->current_buffer;
+}
+
+void /* PRIVATE */
+png_push_read_IDAT(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_IDAT;
+#endif
+   if (!(png_ptr->mode & PNG_HAVE_CHUNK_HEADER))
+   {
+      png_byte chunk_length[4];
+
+      if (png_ptr->buffer_size < 8)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_push_fill_buffer(png_ptr, chunk_length, 4);
+      png_ptr->push_length = png_get_uint_32(chunk_length);
+
+      png_reset_crc(png_ptr);
+      png_crc_read(png_ptr, png_ptr->chunk_name, 4);
+      png_ptr->mode |= PNG_HAVE_CHUNK_HEADER;
+
+      if (png_memcmp(png_ptr->chunk_name, (png_bytep)png_IDAT, 4))
+      {
+         png_ptr->process_mode = PNG_READ_CHUNK_MODE;
+         if (!(png_ptr->flags & PNG_FLAG_ZLIB_FINISHED))
+            png_error(png_ptr, "Not enough compressed data");
+         return;
+      }
+
+      png_ptr->idat_size = png_ptr->push_length;
+   }
+   if (png_ptr->idat_size && png_ptr->save_buffer_size)
+   {
+      png_size_t save_size;
+
+      if (png_ptr->idat_size < (png_uint_32)png_ptr->save_buffer_size)
+      {
+         save_size = (png_size_t)png_ptr->idat_size;
+         /* check for overflow */
+         if((png_uint_32)save_size != png_ptr->idat_size)
+            png_error(png_ptr, "save_size overflowed in pngpread");
+      }
+      else
+         save_size = png_ptr->save_buffer_size;
+
+      png_calculate_crc(png_ptr, png_ptr->save_buffer_ptr, save_size);
+      png_process_IDAT_data(png_ptr, png_ptr->save_buffer_ptr, save_size);
+
+      png_ptr->idat_size -= save_size;
+      png_ptr->buffer_size -= save_size;
+      png_ptr->save_buffer_size -= save_size;
+      png_ptr->save_buffer_ptr += save_size;
+   }
+   if (png_ptr->idat_size && png_ptr->current_buffer_size)
+   {
+      png_size_t save_size;
+
+      if (png_ptr->idat_size < (png_uint_32)png_ptr->current_buffer_size)
+      {
+         save_size = (png_size_t)png_ptr->idat_size;
+         /* check for overflow */
+         if((png_uint_32)save_size != png_ptr->idat_size)
+            png_error(png_ptr, "save_size overflowed in pngpread");
+      }
+      else
+         save_size = png_ptr->current_buffer_size;
+
+      png_calculate_crc(png_ptr, png_ptr->current_buffer_ptr, save_size);
+      png_process_IDAT_data(png_ptr, png_ptr->current_buffer_ptr, save_size);
+
+      png_ptr->idat_size -= save_size;
+      png_ptr->buffer_size -= save_size;
+      png_ptr->current_buffer_size -= save_size;
+      png_ptr->current_buffer_ptr += save_size;
+   }
+   if (!png_ptr->idat_size)
+   {
+      if (png_ptr->buffer_size < 4)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_crc_finish(png_ptr, 0);
+      png_ptr->mode &= ~PNG_HAVE_CHUNK_HEADER;
+   }
+}
+
+void /* PRIVATE */
+png_process_IDAT_data(png_structp png_ptr, png_bytep buffer,
+   png_size_t buffer_length)
+{
+   int ret;
+
+   if ((png_ptr->flags & PNG_FLAG_ZLIB_FINISHED) && buffer_length)
+      png_error(png_ptr, "Extra compression data");
+
+   png_ptr->zstream.next_in = buffer;
+   png_ptr->zstream.avail_in = (uInt)buffer_length;
+   for(;;)
+   {
+      ret = inflate(&png_ptr->zstream, Z_PARTIAL_FLUSH);
+      if (ret != Z_OK)
+      {
+         if (ret == Z_STREAM_END)
+         {
+            if (png_ptr->zstream.avail_in)
+               png_error(png_ptr, "Extra compressed data");
+            if (!(png_ptr->zstream.avail_out))
+            {
+               png_push_process_row(png_ptr);
+            }
+
+            png_ptr->mode |= PNG_AFTER_IDAT;
+            png_ptr->flags |= PNG_FLAG_ZLIB_FINISHED;
+            break;
+         }
+         else if (ret == Z_BUF_ERROR)
+            break;
+         else
+            png_error(png_ptr, "Decompression Error");
+      }
+      if (!(png_ptr->zstream.avail_out))
+      {
+         if ((
+#if defined(PNG_READ_INTERLACING_SUPPORTED)
+             png_ptr->interlaced && png_ptr->pass > 6) ||
+             (!png_ptr->interlaced &&
+#endif
+             png_ptr->row_number == png_ptr->num_rows-1))
+           png_error(png_ptr, "Too much data in IDAT chunks");
+         png_push_process_row(png_ptr);
+         png_ptr->zstream.avail_out = (uInt)png_ptr->irowbytes;
+         png_ptr->zstream.next_out = png_ptr->row_buf;
+      }
+      else
+         break;
+   }
+}
+
+void /* PRIVATE */
+png_push_process_row(png_structp png_ptr)
+{
+   png_ptr->row_info.color_type = png_ptr->color_type;
+   png_ptr->row_info.width = png_ptr->iwidth;
+   png_ptr->row_info.channels = png_ptr->channels;
+   png_ptr->row_info.bit_depth = png_ptr->bit_depth;
+   png_ptr->row_info.pixel_depth = png_ptr->pixel_depth;
+
+   png_ptr->row_info.rowbytes = ((png_ptr->row_info.width *
+      (png_uint_32)png_ptr->row_info.pixel_depth + 7) >> 3);
+
+   png_read_filter_row(png_ptr, &(png_ptr->row_info),
+      png_ptr->row_buf + 1, png_ptr->prev_row + 1,
+      (int)(png_ptr->row_buf[0]));
+
+   png_memcpy_check(png_ptr, png_ptr->prev_row, png_ptr->row_buf,
+      png_ptr->rowbytes + 1);
+
+   if (png_ptr->transformations)
+      png_do_read_transformations(png_ptr);
+
+#if defined(PNG_READ_INTERLACING_SUPPORTED)
+   /* blow up interlaced rows to full size */
+   if (png_ptr->interlaced && (png_ptr->transformations & PNG_INTERLACE))
+   {
+      if (png_ptr->pass < 6)
+/*       old interface (pre-1.0.9):
+         png_do_read_interlace(&(png_ptr->row_info),
+            png_ptr->row_buf + 1, png_ptr->pass, png_ptr->transformations);
+ */
+         png_do_read_interlace(png_ptr);
+
+    switch (png_ptr->pass)
+    {
+         case 0:
+         {
+            int i;
+            for (i = 0; i < 8 && png_ptr->pass == 0; i++)
+            {
+               png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+               png_read_push_finish_row(png_ptr); /* updates png_ptr->pass */
+            }
+            if (png_ptr->pass == 2) /* pass 1 might be empty */
+            {
+               for (i = 0; i < 4 && png_ptr->pass == 2; i++)
+               {
+                  png_push_have_row(png_ptr, png_bytep_NULL);
+                  png_read_push_finish_row(png_ptr);
+               }
+            }
+	    if (png_ptr->pass == 4 && png_ptr->height <= 4)
+	    {
+	        for (i = 0; i < 2 && png_ptr->pass == 4; i++)
+                {
+                   png_push_have_row(png_ptr, png_bytep_NULL);
+                   png_read_push_finish_row(png_ptr);
+                }
+            }
+            if (png_ptr->pass == 6 && png_ptr->height <= 4)
+            {
+                png_push_have_row(png_ptr, png_bytep_NULL);
+                png_read_push_finish_row(png_ptr);
+            }
+            break;
+         }
+         case 1:
+         {
+            int i;
+            for (i = 0; i < 8 && png_ptr->pass == 1; i++)
+            {
+               png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+               png_read_push_finish_row(png_ptr);
+            }
+            if (png_ptr->pass == 2) /* skip top 4 generated rows */
+            {
+               for (i = 0; i < 4 && png_ptr->pass == 2; i++)
+               {
+                  png_push_have_row(png_ptr, png_bytep_NULL);
+                  png_read_push_finish_row(png_ptr);
+               }
+            }
+            break;
+         }
+         case 2:
+         {
+            int i;
+            for (i = 0; i < 4 && png_ptr->pass == 2; i++)
+            {
+               png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+               png_read_push_finish_row(png_ptr);
+            }
+            for (i = 0; i < 4 && png_ptr->pass == 2; i++)
+            {
+               png_push_have_row(png_ptr, png_bytep_NULL);
+               png_read_push_finish_row(png_ptr);
+            }
+            if (png_ptr->pass == 4) /* pass 3 might be empty */
+            {
+               for (i = 0; i < 2 && png_ptr->pass == 4; i++)
+               {
+                  png_push_have_row(png_ptr, png_bytep_NULL);
+                  png_read_push_finish_row(png_ptr);
+               }
+            }
+            break;
+         }
+         case 3:
+         {
+            int i;
+            for (i = 0; i < 4 && png_ptr->pass == 3; i++)
+            {
+               png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+               png_read_push_finish_row(png_ptr);
+            }
+            if (png_ptr->pass == 4) /* skip top two generated rows */
+            {
+               for (i = 0; i < 2 && png_ptr->pass == 4; i++)
+               {
+                  png_push_have_row(png_ptr, png_bytep_NULL);
+                  png_read_push_finish_row(png_ptr);
+               }
+            }
+            break;
+         }
+         case 4:
+         {
+            int i;
+            for (i = 0; i < 2 && png_ptr->pass == 4; i++)
+            {
+               png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+               png_read_push_finish_row(png_ptr);
+            }
+            for (i = 0; i < 2 && png_ptr->pass == 4; i++)
+            {
+               png_push_have_row(png_ptr, png_bytep_NULL);
+               png_read_push_finish_row(png_ptr);
+            }
+            if (png_ptr->pass == 6) /* pass 5 might be empty */
+            {
+               png_push_have_row(png_ptr, png_bytep_NULL);
+               png_read_push_finish_row(png_ptr);
+            }
+            break;
+         }
+         case 5:
+         {
+            int i;
+            for (i = 0; i < 2 && png_ptr->pass == 5; i++)
+            {
+               png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+               png_read_push_finish_row(png_ptr);
+            }
+            if (png_ptr->pass == 6) /* skip top generated row */
+            {
+               png_push_have_row(png_ptr, png_bytep_NULL);
+               png_read_push_finish_row(png_ptr);
+            }
+            break;
+         }
+         case 6:
+         {
+            png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+            png_read_push_finish_row(png_ptr);
+            if (png_ptr->pass != 6)
+               break;
+            png_push_have_row(png_ptr, png_bytep_NULL);
+            png_read_push_finish_row(png_ptr);
+         }
+      }
+   }
+   else
+#endif
+   {
+      png_push_have_row(png_ptr, png_ptr->row_buf + 1);
+      png_read_push_finish_row(png_ptr);
+   }
+}
+
+void /* PRIVATE */
+png_read_push_finish_row(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+   /* start of interlace block */
+   const int FARDATA png_pass_start[] = {0, 4, 0, 2, 0, 1, 0};
+
+   /* offset to next interlace block */
+   const int FARDATA png_pass_inc[] = {8, 8, 4, 4, 2, 2, 1};
+
+   /* start of interlace block in the y direction */
+   const int FARDATA png_pass_ystart[] = {0, 0, 4, 0, 2, 0, 1};
+
+   /* offset to next interlace block in the y direction */
+   const int FARDATA png_pass_yinc[] = {8, 8, 8, 4, 4, 2, 2};
+
+   /* Width of interlace block.  This is not currently used - if you need
+    * it, uncomment it here and in png.h
+   const int FARDATA png_pass_width[] = {8, 4, 4, 2, 2, 1, 1};
+   */
+
+   /* Height of interlace block.  This is not currently used - if you need
+    * it, uncomment it here and in png.h
+   const int FARDATA png_pass_height[] = {8, 8, 4, 4, 2, 2, 1};
+   */
+#endif
+
+   png_ptr->row_number++;
+   if (png_ptr->row_number < png_ptr->num_rows)
+      return;
+
+   if (png_ptr->interlaced)
+   {
+      png_ptr->row_number = 0;
+      png_memset_check(png_ptr, png_ptr->prev_row, 0,
+         png_ptr->rowbytes + 1);
+      do
+      {
+         png_ptr->pass++;
+         if ((png_ptr->pass == 1 && png_ptr->width < 5) ||
+             (png_ptr->pass == 3 && png_ptr->width < 3) ||
+             (png_ptr->pass == 5 && png_ptr->width < 2))
+           png_ptr->pass++;
+
+         if (png_ptr->pass >= 7)
+            break;
+
+         png_ptr->iwidth = (png_ptr->width +
+            png_pass_inc[png_ptr->pass] - 1 -
+            png_pass_start[png_ptr->pass]) /
+            png_pass_inc[png_ptr->pass];
+
+         png_ptr->irowbytes = ((png_ptr->iwidth *
+            png_ptr->pixel_depth + 7) >> 3) + 1;
+
+         if (png_ptr->transformations & PNG_INTERLACE)
+            break;
+
+         png_ptr->num_rows = (png_ptr->height +
+            png_pass_yinc[png_ptr->pass] - 1 -
+            png_pass_ystart[png_ptr->pass]) /
+            png_pass_yinc[png_ptr->pass];
+
+      } while (png_ptr->iwidth == 0 || png_ptr->num_rows == 0);
+   }
+}
+
+#if defined(PNG_READ_tEXt_SUPPORTED)
+void /* PRIVATE */
+png_push_handle_tEXt(png_structp png_ptr, png_infop info_ptr, png_uint_32
+   length)
+{
+   if (!(png_ptr->mode & PNG_HAVE_IHDR) || (png_ptr->mode & PNG_HAVE_IEND))
+      {
+         png_error(png_ptr, "Out of place tEXt");
+         /* to quiet some compiler warnings */
+         if(info_ptr == NULL) return;
+      }
+
+#ifdef PNG_MAX_MALLOC_64K
+   png_ptr->skip_length = 0;  /* This may not be necessary */
+
+   if (length > (png_uint_32)65535L) /* Can't hold entire string in memory */
+   {
+      png_warning(png_ptr, "tEXt chunk too large to fit in memory");
+      png_ptr->skip_length = length - (png_uint_32)65535L;
+      length = (png_uint_32)65535L;
+   }
+#endif
+
+   png_ptr->current_text = (png_charp)png_malloc(png_ptr,
+         (png_uint_32)(length+1));
+   png_ptr->current_text[length] = '\0';
+   png_ptr->current_text_ptr = png_ptr->current_text;
+   png_ptr->current_text_size = (png_size_t)length;
+   png_ptr->current_text_left = (png_size_t)length;
+   png_ptr->process_mode = PNG_READ_tEXt_MODE;
+}
+
+void /* PRIVATE */
+png_push_read_tEXt(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr->buffer_size && png_ptr->current_text_left)
+   {
+      png_size_t text_size;
+
+      if (png_ptr->buffer_size < png_ptr->current_text_left)
+         text_size = png_ptr->buffer_size;
+      else
+         text_size = png_ptr->current_text_left;
+      png_crc_read(png_ptr, (png_bytep)png_ptr->current_text_ptr, text_size);
+      png_ptr->current_text_left -= text_size;
+      png_ptr->current_text_ptr += text_size;
+   }
+   if (!(png_ptr->current_text_left))
+   {
+      png_textp text_ptr;
+      png_charp text;
+      png_charp key;
+
+      if (png_ptr->buffer_size < 4)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_push_crc_finish(png_ptr);
+
+#if defined(PNG_MAX_MALLOC_64K)
+      if (png_ptr->skip_length)
+         return;
+#endif
+
+      key = png_ptr->current_text;
+      png_ptr->current_text = 0;
+
+      for (text = key; *text; text++)
+         /* empty loop */ ;
+
+      if (text != key + png_ptr->current_text_size)
+         text++;
+
+      text_ptr = (png_textp)png_malloc(png_ptr, (png_uint_32)sizeof(png_text));
+      text_ptr->compression = PNG_TEXT_COMPRESSION_NONE;
+      text_ptr->key = key;
+#ifdef PNG_iTXt_SUPPORTED
+      text_ptr->lang = NULL;
+      text_ptr->lang_key = NULL;
+#endif
+      text_ptr->text = text;
+
+      png_set_text(png_ptr, info_ptr, text_ptr, 1);
+
+      png_free(png_ptr, key);
+      png_free(png_ptr, text_ptr);
+   }
+}
+#endif
+
+#if defined(PNG_READ_zTXt_SUPPORTED)
+void /* PRIVATE */
+png_push_handle_zTXt(png_structp png_ptr, png_infop info_ptr, png_uint_32
+   length)
+{
+   if (!(png_ptr->mode & PNG_HAVE_IHDR) || (png_ptr->mode & PNG_HAVE_IEND))
+      {
+         png_error(png_ptr, "Out of place zTXt");
+         /* to quiet some compiler warnings */
+         if(info_ptr == NULL) return;
+      }
+
+#ifdef PNG_MAX_MALLOC_64K
+   /* We can't handle zTXt chunks > 64K, since we don't have enough space
+    * to be able to store the uncompressed data.  Actually, the threshold
+    * is probably around 32K, but it isn't as definite as 64K is.
+    */
+   if (length > (png_uint_32)65535L)
+   {
+      png_warning(png_ptr, "zTXt chunk too large to fit in memory");
+      png_push_crc_skip(png_ptr, length);
+      return;
+   }
+#endif
+
+   png_ptr->current_text = (png_charp)png_malloc(png_ptr,
+       (png_uint_32)(length+1));
+   png_ptr->current_text[length] = '\0';
+   png_ptr->current_text_ptr = png_ptr->current_text;
+   png_ptr->current_text_size = (png_size_t)length;
+   png_ptr->current_text_left = (png_size_t)length;
+   png_ptr->process_mode = PNG_READ_zTXt_MODE;
+}
+
+void /* PRIVATE */
+png_push_read_zTXt(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr->buffer_size && png_ptr->current_text_left)
+   {
+      png_size_t text_size;
+
+      if (png_ptr->buffer_size < (png_uint_32)png_ptr->current_text_left)
+         text_size = png_ptr->buffer_size;
+      else
+         text_size = png_ptr->current_text_left;
+      png_crc_read(png_ptr, (png_bytep)png_ptr->current_text_ptr, text_size);
+      png_ptr->current_text_left -= text_size;
+      png_ptr->current_text_ptr += text_size;
+   }
+   if (!(png_ptr->current_text_left))
+   {
+      png_textp text_ptr;
+      png_charp text;
+      png_charp key;
+      int ret;
+      png_size_t text_size, key_size;
+
+      if (png_ptr->buffer_size < 4)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_push_crc_finish(png_ptr);
+
+      key = png_ptr->current_text;
+      png_ptr->current_text = 0;
+
+      for (text = key; *text; text++)
+         /* empty loop */ ;
+
+      /* zTXt can't have zero text */
+      if (text == key + png_ptr->current_text_size)
+      {
+         png_free(png_ptr, key);
+         return;
+      }
+
+      text++;
+
+      if (*text != PNG_TEXT_COMPRESSION_zTXt) /* check compression byte */
+      {
+         png_free(png_ptr, key);
+         return;
+      }
+
+      text++;
+
+      png_ptr->zstream.next_in = (png_bytep )text;
+      png_ptr->zstream.avail_in = (uInt)(png_ptr->current_text_size -
+         (text - key));
+      png_ptr->zstream.next_out = png_ptr->zbuf;
+      png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+
+      key_size = text - key;
+      text_size = 0;
+      text = NULL;
+      ret = Z_STREAM_END;
+
+      while (png_ptr->zstream.avail_in)
+      {
+         ret = inflate(&png_ptr->zstream, Z_PARTIAL_FLUSH);
+         if (ret != Z_OK && ret != Z_STREAM_END)
+         {
+            inflateReset(&png_ptr->zstream);
+            png_ptr->zstream.avail_in = 0;
+            png_free(png_ptr, key);
+            png_free(png_ptr, text);
+            return;
+         }
+         if (!(png_ptr->zstream.avail_out) || ret == Z_STREAM_END)
+         {
+            if (text == NULL)
+            {
+               text = (png_charp)png_malloc(png_ptr,
+                  (png_uint_32)(png_ptr->zbuf_size - png_ptr->zstream.avail_out
+                     + key_size + 1));
+               png_memcpy(text + key_size, png_ptr->zbuf,
+                  png_ptr->zbuf_size - png_ptr->zstream.avail_out);
+               png_memcpy(text, key, key_size);
+               text_size = key_size + png_ptr->zbuf_size -
+                  png_ptr->zstream.avail_out;
+               *(text + text_size) = '\0';
+            }
+            else
+            {
+               png_charp tmp;
+
+               tmp = text;
+               text = (png_charp)png_malloc(png_ptr, text_size +
+                  (png_uint_32)(png_ptr->zbuf_size - png_ptr->zstream.avail_out
+                   + 1));
+               png_memcpy(text, tmp, text_size);
+               png_free(png_ptr, tmp);
+               png_memcpy(text + text_size, png_ptr->zbuf,
+                  png_ptr->zbuf_size - png_ptr->zstream.avail_out);
+               text_size += png_ptr->zbuf_size - png_ptr->zstream.avail_out;
+               *(text + text_size) = '\0';
+            }
+            if (ret != Z_STREAM_END)
+            {
+               png_ptr->zstream.next_out = png_ptr->zbuf;
+               png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+            }
+         }
+         else
+         {
+            break;
+         }
+
+         if (ret == Z_STREAM_END)
+            break;
+      }
+
+      inflateReset(&png_ptr->zstream);
+      png_ptr->zstream.avail_in = 0;
+
+      if (ret != Z_STREAM_END)
+      {
+         png_free(png_ptr, key);
+         png_free(png_ptr, text);
+         return;
+      }
+
+      png_free(png_ptr, key);
+      key = text;
+      text += key_size;
+
+      text_ptr = (png_textp)png_malloc(png_ptr, (png_uint_32)sizeof(png_text));
+      text_ptr->compression = PNG_TEXT_COMPRESSION_zTXt;
+      text_ptr->key = key;
+#ifdef PNG_iTXt_SUPPORTED
+      text_ptr->lang = NULL;
+      text_ptr->lang_key = NULL;
+#endif
+      text_ptr->text = text;
+
+      png_set_text(png_ptr, info_ptr, text_ptr, 1);
+
+      png_free(png_ptr, key);
+      png_free(png_ptr, text_ptr);
+   }
+}
+#endif
+
+#if defined(PNG_READ_iTXt_SUPPORTED)
+void /* PRIVATE */
+png_push_handle_iTXt(png_structp png_ptr, png_infop info_ptr, png_uint_32
+   length)
+{
+   if (!(png_ptr->mode & PNG_HAVE_IHDR) || (png_ptr->mode & PNG_HAVE_IEND))
+      {
+         png_error(png_ptr, "Out of place iTXt");
+         /* to quiet some compiler warnings */
+         if(info_ptr == NULL) return;
+      }
+
+#ifdef PNG_MAX_MALLOC_64K
+   png_ptr->skip_length = 0;  /* This may not be necessary */
+
+   if (length > (png_uint_32)65535L) /* Can't hold entire string in memory */
+   {
+      png_warning(png_ptr, "iTXt chunk too large to fit in memory");
+      png_ptr->skip_length = length - (png_uint_32)65535L;
+      length = (png_uint_32)65535L;
+   }
+#endif
+
+   png_ptr->current_text = (png_charp)png_malloc(png_ptr,
+         (png_uint_32)(length+1));
+   png_ptr->current_text[length] = '\0';
+   png_ptr->current_text_ptr = png_ptr->current_text;
+   png_ptr->current_text_size = (png_size_t)length;
+   png_ptr->current_text_left = (png_size_t)length;
+   png_ptr->process_mode = PNG_READ_iTXt_MODE;
+}
+
+void /* PRIVATE */
+png_push_read_iTXt(png_structp png_ptr, png_infop info_ptr)
+{
+
+   if (png_ptr->buffer_size && png_ptr->current_text_left)
+   {
+      png_size_t text_size;
+
+      if (png_ptr->buffer_size < png_ptr->current_text_left)
+         text_size = png_ptr->buffer_size;
+      else
+         text_size = png_ptr->current_text_left;
+      png_crc_read(png_ptr, (png_bytep)png_ptr->current_text_ptr, text_size);
+      png_ptr->current_text_left -= text_size;
+      png_ptr->current_text_ptr += text_size;
+   }
+   if (!(png_ptr->current_text_left))
+   {
+      png_textp text_ptr;
+      png_charp key;
+      int comp_flag;
+      png_charp lang;
+      png_charp lang_key;
+      png_charp text;
+
+      if (png_ptr->buffer_size < 4)
+      {
+         png_push_save_buffer(png_ptr);
+         return;
+      }
+
+      png_push_crc_finish(png_ptr);
+
+#if defined(PNG_MAX_MALLOC_64K)
+      if (png_ptr->skip_length)
+         return;
+#endif
+
+      key = png_ptr->current_text;
+      png_ptr->current_text = 0;
+
+      for (lang = key; *lang; lang++)
+         /* empty loop */ ;
+
+      if (lang != key + png_ptr->current_text_size)
+         lang++;
+
+      comp_flag = *lang++;
+      lang++;     /* skip comp_type, always zero */
+
+      for (lang_key = lang; *lang_key; lang_key++)
+         /* empty loop */ ;
+      lang_key++;        /* skip NUL separator */
+
+      for (text = lang_key; *text; text++)
+         /* empty loop */ ;
+
+      if (text != key + png_ptr->current_text_size)
+         text++;
+
+      text_ptr = (png_textp)png_malloc(png_ptr, (png_uint_32)sizeof(png_text));
+      text_ptr->compression = comp_flag + 2;
+      text_ptr->key = key;
+      text_ptr->lang = lang;
+      text_ptr->lang_key = lang_key;
+      text_ptr->text = text;
+      text_ptr->text_length = 0;
+      text_ptr->itxt_length = png_strlen(text);
+
+      png_set_text(png_ptr, info_ptr, text_ptr, 1);
+
+      png_free(png_ptr, text_ptr);
+   }
+}
+#endif
+
+/* This function is called when we haven't found a handler for this
+ * chunk.  If there isn't a problem with the chunk itself (ie a bad chunk
+ * name or a critical chunk), the chunk is (currently) silently ignored.
+ */
+void /* PRIVATE */
+png_push_handle_unknown(png_structp png_ptr, png_infop info_ptr, png_uint_32
+   length)
+{
+   png_uint_32 skip=0;
+   png_check_chunk_name(png_ptr, png_ptr->chunk_name);
+
+   if (!(png_ptr->chunk_name[0] & 0x20))
+   {
+#if defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED)
+      if(png_handle_as_unknown(png_ptr, png_ptr->chunk_name) !=
+           HANDLE_CHUNK_ALWAYS
+#if defined(PNG_READ_USER_CHUNKS_SUPPORTED)
+           && png_ptr->read_user_chunk_fn == NULL
+#endif
+         )
+#endif
+         png_chunk_error(png_ptr, "unknown critical chunk");
+
+      /* to quiet compiler warnings about unused info_ptr */
+      if (info_ptr == NULL)
+         return;
+   }
+
+#if defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED)
+   if (png_ptr->flags & PNG_FLAG_KEEP_UNKNOWN_CHUNKS)
+   {
+       png_unknown_chunk chunk;
+
+#ifdef PNG_MAX_MALLOC_64K
+       if (length > (png_uint_32)65535L)
+       {
+           png_warning(png_ptr, "unknown chunk too large to fit in memory");
+           skip = length - (png_uint_32)65535L;
+           length = (png_uint_32)65535L;
+       }
+#endif
+
+       png_strcpy((png_charp)chunk.name, (png_charp)png_ptr->chunk_name);
+       chunk.data = (png_bytep)png_malloc(png_ptr, length);
+       png_crc_read(png_ptr, chunk.data, length);
+       chunk.size = length;
+#if defined(PNG_READ_USER_CHUNKS_SUPPORTED)
+       if(png_ptr->read_user_chunk_fn != NULL)
+       {
+          /* callback to user unknown chunk handler */
+          if ((*(png_ptr->read_user_chunk_fn)) (png_ptr, &chunk) <= 0)
+          {
+             if (!(png_ptr->chunk_name[0] & 0x20))
+                if(png_handle_as_unknown(png_ptr, png_ptr->chunk_name) !=
+                     HANDLE_CHUNK_ALWAYS)
+                   png_chunk_error(png_ptr, "unknown critical chunk");
+          }
+             png_set_unknown_chunks(png_ptr, info_ptr, &chunk, 1);
+       }
+       else
+#endif
+          png_set_unknown_chunks(png_ptr, info_ptr, &chunk, 1);
+       png_free(png_ptr, chunk.data);
+   }
+   else
+#endif
+      skip=length;
+   png_push_crc_skip(png_ptr, skip);
+}
+
+void /* PRIVATE */
+png_push_have_info(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr->info_fn != NULL)
+      (*(png_ptr->info_fn))(png_ptr, info_ptr);
+}
+
+void /* PRIVATE */
+png_push_have_end(png_structp png_ptr, png_infop info_ptr)
+{
+   if (png_ptr->end_fn != NULL)
+      (*(png_ptr->end_fn))(png_ptr, info_ptr);
+}
+
+void /* PRIVATE */
+png_push_have_row(png_structp png_ptr, png_bytep row)
+{
+   if (png_ptr->row_fn != NULL)
+      (*(png_ptr->row_fn))(png_ptr, row, png_ptr->row_number,
+         (int)png_ptr->pass);
+}
+
+void PNGAPI
+png_progressive_combine_row (png_structp png_ptr,
+   png_bytep old_row, png_bytep new_row)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   const int FARDATA png_pass_dsp_mask[7] =
+      {0xff, 0x0f, 0xff, 0x33, 0xff, 0x55, 0xff};
+#endif
+   if (new_row != NULL)    /* new_row must == png_ptr->row_buf here. */
+      png_combine_row(png_ptr, old_row, png_pass_dsp_mask[png_ptr->pass]);
+}
+
+void PNGAPI
+png_set_progressive_read_fn(png_structp png_ptr, png_voidp progressive_ptr,
+   png_progressive_info_ptr info_fn, png_progressive_row_ptr row_fn,
+   png_progressive_end_ptr end_fn)
+{
+   png_ptr->info_fn = info_fn;
+   png_ptr->row_fn = row_fn;
+   png_ptr->end_fn = end_fn;
+
+   png_set_read_fn(png_ptr, progressive_ptr, png_push_fill_buffer);
+}
+
+png_voidp PNGAPI
+png_get_progressive_ptr(png_structp png_ptr)
+{
+   return png_ptr->io_ptr;
+}
+#endif /* PNG_PROGRESSIVE_READ_SUPPORTED */
diff --git a/png/pngread.c b/png/pngread.c
new file mode 100644
index 000000000..c47df62b5
--- /dev/null
+++ b/png/pngread.c
@@ -0,0 +1,1398 @@
+
+/* pngread.c - read a PNG file
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file contains routines that an application calls directly to
+ * read a PNG file or stream.
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+/* Create a PNG structure for reading, and allocate any memory needed. */
+png_structp PNGAPI
+png_create_read_struct(png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn)
+{
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   return (png_create_read_struct_2(user_png_ver, error_ptr, error_fn,
+      warn_fn, png_voidp_NULL, png_malloc_ptr_NULL, png_free_ptr_NULL));
+}
+
+/* Alternate create PNG structure for reading, and allocate any memory needed. */
+png_structp PNGAPI
+png_create_read_struct_2(png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn, png_voidp mem_ptr,
+   png_malloc_ptr malloc_fn, png_free_ptr free_fn)
+{
+#endif /* PNG_USER_MEM_SUPPORTED */
+
+   png_structp png_ptr;
+
+#ifdef PNG_SETJMP_SUPPORTED
+#ifdef USE_FAR_KEYWORD
+   jmp_buf jmpbuf;
+#endif
+#endif
+
+   int i;
+
+   png_debug(1, "in png_create_read_struct\n");
+#ifdef PNG_USER_MEM_SUPPORTED
+   if ((png_ptr = (png_structp)png_create_struct_2(PNG_STRUCT_PNG,
+      (png_malloc_ptr)malloc_fn, (png_voidp)mem_ptr)) == NULL)
+#else
+   if ((png_ptr = (png_structp)png_create_struct(PNG_STRUCT_PNG)) == NULL)
+#endif
+      return (NULL);
+
+#ifdef PNG_ASSEMBLER_CODE_SUPPORTED
+   png_init_mmx_flags(png_ptr);   /* 1.2.0 addition */
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+#ifdef USE_FAR_KEYWORD
+   if (setjmp(jmpbuf))
+#else
+   if (setjmp(png_ptr->jmpbuf))
+#endif
+   {
+      png_free(png_ptr, png_ptr->zbuf);
+      png_ptr->zbuf=NULL;
+      png_destroy_struct(png_ptr);
+      return (NULL);
+   }
+#ifdef USE_FAR_KEYWORD
+   png_memcpy(png_ptr->jmpbuf,jmpbuf,sizeof(jmp_buf));
+#endif
+#endif
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_set_mem_fn(png_ptr, mem_ptr, malloc_fn, free_fn);
+#endif
+
+   png_set_error_fn(png_ptr, error_ptr, error_fn, warn_fn);
+
+   i=0;
+   do
+   {
+     if(user_png_ver[i] != png_libpng_ver[i])
+        png_ptr->flags |= PNG_FLAG_LIBRARY_MISMATCH;
+   } while (png_libpng_ver[i++]);
+
+   if (png_ptr->flags & PNG_FLAG_LIBRARY_MISMATCH)
+   {
+     /* Libpng 0.90 and later are binary incompatible with libpng 0.89, so
+      * we must recompile any applications that use any older library version.
+      * For versions after libpng 1.0, we will be compatible, so we need
+      * only check the first digit.
+      */
+     if (user_png_ver == NULL || user_png_ver[0] != png_libpng_ver[0] ||
+         (user_png_ver[0] == '1' && user_png_ver[2] != png_libpng_ver[2]) ||
+         (user_png_ver[0] == '0' && user_png_ver[2] < '9'))
+     {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+        char msg[80];
+        if (user_png_ver)
+        {
+          sprintf(msg, "Application was compiled with png.h from libpng-%.20s",
+             user_png_ver);
+          png_warning(png_ptr, msg);
+        }
+        sprintf(msg, "Application  is  running with png.c from libpng-%.20s",
+           png_libpng_ver);
+        png_warning(png_ptr, msg);
+#endif
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+        png_ptr->flags=0;
+#endif
+        png_error(png_ptr,
+           "Incompatible libpng version in application and library");
+     }
+   }
+
+   /* initialize zbuf - compression buffer */
+   png_ptr->zbuf_size = PNG_ZBUF_SIZE;
+   png_ptr->zbuf = (png_bytep)png_malloc(png_ptr,
+     (png_uint_32)png_ptr->zbuf_size);
+   png_ptr->zstream.zalloc = png_zalloc;
+   png_ptr->zstream.zfree = png_zfree;
+   png_ptr->zstream.opaque = (voidpf)png_ptr;
+
+   switch (inflateInit(&png_ptr->zstream))
+   {
+     case Z_OK: /* Do nothing */ break;
+     case Z_MEM_ERROR:
+     case Z_STREAM_ERROR: png_error(png_ptr, "zlib memory error"); break;
+     case Z_VERSION_ERROR: png_error(png_ptr, "zlib version error"); break;
+     default: png_error(png_ptr, "Unknown zlib error");
+   }
+
+   png_ptr->zstream.next_out = png_ptr->zbuf;
+   png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+
+   png_set_read_fn(png_ptr, png_voidp_NULL, png_rw_ptr_NULL);
+
+   return (png_ptr);
+}
+
+/* Initialize PNG structure for reading, and allocate any memory needed.
+   This interface is deprecated in favour of the png_create_read_struct(),
+   and it will eventually disappear. */
+#undef png_read_init
+void PNGAPI
+png_read_init(png_structp png_ptr)
+{
+   /* We only come here via pre-1.0.7-compiled applications */
+   png_read_init_2(png_ptr, "1.0.6 or earlier", 0, 0);
+}
+
+#undef png_read_init_2
+void PNGAPI
+png_read_init_2(png_structp png_ptr, png_const_charp user_png_ver,
+   png_size_t png_struct_size, png_size_t png_info_size)
+{
+   /* We only come here via pre-1.0.12-compiled applications */
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+   if(sizeof(png_struct) > png_struct_size || sizeof(png_info) > png_info_size)
+   {
+      char msg[80];
+      png_ptr->warning_fn=NULL;
+      if (user_png_ver)
+      {
+        sprintf(msg, "Application was compiled with png.h from libpng-%.20s",
+           user_png_ver);
+        png_warning(png_ptr, msg);
+      }
+      sprintf(msg, "Application  is  running with png.c from libpng-%.20s",
+         png_libpng_ver);
+      png_warning(png_ptr, msg);
+   }
+#endif
+   if(sizeof(png_struct) > png_struct_size)
+     {
+       png_ptr->error_fn=NULL;
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+       png_ptr->flags=0;
+#endif
+       png_error(png_ptr,
+       "The png struct allocated by the application for reading is too small.");
+     }
+   if(sizeof(png_info) > png_info_size)
+     {
+       png_ptr->error_fn=NULL;
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+       png_ptr->flags=0;
+#endif
+       png_error(png_ptr,
+         "The info struct allocated by application for reading is too small.");
+     }
+   png_read_init_3(&png_ptr, user_png_ver, png_struct_size);
+}
+
+void PNGAPI
+png_read_init_3(png_structpp ptr_ptr, png_const_charp user_png_ver,
+   png_size_t png_struct_size)
+{
+#ifdef PNG_SETJMP_SUPPORTED
+   jmp_buf tmp_jmp;  /* to save current jump buffer */
+#endif
+
+   int i=0;
+
+   png_structp png_ptr=*ptr_ptr;
+
+   do
+   {
+     if(user_png_ver[i] != png_libpng_ver[i])
+     {
+#ifdef PNG_LEGACY_SUPPORTED
+       png_ptr->flags |= PNG_FLAG_LIBRARY_MISMATCH;
+#else
+       png_ptr->warning_fn=NULL;
+       png_warning(png_ptr,
+        "Application uses deprecated png_read_init() and should be recompiled.");
+       break;
+#endif
+     }
+   } while (png_libpng_ver[i++]);
+
+   png_debug(1, "in png_read_init_3\n");
+
+#ifdef PNG_SETJMP_SUPPORTED
+   /* save jump buffer and error functions */
+   png_memcpy(tmp_jmp, png_ptr->jmpbuf, sizeof (jmp_buf));
+#endif
+
+   if(sizeof(png_struct) > png_struct_size)
+     {
+       png_destroy_struct(png_ptr);
+       *ptr_ptr = (png_structp)png_create_struct(PNG_STRUCT_PNG);
+       png_ptr = *ptr_ptr;
+     }
+
+   /* reset all variables to 0 */
+   png_memset(png_ptr, 0, sizeof (png_struct));
+
+#ifdef PNG_SETJMP_SUPPORTED
+   /* restore jump buffer */
+   png_memcpy(png_ptr->jmpbuf, tmp_jmp, sizeof (jmp_buf));
+#endif
+
+   /* initialize zbuf - compression buffer */
+   png_ptr->zbuf_size = PNG_ZBUF_SIZE;
+   png_ptr->zbuf = (png_bytep)png_malloc(png_ptr,
+     (png_uint_32)png_ptr->zbuf_size);
+   png_ptr->zstream.zalloc = png_zalloc;
+   png_ptr->zstream.zfree = png_zfree;
+   png_ptr->zstream.opaque = (voidpf)png_ptr;
+
+   switch (inflateInit(&png_ptr->zstream))
+   {
+     case Z_OK: /* Do nothing */ break;
+     case Z_MEM_ERROR:
+     case Z_STREAM_ERROR: png_error(png_ptr, "zlib memory"); break;
+     case Z_VERSION_ERROR: png_error(png_ptr, "zlib version"); break;
+     default: png_error(png_ptr, "Unknown zlib error");
+   }
+
+   png_ptr->zstream.next_out = png_ptr->zbuf;
+   png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+
+   png_set_read_fn(png_ptr, png_voidp_NULL, png_rw_ptr_NULL);
+}
+
+/* Read the information before the actual image data.  This has been
+ * changed in v0.90 to allow reading a file that already has the magic
+ * bytes read from the stream.  You can tell libpng how many bytes have
+ * been read from the beginning of the stream (up to the maximum of 8)
+ * via png_set_sig_bytes(), and we will only check the remaining bytes
+ * here.  The application can then have access to the signature bytes we
+ * read if it is determined that this isn't a valid PNG file.
+ */
+void PNGAPI
+png_read_info(png_structp png_ptr, png_infop info_ptr)
+{
+   png_debug(1, "in png_read_info\n");
+   /* save jump buffer and error functions */
+   /* If we haven't checked all of the PNG signature bytes, do so now. */
+   if (png_ptr->sig_bytes < 8)
+   {
+      png_size_t num_checked = png_ptr->sig_bytes,
+                 num_to_check = 8 - num_checked;
+
+      png_read_data(png_ptr, &(info_ptr->signature[num_checked]), num_to_check);
+      png_ptr->sig_bytes = 8;
+
+      if (png_sig_cmp(info_ptr->signature, num_checked, num_to_check))
+      {
+         if (num_checked < 4 &&
+             png_sig_cmp(info_ptr->signature, num_checked, num_to_check - 4))
+            png_error(png_ptr, "Not a PNG file");
+         else
+            png_error(png_ptr, "PNG file corrupted by ASCII conversion");
+      }
+      if (num_checked < 3)
+         png_ptr->mode |= PNG_HAVE_PNG_SIGNATURE;
+   }
+
+   for(;;)
+   {
+#ifdef PNG_USE_LOCAL_ARRAYS
+      PNG_IHDR;
+      PNG_IDAT;
+      PNG_IEND;
+      PNG_PLTE;
+#if defined(PNG_READ_bKGD_SUPPORTED)
+      PNG_bKGD;
+#endif
+#if defined(PNG_READ_cHRM_SUPPORTED)
+      PNG_cHRM;
+#endif
+#if defined(PNG_READ_gAMA_SUPPORTED)
+      PNG_gAMA;
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+      PNG_hIST;
+#endif
+#if defined(PNG_READ_iCCP_SUPPORTED)
+      PNG_iCCP;
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+      PNG_iTXt;
+#endif
+#if defined(PNG_READ_oFFs_SUPPORTED)
+      PNG_oFFs;
+#endif
+#if defined(PNG_READ_pCAL_SUPPORTED)
+      PNG_pCAL;
+#endif
+#if defined(PNG_READ_pHYs_SUPPORTED)
+      PNG_pHYs;
+#endif
+#if defined(PNG_READ_sBIT_SUPPORTED)
+      PNG_sBIT;
+#endif
+#if defined(PNG_READ_sCAL_SUPPORTED)
+      PNG_sCAL;
+#endif
+#if defined(PNG_READ_sPLT_SUPPORTED)
+      PNG_sPLT;
+#endif
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      PNG_sRGB;
+#endif
+#if defined(PNG_READ_tEXt_SUPPORTED)
+      PNG_tEXt;
+#endif
+#if defined(PNG_READ_tIME_SUPPORTED)
+      PNG_tIME;
+#endif
+#if defined(PNG_READ_tRNS_SUPPORTED)
+      PNG_tRNS;
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+      PNG_zTXt;
+#endif
+#endif /* PNG_GLOBAL_ARRAYS */
+      png_byte chunk_length[4];
+      png_uint_32 length;
+
+      png_read_data(png_ptr, chunk_length, 4);
+      length = png_get_uint_32(chunk_length);
+
+      png_reset_crc(png_ptr);
+      png_crc_read(png_ptr, png_ptr->chunk_name, 4);
+
+      png_debug2(0, "Reading %s chunk, length=%lu.\n", png_ptr->chunk_name,
+         length);
+
+      /* This should be a binary subdivision search or a hash for
+       * matching the chunk name rather than a linear search.
+       */
+      if (!png_memcmp(png_ptr->chunk_name, png_IHDR, 4))
+         png_handle_IHDR(png_ptr, info_ptr, length);
+      else if (!png_memcmp(png_ptr->chunk_name, png_IEND, 4))
+         png_handle_IEND(png_ptr, info_ptr, length);
+#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+      else if (png_handle_as_unknown(png_ptr, png_ptr->chunk_name))
+      {
+         if (!png_memcmp(png_ptr->chunk_name, png_IDAT, 4))
+            png_ptr->mode |= PNG_HAVE_IDAT;
+         png_handle_unknown(png_ptr, info_ptr, length);
+         if (!png_memcmp(png_ptr->chunk_name, png_PLTE, 4))
+            png_ptr->mode |= PNG_HAVE_PLTE;
+         else if (!png_memcmp(png_ptr->chunk_name, png_IDAT, 4))
+         {
+            if (!(png_ptr->mode & PNG_HAVE_IHDR))
+               png_error(png_ptr, "Missing IHDR before IDAT");
+            else if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE &&
+                     !(png_ptr->mode & PNG_HAVE_PLTE))
+               png_error(png_ptr, "Missing PLTE before IDAT");
+            break;
+         }
+      }
+#endif
+      else if (!png_memcmp(png_ptr->chunk_name, png_PLTE, 4))
+         png_handle_PLTE(png_ptr, info_ptr, length);
+      else if (!png_memcmp(png_ptr->chunk_name, png_IDAT, 4))
+      {
+         if (!(png_ptr->mode & PNG_HAVE_IHDR))
+            png_error(png_ptr, "Missing IHDR before IDAT");
+         else if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE &&
+                  !(png_ptr->mode & PNG_HAVE_PLTE))
+            png_error(png_ptr, "Missing PLTE before IDAT");
+
+         png_ptr->idat_size = length;
+         png_ptr->mode |= PNG_HAVE_IDAT;
+         break;
+      }
+#if defined(PNG_READ_bKGD_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_bKGD, 4))
+         png_handle_bKGD(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_cHRM_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_cHRM, 4))
+         png_handle_cHRM(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_gAMA_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_gAMA, 4))
+         png_handle_gAMA(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_hIST, 4))
+         png_handle_hIST(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_oFFs_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_oFFs, 4))
+         png_handle_oFFs(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_pCAL_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_pCAL, 4))
+         png_handle_pCAL(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sCAL_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sCAL, 4))
+         png_handle_sCAL(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_pHYs_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_pHYs, 4))
+         png_handle_pHYs(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sBIT_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sBIT, 4))
+         png_handle_sBIT(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sRGB, 4))
+         png_handle_sRGB(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_iCCP_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_iCCP, 4))
+         png_handle_iCCP(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sPLT_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sPLT, 4))
+         png_handle_sPLT(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_tEXt_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_tEXt, 4))
+         png_handle_tEXt(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_tIME_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_tIME, 4))
+         png_handle_tIME(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_tRNS_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_tRNS, 4))
+         png_handle_tRNS(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_zTXt, 4))
+         png_handle_zTXt(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_iTXt, 4))
+         png_handle_iTXt(png_ptr, info_ptr, length);
+#endif
+      else
+         png_handle_unknown(png_ptr, info_ptr, length);
+   }
+}
+
+/* optional call to update the users info_ptr structure */
+void PNGAPI
+png_read_update_info(png_structp png_ptr, png_infop info_ptr)
+{
+   png_debug(1, "in png_read_update_info\n");
+   /* save jump buffer and error functions */
+   if (!(png_ptr->flags & PNG_FLAG_ROW_INIT))
+      png_read_start_row(png_ptr);
+   else
+      png_warning(png_ptr,
+      "Ignoring extra png_read_update_info() call; row buffer not reallocated");
+   png_read_transform_info(png_ptr, info_ptr);
+}
+
+/* Initialize palette, background, etc, after transformations
+ * are set, but before any reading takes place.  This allows
+ * the user to obtain a gamma-corrected palette, for example.
+ * If the user doesn't call this, we will do it ourselves.
+ */
+void PNGAPI
+png_start_read_image(png_structp png_ptr)
+{
+   png_debug(1, "in png_start_read_image\n");
+   /* save jump buffer and error functions */
+   if (!(png_ptr->flags & PNG_FLAG_ROW_INIT))
+      png_read_start_row(png_ptr);
+}
+
+void PNGAPI
+png_read_row(png_structp png_ptr, png_bytep row, png_bytep dsp_row)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_IDAT;
+   const int png_pass_dsp_mask[7] = {0xff, 0x0f, 0xff, 0x33, 0xff, 0x55, 0xff};
+   const int png_pass_mask[7] = {0x80, 0x08, 0x88, 0x22, 0xaa, 0x55, 0xff};
+#endif
+   int ret;
+   png_debug2(1, "in png_read_row (row %lu, pass %d)\n",
+      png_ptr->row_number, png_ptr->pass);
+   /* save jump buffer and error functions */
+   if (!(png_ptr->flags & PNG_FLAG_ROW_INIT))
+      png_read_start_row(png_ptr);
+   if (png_ptr->row_number == 0 && png_ptr->pass == 0)
+   {
+   /* check for transforms that have been set but were defined out */
+#if defined(PNG_WRITE_INVERT_SUPPORTED) && !defined(PNG_READ_INVERT_SUPPORTED)
+   if (png_ptr->transformations & PNG_INVERT_MONO)
+      png_warning(png_ptr, "PNG_READ_INVERT_SUPPORTED is not defined.");
+#endif
+#if defined(PNG_WRITE_FILLER_SUPPORTED) && !defined(PNG_READ_FILLER_SUPPORTED)
+   if (png_ptr->transformations & PNG_FILLER)
+      png_warning(png_ptr, "PNG_READ_FILLER_SUPPORTED is not defined.");
+#endif
+#if defined(PNG_WRITE_PACKSWAP_SUPPORTED) && !defined(PNG_READ_PACKSWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACKSWAP)
+      png_warning(png_ptr, "PNG_READ_PACKSWAP_SUPPORTED is not defined.");
+#endif
+#if defined(PNG_WRITE_PACK_SUPPORTED) && !defined(PNG_READ_PACK_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACK)
+      png_warning(png_ptr, "PNG_READ_PACK_SUPPORTED is not defined.");
+#endif
+#if defined(PNG_WRITE_SHIFT_SUPPORTED) && !defined(PNG_READ_SHIFT_SUPPORTED)
+   if (png_ptr->transformations & PNG_SHIFT)
+      png_warning(png_ptr, "PNG_READ_SHIFT_SUPPORTED is not defined.");
+#endif
+#if defined(PNG_WRITE_BGR_SUPPORTED) && !defined(PNG_READ_BGR_SUPPORTED)
+   if (png_ptr->transformations & PNG_BGR)
+      png_warning(png_ptr, "PNG_READ_BGR_SUPPORTED is not defined.");
+#endif
+#if defined(PNG_WRITE_SWAP_SUPPORTED) && !defined(PNG_READ_SWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_SWAP_BYTES)
+      png_warning(png_ptr, "PNG_READ_SWAP_SUPPORTED is not defined.");
+#endif
+   }
+
+#if defined(PNG_READ_INTERLACING_SUPPORTED)
+   /* if interlaced and we do not need a new row, combine row and return */
+   if (png_ptr->interlaced && (png_ptr->transformations & PNG_INTERLACE))
+   {
+      switch (png_ptr->pass)
+      {
+         case 0:
+            if (png_ptr->row_number & 0x07)
+            {
+               if (dsp_row != NULL)
+                  png_combine_row(png_ptr, dsp_row,
+                     png_pass_dsp_mask[png_ptr->pass]);
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 1:
+            if ((png_ptr->row_number & 0x07) || png_ptr->width < 5)
+            {
+               if (dsp_row != NULL)
+                  png_combine_row(png_ptr, dsp_row,
+                     png_pass_dsp_mask[png_ptr->pass]);
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 2:
+            if ((png_ptr->row_number & 0x07) != 4)
+            {
+               if (dsp_row != NULL && (png_ptr->row_number & 4))
+                  png_combine_row(png_ptr, dsp_row,
+                     png_pass_dsp_mask[png_ptr->pass]);
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 3:
+            if ((png_ptr->row_number & 3) || png_ptr->width < 3)
+            {
+               if (dsp_row != NULL)
+                  png_combine_row(png_ptr, dsp_row,
+                     png_pass_dsp_mask[png_ptr->pass]);
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 4:
+            if ((png_ptr->row_number & 3) != 2)
+            {
+               if (dsp_row != NULL && (png_ptr->row_number & 2))
+                  png_combine_row(png_ptr, dsp_row,
+                     png_pass_dsp_mask[png_ptr->pass]);
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 5:
+            if ((png_ptr->row_number & 1) || png_ptr->width < 2)
+            {
+               if (dsp_row != NULL)
+                  png_combine_row(png_ptr, dsp_row,
+                     png_pass_dsp_mask[png_ptr->pass]);
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 6:
+            if (!(png_ptr->row_number & 1))
+            {
+               png_read_finish_row(png_ptr);
+               return;
+            }
+            break;
+      }
+   }
+#endif
+
+   if (!(png_ptr->mode & PNG_HAVE_IDAT))
+      png_error(png_ptr, "Invalid attempt to read row data");
+
+   png_ptr->zstream.next_out = png_ptr->row_buf;
+   png_ptr->zstream.avail_out = (uInt)png_ptr->irowbytes;
+   do
+   {
+      if (!(png_ptr->zstream.avail_in))
+      {
+         while (!png_ptr->idat_size)
+         {
+            png_byte chunk_length[4];
+
+            png_crc_finish(png_ptr, 0);
+
+            png_read_data(png_ptr, chunk_length, 4);
+            png_ptr->idat_size = png_get_uint_32(chunk_length);
+
+            png_reset_crc(png_ptr);
+            png_crc_read(png_ptr, png_ptr->chunk_name, 4);
+            if (png_memcmp(png_ptr->chunk_name, png_IDAT, 4))
+               png_error(png_ptr, "Not enough image data");
+         }
+         png_ptr->zstream.avail_in = (uInt)png_ptr->zbuf_size;
+         png_ptr->zstream.next_in = png_ptr->zbuf;
+         if (png_ptr->zbuf_size > png_ptr->idat_size)
+            png_ptr->zstream.avail_in = (uInt)png_ptr->idat_size;
+         png_crc_read(png_ptr, png_ptr->zbuf,
+            (png_size_t)png_ptr->zstream.avail_in);
+         png_ptr->idat_size -= png_ptr->zstream.avail_in;
+      }
+      ret = inflate(&png_ptr->zstream, Z_PARTIAL_FLUSH);
+      if (ret == Z_STREAM_END)
+      {
+         if (png_ptr->zstream.avail_out || png_ptr->zstream.avail_in ||
+            png_ptr->idat_size)
+            png_error(png_ptr, "Extra compressed data");
+         png_ptr->mode |= PNG_AFTER_IDAT;
+         png_ptr->flags |= PNG_FLAG_ZLIB_FINISHED;
+         break;
+      }
+      if (ret != Z_OK)
+         png_error(png_ptr, png_ptr->zstream.msg ? png_ptr->zstream.msg :
+                   "Decompression error");
+
+   } while (png_ptr->zstream.avail_out);
+
+   png_ptr->row_info.color_type = png_ptr->color_type;
+   png_ptr->row_info.width = png_ptr->iwidth;
+   png_ptr->row_info.channels = png_ptr->channels;
+   png_ptr->row_info.bit_depth = png_ptr->bit_depth;
+   png_ptr->row_info.pixel_depth = png_ptr->pixel_depth;
+   png_ptr->row_info.rowbytes = ((png_ptr->row_info.width *
+      (png_uint_32)png_ptr->row_info.pixel_depth + 7) >> 3);
+
+   if(png_ptr->row_buf[0])
+   png_read_filter_row(png_ptr, &(png_ptr->row_info),
+      png_ptr->row_buf + 1, png_ptr->prev_row + 1,
+      (int)(png_ptr->row_buf[0]));
+
+   png_memcpy_check(png_ptr, png_ptr->prev_row, png_ptr->row_buf,
+      png_ptr->rowbytes + 1);
+   
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   if((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) &&
+      (png_ptr->filter_type == PNG_INTRAPIXEL_DIFFERENCING))
+   {
+      /* Intrapixel differencing */
+      png_do_read_intrapixel(&(png_ptr->row_info), png_ptr->row_buf + 1);
+   }
+#endif
+
+   if (png_ptr->transformations)
+      png_do_read_transformations(png_ptr);
+
+#if defined(PNG_READ_INTERLACING_SUPPORTED)
+   /* blow up interlaced rows to full size */
+   if (png_ptr->interlaced &&
+      (png_ptr->transformations & PNG_INTERLACE))
+   {
+      if (png_ptr->pass < 6)
+/*       old interface (pre-1.0.9):
+         png_do_read_interlace(&(png_ptr->row_info),
+            png_ptr->row_buf + 1, png_ptr->pass, png_ptr->transformations);
+ */
+         png_do_read_interlace(png_ptr);
+
+      if (dsp_row != NULL)
+         png_combine_row(png_ptr, dsp_row,
+            png_pass_dsp_mask[png_ptr->pass]);
+      if (row != NULL)
+         png_combine_row(png_ptr, row,
+            png_pass_mask[png_ptr->pass]);
+   }
+   else
+#endif
+   {
+      if (row != NULL)
+         png_combine_row(png_ptr, row, 0xff);
+      if (dsp_row != NULL)
+         png_combine_row(png_ptr, dsp_row, 0xff);
+   }
+   png_read_finish_row(png_ptr);
+
+   if (png_ptr->read_row_fn != NULL)
+      (*(png_ptr->read_row_fn))(png_ptr, png_ptr->row_number, png_ptr->pass);
+}
+
+/* Read one or more rows of image data.  If the image is interlaced,
+ * and png_set_interlace_handling() has been called, the rows need to
+ * contain the contents of the rows from the previous pass.  If the
+ * image has alpha or transparency, and png_handle_alpha()[*] has been
+ * called, the rows contents must be initialized to the contents of the
+ * screen.
+ *
+ * "row" holds the actual image, and pixels are placed in it
+ * as they arrive.  If the image is displayed after each pass, it will
+ * appear to "sparkle" in.  "display_row" can be used to display a
+ * "chunky" progressive image, with finer detail added as it becomes
+ * available.  If you do not want this "chunky" display, you may pass
+ * NULL for display_row.  If you do not want the sparkle display, and
+ * you have not called png_handle_alpha(), you may pass NULL for rows.
+ * If you have called png_handle_alpha(), and the image has either an
+ * alpha channel or a transparency chunk, you must provide a buffer for
+ * rows.  In this case, you do not have to provide a display_row buffer
+ * also, but you may.  If the image is not interlaced, or if you have
+ * not called png_set_interlace_handling(), the display_row buffer will
+ * be ignored, so pass NULL to it.
+ *
+ * [*] png_handle_alpha() does not exist yet, as of libpng version 1.2.1
+ */
+
+void PNGAPI
+png_read_rows(png_structp png_ptr, png_bytepp row,
+   png_bytepp display_row, png_uint_32 num_rows)
+{
+   png_uint_32 i;
+   png_bytepp rp;
+   png_bytepp dp;
+
+   png_debug(1, "in png_read_rows\n");
+   /* save jump buffer and error functions */
+   rp = row;
+   dp = display_row;
+   if (rp != NULL && dp != NULL)
+      for (i = 0; i < num_rows; i++)
+      {
+         png_bytep rptr = *rp++;
+         png_bytep dptr = *dp++;
+
+         png_read_row(png_ptr, rptr, dptr);
+      }
+   else if(rp != NULL)
+      for (i = 0; i < num_rows; i++)
+      {
+         png_bytep rptr = *rp;
+         png_read_row(png_ptr, rptr, png_bytep_NULL);
+         rp++;
+      }
+   else if(dp != NULL)
+      for (i = 0; i < num_rows; i++)
+      {
+         png_bytep dptr = *dp;
+         png_read_row(png_ptr, png_bytep_NULL, dptr);
+         dp++;
+      }
+}
+
+/* Read the entire image.  If the image has an alpha channel or a tRNS
+ * chunk, and you have called png_handle_alpha()[*], you will need to
+ * initialize the image to the current image that PNG will be overlaying.
+ * We set the num_rows again here, in case it was incorrectly set in
+ * png_read_start_row() by a call to png_read_update_info() or
+ * png_start_read_image() if png_set_interlace_handling() wasn't called
+ * prior to either of these functions like it should have been.  You can
+ * only call this function once.  If you desire to have an image for
+ * each pass of a interlaced image, use png_read_rows() instead.
+ *
+ * [*] png_handle_alpha() does not exist yet, as of libpng version 1.2.1
+ */
+void PNGAPI
+png_read_image(png_structp png_ptr, png_bytepp image)
+{
+   png_uint_32 i,image_height;
+   int pass, j;
+   png_bytepp rp;
+
+   png_debug(1, "in png_read_image\n");
+   /* save jump buffer and error functions */
+
+#ifdef PNG_READ_INTERLACING_SUPPORTED
+   pass = png_set_interlace_handling(png_ptr);
+#else
+   if (png_ptr->interlaced)
+      png_error(png_ptr,
+        "Cannot read interlaced image -- interlace handler disabled.");
+   pass = 1;
+#endif
+
+
+   image_height=png_ptr->height;
+   png_ptr->num_rows = image_height; /* Make sure this is set correctly */
+
+   for (j = 0; j < pass; j++)
+   {
+      rp = image;
+      for (i = 0; i < image_height; i++)
+      {
+         png_read_row(png_ptr, *rp, png_bytep_NULL);
+         rp++;
+      }
+   }
+}
+
+/* Read the end of the PNG file.  Will not read past the end of the
+ * file, will verify the end is accurate, and will read any comments
+ * or time information at the end of the file, if info is not NULL.
+ */
+void PNGAPI
+png_read_end(png_structp png_ptr, png_infop info_ptr)
+{
+   png_byte chunk_length[4];
+   png_uint_32 length;
+
+   png_debug(1, "in png_read_end\n");
+   /* save jump buffer and error functions */
+   png_crc_finish(png_ptr, 0); /* Finish off CRC from last IDAT chunk */
+
+   do
+   {
+#ifdef PNG_USE_LOCAL_ARRAYS
+      PNG_IHDR;
+      PNG_IDAT;
+      PNG_IEND;
+      PNG_PLTE;
+#if defined(PNG_READ_bKGD_SUPPORTED)
+      PNG_bKGD;
+#endif
+#if defined(PNG_READ_cHRM_SUPPORTED)
+      PNG_cHRM;
+#endif
+#if defined(PNG_READ_gAMA_SUPPORTED)
+      PNG_gAMA;
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+      PNG_hIST;
+#endif
+#if defined(PNG_READ_iCCP_SUPPORTED)
+      PNG_iCCP;
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+      PNG_iTXt;
+#endif
+#if defined(PNG_READ_oFFs_SUPPORTED)
+      PNG_oFFs;
+#endif
+#if defined(PNG_READ_pCAL_SUPPORTED)
+      PNG_pCAL;
+#endif
+#if defined(PNG_READ_pHYs_SUPPORTED)
+      PNG_pHYs;
+#endif
+#if defined(PNG_READ_sBIT_SUPPORTED)
+      PNG_sBIT;
+#endif
+#if defined(PNG_READ_sCAL_SUPPORTED)
+      PNG_sCAL;
+#endif
+#if defined(PNG_READ_sPLT_SUPPORTED)
+      PNG_sPLT;
+#endif
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      PNG_sRGB;
+#endif
+#if defined(PNG_READ_tEXt_SUPPORTED)
+      PNG_tEXt;
+#endif
+#if defined(PNG_READ_tIME_SUPPORTED)
+      PNG_tIME;
+#endif
+#if defined(PNG_READ_tRNS_SUPPORTED)
+      PNG_tRNS;
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+      PNG_zTXt;
+#endif
+#endif /* PNG_GLOBAL_ARRAYS */
+
+      png_read_data(png_ptr, chunk_length, 4);
+      length = png_get_uint_32(chunk_length);
+
+      png_reset_crc(png_ptr);
+      png_crc_read(png_ptr, png_ptr->chunk_name, 4);
+
+      png_debug1(0, "Reading %s chunk.\n", png_ptr->chunk_name);
+
+      if (!png_memcmp(png_ptr->chunk_name, png_IHDR, 4))
+         png_handle_IHDR(png_ptr, info_ptr, length);
+      else if (!png_memcmp(png_ptr->chunk_name, png_IEND, 4))
+         png_handle_IEND(png_ptr, info_ptr, length);
+#ifdef PNG_HANDLE_AS_UNKNOWN_SUPPORTED
+      else if (png_handle_as_unknown(png_ptr, png_ptr->chunk_name))
+      {
+         if (!png_memcmp(png_ptr->chunk_name, png_IDAT, 4))
+         {
+            if (length > 0 || png_ptr->mode & PNG_AFTER_IDAT)
+               png_error(png_ptr, "Too many IDAT's found");
+         }
+         else
+            png_ptr->mode |= PNG_AFTER_IDAT;
+         png_handle_unknown(png_ptr, info_ptr, length);
+         if (!png_memcmp(png_ptr->chunk_name, png_PLTE, 4))
+            png_ptr->mode |= PNG_HAVE_PLTE;
+      }
+#endif
+      else if (!png_memcmp(png_ptr->chunk_name, png_IDAT, 4))
+      {
+         /* Zero length IDATs are legal after the last IDAT has been
+          * read, but not after other chunks have been read.
+          */
+         if (length > 0 || png_ptr->mode & PNG_AFTER_IDAT)
+            png_error(png_ptr, "Too many IDAT's found");
+         png_crc_finish(png_ptr, length);
+      }
+      else if (!png_memcmp(png_ptr->chunk_name, png_PLTE, 4))
+         png_handle_PLTE(png_ptr, info_ptr, length);
+#if defined(PNG_READ_bKGD_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_bKGD, 4))
+         png_handle_bKGD(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_cHRM_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_cHRM, 4))
+         png_handle_cHRM(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_gAMA_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_gAMA, 4))
+         png_handle_gAMA(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_hIST, 4))
+         png_handle_hIST(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_oFFs_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_oFFs, 4))
+         png_handle_oFFs(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_pCAL_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_pCAL, 4))
+         png_handle_pCAL(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sCAL_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sCAL, 4))
+         png_handle_sCAL(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_pHYs_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_pHYs, 4))
+         png_handle_pHYs(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sBIT_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sBIT, 4))
+         png_handle_sBIT(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sRGB, 4))
+         png_handle_sRGB(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_iCCP_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_iCCP, 4))
+         png_handle_iCCP(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_sPLT_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_sPLT, 4))
+         png_handle_sPLT(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_tEXt_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_tEXt, 4))
+         png_handle_tEXt(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_tIME_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_tIME, 4))
+         png_handle_tIME(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_tRNS_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_tRNS, 4))
+         png_handle_tRNS(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_zTXt_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_zTXt, 4))
+         png_handle_zTXt(png_ptr, info_ptr, length);
+#endif
+#if defined(PNG_READ_iTXt_SUPPORTED)
+      else if (!png_memcmp(png_ptr->chunk_name, png_iTXt, 4))
+         png_handle_iTXt(png_ptr, info_ptr, length);
+#endif
+      else
+         png_handle_unknown(png_ptr, info_ptr, length);
+   } while (!(png_ptr->mode & PNG_HAVE_IEND));
+}
+
+/* free all memory used by the read */
+void PNGAPI
+png_destroy_read_struct(png_structpp png_ptr_ptr, png_infopp info_ptr_ptr,
+   png_infopp end_info_ptr_ptr)
+{
+   png_structp png_ptr = NULL;
+   png_infop info_ptr = NULL, end_info_ptr = NULL;
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_free_ptr free_fn = NULL;
+   png_voidp mem_ptr = NULL;
+#endif
+
+   png_debug(1, "in png_destroy_read_struct\n");
+   /* save jump buffer and error functions */
+   if (png_ptr_ptr != NULL)
+      png_ptr = *png_ptr_ptr;
+
+   if (info_ptr_ptr != NULL)
+      info_ptr = *info_ptr_ptr;
+
+   if (end_info_ptr_ptr != NULL)
+      end_info_ptr = *end_info_ptr_ptr;
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   free_fn = png_ptr->free_fn;
+   mem_ptr = png_ptr->mem_ptr;
+#endif
+
+   png_read_destroy(png_ptr, info_ptr, end_info_ptr);
+
+   if (info_ptr != NULL)
+   {
+#if defined(PNG_TEXT_SUPPORTED)
+      png_free_data(png_ptr, info_ptr, PNG_FREE_TEXT, -1);
+#endif
+
+#ifdef PNG_USER_MEM_SUPPORTED
+      png_destroy_struct_2((png_voidp)info_ptr, (png_free_ptr)free_fn,
+          (png_voidp)mem_ptr);
+#else
+      png_destroy_struct((png_voidp)info_ptr);
+#endif
+      *info_ptr_ptr = NULL;
+   }
+
+   if (end_info_ptr != NULL)
+   {
+#if defined(PNG_READ_TEXT_SUPPORTED)
+      png_free_data(png_ptr, end_info_ptr, PNG_FREE_TEXT, -1);
+#endif
+#ifdef PNG_USER_MEM_SUPPORTED
+      png_destroy_struct_2((png_voidp)end_info_ptr, (png_free_ptr)free_fn,
+         (png_voidp)mem_ptr);
+#else
+      png_destroy_struct((png_voidp)end_info_ptr);
+#endif
+      *end_info_ptr_ptr = NULL;
+   }
+
+   if (png_ptr != NULL)
+   {
+#ifdef PNG_USER_MEM_SUPPORTED
+      png_destroy_struct_2((png_voidp)png_ptr, (png_free_ptr)free_fn,
+          (png_voidp)mem_ptr);
+#else
+      png_destroy_struct((png_voidp)png_ptr);
+#endif
+      *png_ptr_ptr = NULL;
+   }
+}
+
+/* free all memory used by the read (old method) */
+void /* PRIVATE */
+png_read_destroy(png_structp png_ptr, png_infop info_ptr, png_infop end_info_ptr)
+{
+#ifdef PNG_SETJMP_SUPPORTED
+   jmp_buf tmp_jmp;
+#endif
+   png_error_ptr error_fn;
+   png_error_ptr warning_fn;
+   png_voidp error_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_free_ptr free_fn;
+#endif
+
+   png_debug(1, "in png_read_destroy\n");
+   /* save jump buffer and error functions */
+   if (info_ptr != NULL)
+      png_info_destroy(png_ptr, info_ptr);
+
+   if (end_info_ptr != NULL)
+      png_info_destroy(png_ptr, end_info_ptr);
+
+   png_free(png_ptr, png_ptr->zbuf);
+   png_free(png_ptr, png_ptr->big_row_buf);
+   png_free(png_ptr, png_ptr->prev_row);
+#if defined(PNG_READ_DITHER_SUPPORTED)
+   png_free(png_ptr, png_ptr->palette_lookup);
+   png_free(png_ptr, png_ptr->dither_index);
+#endif
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   png_free(png_ptr, png_ptr->gamma_table);
+#endif
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   png_free(png_ptr, png_ptr->gamma_from_1);
+   png_free(png_ptr, png_ptr->gamma_to_1);
+#endif
+#ifdef PNG_FREE_ME_SUPPORTED
+   if (png_ptr->free_me & PNG_FREE_PLTE)
+      png_zfree(png_ptr, png_ptr->palette);
+   png_ptr->free_me &= ~PNG_FREE_PLTE;
+#else
+   if (png_ptr->flags & PNG_FLAG_FREE_PLTE)
+      png_zfree(png_ptr, png_ptr->palette);
+   png_ptr->flags &= ~PNG_FLAG_FREE_PLTE;
+#endif
+#if defined(PNG_tRNS_SUPPORTED) || \
+    defined(PNG_READ_EXPAND_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+#ifdef PNG_FREE_ME_SUPPORTED
+   if (png_ptr->free_me & PNG_FREE_TRNS)
+      png_free(png_ptr, png_ptr->trans);
+   png_ptr->free_me &= ~PNG_FREE_TRNS;
+#else
+   if (png_ptr->flags & PNG_FLAG_FREE_TRNS)
+      png_free(png_ptr, png_ptr->trans);
+   png_ptr->flags &= ~PNG_FLAG_FREE_TRNS;
+#endif
+#endif
+#if defined(PNG_READ_hIST_SUPPORTED)
+#ifdef PNG_FREE_ME_SUPPORTED
+   if (png_ptr->free_me & PNG_FREE_HIST)
+      png_free(png_ptr, png_ptr->hist);
+   png_ptr->free_me &= ~PNG_FREE_HIST;
+#else
+   if (png_ptr->flags & PNG_FLAG_FREE_HIST)
+      png_free(png_ptr, png_ptr->hist);
+   png_ptr->flags &= ~PNG_FLAG_FREE_HIST;
+#endif
+#endif
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   if (png_ptr->gamma_16_table != NULL)
+   {
+      int i;
+      int istop = (1 << (8 - png_ptr->gamma_shift));
+      for (i = 0; i < istop; i++)
+      {
+         png_free(png_ptr, png_ptr->gamma_16_table[i]);
+      }
+   png_free(png_ptr, png_ptr->gamma_16_table);
+   }
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   if (png_ptr->gamma_16_from_1 != NULL)
+   {
+      int i;
+      int istop = (1 << (8 - png_ptr->gamma_shift));
+      for (i = 0; i < istop; i++)
+      {
+         png_free(png_ptr, png_ptr->gamma_16_from_1[i]);
+      }
+   png_free(png_ptr, png_ptr->gamma_16_from_1);
+   }
+   if (png_ptr->gamma_16_to_1 != NULL)
+   {
+      int i;
+      int istop = (1 << (8 - png_ptr->gamma_shift));
+      for (i = 0; i < istop; i++)
+      {
+         png_free(png_ptr, png_ptr->gamma_16_to_1[i]);
+      }
+   png_free(png_ptr, png_ptr->gamma_16_to_1);
+   }
+#endif
+#endif
+#if defined(PNG_TIME_RFC1123_SUPPORTED)
+   png_free(png_ptr, png_ptr->time_buffer);
+#endif
+
+   inflateEnd(&png_ptr->zstream);
+#ifdef PNG_PROGRESSIVE_READ_SUPPORTED
+   png_free(png_ptr, png_ptr->save_buffer);
+#endif
+
+   /* Save the important info out of the png_struct, in case it is
+    * being used again.
+    */
+#ifdef PNG_SETJMP_SUPPORTED
+   png_memcpy(tmp_jmp, png_ptr->jmpbuf, sizeof (jmp_buf));
+#endif
+
+   error_fn = png_ptr->error_fn;
+   warning_fn = png_ptr->warning_fn;
+   error_ptr = png_ptr->error_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+   free_fn = png_ptr->free_fn;
+#endif
+
+   png_memset(png_ptr, 0, sizeof (png_struct));
+
+   png_ptr->error_fn = error_fn;
+   png_ptr->warning_fn = warning_fn;
+   png_ptr->error_ptr = error_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_ptr->free_fn = free_fn;
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+   png_memcpy(png_ptr->jmpbuf, tmp_jmp, sizeof (jmp_buf));
+#endif
+
+}
+
+void PNGAPI
+png_set_read_status_fn(png_structp png_ptr, png_read_status_ptr read_row_fn)
+{
+   png_ptr->read_row_fn = read_row_fn;
+}
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+void PNGAPI
+png_read_png(png_structp png_ptr, png_infop info_ptr,
+                           int transforms,
+                           voidp params)
+{
+   int row;
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED)
+   /* invert the alpha channel from opacity to transparency */
+   if (transforms & PNG_TRANSFORM_INVERT_ALPHA)
+       png_set_invert_alpha(png_ptr);
+#endif
+
+   /* The call to png_read_info() gives us all of the information from the
+    * PNG file before the first IDAT (image data chunk).
+    */
+   png_read_info(png_ptr, info_ptr);
+
+   /* -------------- image transformations start here ------------------- */
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+   /* tell libpng to strip 16 bit/color files down to 8 bits/color */
+   if (transforms & PNG_TRANSFORM_STRIP_16)
+       png_set_strip_16(png_ptr);
+#endif
+
+#if defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+   /* Strip alpha bytes from the input data without combining with the
+    * background (not recommended).
+    */
+   if (transforms & PNG_TRANSFORM_STRIP_ALPHA)
+       png_set_strip_alpha(png_ptr);
+#endif
+
+#if defined(PNG_READ_PACK_SUPPORTED) && !defined(PNG_READ_EXPAND_SUPPORTED)
+   /* Extract multiple pixels with bit depths of 1, 2, and 4 from a single
+    * byte into separate bytes (useful for paletted and grayscale images).
+    */
+   if (transforms & PNG_TRANSFORM_PACKING)
+       png_set_packing(png_ptr);
+#endif
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+   /* Change the order of packed pixels to least significant bit first
+    * (not useful if you are using png_set_packing). */
+   if (transforms & PNG_TRANSFORM_PACKSWAP)
+       png_set_packswap(png_ptr);
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+   /* Expand paletted colors into true RGB triplets
+    * Expand grayscale images to full 8 bits from 1, 2, or 4 bits/pixel
+    * Expand paletted or RGB images with transparency to full alpha
+    * channels so the data will be available as RGBA quartets.
+    */
+   if (transforms & PNG_TRANSFORM_EXPAND)
+       if ((png_ptr->bit_depth < 8) ||
+           (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE) ||
+           (png_get_valid(png_ptr, info_ptr, PNG_INFO_tRNS)))
+         png_set_expand(png_ptr);
+#endif
+
+   /* We don't handle background color or gamma transformation or dithering. */
+
+#if defined(PNG_READ_INVERT_SUPPORTED)
+   /* invert monochrome files to have 0 as white and 1 as black */
+   if (transforms & PNG_TRANSFORM_INVERT_MONO)
+       png_set_invert_mono(png_ptr);
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED)
+   /* If you want to shift the pixel values from the range [0,255] or
+    * [0,65535] to the original [0,7] or [0,31], or whatever range the
+    * colors were originally in:
+    */
+   if ((transforms & PNG_TRANSFORM_SHIFT)
+       && png_get_valid(png_ptr, info_ptr, PNG_INFO_sBIT))
+   {
+      png_color_8p sig_bit;
+
+      png_get_sBIT(png_ptr, info_ptr, &sig_bit);
+      png_set_shift(png_ptr, sig_bit);
+   }
+#endif
+
+#if defined(PNG_READ_BGR_SUPPORTED)
+   /* flip the RGB pixels to BGR (or RGBA to BGRA) */
+   if (transforms & PNG_TRANSFORM_BGR)
+       png_set_bgr(png_ptr);
+#endif
+
+#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED)
+   /* swap the RGBA or GA data to ARGB or AG (or BGRA to ABGR) */
+   if (transforms & PNG_TRANSFORM_SWAP_ALPHA)
+       png_set_swap_alpha(png_ptr);
+#endif
+
+#if defined(PNG_READ_SWAP_SUPPORTED)
+   /* swap bytes of 16 bit files to least significant byte first */
+   if (transforms & PNG_TRANSFORM_SWAP_ENDIAN)
+       png_set_swap(png_ptr);
+#endif
+
+   /* We don't handle adding filler bytes */
+
+   /* Optional call to gamma correct and add the background to the palette
+    * and update info structure.  REQUIRED if you are expecting libpng to
+    * update the palette for you (i.e., you selected such a transform above).
+    */
+   png_read_update_info(png_ptr, info_ptr);
+
+   /* -------------- image transformations end here ------------------- */
+
+#ifdef PNG_FREE_ME_SUPPORTED
+   png_free_data(png_ptr, info_ptr, PNG_FREE_ROWS, 0);
+#endif
+   if(info_ptr->row_pointers == NULL)
+   {
+      info_ptr->row_pointers = (png_bytepp)png_malloc(png_ptr,
+         info_ptr->height * sizeof(png_bytep));
+#ifdef PNG_FREE_ME_SUPPORTED
+      info_ptr->free_me |= PNG_FREE_ROWS;
+#endif
+      for (row = 0; row < (int)info_ptr->height; row++)
+      {
+         info_ptr->row_pointers[row] = (png_bytep)png_malloc(png_ptr,
+            png_get_rowbytes(png_ptr, info_ptr));
+      }
+   }
+
+   png_read_image(png_ptr, info_ptr->row_pointers);
+   info_ptr->valid |= PNG_INFO_IDAT;
+
+   /* read rest of file, and get additional chunks in info_ptr - REQUIRED */
+   png_read_end(png_ptr, info_ptr);
+
+   if(transforms == 0 || params == NULL)
+      /* quiet compiler warnings */ return;
+
+}
+#endif
diff --git a/png/pngrio.c b/png/pngrio.c
new file mode 100644
index 000000000..79755b492
--- /dev/null
+++ b/png/pngrio.c
@@ -0,0 +1,161 @@
+
+/* pngrio.c - functions for data input
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file provides a location for all input.  Users who need
+ * special handling are expected to write a function that has the same
+ * arguments as this and performs a similar function, but that possibly
+ * has a different input method.  Note that you shouldn't change this
+ * function, but rather write a replacement function and then make
+ * libpng use it at run time with png_set_read_fn(...).
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+/* Read the data from whatever input you are using.  The default routine
+   reads from a file pointer.  Note that this routine sometimes gets called
+   with very small lengths, so you should implement some kind of simple
+   buffering if you are using unbuffered reads.  This should never be asked
+   to read more then 64K on a 16 bit machine. */
+void /* PRIVATE */
+png_read_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   png_debug1(4,"reading %d bytes\n", (int)length);
+   if (png_ptr->read_data_fn != NULL)
+      (*(png_ptr->read_data_fn))(png_ptr, data, length);
+   else
+      png_error(png_ptr, "Call to NULL read function");
+}
+
+#if !defined(PNG_NO_STDIO)
+/* This is the function that does the actual reading of data.  If you are
+   not reading from a standard C stream, you should create a replacement
+   read_data function and use it at run time with png_set_read_fn(), rather
+   than changing the library. */
+#ifndef USE_FAR_KEYWORD
+static void /* PRIVATE */
+png_default_read_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   png_size_t check;
+
+   /* fread() returns 0 on error, so it is OK to store this in a png_size_t
+    * instead of an int, which is what fread() actually returns.
+    */
+#if defined(_WIN32_WCE)
+   if ( !ReadFile((HANDLE)(png_ptr->io_ptr), data, length, &check, NULL) )
+      check = 0;
+#else
+   check = (png_size_t)fread(data, (png_size_t)1, length,
+      (png_FILE_p)png_ptr->io_ptr);
+#endif
+
+   if (check != length)
+      png_error(png_ptr, "Read Error");
+}
+#else
+/* this is the model-independent version. Since the standard I/O library
+   can't handle far buffers in the medium and small models, we have to copy
+   the data.
+*/
+
+#define NEAR_BUF_SIZE 1024
+#define MIN(a,b) (a <= b ? a : b)
+
+static void /* PRIVATE */
+png_default_read_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   int check;
+   png_byte *n_data;
+   png_FILE_p io_ptr;
+
+   /* Check if data really is near. If so, use usual code. */
+   n_data = (png_byte *)CVT_PTR_NOCHECK(data);
+   io_ptr = (png_FILE_p)CVT_PTR(png_ptr->io_ptr);
+   if ((png_bytep)n_data == data)
+   {
+#if defined(_WIN32_WCE)
+      if ( !ReadFile((HANDLE)(png_ptr->io_ptr), data, length, &check, NULL) )
+         check = 0;
+#else
+      check = fread(n_data, 1, length, io_ptr);
+#endif
+   }
+   else
+   {
+      png_byte buf[NEAR_BUF_SIZE];
+      png_size_t read, remaining, err;
+      check = 0;
+      remaining = length;
+      do
+      {
+         read = MIN(NEAR_BUF_SIZE, remaining);
+#if defined(_WIN32_WCE)
+         if ( !ReadFile((HANDLE)(io_ptr), buf, read, &err, NULL) )
+            err = 0;
+#else
+         err = fread(buf, (png_size_t)1, read, io_ptr);
+#endif
+         png_memcpy(data, buf, read); /* copy far buffer to near buffer */
+         if(err != read)
+            break;
+         else
+            check += err;
+         data += read;
+         remaining -= read;
+      }
+      while (remaining != 0);
+   }
+   if ((png_uint_32)check != (png_uint_32)length)
+      png_error(png_ptr, "read Error");
+}
+#endif
+#endif
+
+/* This function allows the application to supply a new input function
+   for libpng if standard C streams aren't being used.
+
+   This function takes as its arguments:
+   png_ptr      - pointer to a png input data structure
+   io_ptr       - pointer to user supplied structure containing info about
+                  the input functions.  May be NULL.
+   read_data_fn - pointer to a new input function that takes as its
+                  arguments a pointer to a png_struct, a pointer to
+                  a location where input data can be stored, and a 32-bit
+                  unsigned int that is the number of bytes to be read.
+                  To exit and output any fatal error messages the new write
+                  function should call png_error(png_ptr, "Error msg"). */
+void PNGAPI
+png_set_read_fn(png_structp png_ptr, png_voidp io_ptr,
+   png_rw_ptr read_data_fn)
+{
+   png_ptr->io_ptr = io_ptr;
+
+#if !defined(PNG_NO_STDIO)
+   if (read_data_fn != NULL)
+      png_ptr->read_data_fn = read_data_fn;
+   else
+      png_ptr->read_data_fn = png_default_read_data;
+#else
+   png_ptr->read_data_fn = read_data_fn;
+#endif
+
+   /* It is an error to write to a read device */
+   if (png_ptr->write_data_fn != NULL)
+   {
+      png_ptr->write_data_fn = NULL;
+      png_warning(png_ptr,
+         "It's an error to set both read_data_fn and write_data_fn in the ");
+      png_warning(png_ptr,
+         "same structure.  Resetting write_data_fn to NULL.");
+   }
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+   png_ptr->output_flush_fn = NULL;
+#endif
+}
diff --git a/png/pngrtran.c b/png/pngrtran.c
new file mode 100644
index 000000000..0372b8ea9
--- /dev/null
+++ b/png/pngrtran.c
@@ -0,0 +1,4137 @@
+
+/* pngrtran.c - transforms the data in a row for PNG readers
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file contains functions optionally called by an application
+ * in order to tell libpng how to handle data when reading a PNG.
+ * Transformations that are used in both reading and writing are
+ * in pngtrans.c.
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+/* Set the action on getting a CRC error for an ancillary or critical chunk. */
+void PNGAPI
+png_set_crc_action(png_structp png_ptr, int crit_action, int ancil_action)
+{
+   png_debug(1, "in png_set_crc_action\n");
+   /* Tell libpng how we react to CRC errors in critical chunks */
+   switch (crit_action)
+   {
+      case PNG_CRC_NO_CHANGE:                        /* leave setting as is */
+         break;
+      case PNG_CRC_WARN_USE:                               /* warn/use data */
+         png_ptr->flags &= ~PNG_FLAG_CRC_CRITICAL_MASK;
+         png_ptr->flags |= PNG_FLAG_CRC_CRITICAL_USE;
+         break;
+      case PNG_CRC_QUIET_USE:                             /* quiet/use data */
+         png_ptr->flags &= ~PNG_FLAG_CRC_CRITICAL_MASK;
+         png_ptr->flags |= PNG_FLAG_CRC_CRITICAL_USE |
+                           PNG_FLAG_CRC_CRITICAL_IGNORE;
+         break;
+      case PNG_CRC_WARN_DISCARD:    /* not a valid action for critical data */
+         png_warning(png_ptr, "Can't discard critical data on CRC error.");
+      case PNG_CRC_ERROR_QUIT:                                /* error/quit */
+      case PNG_CRC_DEFAULT:
+      default:
+         png_ptr->flags &= ~PNG_FLAG_CRC_CRITICAL_MASK;
+         break;
+   }
+
+   switch (ancil_action)
+   {
+      case PNG_CRC_NO_CHANGE:                       /* leave setting as is */
+         break;
+      case PNG_CRC_WARN_USE:                              /* warn/use data */
+         png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK;
+         png_ptr->flags |= PNG_FLAG_CRC_ANCILLARY_USE;
+         break;
+      case PNG_CRC_QUIET_USE:                            /* quiet/use data */
+         png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK;
+         png_ptr->flags |= PNG_FLAG_CRC_ANCILLARY_USE |
+                           PNG_FLAG_CRC_ANCILLARY_NOWARN;
+         break;
+      case PNG_CRC_ERROR_QUIT:                               /* error/quit */
+         png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK;
+         png_ptr->flags |= PNG_FLAG_CRC_ANCILLARY_NOWARN;
+         break;
+      case PNG_CRC_WARN_DISCARD:                      /* warn/discard data */
+      case PNG_CRC_DEFAULT:
+      default:
+         png_ptr->flags &= ~PNG_FLAG_CRC_ANCILLARY_MASK;
+         break;
+   }
+}
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED) && \
+    defined(PNG_FLOATING_POINT_SUPPORTED)
+/* handle alpha and tRNS via a background color */
+void PNGAPI
+png_set_background(png_structp png_ptr,
+   png_color_16p background_color, int background_gamma_code,
+   int need_expand, double background_gamma)
+{
+   png_debug(1, "in png_set_background\n");
+   if (background_gamma_code == PNG_BACKGROUND_GAMMA_UNKNOWN)
+   {
+      png_warning(png_ptr, "Application must supply a known background gamma");
+      return;
+   }
+
+   png_ptr->transformations |= PNG_BACKGROUND;
+   png_memcpy(&(png_ptr->background), background_color, sizeof(png_color_16));
+   png_ptr->background_gamma = (float)background_gamma;
+   png_ptr->background_gamma_type = (png_byte)(background_gamma_code);
+   png_ptr->transformations |= (need_expand ? PNG_BACKGROUND_EXPAND : 0);
+
+   /* Note:  if need_expand is set and color_type is either RGB or RGB_ALPHA
+    * (in which case need_expand is superfluous anyway), the background color
+    * might actually be gray yet not be flagged as such. This is not a problem
+    * for the current code, which uses PNG_BACKGROUND_IS_GRAY only to
+    * decide when to do the png_do_gray_to_rgb() transformation.
+    */
+   if ((need_expand && !(png_ptr->color_type & PNG_COLOR_MASK_COLOR)) ||
+       (!need_expand && background_color->red == background_color->green &&
+        background_color->red == background_color->blue))
+      png_ptr->mode |= PNG_BACKGROUND_IS_GRAY;
+}
+#endif
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+/* strip 16 bit depth files to 8 bit depth */
+void PNGAPI
+png_set_strip_16(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_strip_16\n");
+   png_ptr->transformations |= PNG_16_TO_8;
+}
+#endif
+
+#if defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+void PNGAPI
+png_set_strip_alpha(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_strip_alpha\n");
+   png_ptr->transformations |= PNG_STRIP_ALPHA;
+}
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+/* Dither file to 8 bit.  Supply a palette, the current number
+ * of elements in the palette, the maximum number of elements
+ * allowed, and a histogram if possible.  If the current number
+ * of colors is greater then the maximum number, the palette will be
+ * modified to fit in the maximum number.  "full_dither" indicates
+ * whether we need a dithering cube set up for RGB images, or if we
+ * simply are reducing the number of colors in a paletted image.
+ */
+
+typedef struct png_dsort_struct
+{
+   struct png_dsort_struct FAR * next;
+   png_byte left;
+   png_byte right;
+} png_dsort;
+typedef png_dsort FAR *       png_dsortp;
+typedef png_dsort FAR * FAR * png_dsortpp;
+
+void PNGAPI
+png_set_dither(png_structp png_ptr, png_colorp palette,
+   int num_palette, int maximum_colors, png_uint_16p histogram,
+   int full_dither)
+{
+   png_debug(1, "in png_set_dither\n");
+   png_ptr->transformations |= PNG_DITHER;
+
+   if (!full_dither)
+   {
+      int i;
+
+      png_ptr->dither_index = (png_bytep)png_malloc(png_ptr,
+         (png_uint_32)(num_palette * sizeof (png_byte)));
+      for (i = 0; i < num_palette; i++)
+         png_ptr->dither_index[i] = (png_byte)i;
+   }
+
+   if (num_palette > maximum_colors)
+   {
+      if (histogram != NULL)
+      {
+         /* This is easy enough, just throw out the least used colors.
+            Perhaps not the best solution, but good enough. */
+
+         int i;
+         png_bytep sort;
+
+         /* initialize an array to sort colors */
+         sort = (png_bytep)png_malloc(png_ptr, (png_uint_32)(num_palette
+            * sizeof (png_byte)));
+
+         /* initialize the sort array */
+         for (i = 0; i < num_palette; i++)
+            sort[i] = (png_byte)i;
+
+         /* Find the least used palette entries by starting a
+            bubble sort, and running it until we have sorted
+            out enough colors.  Note that we don't care about
+            sorting all the colors, just finding which are
+            least used. */
+
+         for (i = num_palette - 1; i >= maximum_colors; i--)
+         {
+            int done; /* to stop early if the list is pre-sorted */
+            int j;
+
+            done = 1;
+            for (j = 0; j < i; j++)
+            {
+               if (histogram[sort[j]] < histogram[sort[j + 1]])
+               {
+                  png_byte t;
+
+                  t = sort[j];
+                  sort[j] = sort[j + 1];
+                  sort[j + 1] = t;
+                  done = 0;
+               }
+            }
+            if (done)
+               break;
+         }
+
+         /* swap the palette around, and set up a table, if necessary */
+         if (full_dither)
+         {
+            int j = num_palette;
+
+            /* put all the useful colors within the max, but don't
+               move the others */
+            for (i = 0; i < maximum_colors; i++)
+            {
+               if ((int)sort[i] >= maximum_colors)
+               {
+                  do
+                     j--;
+                  while ((int)sort[j] >= maximum_colors);
+                  palette[i] = palette[j];
+               }
+            }
+         }
+         else
+         {
+            int j = num_palette;
+
+            /* move all the used colors inside the max limit, and
+               develop a translation table */
+            for (i = 0; i < maximum_colors; i++)
+            {
+               /* only move the colors we need to */
+               if ((int)sort[i] >= maximum_colors)
+               {
+                  png_color tmp_color;
+
+                  do
+                     j--;
+                  while ((int)sort[j] >= maximum_colors);
+
+                  tmp_color = palette[j];
+                  palette[j] = palette[i];
+                  palette[i] = tmp_color;
+                  /* indicate where the color went */
+                  png_ptr->dither_index[j] = (png_byte)i;
+                  png_ptr->dither_index[i] = (png_byte)j;
+               }
+            }
+
+            /* find closest color for those colors we are not using */
+            for (i = 0; i < num_palette; i++)
+            {
+               if ((int)png_ptr->dither_index[i] >= maximum_colors)
+               {
+                  int min_d, k, min_k, d_index;
+
+                  /* find the closest color to one we threw out */
+                  d_index = png_ptr->dither_index[i];
+                  min_d = PNG_COLOR_DIST(palette[d_index], palette[0]);
+                  for (k = 1, min_k = 0; k < maximum_colors; k++)
+                  {
+                     int d;
+
+                     d = PNG_COLOR_DIST(palette[d_index], palette[k]);
+
+                     if (d < min_d)
+                     {
+                        min_d = d;
+                        min_k = k;
+                     }
+                  }
+                  /* point to closest color */
+                  png_ptr->dither_index[i] = (png_byte)min_k;
+               }
+            }
+         }
+         png_free(png_ptr, sort);
+      }
+      else
+      {
+         /* This is much harder to do simply (and quickly).  Perhaps
+            we need to go through a median cut routine, but those
+            don't always behave themselves with only a few colors
+            as input.  So we will just find the closest two colors,
+            and throw out one of them (chosen somewhat randomly).
+            [We don't understand this at all, so if someone wants to
+             work on improving it, be our guest - AED, GRP]
+            */
+         int i;
+         int max_d;
+         int num_new_palette;
+         png_dsortpp hash;
+         png_bytep index_to_palette;
+            /* where the original index currently is in the palette */
+         png_bytep palette_to_index;
+            /* which original index points to this palette color */
+
+         /* initialize palette index arrays */
+         index_to_palette = (png_bytep)png_malloc(png_ptr,
+            (png_uint_32)(num_palette * sizeof (png_byte)));
+         palette_to_index = (png_bytep)png_malloc(png_ptr,
+            (png_uint_32)(num_palette * sizeof (png_byte)));
+
+         /* initialize the sort array */
+         for (i = 0; i < num_palette; i++)
+         {
+            index_to_palette[i] = (png_byte)i;
+            palette_to_index[i] = (png_byte)i;
+         }
+
+         hash = (png_dsortpp)png_malloc(png_ptr, (png_uint_32)(769 *
+            sizeof (png_dsortp)));
+         for (i = 0; i < 769; i++)
+            hash[i] = NULL;
+/*         png_memset(hash, 0, 769 * sizeof (png_dsortp)); */
+
+         num_new_palette = num_palette;
+
+         /* initial wild guess at how far apart the farthest pixel
+            pair we will be eliminating will be.  Larger
+            numbers mean more areas will be allocated, Smaller
+            numbers run the risk of not saving enough data, and
+            having to do this all over again.
+
+            I have not done extensive checking on this number.
+            */
+         max_d = 96;
+
+         while (num_new_palette > maximum_colors)
+         {
+            for (i = 0; i < num_new_palette - 1; i++)
+            {
+               int j;
+
+               for (j = i + 1; j < num_new_palette; j++)
+               {
+                  int d;
+
+                  d = PNG_COLOR_DIST(palette[i], palette[j]);
+
+                  if (d <= max_d)
+                  {
+                     png_dsortp t;
+
+                     t = (png_dsortp)png_malloc(png_ptr, (png_uint_32)(sizeof
+                         (png_dsort)));
+                     t->next = hash[d];
+                     t->left = (png_byte)i;
+                     t->right = (png_byte)j;
+                     hash[d] = t;
+                  }
+               }
+            }
+
+            for (i = 0; i <= max_d; i++)
+            {
+               if (hash[i] != NULL)
+               {
+                  png_dsortp p;
+
+                  for (p = hash[i]; p; p = p->next)
+                  {
+                     if ((int)index_to_palette[p->left] < num_new_palette &&
+                        (int)index_to_palette[p->right] < num_new_palette)
+                     {
+                        int j, next_j;
+
+                        if (num_new_palette & 0x01)
+                        {
+                           j = p->left;
+                           next_j = p->right;
+                        }
+                        else
+                        {
+                           j = p->right;
+                           next_j = p->left;
+                        }
+
+                        num_new_palette--;
+                        palette[index_to_palette[j]] = palette[num_new_palette];
+                        if (!full_dither)
+                        {
+                           int k;
+
+                           for (k = 0; k < num_palette; k++)
+                           {
+                              if (png_ptr->dither_index[k] ==
+                                 index_to_palette[j])
+                                 png_ptr->dither_index[k] =
+                                    index_to_palette[next_j];
+                              if ((int)png_ptr->dither_index[k] ==
+                                 num_new_palette)
+                                 png_ptr->dither_index[k] =
+                                    index_to_palette[j];
+                           }
+                        }
+
+                        index_to_palette[palette_to_index[num_new_palette]] =
+                           index_to_palette[j];
+                        palette_to_index[index_to_palette[j]] =
+                           palette_to_index[num_new_palette];
+
+                        index_to_palette[j] = (png_byte)num_new_palette;
+                        palette_to_index[num_new_palette] = (png_byte)j;
+                     }
+                     if (num_new_palette <= maximum_colors)
+                        break;
+                  }
+                  if (num_new_palette <= maximum_colors)
+                     break;
+               }
+            }
+
+            for (i = 0; i < 769; i++)
+            {
+               if (hash[i] != NULL)
+               {
+                  png_dsortp p = hash[i];
+                  while (p)
+                  {
+                     png_dsortp t;
+
+                     t = p->next;
+                     png_free(png_ptr, p);
+                     p = t;
+                  }
+               }
+               hash[i] = 0;
+            }
+            max_d += 96;
+         }
+         png_free(png_ptr, hash);
+         png_free(png_ptr, palette_to_index);
+         png_free(png_ptr, index_to_palette);
+      }
+      num_palette = maximum_colors;
+   }
+   if (png_ptr->palette == NULL)
+   {
+      png_ptr->palette = palette;
+   }
+   png_ptr->num_palette = (png_uint_16)num_palette;
+
+   if (full_dither)
+   {
+      int i;
+      png_bytep distance;
+      int total_bits = PNG_DITHER_RED_BITS + PNG_DITHER_GREEN_BITS +
+         PNG_DITHER_BLUE_BITS;
+      int num_red = (1 << PNG_DITHER_RED_BITS);
+      int num_green = (1 << PNG_DITHER_GREEN_BITS);
+      int num_blue = (1 << PNG_DITHER_BLUE_BITS);
+      png_size_t num_entries = ((png_size_t)1 << total_bits);
+
+      png_ptr->palette_lookup = (png_bytep )png_malloc(png_ptr,
+         (png_uint_32)(num_entries * sizeof (png_byte)));
+
+      png_memset(png_ptr->palette_lookup, 0, num_entries * sizeof (png_byte));
+
+      distance = (png_bytep)png_malloc(png_ptr, (png_uint_32)(num_entries *
+         sizeof(png_byte)));
+
+      png_memset(distance, 0xff, num_entries * sizeof(png_byte));
+
+      for (i = 0; i < num_palette; i++)
+      {
+         int ir, ig, ib;
+         int r = (palette[i].red >> (8 - PNG_DITHER_RED_BITS));
+         int g = (palette[i].green >> (8 - PNG_DITHER_GREEN_BITS));
+         int b = (palette[i].blue >> (8 - PNG_DITHER_BLUE_BITS));
+
+         for (ir = 0; ir < num_red; ir++)
+         {
+            int dr = abs(ir - r);
+            int index_r = (ir << (PNG_DITHER_BLUE_BITS + PNG_DITHER_GREEN_BITS));
+
+            for (ig = 0; ig < num_green; ig++)
+            {
+               int dg = abs(ig - g);
+               int dt = dr + dg;
+               int dm = ((dr > dg) ? dr : dg);
+               int index_g = index_r | (ig << PNG_DITHER_BLUE_BITS);
+
+               for (ib = 0; ib < num_blue; ib++)
+               {
+                  int d_index = index_g | ib;
+                  int db = abs(ib - b);
+                  int dmax = ((dm > db) ? dm : db);
+                  int d = dmax + dt + db;
+
+                  if (d < (int)distance[d_index])
+                  {
+                     distance[d_index] = (png_byte)d;
+                     png_ptr->palette_lookup[d_index] = (png_byte)i;
+                  }
+               }
+            }
+         }
+      }
+
+      png_free(png_ptr, distance);
+   }
+}
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED) && defined(PNG_FLOATING_POINT_SUPPORTED)
+/* Transform the image from the file_gamma to the screen_gamma.  We
+ * only do transformations on images where the file_gamma and screen_gamma
+ * are not close reciprocals, otherwise it slows things down slightly, and
+ * also needlessly introduces small errors.
+ */
+void PNGAPI
+png_set_gamma(png_structp png_ptr, double scrn_gamma, double file_gamma)
+{
+   png_debug(1, "in png_set_gamma\n");
+   if (fabs(scrn_gamma * file_gamma - 1.0) > PNG_GAMMA_THRESHOLD)
+      png_ptr->transformations |= PNG_GAMMA;
+   png_ptr->gamma = (float)file_gamma;
+   png_ptr->screen_gamma = (float)scrn_gamma;
+}
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+/* Expand paletted images to RGB, expand grayscale images of
+ * less than 8-bit depth to 8-bit depth, and expand tRNS chunks
+ * to alpha channels.
+ */
+void PNGAPI
+png_set_expand(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_expand\n");
+   png_ptr->transformations |= PNG_EXPAND;
+}
+
+/* GRR 19990627:  the following three functions currently are identical
+ *  to png_set_expand().  However, it is entirely reasonable that someone
+ *  might wish to expand an indexed image to RGB but *not* expand a single,
+ *  fully transparent palette entry to a full alpha channel--perhaps instead
+ *  convert tRNS to the grayscale/RGB format (16-bit RGB value), or replace
+ *  the transparent color with a particular RGB value, or drop tRNS entirely.
+ *  IOW, a future version of the library may make the transformations flag
+ *  a bit more fine-grained, with separate bits for each of these three
+ *  functions.
+ *
+ *  More to the point, these functions make it obvious what libpng will be
+ *  doing, whereas "expand" can (and does) mean any number of things.
+ */
+
+/* Expand paletted images to RGB. */
+void PNGAPI
+png_set_palette_to_rgb(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_expand\n");
+   png_ptr->transformations |= PNG_EXPAND;
+}
+
+/* Expand grayscale images of less than 8-bit depth to 8 bits. */
+void PNGAPI
+png_set_gray_1_2_4_to_8(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_expand\n");
+   png_ptr->transformations |= PNG_EXPAND;
+}
+
+/* Expand tRNS chunks to alpha channels. */
+void PNGAPI
+png_set_tRNS_to_alpha(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_expand\n");
+   png_ptr->transformations |= PNG_EXPAND;
+}
+#endif /* defined(PNG_READ_EXPAND_SUPPORTED) */
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+void PNGAPI
+png_set_gray_to_rgb(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_gray_to_rgb\n");
+   png_ptr->transformations |= PNG_GRAY_TO_RGB;
+}
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+#if defined(PNG_FLOATING_POINT_SUPPORTED)
+/* Convert a RGB image to a grayscale of the same width.  This allows us,
+ * for example, to convert a 24 bpp RGB image into an 8 bpp grayscale image.
+ */
+
+void PNGAPI
+png_set_rgb_to_gray(png_structp png_ptr, int error_action, double red,
+   double green)
+{
+      int red_fixed = (int)((float)red*100000.0 + 0.5);
+      int green_fixed = (int)((float)green*100000.0 + 0.5);
+      png_set_rgb_to_gray_fixed(png_ptr, error_action, red_fixed, green_fixed);
+}
+#endif
+
+void PNGAPI
+png_set_rgb_to_gray_fixed(png_structp png_ptr, int error_action,
+   png_fixed_point red, png_fixed_point green)
+{
+   png_debug(1, "in png_set_rgb_to_gray\n");
+   switch(error_action)
+   {
+      case 1: png_ptr->transformations |= PNG_RGB_TO_GRAY;
+              break;
+      case 2: png_ptr->transformations |= PNG_RGB_TO_GRAY_WARN;
+              break;
+      case 3: png_ptr->transformations |= PNG_RGB_TO_GRAY_ERR;
+   }
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+      png_ptr->transformations |= PNG_EXPAND;
+#else
+   {
+      png_warning(png_ptr, "Cannot do RGB_TO_GRAY without EXPAND_SUPPORTED.");
+      png_ptr->transformations &= ~PNG_RGB_TO_GRAY;
+   }
+#endif
+   {
+      png_uint_16 red_int, green_int;
+      if(red < 0 || green < 0)
+      {
+         red_int   =  6968; /* .212671 * 32768 + .5 */
+         green_int = 23434; /* .715160 * 32768 + .5 */
+      }
+      else if(red + green < 100000L)
+      {
+        red_int = (png_uint_16)(((png_uint_32)red*32768L)/100000L);
+        green_int = (png_uint_16)(((png_uint_32)green*32768L)/100000L);
+      }
+      else
+      {
+         png_warning(png_ptr, "ignoring out of range rgb_to_gray coefficients");
+         red_int   =  6968;
+         green_int = 23434;
+      }
+      png_ptr->rgb_to_gray_red_coeff   = red_int;
+      png_ptr->rgb_to_gray_green_coeff = green_int;
+      png_ptr->rgb_to_gray_blue_coeff  = (png_uint_16)(32768-red_int-green_int);
+   }
+}
+#endif
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_LEGACY_SUPPORTED)
+void PNGAPI
+png_set_read_user_transform_fn(png_structp png_ptr, png_user_transform_ptr
+   read_user_transform_fn)
+{
+   png_debug(1, "in png_set_read_user_transform_fn\n");
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED)
+   png_ptr->transformations |= PNG_USER_TRANSFORM;
+   png_ptr->read_user_transform_fn = read_user_transform_fn;
+#endif
+#ifdef PNG_LEGACY_SUPPORTED
+   if(read_user_transform_fn)
+      png_warning(png_ptr,
+        "This version of libpng does not support user transforms");
+#endif
+}
+#endif
+
+/* Initialize everything needed for the read.  This includes modifying
+ * the palette.
+ */
+void /* PRIVATE */
+png_init_read_transformations(png_structp png_ptr)
+{
+   png_debug(1, "in png_init_read_transformations\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if(png_ptr != NULL)
+#endif
+  {
+#if defined(PNG_READ_BACKGROUND_SUPPORTED) || defined(PNG_READ_SHIFT_SUPPORTED) \
+ || defined(PNG_READ_GAMMA_SUPPORTED)
+   int color_type = png_ptr->color_type;
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED) && defined(PNG_READ_BACKGROUND_SUPPORTED)
+   if ((png_ptr->transformations & PNG_BACKGROUND_EXPAND) &&
+       (png_ptr->transformations & PNG_EXPAND))
+   {
+      if (!(color_type & PNG_COLOR_MASK_COLOR))  /* i.e., GRAY or GRAY_ALPHA */
+      {
+         /* expand background chunk. */
+         switch (png_ptr->bit_depth)
+         {
+            case 1:
+               png_ptr->background.gray *= (png_uint_16)0xff;
+               png_ptr->background.red = png_ptr->background.green =
+               png_ptr->background.blue = png_ptr->background.gray;
+               break;
+            case 2:
+               png_ptr->background.gray *= (png_uint_16)0x55;
+               png_ptr->background.red = png_ptr->background.green =
+               png_ptr->background.blue = png_ptr->background.gray;
+               break;
+            case 4:
+               png_ptr->background.gray *= (png_uint_16)0x11;
+               png_ptr->background.red = png_ptr->background.green =
+               png_ptr->background.blue = png_ptr->background.gray;
+               break;
+            case 8:
+            case 16:
+               png_ptr->background.red = png_ptr->background.green =
+               png_ptr->background.blue = png_ptr->background.gray;
+               break;
+         }
+      }
+      else if (color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         png_ptr->background.red   =
+            png_ptr->palette[png_ptr->background.index].red;
+         png_ptr->background.green =
+            png_ptr->palette[png_ptr->background.index].green;
+         png_ptr->background.blue  =
+            png_ptr->palette[png_ptr->background.index].blue;
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED)
+        if (png_ptr->transformations & PNG_INVERT_ALPHA)
+        {
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+           if (!(png_ptr->transformations & PNG_EXPAND))
+#endif
+           {
+           /* invert the alpha channel (in tRNS) unless the pixels are
+              going to be expanded, in which case leave it for later */
+              int i,istop;
+              istop=(int)png_ptr->num_trans;
+              for (i=0; i<istop; i++)
+                 png_ptr->trans[i] = (png_byte)(255 - png_ptr->trans[i]);
+           }
+        }
+#endif
+
+      }
+   }
+#endif
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED) && defined(PNG_READ_GAMMA_SUPPORTED)
+   png_ptr->background_1 = png_ptr->background;
+#endif
+#if defined(PNG_READ_GAMMA_SUPPORTED) && defined(PNG_FLOATING_POINT_SUPPORTED)
+   if (png_ptr->transformations & (PNG_GAMMA | PNG_RGB_TO_GRAY))
+   {
+      png_build_gamma_table(png_ptr);
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+      if (png_ptr->transformations & PNG_BACKGROUND)
+      {
+         if (color_type == PNG_COLOR_TYPE_PALETTE)
+         {
+            png_color back, back_1;
+            png_colorp palette = png_ptr->palette;
+            int num_palette = png_ptr->num_palette;
+            int i;
+            if (png_ptr->background_gamma_type == PNG_BACKGROUND_GAMMA_FILE)
+            {
+               back.red = png_ptr->gamma_table[png_ptr->background.red];
+               back.green = png_ptr->gamma_table[png_ptr->background.green];
+               back.blue = png_ptr->gamma_table[png_ptr->background.blue];
+
+               back_1.red = png_ptr->gamma_to_1[png_ptr->background.red];
+               back_1.green = png_ptr->gamma_to_1[png_ptr->background.green];
+               back_1.blue = png_ptr->gamma_to_1[png_ptr->background.blue];
+            }
+            else
+            {
+               double g, gs;
+
+               switch (png_ptr->background_gamma_type)
+               {
+                  case PNG_BACKGROUND_GAMMA_SCREEN:
+                     g = (png_ptr->screen_gamma);
+                     gs = 1.0;
+                     break;
+                  case PNG_BACKGROUND_GAMMA_FILE:
+                     g = 1.0 / (png_ptr->gamma);
+                     gs = 1.0 / (png_ptr->gamma * png_ptr->screen_gamma);
+                     break;
+                  case PNG_BACKGROUND_GAMMA_UNIQUE:
+                     g = 1.0 / (png_ptr->background_gamma);
+                     gs = 1.0 / (png_ptr->background_gamma *
+                                 png_ptr->screen_gamma);
+                     break;
+                  default:
+                     g = 1.0;    /* back_1 */
+                     gs = 1.0;   /* back */
+               }
+
+               if ( fabs(gs - 1.0) < PNG_GAMMA_THRESHOLD)
+               {
+                  back.red   = (png_byte)png_ptr->background.red;
+                  back.green = (png_byte)png_ptr->background.green;
+                  back.blue  = (png_byte)png_ptr->background.blue;
+               }
+               else
+               {
+                  back.red = (png_byte)(pow(
+                     (double)png_ptr->background.red/255, gs) * 255.0 + .5);
+                  back.green = (png_byte)(pow(
+                     (double)png_ptr->background.green/255, gs) * 255.0 + .5);
+                  back.blue = (png_byte)(pow(
+                     (double)png_ptr->background.blue/255, gs) * 255.0 + .5);
+               }
+
+               back_1.red = (png_byte)(pow(
+                  (double)png_ptr->background.red/255, g) * 255.0 + .5);
+               back_1.green = (png_byte)(pow(
+                  (double)png_ptr->background.green/255, g) * 255.0 + .5);
+               back_1.blue = (png_byte)(pow(
+                  (double)png_ptr->background.blue/255, g) * 255.0 + .5);
+            }
+            for (i = 0; i < num_palette; i++)
+            {
+               if (i < (int)png_ptr->num_trans && png_ptr->trans[i] != 0xff)
+               {
+                  if (png_ptr->trans[i] == 0)
+                  {
+                     palette[i] = back;
+                  }
+                  else /* if (png_ptr->trans[i] != 0xff) */
+                  {
+                     png_byte v, w;
+
+                     v = png_ptr->gamma_to_1[palette[i].red];
+                     png_composite(w, v, png_ptr->trans[i], back_1.red);
+                     palette[i].red = png_ptr->gamma_from_1[w];
+
+                     v = png_ptr->gamma_to_1[palette[i].green];
+                     png_composite(w, v, png_ptr->trans[i], back_1.green);
+                     palette[i].green = png_ptr->gamma_from_1[w];
+
+                     v = png_ptr->gamma_to_1[palette[i].blue];
+                     png_composite(w, v, png_ptr->trans[i], back_1.blue);
+                     palette[i].blue = png_ptr->gamma_from_1[w];
+                  }
+               }
+               else
+               {
+                  palette[i].red = png_ptr->gamma_table[palette[i].red];
+                  palette[i].green = png_ptr->gamma_table[palette[i].green];
+                  palette[i].blue = png_ptr->gamma_table[palette[i].blue];
+               }
+            }
+         }
+         /* if (png_ptr->background_gamma_type!=PNG_BACKGROUND_GAMMA_UNKNOWN)*/
+         else
+         /* color_type != PNG_COLOR_TYPE_PALETTE */
+         {
+            double m = (double)(((png_uint_32)1 << png_ptr->bit_depth) - 1);
+            double g = 1.0;
+            double gs = 1.0;
+
+            switch (png_ptr->background_gamma_type)
+            {
+               case PNG_BACKGROUND_GAMMA_SCREEN:
+                  g = (png_ptr->screen_gamma);
+                  gs = 1.0;
+                  break;
+               case PNG_BACKGROUND_GAMMA_FILE:
+                  g = 1.0 / (png_ptr->gamma);
+                  gs = 1.0 / (png_ptr->gamma * png_ptr->screen_gamma);
+                  break;
+               case PNG_BACKGROUND_GAMMA_UNIQUE:
+                  g = 1.0 / (png_ptr->background_gamma);
+                  gs = 1.0 / (png_ptr->background_gamma *
+                     png_ptr->screen_gamma);
+                  break;
+            }
+
+            if (color_type & PNG_COLOR_MASK_COLOR)
+            {
+               /* RGB or RGBA */
+               png_ptr->background_1.red = (png_uint_16)(pow(
+                  (double)png_ptr->background.red / m, g) * m + .5);
+               png_ptr->background_1.green = (png_uint_16)(pow(
+                  (double)png_ptr->background.green / m, g) * m + .5);
+               png_ptr->background_1.blue = (png_uint_16)(pow(
+                  (double)png_ptr->background.blue / m, g) * m + .5);
+               png_ptr->background.red = (png_uint_16)(pow(
+                  (double)png_ptr->background.red / m, gs) * m + .5);
+               png_ptr->background.green = (png_uint_16)(pow(
+                  (double)png_ptr->background.green / m, gs) * m + .5);
+               png_ptr->background.blue = (png_uint_16)(pow(
+                  (double)png_ptr->background.blue / m, gs) * m + .5);
+            }
+            else
+            {
+               /* GRAY or GRAY ALPHA */
+               png_ptr->background_1.gray = (png_uint_16)(pow(
+                  (double)png_ptr->background.gray / m, g) * m + .5);
+               png_ptr->background.gray = (png_uint_16)(pow(
+                  (double)png_ptr->background.gray / m, gs) * m + .5);
+            }
+         }
+      }
+      else
+      /* transformation does not include PNG_BACKGROUND */
+#endif /* PNG_READ_BACKGROUND_SUPPORTED */
+      if (color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         png_colorp palette = png_ptr->palette;
+         int num_palette = png_ptr->num_palette;
+         int i;
+
+         for (i = 0; i < num_palette; i++)
+         {
+            palette[i].red = png_ptr->gamma_table[palette[i].red];
+            palette[i].green = png_ptr->gamma_table[palette[i].green];
+            palette[i].blue = png_ptr->gamma_table[palette[i].blue];
+         }
+      }
+   }
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   else
+#endif
+#endif /* PNG_READ_GAMMA_SUPPORTED && PNG_FLOATING_POINT_SUPPORTED */
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   /* No GAMMA transformation */
+   if ((png_ptr->transformations & PNG_BACKGROUND) &&
+       (color_type == PNG_COLOR_TYPE_PALETTE))
+   {
+      int i;
+      int istop = (int)png_ptr->num_trans;
+      png_color back;
+      png_colorp palette = png_ptr->palette;
+
+      back.red   = (png_byte)png_ptr->background.red;
+      back.green = (png_byte)png_ptr->background.green;
+      back.blue  = (png_byte)png_ptr->background.blue;
+
+      for (i = 0; i < istop; i++)
+      {
+         if (png_ptr->trans[i] == 0)
+         {
+            palette[i] = back;
+         }
+         else if (png_ptr->trans[i] != 0xff)
+         {
+            /* The png_composite() macro is defined in png.h */
+            png_composite(palette[i].red, palette[i].red,
+               png_ptr->trans[i], back.red);
+            png_composite(palette[i].green, palette[i].green,
+               png_ptr->trans[i], back.green);
+            png_composite(palette[i].blue, palette[i].blue,
+               png_ptr->trans[i], back.blue);
+         }
+      }
+   }
+#endif /* PNG_READ_BACKGROUND_SUPPORTED */
+
+#if defined(PNG_READ_SHIFT_SUPPORTED)
+   if ((png_ptr->transformations & PNG_SHIFT) &&
+      (color_type == PNG_COLOR_TYPE_PALETTE))
+   {
+      png_uint_16 i;
+      png_uint_16 istop = png_ptr->num_palette;
+      int sr = 8 - png_ptr->sig_bit.red;
+      int sg = 8 - png_ptr->sig_bit.green;
+      int sb = 8 - png_ptr->sig_bit.blue;
+
+      if (sr < 0 || sr > 8)
+         sr = 0;
+      if (sg < 0 || sg > 8)
+         sg = 0;
+      if (sb < 0 || sb > 8)
+         sb = 0;
+      for (i = 0; i < istop; i++)
+      {
+         png_ptr->palette[i].red >>= sr;
+         png_ptr->palette[i].green >>= sg;
+         png_ptr->palette[i].blue >>= sb;
+      }
+   }
+#endif  /* PNG_READ_SHIFT_SUPPORTED */
+ }
+#if !defined(PNG_READ_GAMMA_SUPPORTED) && !defined(PNG_READ_SHIFT_SUPPORTED) \
+ && !defined(PNG_READ_BACKGROUND_SUPPORTED)
+   if(png_ptr)
+      return;
+#endif
+}
+
+/* Modify the info structure to reflect the transformations.  The
+ * info should be updated so a PNG file could be written with it,
+ * assuming the transformations result in valid PNG data.
+ */
+void /* PRIVATE */
+png_read_transform_info(png_structp png_ptr, png_infop info_ptr)
+{
+   png_debug(1, "in png_read_transform_info\n");
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+   if (png_ptr->transformations & PNG_EXPAND)
+   {
+      if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         if (png_ptr->num_trans)
+            info_ptr->color_type = PNG_COLOR_TYPE_RGB_ALPHA;
+         else
+            info_ptr->color_type = PNG_COLOR_TYPE_RGB;
+         info_ptr->bit_depth = 8;
+         info_ptr->num_trans = 0;
+      }
+      else
+      {
+         if (png_ptr->num_trans)
+            info_ptr->color_type |= PNG_COLOR_MASK_ALPHA;
+         if (info_ptr->bit_depth < 8)
+            info_ptr->bit_depth = 8;
+         info_ptr->num_trans = 0;
+      }
+   }
+#endif
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   if (png_ptr->transformations & PNG_BACKGROUND)
+   {
+      info_ptr->color_type &= ~PNG_COLOR_MASK_ALPHA;
+      info_ptr->num_trans = 0;
+      info_ptr->background = png_ptr->background;
+   }
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   if (png_ptr->transformations & PNG_GAMMA)
+   {
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+      info_ptr->gamma = png_ptr->gamma;
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+      info_ptr->int_gamma = png_ptr->int_gamma;
+#endif
+   }
+#endif
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+   if ((png_ptr->transformations & PNG_16_TO_8) && (info_ptr->bit_depth == 16))
+      info_ptr->bit_depth = 8;
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+   if (png_ptr->transformations & PNG_DITHER)
+   {
+      if (((info_ptr->color_type == PNG_COLOR_TYPE_RGB) ||
+         (info_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA)) &&
+         png_ptr->palette_lookup && info_ptr->bit_depth == 8)
+      {
+         info_ptr->color_type = PNG_COLOR_TYPE_PALETTE;
+      }
+   }
+#endif
+
+#if defined(PNG_READ_PACK_SUPPORTED)
+   if ((png_ptr->transformations & PNG_PACK) && (info_ptr->bit_depth < 8))
+      info_ptr->bit_depth = 8;
+#endif
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+   if (png_ptr->transformations & PNG_GRAY_TO_RGB)
+      info_ptr->color_type |= PNG_COLOR_MASK_COLOR;
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+   if (png_ptr->transformations & PNG_RGB_TO_GRAY)
+      info_ptr->color_type &= ~PNG_COLOR_MASK_COLOR;
+#endif
+
+   if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      info_ptr->channels = 1;
+   else if (info_ptr->color_type & PNG_COLOR_MASK_COLOR)
+      info_ptr->channels = 3;
+   else
+      info_ptr->channels = 1;
+
+#if defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+   if (png_ptr->transformations & PNG_STRIP_ALPHA)
+      info_ptr->color_type &= ~PNG_COLOR_MASK_ALPHA;
+#endif
+
+   if (info_ptr->color_type & PNG_COLOR_MASK_ALPHA)
+      info_ptr->channels++;
+
+#if defined(PNG_READ_FILLER_SUPPORTED)
+   /* STRIP_ALPHA and FILLER allowed:  MASK_ALPHA bit stripped above */
+   if ((png_ptr->transformations & PNG_FILLER) &&
+       ((info_ptr->color_type == PNG_COLOR_TYPE_RGB) ||
+       (info_ptr->color_type == PNG_COLOR_TYPE_GRAY)))
+   {
+      info_ptr->channels++;
+#if 0 /* if adding a true alpha channel not just filler */
+      info_ptr->color_type |= PNG_COLOR_MASK_ALPHA;
+#endif
+   }
+#endif
+
+#if defined(PNG_USER_TRANSFORM_PTR_SUPPORTED) && \
+defined(PNG_READ_USER_TRANSFORM_SUPPORTED)
+   if(png_ptr->transformations & PNG_USER_TRANSFORM)
+     {
+       if(info_ptr->bit_depth < png_ptr->user_transform_depth)
+         info_ptr->bit_depth = png_ptr->user_transform_depth;
+       if(info_ptr->channels < png_ptr->user_transform_channels)
+         info_ptr->channels = png_ptr->user_transform_channels;
+     }
+#endif
+
+   info_ptr->pixel_depth = (png_byte)(info_ptr->channels *
+      info_ptr->bit_depth);
+   info_ptr->rowbytes = ((info_ptr->width * info_ptr->pixel_depth + 7) >> 3);
+
+#if !defined(PNG_READ_EXPAND_SUPPORTED)
+   if(png_ptr)
+      return;
+#endif
+}
+
+/* Transform the row.  The order of transformations is significant,
+ * and is very touchy.  If you add a transformation, take care to
+ * decide how it fits in with the other transformations here.
+ */
+void /* PRIVATE */
+png_do_read_transformations(png_structp png_ptr)
+{
+   png_debug(1, "in png_do_read_transformations\n");
+#if !defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (png_ptr->row_buf == NULL)
+   {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+      char msg[50];
+
+      sprintf(msg, "NULL row buffer for row %ld, pass %d", png_ptr->row_number,
+         png_ptr->pass);
+      png_error(png_ptr, msg);
+#else
+      png_error(png_ptr, "NULL row buffer");
+#endif
+   }
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+   if (png_ptr->transformations & PNG_EXPAND)
+   {
+      if (png_ptr->row_info.color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         png_do_expand_palette(&(png_ptr->row_info), png_ptr->row_buf + 1,
+            png_ptr->palette, png_ptr->trans, png_ptr->num_trans);
+      }
+      else
+      {
+         if (png_ptr->num_trans)
+            png_do_expand(&(png_ptr->row_info), png_ptr->row_buf + 1,
+               &(png_ptr->trans_values));
+         else
+            png_do_expand(&(png_ptr->row_info), png_ptr->row_buf + 1,
+               NULL);
+      }
+   }
+#endif
+
+#if defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+   if (png_ptr->transformations & PNG_STRIP_ALPHA)
+      png_do_strip_filler(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         PNG_FLAG_FILLER_AFTER);
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+   if (png_ptr->transformations & PNG_RGB_TO_GRAY)
+   {
+      int rgb_error =
+         png_do_rgb_to_gray(png_ptr, &(png_ptr->row_info), png_ptr->row_buf + 1);
+      if(rgb_error)
+      {
+         png_ptr->rgb_to_gray_status=1;
+         if(png_ptr->transformations == PNG_RGB_TO_GRAY_WARN)
+            png_warning(png_ptr, "png_do_rgb_to_gray found nongray pixel");
+         if(png_ptr->transformations == PNG_RGB_TO_GRAY_ERR)
+            png_error(png_ptr, "png_do_rgb_to_gray found nongray pixel");
+      }
+   }
+#endif
+
+/*
+From Andreas Dilger e-mail to png-implement, 26 March 1998:
+
+  In most cases, the "simple transparency" should be done prior to doing
+  gray-to-RGB, or you will have to test 3x as many bytes to check if a
+  pixel is transparent.  You would also need to make sure that the
+  transparency information is upgraded to RGB.
+
+  To summarize, the current flow is:
+  - Gray + simple transparency -> compare 1 or 2 gray bytes and composite
+                                  with background "in place" if transparent,
+                                  convert to RGB if necessary
+  - Gray + alpha -> composite with gray background and remove alpha bytes,
+                                  convert to RGB if necessary
+
+  To support RGB backgrounds for gray images we need:
+  - Gray + simple transparency -> convert to RGB + simple transparency, compare
+                                  3 or 6 bytes and composite with background
+                                  "in place" if transparent (3x compare/pixel
+                                  compared to doing composite with gray bkgrnd)
+  - Gray + alpha -> convert to RGB + alpha, composite with background and
+                                  remove alpha bytes (3x float operations/pixel
+                                  compared with composite on gray background)
+
+  Greg's change will do this.  The reason it wasn't done before is for
+  performance, as this increases the per-pixel operations.  If we would check
+  in advance if the background was gray or RGB, and position the gray-to-RGB
+  transform appropriately, then it would save a lot of work/time.
+ */
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+   /* if gray -> RGB, do so now only if background is non-gray; else do later
+    * for performance reasons */
+   if ((png_ptr->transformations & PNG_GRAY_TO_RGB) &&
+       !(png_ptr->mode & PNG_BACKGROUND_IS_GRAY))
+      png_do_gray_to_rgb(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   if ((png_ptr->transformations & PNG_BACKGROUND) &&
+      ((png_ptr->num_trans != 0 ) ||
+      (png_ptr->color_type & PNG_COLOR_MASK_ALPHA)))
+      png_do_background(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         &(png_ptr->trans_values), &(png_ptr->background)
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+         , &(png_ptr->background_1),
+         png_ptr->gamma_table, png_ptr->gamma_from_1,
+         png_ptr->gamma_to_1, png_ptr->gamma_16_table,
+         png_ptr->gamma_16_from_1, png_ptr->gamma_16_to_1,
+         png_ptr->gamma_shift
+#endif
+);
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   if ((png_ptr->transformations & PNG_GAMMA) &&
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+      !((png_ptr->transformations & PNG_BACKGROUND) &&
+      ((png_ptr->num_trans != 0) ||
+      (png_ptr->color_type & PNG_COLOR_MASK_ALPHA))) &&
+#endif
+      (png_ptr->color_type != PNG_COLOR_TYPE_PALETTE))
+      png_do_gamma(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         png_ptr->gamma_table, png_ptr->gamma_16_table,
+         png_ptr->gamma_shift);
+#endif
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+   if (png_ptr->transformations & PNG_16_TO_8)
+      png_do_chop(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+   if (png_ptr->transformations & PNG_DITHER)
+   {
+      png_do_dither((png_row_infop)&(png_ptr->row_info), png_ptr->row_buf + 1,
+         png_ptr->palette_lookup, png_ptr->dither_index);
+      if(png_ptr->row_info.rowbytes == (png_uint_32)0)
+         png_error(png_ptr, "png_do_dither returned rowbytes=0");
+   }
+#endif
+
+#if defined(PNG_READ_INVERT_SUPPORTED)
+   if (png_ptr->transformations & PNG_INVERT_MONO)
+      png_do_invert(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED)
+   if (png_ptr->transformations & PNG_SHIFT)
+      png_do_unshift(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         &(png_ptr->shift));
+#endif
+
+#if defined(PNG_READ_PACK_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACK)
+      png_do_unpack(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_BGR_SUPPORTED)
+   if (png_ptr->transformations & PNG_BGR)
+      png_do_bgr(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACKSWAP)
+      png_do_packswap(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+   /* if gray -> RGB, do so now only if we did not do so above */
+   if ((png_ptr->transformations & PNG_GRAY_TO_RGB) &&
+       (png_ptr->mode & PNG_BACKGROUND_IS_GRAY))
+      png_do_gray_to_rgb(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_FILLER_SUPPORTED)
+   if (png_ptr->transformations & PNG_FILLER)
+      png_do_read_filler(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         (png_uint_32)png_ptr->filler, png_ptr->flags);
+#endif
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED)
+   if (png_ptr->transformations & PNG_INVERT_ALPHA)
+      png_do_read_invert_alpha(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED)
+   if (png_ptr->transformations & PNG_SWAP_ALPHA)
+      png_do_read_swap_alpha(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_SWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_SWAP_BYTES)
+      png_do_swap(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED)
+   if (png_ptr->transformations & PNG_USER_TRANSFORM)
+    {
+      if(png_ptr->read_user_transform_fn != NULL)
+        (*(png_ptr->read_user_transform_fn)) /* user read transform function */
+          (png_ptr,                    /* png_ptr */
+           &(png_ptr->row_info),       /* row_info:     */
+             /*  png_uint_32 width;          width of row */
+             /*  png_uint_32 rowbytes;       number of bytes in row */
+             /*  png_byte color_type;        color type of pixels */
+             /*  png_byte bit_depth;         bit depth of samples */
+             /*  png_byte channels;          number of channels (1-4) */
+             /*  png_byte pixel_depth;       bits per pixel (depth*channels) */
+           png_ptr->row_buf + 1);      /* start of pixel data for row */
+#if defined(PNG_USER_TRANSFORM_PTR_SUPPORTED)
+      if(png_ptr->user_transform_depth)
+         png_ptr->row_info.bit_depth = png_ptr->user_transform_depth;
+      if(png_ptr->user_transform_channels)
+         png_ptr->row_info.channels = png_ptr->user_transform_channels;
+#endif
+      png_ptr->row_info.pixel_depth = (png_byte)(png_ptr->row_info.bit_depth *
+         png_ptr->row_info.channels);
+      png_ptr->row_info.rowbytes = (png_ptr->row_info.width *
+         png_ptr->row_info.pixel_depth+7)>>3;
+   }
+#endif
+
+}
+
+#if defined(PNG_READ_PACK_SUPPORTED)
+/* Unpack pixels of 1, 2, or 4 bits per pixel into 1 byte per pixel,
+ * without changing the actual values.  Thus, if you had a row with
+ * a bit depth of 1, you would end up with bytes that only contained
+ * the numbers 0 or 1.  If you would rather they contain 0 and 255, use
+ * png_do_shift() after this.
+ */
+void /* PRIVATE */
+png_do_unpack(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_unpack\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL && row_info->bit_depth < 8)
+#else
+   if (row_info->bit_depth < 8)
+#endif
+   {
+      png_uint_32 i;
+      png_uint_32 row_width=row_info->width;
+
+      switch (row_info->bit_depth)
+      {
+         case 1:
+         {
+            png_bytep sp = row + (png_size_t)((row_width - 1) >> 3);
+            png_bytep dp = row + (png_size_t)row_width - 1;
+            png_uint_32 shift = 7 - (int)((row_width + 7) & 0x07);
+            for (i = 0; i < row_width; i++)
+            {
+               *dp = (png_byte)((*sp >> shift) & 0x01);
+               if (shift == 7)
+               {
+                  shift = 0;
+                  sp--;
+               }
+               else
+                  shift++;
+
+               dp--;
+            }
+            break;
+         }
+         case 2:
+         {
+
+            png_bytep sp = row + (png_size_t)((row_width - 1) >> 2);
+            png_bytep dp = row + (png_size_t)row_width - 1;
+            png_uint_32 shift = (int)((3 - ((row_width + 3) & 0x03)) << 1);
+            for (i = 0; i < row_width; i++)
+            {
+               *dp = (png_byte)((*sp >> shift) & 0x03);
+               if (shift == 6)
+               {
+                  shift = 0;
+                  sp--;
+               }
+               else
+                  shift += 2;
+
+               dp--;
+            }
+            break;
+         }
+         case 4:
+         {
+            png_bytep sp = row + (png_size_t)((row_width - 1) >> 1);
+            png_bytep dp = row + (png_size_t)row_width - 1;
+            png_uint_32 shift = (int)((1 - ((row_width + 1) & 0x01)) << 2);
+            for (i = 0; i < row_width; i++)
+            {
+               *dp = (png_byte)((*sp >> shift) & 0x0f);
+               if (shift == 4)
+               {
+                  shift = 0;
+                  sp--;
+               }
+               else
+                  shift = 4;
+
+               dp--;
+            }
+            break;
+         }
+      }
+      row_info->bit_depth = 8;
+      row_info->pixel_depth = (png_byte)(8 * row_info->channels);
+      row_info->rowbytes = row_width * row_info->channels;
+   }
+}
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED)
+/* Reverse the effects of png_do_shift.  This routine merely shifts the
+ * pixels back to their significant bits values.  Thus, if you have
+ * a row of bit depth 8, but only 5 are significant, this will shift
+ * the values back to 0 through 31.
+ */
+void /* PRIVATE */
+png_do_unshift(png_row_infop row_info, png_bytep row, png_color_8p sig_bits)
+{
+   png_debug(1, "in png_do_unshift\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL && sig_bits != NULL &&
+#endif
+       row_info->color_type != PNG_COLOR_TYPE_PALETTE)
+   {
+      int shift[4];
+      int channels = 0;
+      int c;
+      png_uint_16 value = 0;
+      png_uint_32 row_width = row_info->width;
+
+      if (row_info->color_type & PNG_COLOR_MASK_COLOR)
+      {
+         shift[channels++] = row_info->bit_depth - sig_bits->red;
+         shift[channels++] = row_info->bit_depth - sig_bits->green;
+         shift[channels++] = row_info->bit_depth - sig_bits->blue;
+      }
+      else
+      {
+         shift[channels++] = row_info->bit_depth - sig_bits->gray;
+      }
+      if (row_info->color_type & PNG_COLOR_MASK_ALPHA)
+      {
+         shift[channels++] = row_info->bit_depth - sig_bits->alpha;
+      }
+
+      for (c = 0; c < channels; c++)
+      {
+         if (shift[c] <= 0)
+            shift[c] = 0;
+         else
+            value = 1;
+      }
+
+      if (!value)
+         return;
+
+      switch (row_info->bit_depth)
+      {
+         case 2:
+         {
+            png_bytep bp;
+            png_uint_32 i;
+            png_uint_32 istop = row_info->rowbytes;
+
+            for (bp = row, i = 0; i < istop; i++)
+            {
+               *bp >>= 1;
+               *bp++ &= 0x55;
+            }
+            break;
+         }
+         case 4:
+         {
+            png_bytep bp = row;
+            png_uint_32 i;
+            png_uint_32 istop = row_info->rowbytes;
+            png_byte mask = (png_byte)((((int)0xf0 >> shift[0]) & (int)0xf0) |
+               (png_byte)((int)0xf >> shift[0]));
+
+            for (i = 0; i < istop; i++)
+            {
+               *bp >>= shift[0];
+               *bp++ &= mask;
+            }
+            break;
+         }
+         case 8:
+         {
+            png_bytep bp = row;
+            png_uint_32 i;
+            png_uint_32 istop = row_width * channels;
+
+            for (i = 0; i < istop; i++)
+            {
+               *bp++ >>= shift[i%channels];
+            }
+            break;
+         }
+         case 16:
+         {
+            png_bytep bp = row;
+            png_uint_32 i;
+            png_uint_32 istop = channels * row_width;
+
+            for (i = 0; i < istop; i++)
+            {
+               value = (png_uint_16)((*bp << 8) + *(bp + 1));
+               value >>= shift[i%channels];
+               *bp++ = (png_byte)(value >> 8);
+               *bp++ = (png_byte)(value & 0xff);
+            }
+            break;
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_16_TO_8_SUPPORTED)
+/* chop rows of bit depth 16 down to 8 */
+void /* PRIVATE */
+png_do_chop(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_chop\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL && row_info->bit_depth == 16)
+#else
+   if (row_info->bit_depth == 16)
+#endif
+   {
+      png_bytep sp = row;
+      png_bytep dp = row;
+      png_uint_32 i;
+      png_uint_32 istop = row_info->width * row_info->channels;
+
+      for (i = 0; i<istop; i++, sp += 2, dp++)
+      {
+#if defined(PNG_READ_16_TO_8_ACCURATE_SCALE_SUPPORTED)
+      /* This does a more accurate scaling of the 16-bit color
+       * value, rather than a simple low-byte truncation.
+       *
+       * What the ideal calculation should be:
+       *   *dp = (((((png_uint_32)(*sp) << 8) |
+       *          (png_uint_32)(*(sp + 1))) * 255 + 127) / (png_uint_32)65535L;
+       *
+       * GRR: no, I think this is what it really should be:
+       *   *dp = (((((png_uint_32)(*sp) << 8) |
+       *           (png_uint_32)(*(sp + 1))) + 128L) / (png_uint_32)257L;
+       *
+       * GRR: here's the exact calculation with shifts:
+       *   temp = (((png_uint_32)(*sp) << 8) | (png_uint_32)(*(sp + 1))) + 128L;
+       *   *dp = (temp - (temp >> 8)) >> 8;
+       *
+       * Approximate calculation with shift/add instead of multiply/divide:
+       *   *dp = ((((png_uint_32)(*sp) << 8) |
+       *          (png_uint_32)((int)(*(sp + 1)) - *sp)) + 128) >> 8;
+       *
+       * What we actually do to avoid extra shifting and conversion:
+       */
+
+         *dp = *sp + ((((int)(*(sp + 1)) - *sp) > 128) ? 1 : 0);
+#else
+       /* Simply discard the low order byte */
+         *dp = *sp;
+#endif
+      }
+      row_info->bit_depth = 8;
+      row_info->pixel_depth = (png_byte)(8 * row_info->channels);
+      row_info->rowbytes = row_info->width * row_info->channels;
+   }
+}
+#endif
+
+#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED)
+void /* PRIVATE */
+png_do_read_swap_alpha(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_read_swap_alpha\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+      png_uint_32 row_width = row_info->width;
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+      {
+         /* This converts from RGBA to ARGB */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_byte save;
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               save = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = save;
+            }
+         }
+         /* This converts from RRGGBBAA to AARRGGBB */
+         else
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_byte save[2];
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               save[0] = *(--sp);
+               save[1] = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = save[0];
+               *(--dp) = save[1];
+            }
+         }
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      {
+         /* This converts from GA to AG */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_byte save;
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               save = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = save;
+            }
+         }
+         /* This converts from GGAA to AAGG */
+         else
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_byte save[2];
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               save[0] = *(--sp);
+               save[1] = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = save[0];
+               *(--dp) = save[1];
+            }
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED)
+void /* PRIVATE */
+png_do_read_invert_alpha(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_read_invert_alpha\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+      png_uint_32 row_width = row_info->width;
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+      {
+         /* This inverts the alpha channel in RGBA */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = (png_byte)(255 - *(--sp));
+
+/*             This does nothing:
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               We can replace it with:
+*/
+               sp-=3;
+               dp=sp;
+            }
+         }
+         /* This inverts the alpha channel in RRGGBBAA */
+         else
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = (png_byte)(255 - *(--sp));
+               *(--dp) = (png_byte)(255 - *(--sp));
+
+/*             This does nothing:
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               We can replace it with:
+*/
+               sp-=6;
+               dp=sp;
+            }
+         }
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      {
+         /* This inverts the alpha channel in GA */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = (png_byte)(255 - *(--sp));
+               *(--dp) = *(--sp);
+            }
+         }
+         /* This inverts the alpha channel in GGAA */
+         else
+         {
+            png_bytep sp  = row + row_info->rowbytes;
+            png_bytep dp = sp;
+            png_uint_32 i;
+
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = (png_byte)(255 - *(--sp));
+               *(--dp) = (png_byte)(255 - *(--sp));
+/*
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+*/
+               sp-=2;
+               dp=sp;
+            }
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_FILLER_SUPPORTED)
+/* Add filler channel if we have RGB color */
+void /* PRIVATE */
+png_do_read_filler(png_row_infop row_info, png_bytep row,
+   png_uint_32 filler, png_uint_32 flags)
+{
+   png_uint_32 i;
+   png_uint_32 row_width = row_info->width;
+
+   png_byte hi_filler = (png_byte)((filler>>8) & 0xff);
+   png_byte lo_filler = (png_byte)(filler & 0xff);
+
+   png_debug(1, "in png_do_read_filler\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL  && row_info != NULL &&
+#endif
+       row_info->color_type == PNG_COLOR_TYPE_GRAY)
+   {
+      if(row_info->bit_depth == 8)
+      {
+         /* This changes the data from G to GX */
+         if (flags & PNG_FLAG_FILLER_AFTER)
+         {
+            png_bytep sp = row + (png_size_t)row_width;
+            png_bytep dp =  sp + (png_size_t)row_width;
+            for (i = 1; i < row_width; i++)
+            {
+               *(--dp) = lo_filler;
+               *(--dp) = *(--sp);
+            }
+            *(--dp) = lo_filler;
+            row_info->channels = 2;
+            row_info->pixel_depth = 16;
+            row_info->rowbytes = row_width * 2;
+         }
+      /* This changes the data from G to XG */
+         else
+         {
+            png_bytep sp = row + (png_size_t)row_width;
+            png_bytep dp = sp  + (png_size_t)row_width;
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = *(--sp);
+               *(--dp) = lo_filler;
+            }
+            row_info->channels = 2;
+            row_info->pixel_depth = 16;
+            row_info->rowbytes = row_width * 2;
+         }
+      }
+      else if(row_info->bit_depth == 16)
+      {
+         /* This changes the data from GG to GGXX */
+         if (flags & PNG_FLAG_FILLER_AFTER)
+         {
+            png_bytep sp = row + (png_size_t)row_width;
+            png_bytep dp = sp  + (png_size_t)row_width;
+            for (i = 1; i < row_width; i++)
+            {
+               *(--dp) = hi_filler;
+               *(--dp) = lo_filler;
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+            }
+            *(--dp) = hi_filler;
+            *(--dp) = lo_filler;
+            row_info->channels = 2;
+            row_info->pixel_depth = 32;
+            row_info->rowbytes = row_width * 4;
+         }
+         /* This changes the data from GG to XXGG */
+         else
+         {
+            png_bytep sp = row + (png_size_t)row_width;
+            png_bytep dp = sp  + (png_size_t)row_width;
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = hi_filler;
+               *(--dp) = lo_filler;
+            }
+            row_info->channels = 2;
+            row_info->pixel_depth = 32;
+            row_info->rowbytes = row_width * 4;
+         }
+      }
+   } /* COLOR_TYPE == GRAY */
+   else if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+   {
+      if(row_info->bit_depth == 8)
+      {
+         /* This changes the data from RGB to RGBX */
+         if (flags & PNG_FLAG_FILLER_AFTER)
+         {
+            png_bytep sp = row + (png_size_t)row_width * 3;
+            png_bytep dp = sp  + (png_size_t)row_width;
+            for (i = 1; i < row_width; i++)
+            {
+               *(--dp) = lo_filler;
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+            }
+            *(--dp) = lo_filler;
+            row_info->channels = 4;
+            row_info->pixel_depth = 32;
+            row_info->rowbytes = row_width * 4;
+         }
+      /* This changes the data from RGB to XRGB */
+         else
+         {
+            png_bytep sp = row + (png_size_t)row_width * 3;
+            png_bytep dp = sp + (png_size_t)row_width;
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = lo_filler;
+            }
+            row_info->channels = 4;
+            row_info->pixel_depth = 32;
+            row_info->rowbytes = row_width * 4;
+         }
+      }
+      else if(row_info->bit_depth == 16)
+      {
+         /* This changes the data from RRGGBB to RRGGBBXX */
+         if (flags & PNG_FLAG_FILLER_AFTER)
+         {
+            png_bytep sp = row + (png_size_t)row_width * 6;
+            png_bytep dp = sp  + (png_size_t)row_width * 2;
+            for (i = 1; i < row_width; i++)
+            {
+               *(--dp) = hi_filler;
+               *(--dp) = lo_filler;
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+            }
+            *(--dp) = hi_filler;
+            *(--dp) = lo_filler;
+            row_info->channels = 4;
+            row_info->pixel_depth = 64;
+            row_info->rowbytes = row_width * 8;
+         }
+         /* This changes the data from RRGGBB to XXRRGGBB */
+         else
+         {
+            png_bytep sp = row + (png_size_t)row_width * 6;
+            png_bytep dp = sp  + (png_size_t)row_width * 2;
+            for (i = 0; i < row_width; i++)
+            {
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = *(--sp);
+               *(--dp) = hi_filler;
+               *(--dp) = lo_filler;
+            }
+            row_info->channels = 4;
+            row_info->pixel_depth = 64;
+            row_info->rowbytes = row_width * 8;
+         }
+      }
+   } /* COLOR_TYPE == RGB */
+}
+#endif
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+/* expand grayscale files to RGB, with or without alpha */
+void /* PRIVATE */
+png_do_gray_to_rgb(png_row_infop row_info, png_bytep row)
+{
+   png_uint_32 i;
+   png_uint_32 row_width = row_info->width;
+
+   png_debug(1, "in png_do_gray_to_rgb\n");
+   if (row_info->bit_depth >= 8 &&
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+      !(row_info->color_type & PNG_COLOR_MASK_COLOR))
+   {
+      if (row_info->color_type == PNG_COLOR_TYPE_GRAY)
+      {
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp = row + (png_size_t)row_width - 1;
+            png_bytep dp = sp  + (png_size_t)row_width * 2;
+            for (i = 0; i < row_width; i++)
+            {
+               *(dp--) = *sp;
+               *(dp--) = *sp;
+               *(dp--) = *(sp--);
+            }
+         }
+         else
+         {
+            png_bytep sp = row + (png_size_t)row_width * 2 - 1;
+            png_bytep dp = sp  + (png_size_t)row_width * 4;
+            for (i = 0; i < row_width; i++)
+            {
+               *(dp--) = *sp;
+               *(dp--) = *(sp - 1);
+               *(dp--) = *sp;
+               *(dp--) = *(sp - 1);
+               *(dp--) = *(sp--);
+               *(dp--) = *(sp--);
+            }
+         }
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      {
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp = row + (png_size_t)row_width * 2 - 1;
+            png_bytep dp = sp  + (png_size_t)row_width * 2;
+            for (i = 0; i < row_width; i++)
+            {
+               *(dp--) = *(sp--);
+               *(dp--) = *sp;
+               *(dp--) = *sp;
+               *(dp--) = *(sp--);
+            }
+         }
+         else
+         {
+            png_bytep sp = row + (png_size_t)row_width * 4 - 1;
+            png_bytep dp = sp  + (png_size_t)row_width * 4;
+            for (i = 0; i < row_width; i++)
+            {
+               *(dp--) = *(sp--);
+               *(dp--) = *(sp--);
+               *(dp--) = *sp;
+               *(dp--) = *(sp - 1);
+               *(dp--) = *sp;
+               *(dp--) = *(sp - 1);
+               *(dp--) = *(sp--);
+               *(dp--) = *(sp--);
+            }
+         }
+      }
+      row_info->channels += (png_byte)2;
+      row_info->color_type |= PNG_COLOR_MASK_COLOR;
+      row_info->pixel_depth = (png_byte)(row_info->channels *
+         row_info->bit_depth);
+      row_info->rowbytes = ((row_width *
+         row_info->pixel_depth + 7) >> 3);
+   }
+}
+#endif
+
+#if defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+/* reduce RGB files to grayscale, with or without alpha
+ * using the equation given in Poynton's ColorFAQ at
+ * <http://www.inforamp.net/~poynton/>
+ * Copyright (c) 1998-01-04 Charles Poynton poynton@inforamp.net
+ *
+ *     Y = 0.212671 * R + 0.715160 * G + 0.072169 * B
+ *
+ *  We approximate this with
+ *
+ *     Y = 0.21268 * R    + 0.7151 * G    + 0.07217 * B
+ *
+ *  which can be expressed with integers as
+ *
+ *     Y = (6969 * R + 23434 * G + 2365 * B)/32768
+ *
+ *  The calculation is to be done in a linear colorspace.
+ *
+ *  Other integer coefficents can be used via png_set_rgb_to_gray().
+ */
+int /* PRIVATE */
+png_do_rgb_to_gray(png_structp png_ptr, png_row_infop row_info, png_bytep row)
+
+{
+   png_uint_32 i;
+
+   png_uint_32 row_width = row_info->width;
+   int rgb_error = 0;
+
+   png_debug(1, "in png_do_rgb_to_gray\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+      (row_info->color_type & PNG_COLOR_MASK_COLOR))
+   {
+      png_uint_32 rc = png_ptr->rgb_to_gray_red_coeff;
+      png_uint_32 gc = png_ptr->rgb_to_gray_green_coeff;
+      png_uint_32 bc = png_ptr->rgb_to_gray_blue_coeff;
+
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+      {
+         if (row_info->bit_depth == 8)
+         {
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+            if (png_ptr->gamma_from_1 != NULL && png_ptr->gamma_to_1 != NULL)
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+
+               for (i = 0; i < row_width; i++)
+               {
+                  png_byte red   = png_ptr->gamma_to_1[*(sp++)];
+                  png_byte green = png_ptr->gamma_to_1[*(sp++)];
+                  png_byte blue  = png_ptr->gamma_to_1[*(sp++)];
+                  if(red != green || red != blue)
+                  {
+                     rgb_error |= 1;
+                     *(dp++) = png_ptr->gamma_from_1[
+                       (rc*red+gc*green+bc*blue)>>15];
+                  }
+                  else
+                     *(dp++) = *(sp-1);
+               }
+            }
+            else
+#endif
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_byte red   = *(sp++);
+                  png_byte green = *(sp++);
+                  png_byte blue  = *(sp++);
+                  if(red != green || red != blue)
+                  {
+                     rgb_error |= 1;
+                     *(dp++) = (png_byte)((rc*red+gc*green+bc*blue)>>15);
+                  }
+                  else
+                     *(dp++) = *(sp-1);
+               }
+            }
+         }
+
+         else /* RGB bit_depth == 16 */
+         {
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+            if (png_ptr->gamma_16_to_1 != NULL &&
+                png_ptr->gamma_16_from_1 != NULL)
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 red, green, blue, w;
+
+                  red   = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+                  green = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+                  blue  = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+
+                  if(red == green && red == blue)
+                     w = red;
+                  else
+                  {
+                     png_uint_16 red_1   = png_ptr->gamma_16_to_1[(red&0xff) >>
+                                  png_ptr->gamma_shift][red>>8];
+                     png_uint_16 green_1 = png_ptr->gamma_16_to_1[(green&0xff) >>
+                                  png_ptr->gamma_shift][green>>8];
+                     png_uint_16 blue_1  = png_ptr->gamma_16_to_1[(blue&0xff) >>
+                                  png_ptr->gamma_shift][blue>>8];
+                     png_uint_16 gray16  = (png_uint_16)((rc*red_1 + gc*green_1
+                                  + bc*blue_1)>>15);
+                     w = png_ptr->gamma_16_from_1[(gray16&0xff) >>
+                         png_ptr->gamma_shift][gray16 >> 8];
+                     rgb_error |= 1;
+                  }
+
+                  *(dp++) = (png_byte)((w>>8) & 0xff);
+                  *(dp++) = (png_byte)(w & 0xff);
+               }
+            }
+            else
+#endif
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 red, green, blue, gray16;
+
+                  red   = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+                  green = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+                  blue  = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+
+                  if(red != green || red != blue)
+                     rgb_error |= 1;
+                  gray16  = (png_uint_16)((rc*red + gc*green + bc*blue)>>15);
+                  *(dp++) = (png_byte)((gray16>>8) & 0xff);
+                  *(dp++) = (png_byte)(gray16 & 0xff);
+               }
+            }
+         }
+      }
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+      {
+         if (row_info->bit_depth == 8)
+         {
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+            if (png_ptr->gamma_from_1 != NULL && png_ptr->gamma_to_1 != NULL)
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_byte red   = png_ptr->gamma_to_1[*(sp++)];
+                  png_byte green = png_ptr->gamma_to_1[*(sp++)];
+                  png_byte blue  = png_ptr->gamma_to_1[*(sp++)];
+                  if(red != green || red != blue)
+                     rgb_error |= 1;
+                  *(dp++) =  png_ptr->gamma_from_1
+                             [(rc*red + gc*green + bc*blue)>>15];
+                  *(dp++) = *(sp++);  /* alpha */
+               }
+            }
+            else
+#endif
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_byte red   = *(sp++);
+                  png_byte green = *(sp++);
+                  png_byte blue  = *(sp++);
+                  if(red != green || red != blue)
+                     rgb_error |= 1;
+                  *(dp++) =  (png_byte)((gc*red + gc*green + bc*blue)>>8);
+                  *(dp++) = *(sp++);  /* alpha */
+               }
+            }
+         }
+         else /* RGBA bit_depth == 16 */
+         {
+#if defined(PNG_READ_GAMMA_SUPPORTED) || defined(PNG_READ_BACKGROUND_SUPPORTED)
+            if (png_ptr->gamma_16_to_1 != NULL &&
+                png_ptr->gamma_16_from_1 != NULL)
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 red, green, blue, w;
+
+                  red   = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+                  green = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+                  blue  = (png_uint_16)(((*(sp))<<8) | *(sp+1)); sp+=2;
+
+                  if(red == green && red == blue)
+                     w = red;
+                  else
+                  {
+                     png_uint_16 red_1   = png_ptr->gamma_16_to_1[(red&0xff) >>
+                                  png_ptr->gamma_shift][red>>8];
+                     png_uint_16 green_1 = png_ptr->gamma_16_to_1[(green&0xff) >>
+                                  png_ptr->gamma_shift][green>>8];
+                     png_uint_16 blue_1  = png_ptr->gamma_16_to_1[(blue&0xff) >>
+                                  png_ptr->gamma_shift][blue>>8];
+                     png_uint_16 gray16  = (png_uint_16)((rc * red_1
+                                  + gc * green_1 + bc * blue_1)>>15);
+                     w = png_ptr->gamma_16_from_1[(gray16&0xff) >>
+                         png_ptr->gamma_shift][gray16 >> 8];
+                     rgb_error |= 1;
+                  }
+
+                  *(dp++) = (png_byte)((w>>8) & 0xff);
+                  *(dp++) = (png_byte)(w & 0xff);
+                  *(dp++) = *(sp++);  /* alpha */
+                  *(dp++) = *(sp++);
+               }
+            }
+            else
+#endif
+            {
+               png_bytep sp = row;
+               png_bytep dp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 red, green, blue, gray16;
+                  red   = (png_uint_16)((*(sp)<<8) | *(sp+1)); sp+=2;
+                  green = (png_uint_16)((*(sp)<<8) | *(sp+1)); sp+=2;
+                  blue  = (png_uint_16)((*(sp)<<8) | *(sp+1)); sp+=2;
+                  if(red != green || red != blue)
+                     rgb_error |= 1;
+                  gray16  = (png_uint_16)((rc*red + gc*green + bc*blue)>>15);
+                  *(dp++) = (png_byte)((gray16>>8) & 0xff);
+                  *(dp++) = (png_byte)(gray16 & 0xff);
+                  *(dp++) = *(sp++);  /* alpha */
+                  *(dp++) = *(sp++);
+               }
+            }
+         }
+      }
+   row_info->channels -= (png_byte)2;
+      row_info->color_type &= ~PNG_COLOR_MASK_COLOR;
+      row_info->pixel_depth = (png_byte)(row_info->channels *
+         row_info->bit_depth);
+      row_info->rowbytes = ((row_width *
+         row_info->pixel_depth + 7) >> 3);
+   }
+   return rgb_error;
+}
+#endif
+
+/* Build a grayscale palette.  Palette is assumed to be 1 << bit_depth
+ * large of png_color.  This lets grayscale images be treated as
+ * paletted.  Most useful for gamma correction and simplification
+ * of code.
+ */
+void /* PRIVATE */
+png_build_grayscale_palette(int bit_depth, png_colorp palette)
+{
+   int num_palette;
+   int color_inc;
+   int i;
+   int v;
+
+   png_debug(1, "in png_do_build_grayscale_palette\n");
+   if (palette == NULL)
+      return;
+
+   switch (bit_depth)
+   {
+      case 1:
+         num_palette = 2;
+         color_inc = 0xff;
+         break;
+      case 2:
+         num_palette = 4;
+         color_inc = 0x55;
+         break;
+      case 4:
+         num_palette = 16;
+         color_inc = 0x11;
+         break;
+      case 8:
+         num_palette = 256;
+         color_inc = 1;
+         break;
+      default:
+         num_palette = 0;
+         color_inc = 0;
+         break;
+   }
+
+   for (i = 0, v = 0; i < num_palette; i++, v += color_inc)
+   {
+      palette[i].red = (png_byte)v;
+      palette[i].green = (png_byte)v;
+      palette[i].blue = (png_byte)v;
+   }
+}
+
+/* This function is currently unused.  Do we really need it? */
+#if defined(PNG_READ_DITHER_SUPPORTED) && defined(PNG_CORRECT_PALETTE_SUPPORTED)
+void /* PRIVATE */
+png_correct_palette(png_structp png_ptr, png_colorp palette,
+   int num_palette)
+{
+   png_debug(1, "in png_correct_palette\n");
+#if defined(PNG_READ_BACKGROUND_SUPPORTED) && \
+    defined(PNG_READ_GAMMA_SUPPORTED) && defined(PNG_FLOATING_POINT_SUPPORTED)
+   if (png_ptr->transformations & (PNG_GAMMA | PNG_BACKGROUND))
+   {
+      png_color back, back_1;
+
+      if (png_ptr->background_gamma_type == PNG_BACKGROUND_GAMMA_FILE)
+      {
+         back.red = png_ptr->gamma_table[png_ptr->background.red];
+         back.green = png_ptr->gamma_table[png_ptr->background.green];
+         back.blue = png_ptr->gamma_table[png_ptr->background.blue];
+
+         back_1.red = png_ptr->gamma_to_1[png_ptr->background.red];
+         back_1.green = png_ptr->gamma_to_1[png_ptr->background.green];
+         back_1.blue = png_ptr->gamma_to_1[png_ptr->background.blue];
+      }
+      else
+      {
+         double g;
+
+         g = 1.0 / (png_ptr->background_gamma * png_ptr->screen_gamma);
+
+         if (png_ptr->background_gamma_type == PNG_BACKGROUND_GAMMA_SCREEN ||
+             fabs(g - 1.0) < PNG_GAMMA_THRESHOLD)
+         {
+            back.red = png_ptr->background.red;
+            back.green = png_ptr->background.green;
+            back.blue = png_ptr->background.blue;
+         }
+         else
+         {
+            back.red =
+               (png_byte)(pow((double)png_ptr->background.red/255, g) *
+                255.0 + 0.5);
+            back.green =
+               (png_byte)(pow((double)png_ptr->background.green/255, g) *
+                255.0 + 0.5);
+            back.blue =
+               (png_byte)(pow((double)png_ptr->background.blue/255, g) *
+                255.0 + 0.5);
+         }
+
+         g = 1.0 / png_ptr->background_gamma;
+
+         back_1.red =
+            (png_byte)(pow((double)png_ptr->background.red/255, g) *
+             255.0 + 0.5);
+         back_1.green =
+            (png_byte)(pow((double)png_ptr->background.green/255, g) *
+             255.0 + 0.5);
+         back_1.blue =
+            (png_byte)(pow((double)png_ptr->background.blue/255, g) *
+             255.0 + 0.5);
+      }
+
+      if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         png_uint_32 i;
+
+         for (i = 0; i < (png_uint_32)num_palette; i++)
+         {
+            if (i < png_ptr->num_trans && png_ptr->trans[i] == 0)
+            {
+               palette[i] = back;
+            }
+            else if (i < png_ptr->num_trans && png_ptr->trans[i] != 0xff)
+            {
+               png_byte v, w;
+
+               v = png_ptr->gamma_to_1[png_ptr->palette[i].red];
+               png_composite(w, v, png_ptr->trans[i], back_1.red);
+               palette[i].red = png_ptr->gamma_from_1[w];
+
+               v = png_ptr->gamma_to_1[png_ptr->palette[i].green];
+               png_composite(w, v, png_ptr->trans[i], back_1.green);
+               palette[i].green = png_ptr->gamma_from_1[w];
+
+               v = png_ptr->gamma_to_1[png_ptr->palette[i].blue];
+               png_composite(w, v, png_ptr->trans[i], back_1.blue);
+               palette[i].blue = png_ptr->gamma_from_1[w];
+            }
+            else
+            {
+               palette[i].red = png_ptr->gamma_table[palette[i].red];
+               palette[i].green = png_ptr->gamma_table[palette[i].green];
+               palette[i].blue = png_ptr->gamma_table[palette[i].blue];
+            }
+         }
+      }
+      else
+      {
+         int i;
+
+         for (i = 0; i < num_palette; i++)
+         {
+            if (palette[i].red == (png_byte)png_ptr->trans_values.gray)
+            {
+               palette[i] = back;
+            }
+            else
+            {
+               palette[i].red = png_ptr->gamma_table[palette[i].red];
+               palette[i].green = png_ptr->gamma_table[palette[i].green];
+               palette[i].blue = png_ptr->gamma_table[palette[i].blue];
+            }
+         }
+      }
+   }
+   else
+#endif
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   if (png_ptr->transformations & PNG_GAMMA)
+   {
+      int i;
+
+      for (i = 0; i < num_palette; i++)
+      {
+         palette[i].red = png_ptr->gamma_table[palette[i].red];
+         palette[i].green = png_ptr->gamma_table[palette[i].green];
+         palette[i].blue = png_ptr->gamma_table[palette[i].blue];
+      }
+   }
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   else
+#endif
+#endif
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+   if (png_ptr->transformations & PNG_BACKGROUND)
+   {
+      if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         png_color back;
+
+         back.red   = (png_byte)png_ptr->background.red;
+         back.green = (png_byte)png_ptr->background.green;
+         back.blue  = (png_byte)png_ptr->background.blue;
+
+         for (i = 0; i < (int)png_ptr->num_trans; i++)
+         {
+            if (png_ptr->trans[i] == 0)
+            {
+               palette[i].red = back.red;
+               palette[i].green = back.green;
+               palette[i].blue = back.blue;
+            }
+            else if (png_ptr->trans[i] != 0xff)
+            {
+               png_composite(palette[i].red, png_ptr->palette[i].red,
+                  png_ptr->trans[i], back.red);
+               png_composite(palette[i].green, png_ptr->palette[i].green,
+                  png_ptr->trans[i], back.green);
+               png_composite(palette[i].blue, png_ptr->palette[i].blue,
+                  png_ptr->trans[i], back.blue);
+            }
+         }
+      }
+      else /* assume grayscale palette (what else could it be?) */
+      {
+         int i;
+
+         for (i = 0; i < num_palette; i++)
+         {
+            if (i == (png_byte)png_ptr->trans_values.gray)
+            {
+               palette[i].red = (png_byte)png_ptr->background.red;
+               palette[i].green = (png_byte)png_ptr->background.green;
+               palette[i].blue = (png_byte)png_ptr->background.blue;
+            }
+         }
+      }
+   }
+#endif
+}
+#endif
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED)
+/* Replace any alpha or transparency with the supplied background color.
+ * "background" is already in the screen gamma, while "background_1" is
+ * at a gamma of 1.0.  Paletted files have already been taken care of.
+ */
+void /* PRIVATE */
+png_do_background(png_row_infop row_info, png_bytep row,
+   png_color_16p trans_values, png_color_16p background
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+   , png_color_16p background_1,
+   png_bytep gamma_table, png_bytep gamma_from_1, png_bytep gamma_to_1,
+   png_uint_16pp gamma_16, png_uint_16pp gamma_16_from_1,
+   png_uint_16pp gamma_16_to_1, int gamma_shift
+#endif
+   )
+{
+   png_bytep sp, dp;
+   png_uint_32 i;
+   png_uint_32 row_width=row_info->width;
+   int shift;
+
+   png_debug(1, "in png_do_background\n");
+   if (background != NULL &&
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+      (!(row_info->color_type & PNG_COLOR_MASK_ALPHA) ||
+      (row_info->color_type != PNG_COLOR_TYPE_PALETTE && trans_values)))
+   {
+      switch (row_info->color_type)
+      {
+         case PNG_COLOR_TYPE_GRAY:
+         {
+            switch (row_info->bit_depth)
+            {
+               case 1:
+               {
+                  sp = row;
+                  shift = 7;
+                  for (i = 0; i < row_width; i++)
+                  {
+                     if ((png_uint_16)((*sp >> shift) & 0x01)
+                        == trans_values->gray)
+                     {
+                        *sp &= (png_byte)((0x7f7f >> (7 - shift)) & 0xff);
+                        *sp |= (png_byte)(background->gray << shift);
+                     }
+                     if (!shift)
+                     {
+                        shift = 7;
+                        sp++;
+                     }
+                     else
+                        shift--;
+                  }
+                  break;
+               }
+               case 2:
+               {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                  if (gamma_table != NULL)
+                  {
+                     sp = row;
+                     shift = 6;
+                     for (i = 0; i < row_width; i++)
+                     {
+                        if ((png_uint_16)((*sp >> shift) & 0x03)
+                            == trans_values->gray)
+                        {
+                           *sp &= (png_byte)((0x3f3f >> (6 - shift)) & 0xff);
+                           *sp |= (png_byte)(background->gray << shift);
+                        }
+                        else
+                        {
+                           png_byte p = (png_byte)((*sp >> shift) & 0x03);
+                           png_byte g = (png_byte)((gamma_table [p | (p << 2) |
+                               (p << 4) | (p << 6)] >> 6) & 0x03);
+                           *sp &= (png_byte)((0x3f3f >> (6 - shift)) & 0xff);
+                           *sp |= (png_byte)(g << shift);
+                        }
+                        if (!shift)
+                        {
+                           shift = 6;
+                           sp++;
+                        }
+                        else
+                           shift -= 2;
+                     }
+                  }
+                  else
+#endif
+                  {
+                     sp = row;
+                     shift = 6;
+                     for (i = 0; i < row_width; i++)
+                     {
+                        if ((png_uint_16)((*sp >> shift) & 0x03)
+                            == trans_values->gray)
+                        {
+                           *sp &= (png_byte)((0x3f3f >> (6 - shift)) & 0xff);
+                           *sp |= (png_byte)(background->gray << shift);
+                        }
+                        if (!shift)
+                        {
+                           shift = 6;
+                           sp++;
+                        }
+                        else
+                           shift -= 2;
+                     }
+                  }
+                  break;
+               }
+               case 4:
+               {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                  if (gamma_table != NULL)
+                  {
+                     sp = row;
+                     shift = 4;
+                     for (i = 0; i < row_width; i++)
+                     {
+                        if ((png_uint_16)((*sp >> shift) & 0x0f)
+                            == trans_values->gray)
+                        {
+                           *sp &= (png_byte)((0xf0f >> (4 - shift)) & 0xff);
+                           *sp |= (png_byte)(background->gray << shift);
+                        }
+                        else
+                        {
+                           png_byte p = (png_byte)((*sp >> shift) & 0x0f);
+                           png_byte g = (png_byte)((gamma_table[p |
+                             (p << 4)] >> 4) & 0x0f);
+                           *sp &= (png_byte)((0xf0f >> (4 - shift)) & 0xff);
+                           *sp |= (png_byte)(g << shift);
+                        }
+                        if (!shift)
+                        {
+                           shift = 4;
+                           sp++;
+                        }
+                        else
+                           shift -= 4;
+                     }
+                  }
+                  else
+#endif
+                  {
+                     sp = row;
+                     shift = 4;
+                     for (i = 0; i < row_width; i++)
+                     {
+                        if ((png_uint_16)((*sp >> shift) & 0x0f)
+                            == trans_values->gray)
+                        {
+                           *sp &= (png_byte)((0xf0f >> (4 - shift)) & 0xff);
+                           *sp |= (png_byte)(background->gray << shift);
+                        }
+                        if (!shift)
+                        {
+                           shift = 4;
+                           sp++;
+                        }
+                        else
+                           shift -= 4;
+                     }
+                  }
+                  break;
+               }
+               case 8:
+               {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                  if (gamma_table != NULL)
+                  {
+                     sp = row;
+                     for (i = 0; i < row_width; i++, sp++)
+                     {
+                        if (*sp == trans_values->gray)
+                        {
+                           *sp = (png_byte)background->gray;
+                        }
+                        else
+                        {
+                           *sp = gamma_table[*sp];
+                        }
+                     }
+                  }
+                  else
+#endif
+                  {
+                     sp = row;
+                     for (i = 0; i < row_width; i++, sp++)
+                     {
+                        if (*sp == trans_values->gray)
+                        {
+                           *sp = (png_byte)background->gray;
+                        }
+                     }
+                  }
+                  break;
+               }
+               case 16:
+               {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                  if (gamma_16 != NULL)
+                  {
+                     sp = row;
+                     for (i = 0; i < row_width; i++, sp += 2)
+                     {
+                        png_uint_16 v;
+
+                        v = (png_uint_16)(((*sp) << 8) + *(sp + 1));
+                        if (v == trans_values->gray)
+                        {
+                           /* background is already in screen gamma */
+                           *sp = (png_byte)((background->gray >> 8) & 0xff);
+                           *(sp + 1) = (png_byte)(background->gray & 0xff);
+                        }
+                        else
+                        {
+                           v = gamma_16[*(sp + 1) >> gamma_shift][*sp];
+                           *sp = (png_byte)((v >> 8) & 0xff);
+                           *(sp + 1) = (png_byte)(v & 0xff);
+                        }
+                     }
+                  }
+                  else
+#endif
+                  {
+                     sp = row;
+                     for (i = 0; i < row_width; i++, sp += 2)
+                     {
+                        png_uint_16 v;
+
+                        v = (png_uint_16)(((*sp) << 8) + *(sp + 1));
+                        if (v == trans_values->gray)
+                        {
+                           *sp = (png_byte)((background->gray >> 8) & 0xff);
+                           *(sp + 1) = (png_byte)(background->gray & 0xff);
+                        }
+                     }
+                  }
+                  break;
+               }
+            }
+            break;
+         }
+         case PNG_COLOR_TYPE_RGB:
+         {
+            if (row_info->bit_depth == 8)
+            {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+               if (gamma_table != NULL)
+               {
+                  sp = row;
+                  for (i = 0; i < row_width; i++, sp += 3)
+                  {
+                     if (*sp == trans_values->red &&
+                        *(sp + 1) == trans_values->green &&
+                        *(sp + 2) == trans_values->blue)
+                     {
+                        *sp = (png_byte)background->red;
+                        *(sp + 1) = (png_byte)background->green;
+                        *(sp + 2) = (png_byte)background->blue;
+                     }
+                     else
+                     {
+                        *sp = gamma_table[*sp];
+                        *(sp + 1) = gamma_table[*(sp + 1)];
+                        *(sp + 2) = gamma_table[*(sp + 2)];
+                     }
+                  }
+               }
+               else
+#endif
+               {
+                  sp = row;
+                  for (i = 0; i < row_width; i++, sp += 3)
+                  {
+                     if (*sp == trans_values->red &&
+                        *(sp + 1) == trans_values->green &&
+                        *(sp + 2) == trans_values->blue)
+                     {
+                        *sp = (png_byte)background->red;
+                        *(sp + 1) = (png_byte)background->green;
+                        *(sp + 2) = (png_byte)background->blue;
+                     }
+                  }
+               }
+            }
+            else /* if (row_info->bit_depth == 16) */
+            {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+               if (gamma_16 != NULL)
+               {
+                  sp = row;
+                  for (i = 0; i < row_width; i++, sp += 6)
+                  {
+                     png_uint_16 r = (png_uint_16)(((*sp) << 8) + *(sp + 1));
+                     png_uint_16 g = (png_uint_16)(((*(sp+2)) << 8) + *(sp+3));
+                     png_uint_16 b = (png_uint_16)(((*(sp+4)) << 8) + *(sp+5));
+                     if (r == trans_values->red && g == trans_values->green &&
+                        b == trans_values->blue)
+                     {
+                        /* background is already in screen gamma */
+                        *sp = (png_byte)((background->red >> 8) & 0xff);
+                        *(sp + 1) = (png_byte)(background->red & 0xff);
+                        *(sp + 2) = (png_byte)((background->green >> 8) & 0xff);
+                        *(sp + 3) = (png_byte)(background->green & 0xff);
+                        *(sp + 4) = (png_byte)((background->blue >> 8) & 0xff);
+                        *(sp + 5) = (png_byte)(background->blue & 0xff);
+                     }
+                     else
+                     {
+                        png_uint_16 v = gamma_16[*(sp + 1) >> gamma_shift][*sp];
+                        *sp = (png_byte)((v >> 8) & 0xff);
+                        *(sp + 1) = (png_byte)(v & 0xff);
+                        v = gamma_16[*(sp + 3) >> gamma_shift][*(sp + 2)];
+                        *(sp + 2) = (png_byte)((v >> 8) & 0xff);
+                        *(sp + 3) = (png_byte)(v & 0xff);
+                        v = gamma_16[*(sp + 5) >> gamma_shift][*(sp + 4)];
+                        *(sp + 4) = (png_byte)((v >> 8) & 0xff);
+                        *(sp + 5) = (png_byte)(v & 0xff);
+                     }
+                  }
+               }
+               else
+#endif
+               {
+                  sp = row;
+                  for (i = 0; i < row_width; i++, sp += 6)
+                  {
+                     png_uint_16 r = (png_uint_16)(((*sp) << 8) + *(sp+1));
+                     png_uint_16 g = (png_uint_16)(((*(sp+2)) << 8) + *(sp+3));
+                     png_uint_16 b = (png_uint_16)(((*(sp+4)) << 8) + *(sp+5));
+
+                     if (r == trans_values->red && g == trans_values->green &&
+                        b == trans_values->blue)
+                     {
+                        *sp = (png_byte)((background->red >> 8) & 0xff);
+                        *(sp + 1) = (png_byte)(background->red & 0xff);
+                        *(sp + 2) = (png_byte)((background->green >> 8) & 0xff);
+                        *(sp + 3) = (png_byte)(background->green & 0xff);
+                        *(sp + 4) = (png_byte)((background->blue >> 8) & 0xff);
+                        *(sp + 5) = (png_byte)(background->blue & 0xff);
+                     }
+                  }
+               }
+            }
+            break;
+         }
+         case PNG_COLOR_TYPE_GRAY_ALPHA:
+         {
+            if (row_info->bit_depth == 8)
+            {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+               if (gamma_to_1 != NULL && gamma_from_1 != NULL &&
+                   gamma_table != NULL)
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 2, dp++)
+                  {
+                     png_uint_16 a = *(sp + 1);
+
+                     if (a == 0xff)
+                     {
+                        *dp = gamma_table[*sp];
+                     }
+                     else if (a == 0)
+                     {
+                        /* background is already in screen gamma */
+                        *dp = (png_byte)background->gray;
+                     }
+                     else
+                     {
+                        png_byte v, w;
+
+                        v = gamma_to_1[*sp];
+                        png_composite(w, v, a, background_1->gray);
+                        *dp = gamma_from_1[w];
+                     }
+                  }
+               }
+               else
+#endif
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 2, dp++)
+                  {
+                     png_byte a = *(sp + 1);
+
+                     if (a == 0xff)
+                     {
+                        *dp = *sp;
+                     }
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                     else if (a == 0)
+                     {
+                        *dp = (png_byte)background->gray;
+                     }
+                     else
+                     {
+                        png_composite(*dp, *sp, a, background_1->gray);
+                     }
+#else
+                     *dp = (png_byte)background->gray;
+#endif
+                  }
+               }
+            }
+            else /* if (png_ptr->bit_depth == 16) */
+            {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+               if (gamma_16 != NULL && gamma_16_from_1 != NULL &&
+                   gamma_16_to_1 != NULL)
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 4, dp += 2)
+                  {
+                     png_uint_16 a = (png_uint_16)(((*(sp+2)) << 8) + *(sp+3));
+
+                     if (a == (png_uint_16)0xffff)
+                     {
+                        png_uint_16 v;
+
+                        v = gamma_16[*(sp + 1) >> gamma_shift][*sp];
+                        *dp = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(v & 0xff);
+                     }
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                     else if (a == 0)
+#else
+                     else
+#endif
+                     {
+                        /* background is already in screen gamma */
+                        *dp = (png_byte)((background->gray >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(background->gray & 0xff);
+                     }
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                     else
+                     {
+                        png_uint_16 g, v, w;
+
+                        g = gamma_16_to_1[*(sp + 1) >> gamma_shift][*sp];
+                        png_composite_16(v, g, a, background_1->gray);
+                        w = gamma_16_from_1[(v&0xff) >> gamma_shift][v >> 8];
+                        *dp = (png_byte)((w >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(w & 0xff);
+                     }
+#endif
+                  }
+               }
+               else
+#endif
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 4, dp += 2)
+                  {
+                     png_uint_16 a = (png_uint_16)(((*(sp+2)) << 8) + *(sp+3));
+                     if (a == (png_uint_16)0xffff)
+                     {
+                        png_memcpy(dp, sp, 2);
+                     }
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                     else if (a == 0)
+#else
+                     else
+#endif
+                     {
+                        *dp = (png_byte)((background->gray >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(background->gray & 0xff);
+                     }
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+                     else
+                     {
+                        png_uint_16 g, v;
+
+                        g = (png_uint_16)(((*sp) << 8) + *(sp + 1));
+                        png_composite_16(v, g, a, background_1->gray);
+                        *dp = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(v & 0xff);
+                     }
+#endif
+                  }
+               }
+            }
+            break;
+         }
+         case PNG_COLOR_TYPE_RGB_ALPHA:
+         {
+            if (row_info->bit_depth == 8)
+            {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+               if (gamma_to_1 != NULL && gamma_from_1 != NULL &&
+                   gamma_table != NULL)
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 4, dp += 3)
+                  {
+                     png_byte a = *(sp + 3);
+
+                     if (a == 0xff)
+                     {
+                        *dp = gamma_table[*sp];
+                        *(dp + 1) = gamma_table[*(sp + 1)];
+                        *(dp + 2) = gamma_table[*(sp + 2)];
+                     }
+                     else if (a == 0)
+                     {
+                        /* background is already in screen gamma */
+                        *dp = (png_byte)background->red;
+                        *(dp + 1) = (png_byte)background->green;
+                        *(dp + 2) = (png_byte)background->blue;
+                     }
+                     else
+                     {
+                        png_byte v, w;
+
+                        v = gamma_to_1[*sp];
+                        png_composite(w, v, a, background_1->red);
+                        *dp = gamma_from_1[w];
+                        v = gamma_to_1[*(sp + 1)];
+                        png_composite(w, v, a, background_1->green);
+                        *(dp + 1) = gamma_from_1[w];
+                        v = gamma_to_1[*(sp + 2)];
+                        png_composite(w, v, a, background_1->blue);
+                        *(dp + 2) = gamma_from_1[w];
+                     }
+                  }
+               }
+               else
+#endif
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 4, dp += 3)
+                  {
+                     png_byte a = *(sp + 3);
+
+                     if (a == 0xff)
+                     {
+                        *dp = *sp;
+                        *(dp + 1) = *(sp + 1);
+                        *(dp + 2) = *(sp + 2);
+                     }
+                     else if (a == 0)
+                     {
+                        *dp = (png_byte)background->red;
+                        *(dp + 1) = (png_byte)background->green;
+                        *(dp + 2) = (png_byte)background->blue;
+                     }
+                     else
+                     {
+                        png_composite(*dp, *sp, a, background->red);
+                        png_composite(*(dp + 1), *(sp + 1), a,
+                           background->green);
+                        png_composite(*(dp + 2), *(sp + 2), a,
+                           background->blue);
+                     }
+                  }
+               }
+            }
+            else /* if (row_info->bit_depth == 16) */
+            {
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+               if (gamma_16 != NULL && gamma_16_from_1 != NULL &&
+                   gamma_16_to_1 != NULL)
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 8, dp += 6)
+                  {
+                     png_uint_16 a = (png_uint_16)(((png_uint_16)(*(sp + 6))
+                         << 8) + (png_uint_16)(*(sp + 7)));
+                     if (a == (png_uint_16)0xffff)
+                     {
+                        png_uint_16 v;
+
+                        v = gamma_16[*(sp + 1) >> gamma_shift][*sp];
+                        *dp = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(v & 0xff);
+                        v = gamma_16[*(sp + 3) >> gamma_shift][*(sp + 2)];
+                        *(dp + 2) = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 3) = (png_byte)(v & 0xff);
+                        v = gamma_16[*(sp + 5) >> gamma_shift][*(sp + 4)];
+                        *(dp + 4) = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 5) = (png_byte)(v & 0xff);
+                     }
+                     else if (a == 0)
+                     {
+                        /* background is already in screen gamma */
+                        *dp = (png_byte)((background->red >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(background->red & 0xff);
+                        *(dp + 2) = (png_byte)((background->green >> 8) & 0xff);
+                        *(dp + 3) = (png_byte)(background->green & 0xff);
+                        *(dp + 4) = (png_byte)((background->blue >> 8) & 0xff);
+                        *(dp + 5) = (png_byte)(background->blue & 0xff);
+                     }
+                     else
+                     {
+                        png_uint_16 v, w, x;
+
+                        v = gamma_16_to_1[*(sp + 1) >> gamma_shift][*sp];
+                        png_composite_16(w, v, a, background->red);
+                        x = gamma_16_from_1[((w&0xff) >> gamma_shift)][w >> 8];
+                        *dp = (png_byte)((x >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(x & 0xff);
+                        v = gamma_16_to_1[*(sp + 3) >> gamma_shift][*(sp + 2)];
+                        png_composite_16(w, v, a, background->green);
+                        x = gamma_16_from_1[((w&0xff) >> gamma_shift)][w >> 8];
+                        *(dp + 2) = (png_byte)((x >> 8) & 0xff);
+                        *(dp + 3) = (png_byte)(x & 0xff);
+                        v = gamma_16_to_1[*(sp + 5) >> gamma_shift][*(sp + 4)];
+                        png_composite_16(w, v, a, background->blue);
+                        x = gamma_16_from_1[(w & 0xff) >> gamma_shift][w >> 8];
+                        *(dp + 4) = (png_byte)((x >> 8) & 0xff);
+                        *(dp + 5) = (png_byte)(x & 0xff);
+                     }
+                  }
+               }
+               else
+#endif
+               {
+                  sp = row;
+                  dp = row;
+                  for (i = 0; i < row_width; i++, sp += 8, dp += 6)
+                  {
+                     png_uint_16 a = (png_uint_16)(((png_uint_16)(*(sp + 6))
+                        << 8) + (png_uint_16)(*(sp + 7)));
+                     if (a == (png_uint_16)0xffff)
+                     {
+                        png_memcpy(dp, sp, 6);
+                     }
+                     else if (a == 0)
+                     {
+                        *dp = (png_byte)((background->red >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(background->red & 0xff);
+                        *(dp + 2) = (png_byte)((background->green >> 8) & 0xff);
+                        *(dp + 3) = (png_byte)(background->green & 0xff);
+                        *(dp + 4) = (png_byte)((background->blue >> 8) & 0xff);
+                        *(dp + 5) = (png_byte)(background->blue & 0xff);
+                     }
+                     else
+                     {
+                        png_uint_16 v;
+
+                        png_uint_16 r = (png_uint_16)(((*sp) << 8) + *(sp + 1));
+                        png_uint_16 g = (png_uint_16)(((*(sp + 2)) << 8)
+                            + *(sp + 3));
+                        png_uint_16 b = (png_uint_16)(((*(sp + 4)) << 8)
+                            + *(sp + 5));
+
+                        png_composite_16(v, r, a, background->red);
+                        *dp = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 1) = (png_byte)(v & 0xff);
+                        png_composite_16(v, g, a, background->green);
+                        *(dp + 2) = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 3) = (png_byte)(v & 0xff);
+                        png_composite_16(v, b, a, background->blue);
+                        *(dp + 4) = (png_byte)((v >> 8) & 0xff);
+                        *(dp + 5) = (png_byte)(v & 0xff);
+                     }
+                  }
+               }
+            }
+            break;
+         }
+      }
+
+      if (row_info->color_type & PNG_COLOR_MASK_ALPHA)
+      {
+         row_info->color_type &= ~PNG_COLOR_MASK_ALPHA;
+         row_info->channels--;
+         row_info->pixel_depth = (png_byte)(row_info->channels *
+            row_info->bit_depth);
+         row_info->rowbytes = ((row_width *
+            row_info->pixel_depth + 7) >> 3);
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+/* Gamma correct the image, avoiding the alpha channel.  Make sure
+ * you do this after you deal with the transparency issue on grayscale
+ * or RGB images. If your bit depth is 8, use gamma_table, if it
+ * is 16, use gamma_16_table and gamma_shift.  Build these with
+ * build_gamma_table().
+ */
+void /* PRIVATE */
+png_do_gamma(png_row_infop row_info, png_bytep row,
+   png_bytep gamma_table, png_uint_16pp gamma_16_table,
+   int gamma_shift)
+{
+   png_bytep sp;
+   png_uint_32 i;
+   png_uint_32 row_width=row_info->width;
+
+   png_debug(1, "in png_do_gamma\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       ((row_info->bit_depth <= 8 && gamma_table != NULL) ||
+        (row_info->bit_depth == 16 && gamma_16_table != NULL)))
+   {
+      switch (row_info->color_type)
+      {
+         case PNG_COLOR_TYPE_RGB:
+         {
+            if (row_info->bit_depth == 8)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  *sp = gamma_table[*sp];
+                  sp++;
+                  *sp = gamma_table[*sp];
+                  sp++;
+                  *sp = gamma_table[*sp];
+                  sp++;
+               }
+            }
+            else /* if (row_info->bit_depth == 16) */
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 v;
+
+                  v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 2;
+                  v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 2;
+                  v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 2;
+               }
+            }
+            break;
+         }
+         case PNG_COLOR_TYPE_RGB_ALPHA:
+         {
+            if (row_info->bit_depth == 8)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  *sp = gamma_table[*sp];
+                  sp++;
+                  *sp = gamma_table[*sp];
+                  sp++;
+                  *sp = gamma_table[*sp];
+                  sp++;
+                  sp++;
+               }
+            }
+            else /* if (row_info->bit_depth == 16) */
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 2;
+                  v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 2;
+                  v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 4;
+               }
+            }
+            break;
+         }
+         case PNG_COLOR_TYPE_GRAY_ALPHA:
+         {
+            if (row_info->bit_depth == 8)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  *sp = gamma_table[*sp];
+                  sp += 2;
+               }
+            }
+            else /* if (row_info->bit_depth == 16) */
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 4;
+               }
+            }
+            break;
+         }
+         case PNG_COLOR_TYPE_GRAY:
+         {
+            if (row_info->bit_depth == 2)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i += 4)
+               {
+                  int a = *sp & 0xc0;
+                  int b = *sp & 0x30;
+                  int c = *sp & 0x0c;
+                  int d = *sp & 0x03;
+
+                  *sp = (png_byte)(
+                        ((((int)gamma_table[a|(a>>2)|(a>>4)|(a>>6)])   ) & 0xc0)|
+                        ((((int)gamma_table[(b<<2)|b|(b>>2)|(b>>4)])>>2) & 0x30)|
+                        ((((int)gamma_table[(c<<4)|(c<<2)|c|(c>>2)])>>4) & 0x0c)|
+                        ((((int)gamma_table[(d<<6)|(d<<4)|(d<<2)|d])>>6) ));
+                  sp++;
+               }
+            }
+            if (row_info->bit_depth == 4)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i += 2)
+               {
+                  int msb = *sp & 0xf0;
+                  int lsb = *sp & 0x0f;
+
+                  *sp = (png_byte)((((int)gamma_table[msb | (msb >> 4)]) & 0xf0)
+                          | (((int)gamma_table[(lsb << 4) | lsb]) >> 4));
+                  sp++;
+               }
+            }
+            else if (row_info->bit_depth == 8)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  *sp = gamma_table[*sp];
+                  sp++;
+               }
+            }
+            else if (row_info->bit_depth == 16)
+            {
+               sp = row;
+               for (i = 0; i < row_width; i++)
+               {
+                  png_uint_16 v = gamma_16_table[*(sp + 1) >> gamma_shift][*sp];
+                  *sp = (png_byte)((v >> 8) & 0xff);
+                  *(sp + 1) = (png_byte)(v & 0xff);
+                  sp += 2;
+               }
+            }
+            break;
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+/* Expands a palette row to an RGB or RGBA row depending
+ * upon whether you supply trans and num_trans.
+ */
+void /* PRIVATE */
+png_do_expand_palette(png_row_infop row_info, png_bytep row,
+   png_colorp palette, png_bytep trans, int num_trans)
+{
+   int shift, value;
+   png_bytep sp, dp;
+   png_uint_32 i;
+   png_uint_32 row_width=row_info->width;
+
+   png_debug(1, "in png_do_expand_palette\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       row_info->color_type == PNG_COLOR_TYPE_PALETTE)
+   {
+      if (row_info->bit_depth < 8)
+      {
+         switch (row_info->bit_depth)
+         {
+            case 1:
+            {
+               sp = row + (png_size_t)((row_width - 1) >> 3);
+               dp = row + (png_size_t)row_width - 1;
+               shift = 7 - (int)((row_width + 7) & 0x07);
+               for (i = 0; i < row_width; i++)
+               {
+                  if ((*sp >> shift) & 0x01)
+                     *dp = 1;
+                  else
+                     *dp = 0;
+                  if (shift == 7)
+                  {
+                     shift = 0;
+                     sp--;
+                  }
+                  else
+                     shift++;
+
+                  dp--;
+               }
+               break;
+            }
+            case 2:
+            {
+               sp = row + (png_size_t)((row_width - 1) >> 2);
+               dp = row + (png_size_t)row_width - 1;
+               shift = (int)((3 - ((row_width + 3) & 0x03)) << 1);
+               for (i = 0; i < row_width; i++)
+               {
+                  value = (*sp >> shift) & 0x03;
+                  *dp = (png_byte)value;
+                  if (shift == 6)
+                  {
+                     shift = 0;
+                     sp--;
+                  }
+                  else
+                     shift += 2;
+
+                  dp--;
+               }
+               break;
+            }
+            case 4:
+            {
+               sp = row + (png_size_t)((row_width - 1) >> 1);
+               dp = row + (png_size_t)row_width - 1;
+               shift = (int)((row_width & 0x01) << 2);
+               for (i = 0; i < row_width; i++)
+               {
+                  value = (*sp >> shift) & 0x0f;
+                  *dp = (png_byte)value;
+                  if (shift == 4)
+                  {
+                     shift = 0;
+                     sp--;
+                  }
+                  else
+                     shift += 4;
+
+                  dp--;
+               }
+               break;
+            }
+         }
+         row_info->bit_depth = 8;
+         row_info->pixel_depth = 8;
+         row_info->rowbytes = row_width;
+      }
+      switch (row_info->bit_depth)
+      {
+         case 8:
+         {
+            if (trans != NULL)
+            {
+               sp = row + (png_size_t)row_width - 1;
+               dp = row + (png_size_t)(row_width << 2) - 1;
+
+               for (i = 0; i < row_width; i++)
+               {
+                  if ((int)(*sp) >= num_trans)
+                     *dp-- = 0xff;
+                  else
+                     *dp-- = trans[*sp];
+                  *dp-- = palette[*sp].blue;
+                  *dp-- = palette[*sp].green;
+                  *dp-- = palette[*sp].red;
+                  sp--;
+               }
+               row_info->bit_depth = 8;
+               row_info->pixel_depth = 32;
+               row_info->rowbytes = row_width * 4;
+               row_info->color_type = 6;
+               row_info->channels = 4;
+            }
+            else
+            {
+               sp = row + (png_size_t)row_width - 1;
+               dp = row + (png_size_t)(row_width * 3) - 1;
+
+               for (i = 0; i < row_width; i++)
+               {
+                  *dp-- = palette[*sp].blue;
+                  *dp-- = palette[*sp].green;
+                  *dp-- = palette[*sp].red;
+                  sp--;
+               }
+               row_info->bit_depth = 8;
+               row_info->pixel_depth = 24;
+               row_info->rowbytes = row_width * 3;
+               row_info->color_type = 2;
+               row_info->channels = 3;
+            }
+            break;
+         }
+      }
+   }
+}
+
+/* If the bit depth < 8, it is expanded to 8.  Also, if the
+ * transparency value is supplied, an alpha channel is built.
+ */
+void /* PRIVATE */
+png_do_expand(png_row_infop row_info, png_bytep row,
+   png_color_16p trans_value)
+{
+   int shift, value;
+   png_bytep sp, dp;
+   png_uint_32 i;
+   png_uint_32 row_width=row_info->width;
+
+   png_debug(1, "in png_do_expand\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+      if (row_info->color_type == PNG_COLOR_TYPE_GRAY)
+      {
+         png_uint_16 gray = (png_uint_16)(trans_value ? trans_value->gray : 0);
+
+         if (row_info->bit_depth < 8)
+         {
+            switch (row_info->bit_depth)
+            {
+               case 1:
+               {
+                  gray = (png_uint_16)(gray*0xff);
+                  sp = row + (png_size_t)((row_width - 1) >> 3);
+                  dp = row + (png_size_t)row_width - 1;
+                  shift = 7 - (int)((row_width + 7) & 0x07);
+                  for (i = 0; i < row_width; i++)
+                  {
+                     if ((*sp >> shift) & 0x01)
+                        *dp = 0xff;
+                     else
+                        *dp = 0;
+                     if (shift == 7)
+                     {
+                        shift = 0;
+                        sp--;
+                     }
+                     else
+                        shift++;
+
+                     dp--;
+                  }
+                  break;
+               }
+               case 2:
+               {
+                  gray = (png_uint_16)(gray*0x55);
+                  sp = row + (png_size_t)((row_width - 1) >> 2);
+                  dp = row + (png_size_t)row_width - 1;
+                  shift = (int)((3 - ((row_width + 3) & 0x03)) << 1);
+                  for (i = 0; i < row_width; i++)
+                  {
+                     value = (*sp >> shift) & 0x03;
+                     *dp = (png_byte)(value | (value << 2) | (value << 4) |
+                        (value << 6));
+                     if (shift == 6)
+                     {
+                        shift = 0;
+                        sp--;
+                     }
+                     else
+                        shift += 2;
+
+                     dp--;
+                  }
+                  break;
+               }
+               case 4:
+               {
+                  gray = (png_uint_16)(gray*0x11);
+                  sp = row + (png_size_t)((row_width - 1) >> 1);
+                  dp = row + (png_size_t)row_width - 1;
+                  shift = (int)((1 - ((row_width + 1) & 0x01)) << 2);
+                  for (i = 0; i < row_width; i++)
+                  {
+                     value = (*sp >> shift) & 0x0f;
+                     *dp = (png_byte)(value | (value << 4));
+                     if (shift == 4)
+                     {
+                        shift = 0;
+                        sp--;
+                     }
+                     else
+                        shift = 4;
+
+                     dp--;
+                  }
+                  break;
+               }
+            }
+            row_info->bit_depth = 8;
+            row_info->pixel_depth = 8;
+            row_info->rowbytes = row_width;
+         }
+
+         if (trans_value != NULL)
+         {
+            if (row_info->bit_depth == 8)
+            {
+               sp = row + (png_size_t)row_width - 1;
+               dp = row + (png_size_t)(row_width << 1) - 1;
+               for (i = 0; i < row_width; i++)
+               {
+                  if (*sp == gray)
+                     *dp-- = 0;
+                  else
+                     *dp-- = 0xff;
+                  *dp-- = *sp--;
+               }
+            }
+            else if (row_info->bit_depth == 16)
+            {
+               sp = row + row_info->rowbytes - 1;
+               dp = row + (row_info->rowbytes << 1) - 1;
+               for (i = 0; i < row_width; i++)
+               {
+                  if (((png_uint_16)*(sp) |
+                     ((png_uint_16)*(sp - 1) << 8)) == gray)
+                  {
+                     *dp-- = 0;
+                     *dp-- = 0;
+                  }
+                  else
+                  {
+                     *dp-- = 0xff;
+                     *dp-- = 0xff;
+                  }
+                  *dp-- = *sp--;
+                  *dp-- = *sp--;
+               }
+            }
+            row_info->color_type = PNG_COLOR_TYPE_GRAY_ALPHA;
+            row_info->channels = 2;
+            row_info->pixel_depth = (png_byte)(row_info->bit_depth << 1);
+            row_info->rowbytes =
+               ((row_width * row_info->pixel_depth) >> 3);
+         }
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_RGB && trans_value)
+      {
+         if (row_info->bit_depth == 8)
+         {
+            sp = row + (png_size_t)row_info->rowbytes - 1;
+            dp = row + (png_size_t)(row_width << 2) - 1;
+            for (i = 0; i < row_width; i++)
+            {
+               if (*(sp - 2) == trans_value->red &&
+                  *(sp - 1) == trans_value->green &&
+                  *(sp - 0) == trans_value->blue)
+                  *dp-- = 0;
+               else
+                  *dp-- = 0xff;
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+            }
+         }
+         else if (row_info->bit_depth == 16)
+         {
+            sp = row + row_info->rowbytes - 1;
+            dp = row + (png_size_t)(row_width << 3) - 1;
+            for (i = 0; i < row_width; i++)
+            {
+               if ((((png_uint_16)*(sp - 4) |
+                  ((png_uint_16)*(sp - 5) << 8)) == trans_value->red) &&
+                  (((png_uint_16)*(sp - 2) |
+                  ((png_uint_16)*(sp - 3) << 8)) == trans_value->green) &&
+                  (((png_uint_16)*(sp - 0) |
+                  ((png_uint_16)*(sp - 1) << 8)) == trans_value->blue))
+               {
+                  *dp-- = 0;
+                  *dp-- = 0;
+               }
+               else
+               {
+                  *dp-- = 0xff;
+                  *dp-- = 0xff;
+               }
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+               *dp-- = *sp--;
+            }
+         }
+         row_info->color_type = PNG_COLOR_TYPE_RGB_ALPHA;
+         row_info->channels = 4;
+         row_info->pixel_depth = (png_byte)(row_info->bit_depth << 2);
+         row_info->rowbytes =
+            ((row_width * row_info->pixel_depth) >> 3);
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_DITHER_SUPPORTED)
+void /* PRIVATE */
+png_do_dither(png_row_infop row_info, png_bytep row,
+    png_bytep palette_lookup, png_bytep dither_lookup)
+{
+   png_bytep sp, dp;
+   png_uint_32 i;
+   png_uint_32 row_width=row_info->width;
+
+   png_debug(1, "in png_do_dither\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB &&
+         palette_lookup && row_info->bit_depth == 8)
+      {
+         int r, g, b, p;
+         sp = row;
+         dp = row;
+         for (i = 0; i < row_width; i++)
+         {
+            r = *sp++;
+            g = *sp++;
+            b = *sp++;
+
+            /* this looks real messy, but the compiler will reduce
+               it down to a reasonable formula.  For example, with
+               5 bits per color, we get:
+               p = (((r >> 3) & 0x1f) << 10) |
+                  (((g >> 3) & 0x1f) << 5) |
+                  ((b >> 3) & 0x1f);
+               */
+            p = (((r >> (8 - PNG_DITHER_RED_BITS)) &
+               ((1 << PNG_DITHER_RED_BITS) - 1)) <<
+               (PNG_DITHER_GREEN_BITS + PNG_DITHER_BLUE_BITS)) |
+               (((g >> (8 - PNG_DITHER_GREEN_BITS)) &
+               ((1 << PNG_DITHER_GREEN_BITS) - 1)) <<
+               (PNG_DITHER_BLUE_BITS)) |
+               ((b >> (8 - PNG_DITHER_BLUE_BITS)) &
+               ((1 << PNG_DITHER_BLUE_BITS) - 1));
+
+            *dp++ = palette_lookup[p];
+         }
+         row_info->color_type = PNG_COLOR_TYPE_PALETTE;
+         row_info->channels = 1;
+         row_info->pixel_depth = row_info->bit_depth;
+         row_info->rowbytes =
+             ((row_width * row_info->pixel_depth + 7) >> 3);
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA &&
+         palette_lookup != NULL && row_info->bit_depth == 8)
+      {
+         int r, g, b, p;
+         sp = row;
+         dp = row;
+         for (i = 0; i < row_width; i++)
+         {
+            r = *sp++;
+            g = *sp++;
+            b = *sp++;
+            sp++;
+
+            p = (((r >> (8 - PNG_DITHER_RED_BITS)) &
+               ((1 << PNG_DITHER_RED_BITS) - 1)) <<
+               (PNG_DITHER_GREEN_BITS + PNG_DITHER_BLUE_BITS)) |
+               (((g >> (8 - PNG_DITHER_GREEN_BITS)) &
+               ((1 << PNG_DITHER_GREEN_BITS) - 1)) <<
+               (PNG_DITHER_BLUE_BITS)) |
+               ((b >> (8 - PNG_DITHER_BLUE_BITS)) &
+               ((1 << PNG_DITHER_BLUE_BITS) - 1));
+
+            *dp++ = palette_lookup[p];
+         }
+         row_info->color_type = PNG_COLOR_TYPE_PALETTE;
+         row_info->channels = 1;
+         row_info->pixel_depth = row_info->bit_depth;
+         row_info->rowbytes =
+            ((row_width * row_info->pixel_depth + 7) >> 3);
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_PALETTE &&
+         dither_lookup && row_info->bit_depth == 8)
+      {
+         sp = row;
+         for (i = 0; i < row_width; i++, sp++)
+         {
+            *sp = dither_lookup[*sp];
+         }
+      }
+   }
+}
+#endif
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+#if defined(PNG_READ_GAMMA_SUPPORTED)
+static int png_gamma_shift[] =
+   {0x10, 0x21, 0x42, 0x84, 0x110, 0x248, 0x550, 0xff0};
+
+/* We build the 8- or 16-bit gamma tables here.  Note that for 16-bit
+ * tables, we don't make a full table if we are reducing to 8-bit in
+ * the future.  Note also how the gamma_16 tables are segmented so that
+ * we don't need to allocate > 64K chunks for a full 16-bit table.
+ */
+void /* PRIVATE */
+png_build_gamma_table(png_structp png_ptr)
+{
+  png_debug(1, "in png_build_gamma_table\n");
+  if(png_ptr->gamma != 0.0)
+  {
+   if (png_ptr->bit_depth <= 8)
+   {
+      int i;
+      double g;
+
+      if (png_ptr->screen_gamma > .000001)
+         g = 1.0 / (png_ptr->gamma * png_ptr->screen_gamma);
+      else
+         g = 1.0;
+
+      png_ptr->gamma_table = (png_bytep)png_malloc(png_ptr,
+         (png_uint_32)256);
+
+      for (i = 0; i < 256; i++)
+      {
+         png_ptr->gamma_table[i] = (png_byte)(pow((double)i / 255.0,
+            g) * 255.0 + .5);
+      }
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED) || \
+    defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+      if (png_ptr->transformations & ((PNG_BACKGROUND) | PNG_RGB_TO_GRAY))
+      {
+
+         g = 1.0 / (png_ptr->gamma);
+
+         png_ptr->gamma_to_1 = (png_bytep)png_malloc(png_ptr,
+            (png_uint_32)256);
+
+         for (i = 0; i < 256; i++)
+         {
+            png_ptr->gamma_to_1[i] = (png_byte)(pow((double)i / 255.0,
+               g) * 255.0 + .5);
+         }
+
+
+         png_ptr->gamma_from_1 = (png_bytep)png_malloc(png_ptr,
+            (png_uint_32)256);
+
+         if(png_ptr->screen_gamma > 0.000001)
+            g = 1.0 / png_ptr->screen_gamma;
+         else
+            g = png_ptr->gamma;   /* probably doing rgb_to_gray */
+
+         for (i = 0; i < 256; i++)
+         {
+            png_ptr->gamma_from_1[i] = (png_byte)(pow((double)i / 255.0,
+               g) * 255.0 + .5);
+
+         }
+      }
+#endif /* PNG_READ_BACKGROUND_SUPPORTED || PNG_RGB_TO_GRAY_SUPPORTED */
+   }
+   else
+   {
+      double g;
+      int i, j, shift, num;
+      int sig_bit;
+      png_uint_32 ig;
+
+      if (png_ptr->color_type & PNG_COLOR_MASK_COLOR)
+      {
+         sig_bit = (int)png_ptr->sig_bit.red;
+         if ((int)png_ptr->sig_bit.green > sig_bit)
+            sig_bit = png_ptr->sig_bit.green;
+         if ((int)png_ptr->sig_bit.blue > sig_bit)
+            sig_bit = png_ptr->sig_bit.blue;
+      }
+      else
+      {
+         sig_bit = (int)png_ptr->sig_bit.gray;
+      }
+
+      if (sig_bit > 0)
+         shift = 16 - sig_bit;
+      else
+         shift = 0;
+
+      if (png_ptr->transformations & PNG_16_TO_8)
+      {
+         if (shift < (16 - PNG_MAX_GAMMA_8))
+            shift = (16 - PNG_MAX_GAMMA_8);
+      }
+
+      if (shift > 8)
+         shift = 8;
+      if (shift < 0)
+         shift = 0;
+
+      png_ptr->gamma_shift = (png_byte)shift;
+
+      num = (1 << (8 - shift));
+
+      if (png_ptr->screen_gamma > .000001)
+         g = 1.0 / (png_ptr->gamma * png_ptr->screen_gamma);
+      else
+         g = 1.0;
+
+      png_ptr->gamma_16_table = (png_uint_16pp)png_malloc(png_ptr,
+         (png_uint_32)(num * sizeof (png_uint_16p)));
+
+      if (png_ptr->transformations & (PNG_16_TO_8 | PNG_BACKGROUND))
+      {
+         double fin, fout;
+         png_uint_32 last, max;
+
+         for (i = 0; i < num; i++)
+         {
+            png_ptr->gamma_16_table[i] = (png_uint_16p)png_malloc(png_ptr,
+               (png_uint_32)(256 * sizeof (png_uint_16)));
+         }
+
+         g = 1.0 / g;
+         last = 0;
+         for (i = 0; i < 256; i++)
+         {
+            fout = ((double)i + 0.5) / 256.0;
+            fin = pow(fout, g);
+            max = (png_uint_32)(fin * (double)((png_uint_32)num << 8));
+            while (last <= max)
+            {
+               png_ptr->gamma_16_table[(int)(last & (0xff >> shift))]
+                  [(int)(last >> (8 - shift))] = (png_uint_16)(
+                  (png_uint_16)i | ((png_uint_16)i << 8));
+               last++;
+            }
+         }
+         while (last < ((png_uint_32)num << 8))
+         {
+            png_ptr->gamma_16_table[(int)(last & (0xff >> shift))]
+               [(int)(last >> (8 - shift))] = (png_uint_16)65535L;
+            last++;
+         }
+      }
+      else
+      {
+         for (i = 0; i < num; i++)
+         {
+            png_ptr->gamma_16_table[i] = (png_uint_16p)png_malloc(png_ptr,
+               (png_uint_32)(256 * sizeof (png_uint_16)));
+
+            ig = (((png_uint_32)i * (png_uint_32)png_gamma_shift[shift]) >> 4);
+            for (j = 0; j < 256; j++)
+            {
+               png_ptr->gamma_16_table[i][j] =
+                  (png_uint_16)(pow((double)(ig + ((png_uint_32)j << 8)) /
+                     65535.0, g) * 65535.0 + .5);
+            }
+         }
+      }
+
+#if defined(PNG_READ_BACKGROUND_SUPPORTED) || \
+    defined(PNG_READ_RGB_TO_GRAY_SUPPORTED)
+      if (png_ptr->transformations & (PNG_BACKGROUND | PNG_RGB_TO_GRAY))
+      {
+
+         g = 1.0 / (png_ptr->gamma);
+
+         png_ptr->gamma_16_to_1 = (png_uint_16pp)png_malloc(png_ptr,
+            (png_uint_32)(num * sizeof (png_uint_16p )));
+
+         for (i = 0; i < num; i++)
+         {
+            png_ptr->gamma_16_to_1[i] = (png_uint_16p)png_malloc(png_ptr,
+               (png_uint_32)(256 * sizeof (png_uint_16)));
+
+            ig = (((png_uint_32)i *
+               (png_uint_32)png_gamma_shift[shift]) >> 4);
+            for (j = 0; j < 256; j++)
+            {
+               png_ptr->gamma_16_to_1[i][j] =
+                  (png_uint_16)(pow((double)(ig + ((png_uint_32)j << 8)) /
+                     65535.0, g) * 65535.0 + .5);
+            }
+         }
+
+         if(png_ptr->screen_gamma > 0.000001)
+            g = 1.0 / png_ptr->screen_gamma;
+         else
+            g = png_ptr->gamma;   /* probably doing rgb_to_gray */
+
+         png_ptr->gamma_16_from_1 = (png_uint_16pp)png_malloc(png_ptr,
+            (png_uint_32)(num * sizeof (png_uint_16p)));
+
+         for (i = 0; i < num; i++)
+         {
+            png_ptr->gamma_16_from_1[i] = (png_uint_16p)png_malloc(png_ptr,
+               (png_uint_32)(256 * sizeof (png_uint_16)));
+
+            ig = (((png_uint_32)i *
+               (png_uint_32)png_gamma_shift[shift]) >> 4);
+            for (j = 0; j < 256; j++)
+            {
+               png_ptr->gamma_16_from_1[i][j] =
+                  (png_uint_16)(pow((double)(ig + ((png_uint_32)j << 8)) /
+                     65535.0, g) * 65535.0 + .5);
+            }
+         }
+      }
+#endif /* PNG_READ_BACKGROUND_SUPPORTED || PNG_RGB_TO_GRAY_SUPPORTED */
+   }
+ }
+}
+#endif
+/* To do: install integer version of png_build_gamma_table here */
+#endif
+
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+/* undoes intrapixel differencing  */
+void /* PRIVATE */
+png_do_read_intrapixel(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_read_intrapixel\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       (row_info->color_type & PNG_COLOR_MASK_COLOR))
+   {
+      int bytes_per_pixel;
+      png_uint_32 row_width = row_info->width;
+      if (row_info->bit_depth == 8)
+      {
+         png_bytep rp;
+         png_uint_32 i;
+
+         if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+            bytes_per_pixel = 3;
+         else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+            bytes_per_pixel = 4;
+         else
+            return;
+
+         for (i = 0, rp = row; i < row_width; i++, rp += bytes_per_pixel)
+         {
+            *(rp) = (png_byte)((256 + *rp + *(rp+1))&0xff);
+            *(rp+2) = (png_byte)((256 + *(rp+2) + *(rp+1))&0xff);
+         }
+      }
+      else if (row_info->bit_depth == 16)
+      {
+         png_bytep rp;
+         png_uint_32 i;
+
+         if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+            bytes_per_pixel = 6;
+         else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+            bytes_per_pixel = 8;
+         else
+            return;
+
+         for (i = 0, rp = row; i < row_width; i++, rp += bytes_per_pixel)
+         {
+            png_uint_32 s0=*(rp  )<<8 | *(rp+1);
+            png_uint_32 s1=*(rp+2)<<8 | *(rp+3);
+            png_uint_32 s2=*(rp+4)<<8 | *(rp+5);
+            png_uint_32 red=(65536+s0+s1)&0xffff;
+            png_uint_32 blue=(65536+s2+s1)&0xffff;
+            *(rp  ) = (png_byte)((red>>8)&0xff);
+            *(rp+1) = (png_byte)(red&0xff);
+            *(rp+4) = (png_byte)((blue>>8)&0xff);
+            *(rp+5) = (png_byte)(blue&0xff);
+         }
+      }
+   }
+}
+#endif /* PNG_MNG_FEATURES_SUPPORTED */
diff --git a/png/pngrutil.c b/png/pngrutil.c
new file mode 100644
index 000000000..dca7c655d
--- /dev/null
+++ b/png/pngrutil.c
@@ -0,0 +1,3001 @@
+
+/* pngrutil.c - utilities to read a PNG file
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file contains routines that are only called from within
+ * libpng itself during the course of reading an image.
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+#if defined(_WIN32_WCE)
+/* strtod() function is not supported on WindowsCE */
+#  ifdef PNG_FLOATING_POINT_SUPPORTED
+__inline double strtod(const char *nptr, char **endptr)
+{
+   double result = 0;
+   int len;
+   wchar_t *str, *end;
+
+   len = MultiByteToWideChar(CP_ACP, 0, nptr, -1, NULL, 0);
+   str = (wchar_t *)malloc(len * sizeof(wchar_t));
+   if ( NULL != str )
+   {
+      MultiByteToWideChar(CP_ACP, 0, nptr, -1, str, len);
+      result = wcstod(str, &end);
+      len = WideCharToMultiByte(CP_ACP, 0, end, -1, NULL, 0, NULL, NULL);
+      *endptr = (char *)nptr + (png_strlen(nptr) - len + 1);
+      free(str);
+   }
+   return result;
+}
+#  endif
+#endif
+
+#ifndef PNG_READ_BIG_ENDIAN_SUPPORTED
+/* Grab an unsigned 32-bit integer from a buffer in big-endian format. */
+png_uint_32 /* PRIVATE */
+png_get_uint_32(png_bytep buf)
+{
+   png_uint_32 i = ((png_uint_32)(*buf) << 24) +
+      ((png_uint_32)(*(buf + 1)) << 16) +
+      ((png_uint_32)(*(buf + 2)) << 8) +
+      (png_uint_32)(*(buf + 3));
+
+   return (i);
+}
+
+#if defined(PNG_READ_pCAL_SUPPORTED) || defined(PNG_READ_oFFs_SUPPORTED)
+/* Grab a signed 32-bit integer from a buffer in big-endian format.  The
+ * data is stored in the PNG file in two's complement format, and it is
+ * assumed that the machine format for signed integers is the same. */
+png_int_32 /* PRIVATE */
+png_get_int_32(png_bytep buf)
+{
+   png_int_32 i = ((png_int_32)(*buf) << 24) +
+      ((png_int_32)(*(buf + 1)) << 16) +
+      ((png_int_32)(*(buf + 2)) << 8) +
+      (png_int_32)(*(buf + 3));
+
+   return (i);
+}
+#endif /* PNG_READ_pCAL_SUPPORTED */
+
+/* Grab an unsigned 16-bit integer from a buffer in big-endian format. */
+png_uint_16 /* PRIVATE */
+png_get_uint_16(png_bytep buf)
+{
+   png_uint_16 i = (png_uint_16)(((png_uint_16)(*buf) << 8) +
+      (png_uint_16)(*(buf + 1)));
+
+   return (i);
+}
+#endif /* PNG_READ_BIG_ENDIAN_SUPPORTED */
+
+/* Read data, and (optionally) run it through the CRC. */
+void /* PRIVATE */
+png_crc_read(png_structp png_ptr, png_bytep buf, png_size_t length)
+{
+   png_read_data(png_ptr, buf, length);
+   png_calculate_crc(png_ptr, buf, length);
+}
+
+/* Optionally skip data and then check the CRC.  Depending on whether we
+   are reading a ancillary or critical chunk, and how the program has set
+   things up, we may calculate the CRC on the data and print a message.
+   Returns '1' if there was a CRC error, '0' otherwise. */
+int /* PRIVATE */
+png_crc_finish(png_structp png_ptr, png_uint_32 skip)
+{
+   png_size_t i;
+   png_size_t istop = png_ptr->zbuf_size;
+
+   for (i = (png_size_t)skip; i > istop; i -= istop)
+   {
+      png_crc_read(png_ptr, png_ptr->zbuf, png_ptr->zbuf_size);
+   }
+   if (i)
+   {
+      png_crc_read(png_ptr, png_ptr->zbuf, i);
+   }
+
+   if (png_crc_error(png_ptr))
+   {
+      if (((png_ptr->chunk_name[0] & 0x20) &&                /* Ancillary */
+           !(png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_NOWARN)) ||
+          (!(png_ptr->chunk_name[0] & 0x20) &&             /* Critical  */
+          (png_ptr->flags & PNG_FLAG_CRC_CRITICAL_USE)))
+      {
+         png_chunk_warning(png_ptr, "CRC error");
+      }
+      else
+      {
+         png_chunk_error(png_ptr, "CRC error");
+      }
+      return (1);
+   }
+
+   return (0);
+}
+
+/* Compare the CRC stored in the PNG file with that calculated by libpng from
+   the data it has read thus far. */
+int /* PRIVATE */
+png_crc_error(png_structp png_ptr)
+{
+   png_byte crc_bytes[4];
+   png_uint_32 crc;
+   int need_crc = 1;
+
+   if (png_ptr->chunk_name[0] & 0x20)                     /* ancillary */
+   {
+      if ((png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_MASK) ==
+          (PNG_FLAG_CRC_ANCILLARY_USE | PNG_FLAG_CRC_ANCILLARY_NOWARN))
+         need_crc = 0;
+   }
+   else                                                    /* critical */
+   {
+      if (png_ptr->flags & PNG_FLAG_CRC_CRITICAL_IGNORE)
+         need_crc = 0;
+   }
+
+   png_read_data(png_ptr, crc_bytes, 4);
+
+   if (need_crc)
+   {
+      crc = png_get_uint_32(crc_bytes);
+      return ((int)(crc != png_ptr->crc));
+   }
+   else
+      return (0);
+}
+
+#if defined(PNG_READ_zTXt_SUPPORTED) || defined(PNG_READ_iTXt_SUPPORTED) || \
+    defined(PNG_READ_iCCP_SUPPORTED)
+/*
+ * Decompress trailing data in a chunk.  The assumption is that chunkdata
+ * points at an allocated area holding the contents of a chunk with a
+ * trailing compressed part.  What we get back is an allocated area
+ * holding the original prefix part and an uncompressed version of the
+ * trailing part (the malloc area passed in is freed).
+ */
+png_charp /* PRIVATE */
+png_decompress_chunk(png_structp png_ptr, int comp_type,
+                              png_charp chunkdata, png_size_t chunklength,
+                              png_size_t prefix_size, png_size_t *newlength)
+{
+   static char msg[] = "Error decoding compressed text";
+   png_charp text = NULL;
+   png_size_t text_size;
+
+   if (comp_type == PNG_COMPRESSION_TYPE_BASE)
+   {
+      int ret = Z_OK;
+      png_ptr->zstream.next_in = (png_bytep)(chunkdata + prefix_size);
+      png_ptr->zstream.avail_in = (uInt)(chunklength - prefix_size);
+      png_ptr->zstream.next_out = png_ptr->zbuf;
+      png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+
+      text_size = 0;
+      text = NULL;
+
+      while (png_ptr->zstream.avail_in)
+      {
+         ret = inflate(&png_ptr->zstream, Z_PARTIAL_FLUSH);
+         if (ret != Z_OK && ret != Z_STREAM_END)
+         {
+            if (png_ptr->zstream.msg != NULL)
+               png_warning(png_ptr, png_ptr->zstream.msg);
+            else
+               png_warning(png_ptr, msg);
+            inflateReset(&png_ptr->zstream);
+            png_ptr->zstream.avail_in = 0;
+
+            if (text ==  NULL)
+            {
+               text_size = prefix_size + sizeof(msg) + 1;
+               text = (png_charp)png_malloc(png_ptr, text_size);
+               png_memcpy(text, chunkdata, prefix_size);
+            }
+
+            text[text_size - 1] = 0x00;
+
+            /* Copy what we can of the error message into the text chunk */
+            text_size = (png_size_t)(chunklength - (text - chunkdata) - 1);
+            text_size = sizeof(msg) > text_size ? text_size : sizeof(msg);
+            png_memcpy(text + prefix_size, msg, text_size + 1);
+            break;
+         }
+         if (!png_ptr->zstream.avail_out || ret == Z_STREAM_END)
+         {
+            if (text == NULL)
+            {
+               text_size = prefix_size +
+                   png_ptr->zbuf_size - png_ptr->zstream.avail_out;
+               text = (png_charp)png_malloc(png_ptr, text_size + 1);
+               png_memcpy(text + prefix_size, png_ptr->zbuf,
+                    text_size - prefix_size);
+               png_memcpy(text, chunkdata, prefix_size);
+               *(text + text_size) = 0x00;
+            }
+            else
+            {
+               png_charp tmp;
+
+               tmp = text;
+               text = (png_charp)png_malloc(png_ptr, (png_uint_32)(text_size +
+                  png_ptr->zbuf_size - png_ptr->zstream.avail_out + 1));
+               png_memcpy(text, tmp, text_size);
+               png_free(png_ptr, tmp);
+               png_memcpy(text + text_size, png_ptr->zbuf,
+                  (png_ptr->zbuf_size - png_ptr->zstream.avail_out));
+               text_size += png_ptr->zbuf_size - png_ptr->zstream.avail_out;
+               *(text + text_size) = 0x00;
+            }
+            if (ret == Z_STREAM_END)
+               break;
+            else
+            {
+               png_ptr->zstream.next_out = png_ptr->zbuf;
+               png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+            }
+         }
+      }
+      if (ret != Z_STREAM_END)
+      {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+         char umsg[50];
+
+         if (ret == Z_BUF_ERROR)
+            sprintf(umsg,"Buffer error in compressed datastream in %s chunk",
+                png_ptr->chunk_name);
+         else if (ret == Z_DATA_ERROR)
+            sprintf(umsg,"Data error in compressed datastream in %s chunk",
+                png_ptr->chunk_name);
+         else
+            sprintf(umsg,"Incomplete compressed datastream in %s chunk",
+                png_ptr->chunk_name);
+         png_warning(png_ptr, umsg);
+#else
+         png_warning(png_ptr,
+            "Incomplete compressed datastream in chunk other than IDAT");
+#endif
+         text_size=prefix_size;
+         if (text ==  NULL)
+         {
+            text = (png_charp)png_malloc(png_ptr, text_size+1);
+            png_memcpy(text, chunkdata, prefix_size);
+         }
+         *(text + text_size) = 0x00;
+      }
+
+      inflateReset(&png_ptr->zstream);
+      png_ptr->zstream.avail_in = 0;
+
+      png_free(png_ptr, chunkdata);
+      chunkdata = text;
+      *newlength=text_size;
+   }
+   else /* if (comp_type != PNG_COMPRESSION_TYPE_BASE) */
+   {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+      char umsg[50];
+
+      sprintf(umsg, "Unknown zTXt compression type %d", comp_type);
+      png_warning(png_ptr, umsg);
+#else
+      png_warning(png_ptr, "Unknown zTXt compression type");
+#endif
+
+      *(chunkdata + prefix_size) = 0x00;
+      *newlength=prefix_size;
+   }
+
+   return chunkdata;
+}
+#endif
+
+/* read and check the IDHR chunk */
+void /* PRIVATE */
+png_handle_IHDR(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_byte buf[13];
+   png_uint_32 width, height;
+   int bit_depth, color_type, compression_type, filter_type;
+   int interlace_type;
+
+   png_debug(1, "in png_handle_IHDR\n");
+
+   if (png_ptr->mode & PNG_HAVE_IHDR)
+      png_error(png_ptr, "Out of place IHDR");
+
+   /* check the length */
+   if (length != 13)
+      png_error(png_ptr, "Invalid IHDR chunk");
+
+   png_ptr->mode |= PNG_HAVE_IHDR;
+
+   png_crc_read(png_ptr, buf, 13);
+   png_crc_finish(png_ptr, 0);
+
+   width = png_get_uint_32(buf);
+   height = png_get_uint_32(buf + 4);
+   bit_depth = buf[8];
+   color_type = buf[9];
+   compression_type = buf[10];
+   filter_type = buf[11];
+   interlace_type = buf[12];
+
+
+   /* set internal variables */
+   png_ptr->width = width;
+   png_ptr->height = height;
+   png_ptr->bit_depth = (png_byte)bit_depth;
+   png_ptr->interlaced = (png_byte)interlace_type;
+   png_ptr->color_type = (png_byte)color_type;
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   png_ptr->filter_type = (png_byte)filter_type;
+#endif
+
+   /* find number of channels */
+   switch (png_ptr->color_type)
+   {
+      case PNG_COLOR_TYPE_GRAY:
+      case PNG_COLOR_TYPE_PALETTE:
+         png_ptr->channels = 1;
+         break;
+      case PNG_COLOR_TYPE_RGB:
+         png_ptr->channels = 3;
+         break;
+      case PNG_COLOR_TYPE_GRAY_ALPHA:
+         png_ptr->channels = 2;
+         break;
+      case PNG_COLOR_TYPE_RGB_ALPHA:
+         png_ptr->channels = 4;
+         break;
+   }
+
+   /* set up other useful info */
+   png_ptr->pixel_depth = (png_byte)(png_ptr->bit_depth *
+   png_ptr->channels);
+   png_ptr->rowbytes = ((png_ptr->width *
+      (png_uint_32)png_ptr->pixel_depth + 7) >> 3);
+   png_debug1(3,"bit_depth = %d\n", png_ptr->bit_depth);
+   png_debug1(3,"channels = %d\n", png_ptr->channels);
+   png_debug1(3,"rowbytes = %lu\n", png_ptr->rowbytes);
+   png_set_IHDR(png_ptr, info_ptr, width, height, bit_depth,
+      color_type, interlace_type, compression_type, filter_type);
+}
+
+/* read and check the palette */
+void /* PRIVATE */
+png_handle_PLTE(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_color palette[PNG_MAX_PALETTE_LENGTH];
+   int num, i;
+#ifndef PNG_NO_POINTER_INDEXING
+   png_colorp pal_ptr;
+#endif
+
+   png_debug(1, "in png_handle_PLTE\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before PLTE");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid PLTE after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->mode & PNG_HAVE_PLTE)
+      png_error(png_ptr, "Duplicate PLTE chunk");
+
+   png_ptr->mode |= PNG_HAVE_PLTE;
+
+   if (!(png_ptr->color_type&PNG_COLOR_MASK_COLOR))
+   {
+      png_warning(png_ptr,
+        "Ignoring PLTE chunk in grayscale PNG");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+#if !defined(PNG_READ_OPT_PLTE_SUPPORTED)
+   if (png_ptr->color_type != PNG_COLOR_TYPE_PALETTE)
+   {
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+#endif
+
+   if (length > 3*PNG_MAX_PALETTE_LENGTH || length % 3)
+   {
+      if (png_ptr->color_type != PNG_COLOR_TYPE_PALETTE)
+      {
+         png_warning(png_ptr, "Invalid palette chunk");
+         png_crc_finish(png_ptr, length);
+         return;
+      }
+      else
+      {
+         png_error(png_ptr, "Invalid palette chunk");
+      }
+   }
+
+   num = (int)length / 3;
+
+#ifndef PNG_NO_POINTER_INDEXING
+   for (i = 0, pal_ptr = palette; i < num; i++, pal_ptr++)
+   {
+      png_byte buf[3];
+
+      png_crc_read(png_ptr, buf, 3);
+      pal_ptr->red = buf[0];
+      pal_ptr->green = buf[1];
+      pal_ptr->blue = buf[2];
+   }
+#else
+   for (i = 0; i < num; i++)
+   {
+      png_byte buf[3];
+
+      png_crc_read(png_ptr, buf, 3);
+      /* don't depend upon png_color being any order */
+      palette[i].red = buf[0];
+      palette[i].green = buf[1];
+      palette[i].blue = buf[2];
+   }
+#endif
+
+   /* If we actually NEED the PLTE chunk (ie for a paletted image), we do
+      whatever the normal CRC configuration tells us.  However, if we
+      have an RGB image, the PLTE can be considered ancillary, so
+      we will act as though it is. */
+#if !defined(PNG_READ_OPT_PLTE_SUPPORTED)
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+#endif
+   {
+      png_crc_finish(png_ptr, 0);
+   }
+#if !defined(PNG_READ_OPT_PLTE_SUPPORTED)
+   else if (png_crc_error(png_ptr))  /* Only if we have a CRC error */
+   {
+      /* If we don't want to use the data from an ancillary chunk,
+         we have two options: an error abort, or a warning and we
+         ignore the data in this chunk (which should be OK, since
+         it's considered ancillary for a RGB or RGBA image). */
+      if (!(png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_USE))
+      {
+         if (png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_NOWARN)
+         {
+            png_chunk_error(png_ptr, "CRC error");
+         }
+         else
+         {
+            png_chunk_warning(png_ptr, "CRC error");
+            return;
+         }
+      }
+      /* Otherwise, we (optionally) emit a warning and use the chunk. */
+      else if (!(png_ptr->flags & PNG_FLAG_CRC_ANCILLARY_NOWARN))
+      {
+         png_chunk_warning(png_ptr, "CRC error");
+      }
+   }
+#endif
+
+   png_set_PLTE(png_ptr, info_ptr, palette, num);
+
+#if defined(PNG_READ_tRNS_SUPPORTED)
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+   {
+      if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_tRNS))
+      {
+         if (png_ptr->num_trans > (png_uint_16)num)
+         {
+            png_warning(png_ptr, "Truncating incorrect tRNS chunk length");
+            png_ptr->num_trans = (png_uint_16)num;
+         }
+         if (info_ptr->num_trans > (png_uint_16)num)
+         {
+            png_warning(png_ptr, "Truncating incorrect info tRNS chunk length");
+            info_ptr->num_trans = (png_uint_16)num;
+         }
+      }
+   }
+#endif
+
+}
+
+void /* PRIVATE */
+png_handle_IEND(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_debug(1, "in png_handle_IEND\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR) || !(png_ptr->mode & PNG_HAVE_IDAT))
+   {
+      png_error(png_ptr, "No image in file");
+
+      info_ptr = info_ptr; /* quiet compiler warnings about unused info_ptr */
+   }
+
+   png_ptr->mode |= (PNG_AFTER_IDAT | PNG_HAVE_IEND);
+
+   if (length != 0)
+   {
+      png_warning(png_ptr, "Incorrect IEND chunk length");
+   }
+   png_crc_finish(png_ptr, length);
+}
+
+#if defined(PNG_READ_gAMA_SUPPORTED)
+void /* PRIVATE */
+png_handle_gAMA(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_fixed_point igamma;
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   float file_gamma;
+#endif
+   png_byte buf[4];
+
+   png_debug(1, "in png_handle_gAMA\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before gAMA");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid gAMA after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->mode & PNG_HAVE_PLTE)
+      /* Should be an error, but we can cope with it */
+      png_warning(png_ptr, "Out of place gAMA chunk");
+
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_gAMA)
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      && !(info_ptr->valid & PNG_INFO_sRGB)
+#endif
+      )
+   {
+      png_warning(png_ptr, "Duplicate gAMA chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (length != 4)
+   {
+      png_warning(png_ptr, "Incorrect gAMA chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 4);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   igamma = (png_fixed_point)png_get_uint_32(buf);
+   /* check for zero gamma */
+   if (igamma == 0)
+      {
+         png_warning(png_ptr,
+           "Ignoring gAMA chunk with gamma=0");
+         return;
+      }
+
+#if defined(PNG_READ_sRGB_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_sRGB)
+      if(igamma < 45000L || igamma > 46000L)
+      {
+         png_warning(png_ptr,
+           "Ignoring incorrect gAMA value when sRGB is also present");
+#ifndef PNG_NO_CONSOLE_IO
+         fprintf(stderr, "gamma = (%d/100000)\n", (int)igamma);
+#endif
+         return;
+      }
+#endif /* PNG_READ_sRGB_SUPPORTED */
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   file_gamma = (float)igamma / (float)100000.0;
+#  ifdef PNG_READ_GAMMA_SUPPORTED
+     png_ptr->gamma = file_gamma;
+#  endif
+     png_set_gAMA(png_ptr, info_ptr, file_gamma);
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_set_gAMA_fixed(png_ptr, info_ptr, igamma);
+#endif
+}
+#endif
+
+#if defined(PNG_READ_sBIT_SUPPORTED)
+void /* PRIVATE */
+png_handle_sBIT(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_size_t truelen;
+   png_byte buf[4];
+
+   png_debug(1, "in png_handle_sBIT\n");
+
+   buf[0] = buf[1] = buf[2] = buf[3] = 0;
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before sBIT");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid sBIT after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->mode & PNG_HAVE_PLTE)
+   {
+      /* Should be an error, but we can cope with it */
+      png_warning(png_ptr, "Out of place sBIT chunk");
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_sBIT))
+   {
+      png_warning(png_ptr, "Duplicate sBIT chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      truelen = 3;
+   else
+      truelen = (png_size_t)png_ptr->channels;
+
+   if (length != truelen)
+   {
+      png_warning(png_ptr, "Incorrect sBIT chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, truelen);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   if (png_ptr->color_type & PNG_COLOR_MASK_COLOR)
+   {
+      png_ptr->sig_bit.red = buf[0];
+      png_ptr->sig_bit.green = buf[1];
+      png_ptr->sig_bit.blue = buf[2];
+      png_ptr->sig_bit.alpha = buf[3];
+   }
+   else
+   {
+      png_ptr->sig_bit.gray = buf[0];
+      png_ptr->sig_bit.red = buf[0];
+      png_ptr->sig_bit.green = buf[0];
+      png_ptr->sig_bit.blue = buf[0];
+      png_ptr->sig_bit.alpha = buf[1];
+   }
+   png_set_sBIT(png_ptr, info_ptr, &(png_ptr->sig_bit));
+}
+#endif
+
+#if defined(PNG_READ_cHRM_SUPPORTED)
+void /* PRIVATE */
+png_handle_cHRM(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_byte buf[4];
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   float white_x, white_y, red_x, red_y, green_x, green_y, blue_x, blue_y;
+#endif
+   png_fixed_point int_x_white, int_y_white, int_x_red, int_y_red, int_x_green,
+      int_y_green, int_x_blue, int_y_blue;
+
+   png_debug(1, "in png_handle_cHRM\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before cHRM");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid cHRM after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->mode & PNG_HAVE_PLTE)
+      /* Should be an error, but we can cope with it */
+      png_warning(png_ptr, "Missing PLTE before cHRM");
+
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_cHRM)
+#if defined(PNG_READ_sRGB_SUPPORTED)
+      && !(info_ptr->valid & PNG_INFO_sRGB)
+#endif
+      )
+   {
+      png_warning(png_ptr, "Duplicate cHRM chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (length != 32)
+   {
+      png_warning(png_ptr, "Incorrect cHRM chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 4);
+   int_x_white = (png_fixed_point)png_get_uint_32(buf);
+
+   png_crc_read(png_ptr, buf, 4);
+   int_y_white = (png_fixed_point)png_get_uint_32(buf);
+
+   if (int_x_white > 80000L || int_y_white > 80000L ||
+      int_x_white + int_y_white > 100000L)
+   {
+      png_warning(png_ptr, "Invalid cHRM white point");
+      png_crc_finish(png_ptr, 24);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 4);
+   int_x_red = (png_fixed_point)png_get_uint_32(buf);
+
+   png_crc_read(png_ptr, buf, 4);
+   int_y_red = (png_fixed_point)png_get_uint_32(buf);
+
+   if (int_x_red > 80000L || int_y_red > 80000L ||
+      int_x_red + int_y_red > 100000L)
+   {
+      png_warning(png_ptr, "Invalid cHRM red point");
+      png_crc_finish(png_ptr, 16);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 4);
+   int_x_green = (png_fixed_point)png_get_uint_32(buf);
+
+   png_crc_read(png_ptr, buf, 4);
+   int_y_green = (png_fixed_point)png_get_uint_32(buf);
+
+   if (int_x_green > 80000L || int_y_green > 80000L ||
+      int_x_green + int_y_green > 100000L)
+   {
+      png_warning(png_ptr, "Invalid cHRM green point");
+      png_crc_finish(png_ptr, 8);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 4);
+   int_x_blue = (png_fixed_point)png_get_uint_32(buf);
+
+   png_crc_read(png_ptr, buf, 4);
+   int_y_blue = (png_fixed_point)png_get_uint_32(buf);
+
+   if (int_x_blue > 80000L || int_y_blue > 80000L ||
+      int_x_blue + int_y_blue > 100000L)
+   {
+      png_warning(png_ptr, "Invalid cHRM blue point");
+      png_crc_finish(png_ptr, 0);
+      return;
+   }
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   white_x = (float)int_x_white / (float)100000.0;
+   white_y = (float)int_y_white / (float)100000.0;
+   red_x   = (float)int_x_red   / (float)100000.0;
+   red_y   = (float)int_y_red   / (float)100000.0;
+   green_x = (float)int_x_green / (float)100000.0;
+   green_y = (float)int_y_green / (float)100000.0;
+   blue_x  = (float)int_x_blue  / (float)100000.0;
+   blue_y  = (float)int_y_blue  / (float)100000.0;
+#endif
+
+#if defined(PNG_READ_sRGB_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_sRGB)
+      {
+      if (abs(int_x_white - 31270L) > 1000 ||
+          abs(int_y_white - 32900L) > 1000 ||
+          abs(int_x_red   - 64000L) > 1000 ||
+          abs(int_y_red   - 33000L) > 1000 ||
+          abs(int_x_green - 30000L) > 1000 ||
+          abs(int_y_green - 60000L) > 1000 ||
+          abs(int_x_blue  - 15000L) > 1000 ||
+          abs(int_y_blue  -  6000L) > 1000)
+         {
+
+            png_warning(png_ptr,
+              "Ignoring incorrect cHRM value when sRGB is also present");
+#ifndef PNG_NO_CONSOLE_IO
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+            fprintf(stderr,"wx=%f, wy=%f, rx=%f, ry=%f\n",
+               white_x, white_y, red_x, red_y);
+            fprintf(stderr,"gx=%f, gy=%f, bx=%f, by=%f\n",
+               green_x, green_y, blue_x, blue_y);
+#else
+            fprintf(stderr,"wx=%ld, wy=%ld, rx=%ld, ry=%ld\n",
+               int_x_white, int_y_white, int_x_red, int_y_red);
+            fprintf(stderr,"gx=%ld, gy=%ld, bx=%ld, by=%ld\n",
+               int_x_green, int_y_green, int_x_blue, int_y_blue);
+#endif
+#endif /* PNG_NO_CONSOLE_IO */
+         }
+         png_crc_finish(png_ptr, 0);
+         return;
+      }
+#endif /* PNG_READ_sRGB_SUPPORTED */
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   png_set_cHRM(png_ptr, info_ptr,
+      white_x, white_y, red_x, red_y, green_x, green_y, blue_x, blue_y);
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_set_cHRM_fixed(png_ptr, info_ptr,
+      int_x_white, int_y_white, int_x_red, int_y_red, int_x_green,
+      int_y_green, int_x_blue, int_y_blue);
+#endif
+   if (png_crc_finish(png_ptr, 0))
+      return;
+}
+#endif
+
+#if defined(PNG_READ_sRGB_SUPPORTED)
+void /* PRIVATE */
+png_handle_sRGB(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   int intent;
+   png_byte buf[1];
+
+   png_debug(1, "in png_handle_sRGB\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before sRGB");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid sRGB after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->mode & PNG_HAVE_PLTE)
+      /* Should be an error, but we can cope with it */
+      png_warning(png_ptr, "Out of place sRGB chunk");
+
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_sRGB))
+   {
+      png_warning(png_ptr, "Duplicate sRGB chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (length != 1)
+   {
+      png_warning(png_ptr, "Incorrect sRGB chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 1);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   intent = buf[0];
+   /* check for bad intent */
+   if (intent >= PNG_sRGB_INTENT_LAST)
+   {
+      png_warning(png_ptr, "Unknown sRGB intent");
+      return;
+   }
+
+#if defined(PNG_READ_gAMA_SUPPORTED) && defined(PNG_READ_GAMMA_SUPPORTED)
+   if ((info_ptr->valid & PNG_INFO_gAMA))
+   {
+   int igamma;
+#ifdef PNG_FIXED_POINT_SUPPORTED
+      igamma=(int)info_ptr->int_gamma;
+#else
+#  ifdef PNG_FLOATING_POINT_SUPPORTED
+      igamma=(int)(info_ptr->gamma * 100000.);
+#  endif
+#endif
+#if 0 && defined(PNG_cHRM_SUPPORTED) && !defined(PNG_FIXED_POINT_SUPPORTED)
+/* We need to define these here because they aren't in png.h */
+   png_fixed_point int_x_white;
+   png_fixed_point int_y_white;
+   png_fixed_point int_x_red;
+   png_fixed_point int_y_red;
+   png_fixed_point int_x_green;
+   png_fixed_point int_y_green;
+   png_fixed_point int_x_blue;
+   png_fixed_point int_y_blue;
+#endif
+      if(igamma < 45000L || igamma > 46000L)
+      {
+         png_warning(png_ptr,
+           "Ignoring incorrect gAMA value when sRGB is also present");
+#ifndef PNG_NO_CONSOLE_IO
+#  ifdef PNG_FIXED_POINT_SUPPORTED
+         fprintf(stderr,"incorrect gamma=(%d/100000)\n",(int)png_ptr->int_gamma);
+#  else
+#    ifdef PNG_FLOATING_POINT_SUPPORTED
+         fprintf(stderr,"incorrect gamma=%f\n",png_ptr->gamma);
+#    endif
+#  endif
+#endif
+      }
+   }
+#endif /* PNG_READ_gAMA_SUPPORTED */
+
+#ifdef PNG_READ_cHRM_SUPPORTED
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   if (info_ptr->valid & PNG_INFO_cHRM)
+      if (abs(info_ptr->int_x_white - 31270L) > 1000 ||
+          abs(info_ptr->int_y_white - 32900L) > 1000 ||
+          abs(info_ptr->int_x_red   - 64000L) > 1000 ||
+          abs(info_ptr->int_y_red   - 33000L) > 1000 ||
+          abs(info_ptr->int_x_green - 30000L) > 1000 ||
+          abs(info_ptr->int_y_green - 60000L) > 1000 ||
+          abs(info_ptr->int_x_blue  - 15000L) > 1000 ||
+          abs(info_ptr->int_y_blue  -  6000L) > 1000)
+         {
+            png_warning(png_ptr,
+              "Ignoring incorrect cHRM value when sRGB is also present");
+         }
+#endif /* PNG_FIXED_POINT_SUPPORTED */
+#endif /* PNG_READ_cHRM_SUPPORTED */
+
+   png_set_sRGB_gAMA_and_cHRM(png_ptr, info_ptr, intent);
+}
+#endif /* PNG_READ_sRGB_SUPPORTED */
+
+#if defined(PNG_READ_iCCP_SUPPORTED)
+void /* PRIVATE */
+png_handle_iCCP(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+/* Note: this does not properly handle chunks that are > 64K under DOS */
+{
+   png_charp chunkdata;
+   png_byte compression_type;
+   png_charp profile;
+   png_uint_32 skip = 0;
+   png_uint_32 profile_size = 0;
+   png_uint_32 profile_length = 0;
+   png_size_t slength, prefix_length, data_length;
+
+   png_debug(1, "in png_handle_iCCP\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before iCCP");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid iCCP after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->mode & PNG_HAVE_PLTE)
+      /* Should be an error, but we can cope with it */
+      png_warning(png_ptr, "Out of place iCCP chunk");
+
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_iCCP))
+   {
+      png_warning(png_ptr, "Duplicate iCCP chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+#ifdef PNG_MAX_MALLOC_64K
+   if (length > (png_uint_32)65535L)
+   {
+      png_warning(png_ptr, "iCCP chunk too large to fit in memory");
+      skip = length - (png_uint_32)65535L;
+      length = (png_uint_32)65535L;
+   }
+#endif
+
+   chunkdata = (png_charp)png_malloc(png_ptr, length + 1);
+   slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)chunkdata, slength);
+
+   if (png_crc_finish(png_ptr, skip))
+   {
+      png_free(png_ptr, chunkdata);
+      return;
+   }
+
+   chunkdata[slength] = 0x00;
+
+   for (profile = chunkdata; *profile; profile++)
+      /* empty loop to find end of name */ ;
+
+   ++profile;
+
+   /* there should be at least one zero (the compression type byte)
+      following the separator, and we should be on it  */
+   if ( profile >= chunkdata + slength)
+   {
+      png_free(png_ptr, chunkdata);
+      png_warning(png_ptr, "Malformed iCCP chunk");
+      return;
+   }
+
+   /* compression_type should always be zero */
+   compression_type = *profile++;
+   if (compression_type)
+   {
+      png_warning(png_ptr, "Ignoring nonzero compression type in iCCP chunk");
+      compression_type=0x00;  /* Reset it to zero (libpng-1.0.6 through 1.0.8
+                                 wrote nonzero) */
+   }
+
+   prefix_length = profile - chunkdata;
+   chunkdata = png_decompress_chunk(png_ptr, compression_type, chunkdata,
+                                    slength, prefix_length, &data_length);
+
+   profile_length = data_length - prefix_length;
+
+   if ( prefix_length > data_length || profile_length < 4)
+   {
+      png_free(png_ptr, chunkdata);
+      png_warning(png_ptr, "Profile size field missing from iCCP chunk");
+      return;
+   }
+
+   /* Check the profile_size recorded in the first 32 bits of the ICC profile */
+   profile_size = ((*(chunkdata+prefix_length))<<24) |
+                  ((*(chunkdata+prefix_length+1))<<16) |
+                  ((*(chunkdata+prefix_length+2))<< 8) |
+                  ((*(chunkdata+prefix_length+3))    );
+
+   if(profile_size < profile_length)
+      profile_length = profile_size;
+
+   if(profile_size > profile_length)
+   {
+      png_free(png_ptr, chunkdata);
+      png_warning(png_ptr, "Ignoring truncated iCCP profile.\n");
+      return;
+   }
+
+   png_set_iCCP(png_ptr, info_ptr, chunkdata, compression_type,
+                chunkdata + prefix_length, profile_length);
+   png_free(png_ptr, chunkdata);
+}
+#endif /* PNG_READ_iCCP_SUPPORTED */
+
+#if defined(PNG_READ_sPLT_SUPPORTED)
+void /* PRIVATE */
+png_handle_sPLT(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+/* Note: this does not properly handle chunks that are > 64K under DOS */
+{
+   png_bytep chunkdata;
+   png_bytep entry_start;
+   png_sPLT_t new_palette;
+#ifdef PNG_NO_POINTER_INDEXING
+   png_sPLT_entryp pp;
+#endif
+   int data_length, entry_size, i;
+   png_uint_32 skip = 0;
+   png_size_t slength;
+
+   png_debug(1, "in png_handle_sPLT\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before sPLT");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid sPLT after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+#ifdef PNG_MAX_MALLOC_64K
+   if (length > (png_uint_32)65535L)
+   {
+      png_warning(png_ptr, "sPLT chunk too large to fit in memory");
+      skip = length - (png_uint_32)65535L;
+      length = (png_uint_32)65535L;
+   }
+#endif
+
+   chunkdata = (png_bytep)png_malloc(png_ptr, length + 1);
+   slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)chunkdata, slength);
+
+   if (png_crc_finish(png_ptr, skip))
+   {
+      png_free(png_ptr, chunkdata);
+      return;
+   }
+
+   chunkdata[slength] = 0x00;
+
+   for (entry_start = chunkdata; *entry_start; entry_start++)
+      /* empty loop to find end of name */ ;
+   ++entry_start;
+
+   /* a sample depth should follow the separator, and we should be on it  */
+   if (entry_start > chunkdata + slength)
+   {
+      png_free(png_ptr, chunkdata);
+      png_warning(png_ptr, "malformed sPLT chunk");
+      return;
+   }
+
+   new_palette.depth = *entry_start++;
+   entry_size = (new_palette.depth == 8 ? 6 : 10);
+   data_length = (slength - (entry_start - chunkdata));
+
+   /* integrity-check the data length */
+   if (data_length % entry_size)
+   {
+      png_free(png_ptr, chunkdata);
+      png_warning(png_ptr, "sPLT chunk has bad length");
+      return;
+   }
+
+   new_palette.nentries = data_length / entry_size;
+   new_palette.entries = (png_sPLT_entryp)png_malloc(
+       png_ptr, new_palette.nentries * sizeof(png_sPLT_entry));
+
+#ifndef PNG_NO_POINTER_INDEXING
+   for (i = 0; i < new_palette.nentries; i++)
+   {
+      png_sPLT_entryp pp = new_palette.entries + i;
+
+      if (new_palette.depth == 8)
+      {
+          pp->red = *entry_start++;
+          pp->green = *entry_start++;
+          pp->blue = *entry_start++;
+          pp->alpha = *entry_start++;
+      }
+      else
+      {
+          pp->red   = png_get_uint_16(entry_start); entry_start += 2;
+          pp->green = png_get_uint_16(entry_start); entry_start += 2;
+          pp->blue  = png_get_uint_16(entry_start); entry_start += 2;
+          pp->alpha = png_get_uint_16(entry_start); entry_start += 2;
+      }
+      pp->frequency = png_get_uint_16(entry_start); entry_start += 2;
+   }
+#else
+   pp = new_palette.entries;
+   for (i = 0; i < new_palette.nentries; i++)
+   {
+
+      if (new_palette.depth == 8)
+      {
+          pp[i].red   = *entry_start++;
+          pp[i].green = *entry_start++;
+          pp[i].blue  = *entry_start++;
+          pp[i].alpha = *entry_start++;
+      }
+      else
+      {
+          pp[i].red   = png_get_uint_16(entry_start); entry_start += 2;
+          pp[i].green = png_get_uint_16(entry_start); entry_start += 2;
+          pp[i].blue  = png_get_uint_16(entry_start); entry_start += 2;
+          pp[i].alpha = png_get_uint_16(entry_start); entry_start += 2;
+      }
+      pp->frequency = png_get_uint_16(entry_start); entry_start += 2;
+   }
+#endif
+
+   /* discard all chunk data except the name and stash that */
+   new_palette.name = (png_charp)chunkdata;
+
+   png_set_sPLT(png_ptr, info_ptr, &new_palette, 1);
+
+   png_free(png_ptr, chunkdata);
+   png_free(png_ptr, new_palette.entries);
+}
+#endif /* PNG_READ_sPLT_SUPPORTED */
+
+#if defined(PNG_READ_tRNS_SUPPORTED)
+void /* PRIVATE */
+png_handle_tRNS(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_byte	readbuf[PNG_MAX_PALETTE_LENGTH];
+
+   png_debug(1, "in png_handle_tRNS\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before tRNS");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid tRNS after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_tRNS))
+   {
+      png_warning(png_ptr, "Duplicate tRNS chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+   {
+      if (!(png_ptr->mode & PNG_HAVE_PLTE))
+      {
+         /* Should be an error, but we can cope with it */
+         png_warning(png_ptr, "Missing PLTE before tRNS");
+      }
+      else if (length > (png_uint_32)png_ptr->num_palette)
+      {
+         png_warning(png_ptr, "Incorrect tRNS chunk length");
+         png_crc_finish(png_ptr, length);
+         return;
+      }
+      if (length == 0)
+      {
+         png_warning(png_ptr, "Zero length tRNS chunk");
+         png_crc_finish(png_ptr, length);
+         return;
+      }
+
+      png_crc_read(png_ptr, readbuf, (png_size_t)length);
+      png_ptr->num_trans = (png_uint_16)length;
+   }
+   else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB)
+   {
+      png_byte buf[6];
+
+      if (length != 6)
+      {
+         png_warning(png_ptr, "Incorrect tRNS chunk length");
+         png_crc_finish(png_ptr, length);
+         return;
+      }
+
+      png_crc_read(png_ptr, buf, (png_size_t)length);
+      png_ptr->num_trans = 1;
+      png_ptr->trans_values.red = png_get_uint_16(buf);
+      png_ptr->trans_values.green = png_get_uint_16(buf + 2);
+      png_ptr->trans_values.blue = png_get_uint_16(buf + 4);
+   }
+   else if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY)
+   {
+      png_byte buf[6];
+
+      if (length != 2)
+      {
+         png_warning(png_ptr, "Incorrect tRNS chunk length");
+         png_crc_finish(png_ptr, length);
+         return;
+      }
+
+      png_crc_read(png_ptr, buf, 2);
+      png_ptr->num_trans = 1;
+      png_ptr->trans_values.gray = png_get_uint_16(buf);
+   }
+   else
+   {
+      png_warning(png_ptr, "tRNS chunk not allowed with alpha channel");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   png_set_tRNS(png_ptr, info_ptr, readbuf, png_ptr->num_trans,
+      &(png_ptr->trans_values));
+}
+#endif
+
+#if defined(PNG_READ_bKGD_SUPPORTED)
+void /* PRIVATE */
+png_handle_bKGD(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_size_t truelen;
+   png_byte buf[6];
+
+   png_debug(1, "in png_handle_bKGD\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before bKGD");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid bKGD after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE &&
+            !(png_ptr->mode & PNG_HAVE_PLTE))
+   {
+      png_warning(png_ptr, "Missing PLTE before bKGD");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_bKGD))
+   {
+      png_warning(png_ptr, "Duplicate bKGD chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      truelen = 1;
+   else if (png_ptr->color_type & PNG_COLOR_MASK_COLOR)
+      truelen = 6;
+   else
+      truelen = 2;
+
+   if (length != truelen)
+   {
+      png_warning(png_ptr, "Incorrect bKGD chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, truelen);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   /* We convert the index value into RGB components so that we can allow
+    * arbitrary RGB values for background when we have transparency, and
+    * so it is easy to determine the RGB values of the background color
+    * from the info_ptr struct. */
+   if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+   {
+      png_ptr->background.index = buf[0];
+      if(info_ptr->num_palette)
+      {
+          if(buf[0] > info_ptr->num_palette)
+          {
+             png_warning(png_ptr, "Incorrect bKGD chunk index value");
+             return;
+          }
+          png_ptr->background.red =
+             (png_uint_16)png_ptr->palette[buf[0]].red;
+          png_ptr->background.green =
+             (png_uint_16)png_ptr->palette[buf[0]].green;
+          png_ptr->background.blue =
+             (png_uint_16)png_ptr->palette[buf[0]].blue;
+      }
+   }
+   else if (!(png_ptr->color_type & PNG_COLOR_MASK_COLOR)) /* GRAY */
+   {
+      png_ptr->background.red =
+      png_ptr->background.green =
+      png_ptr->background.blue =
+      png_ptr->background.gray = png_get_uint_16(buf);
+   }
+   else
+   {
+      png_ptr->background.red = png_get_uint_16(buf);
+      png_ptr->background.green = png_get_uint_16(buf + 2);
+      png_ptr->background.blue = png_get_uint_16(buf + 4);
+   }
+
+   png_set_bKGD(png_ptr, info_ptr, &(png_ptr->background));
+}
+#endif
+
+#if defined(PNG_READ_hIST_SUPPORTED)
+void /* PRIVATE */
+png_handle_hIST(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   int num, i;
+   png_uint_16	readbuf[PNG_MAX_PALETTE_LENGTH];
+
+   png_debug(1, "in png_handle_hIST\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before hIST");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid hIST after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (!(png_ptr->mode & PNG_HAVE_PLTE))
+   {
+      png_warning(png_ptr, "Missing PLTE before hIST");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_hIST))
+   {
+      png_warning(png_ptr, "Duplicate hIST chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   num = (int)length / 2 ;
+   if (num != png_ptr->num_palette)
+   {
+      png_warning(png_ptr, "Incorrect hIST chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   for (i = 0; i < num; i++)
+   {
+      png_byte buf[2];
+
+      png_crc_read(png_ptr, buf, 2);
+      readbuf[i] = png_get_uint_16(buf);
+   }
+
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   png_set_hIST(png_ptr, info_ptr, readbuf);
+}
+#endif
+
+#if defined(PNG_READ_pHYs_SUPPORTED)
+void /* PRIVATE */
+png_handle_pHYs(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_byte buf[9];
+   png_uint_32 res_x, res_y;
+   int unit_type;
+
+   png_debug(1, "in png_handle_pHYs\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before pHYs");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid pHYs after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_pHYs))
+   {
+      png_warning(png_ptr, "Duplicate pHYs chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (length != 9)
+   {
+      png_warning(png_ptr, "Incorrect pHYs chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 9);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   res_x = png_get_uint_32(buf);
+   res_y = png_get_uint_32(buf + 4);
+   unit_type = buf[8];
+   png_set_pHYs(png_ptr, info_ptr, res_x, res_y, unit_type);
+}
+#endif
+
+#if defined(PNG_READ_oFFs_SUPPORTED)
+void /* PRIVATE */
+png_handle_oFFs(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_byte buf[9];
+   png_int_32 offset_x, offset_y;
+   int unit_type;
+
+   png_debug(1, "in png_handle_oFFs\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before oFFs");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid oFFs after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_oFFs))
+   {
+      png_warning(png_ptr, "Duplicate oFFs chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (length != 9)
+   {
+      png_warning(png_ptr, "Incorrect oFFs chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 9);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   offset_x = png_get_int_32(buf);
+   offset_y = png_get_int_32(buf + 4);
+   unit_type = buf[8];
+   png_set_oFFs(png_ptr, info_ptr, offset_x, offset_y, unit_type);
+}
+#endif
+
+#if defined(PNG_READ_pCAL_SUPPORTED)
+/* read the pCAL chunk (described in the PNG Extensions document) */
+void /* PRIVATE */
+png_handle_pCAL(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_charp purpose;
+   png_int_32 X0, X1;
+   png_byte type, nparams;
+   png_charp buf, units, endptr;
+   png_charpp params;
+   png_size_t slength;
+   int i;
+
+   png_debug(1, "in png_handle_pCAL\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before pCAL");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid pCAL after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_pCAL))
+   {
+      png_warning(png_ptr, "Duplicate pCAL chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_debug1(2, "Allocating and reading pCAL chunk data (%lu bytes)\n",
+      length + 1);
+   purpose = (png_charp)png_malloc(png_ptr, length + 1);
+   slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)purpose, slength);
+
+   if (png_crc_finish(png_ptr, 0))
+   {
+      png_free(png_ptr, purpose);
+      return;
+   }
+
+   purpose[slength] = 0x00; /* null terminate the last string */
+
+   png_debug(3, "Finding end of pCAL purpose string\n");
+   for (buf = purpose; *buf; buf++)
+      /* empty loop */ ;
+
+   endptr = purpose + slength;
+
+   /* We need to have at least 12 bytes after the purpose string
+      in order to get the parameter information. */
+   if (endptr <= buf + 12)
+   {
+      png_warning(png_ptr, "Invalid pCAL data");
+      png_free(png_ptr, purpose);
+      return;
+   }
+
+   png_debug(3, "Reading pCAL X0, X1, type, nparams, and units\n");
+   X0 = png_get_int_32((png_bytep)buf+1);
+   X1 = png_get_int_32((png_bytep)buf+5);
+   type = buf[9];
+   nparams = buf[10];
+   units = buf + 11;
+
+   png_debug(3, "Checking pCAL equation type and number of parameters\n");
+   /* Check that we have the right number of parameters for known
+      equation types. */
+   if ((type == PNG_EQUATION_LINEAR && nparams != 2) ||
+       (type == PNG_EQUATION_BASE_E && nparams != 3) ||
+       (type == PNG_EQUATION_ARBITRARY && nparams != 3) ||
+       (type == PNG_EQUATION_HYPERBOLIC && nparams != 4))
+   {
+      png_warning(png_ptr, "Invalid pCAL parameters for equation type");
+      png_free(png_ptr, purpose);
+      return;
+   }
+   else if (type >= PNG_EQUATION_LAST)
+   {
+      png_warning(png_ptr, "Unrecognized equation type for pCAL chunk");
+   }
+
+   for (buf = units; *buf; buf++)
+      /* Empty loop to move past the units string. */ ;
+
+   png_debug(3, "Allocating pCAL parameters array\n");
+   params = (png_charpp)png_malloc(png_ptr, (png_uint_32)(nparams
+      *sizeof(png_charp))) ;
+
+   /* Get pointers to the start of each parameter string. */
+   for (i = 0; i < (int)nparams; i++)
+   {
+      buf++; /* Skip the null string terminator from previous parameter. */
+
+      png_debug1(3, "Reading pCAL parameter %d\n", i);
+      for (params[i] = buf; *buf != 0x00 && buf <= endptr; buf++)
+         /* Empty loop to move past each parameter string */ ;
+
+      /* Make sure we haven't run out of data yet */
+      if (buf > endptr)
+      {
+         png_warning(png_ptr, "Invalid pCAL data");
+         png_free(png_ptr, purpose);
+         png_free(png_ptr, params);
+         return;
+      }
+   }
+
+   png_set_pCAL(png_ptr, info_ptr, purpose, X0, X1, type, nparams,
+      units, params);
+
+   png_free(png_ptr, purpose);
+   png_free(png_ptr, params);
+}
+#endif
+
+#if defined(PNG_READ_sCAL_SUPPORTED)
+/* read the sCAL chunk */
+void /* PRIVATE */
+png_handle_sCAL(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_charp buffer, ep;
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   double width, height;
+   png_charp vp;
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_charp swidth, sheight;
+#endif
+#endif
+   png_size_t slength;
+
+   png_debug(1, "in png_handle_sCAL\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before sCAL");
+   else if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+      png_warning(png_ptr, "Invalid sCAL after IDAT");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_sCAL))
+   {
+      png_warning(png_ptr, "Duplicate sCAL chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_debug1(2, "Allocating and reading sCAL chunk data (%lu bytes)\n",
+      length + 1);
+   buffer = (png_charp)png_malloc(png_ptr, length + 1);
+   slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)buffer, slength);
+
+   if (png_crc_finish(png_ptr, 0))
+   {
+      png_free(png_ptr, buffer);
+      return;
+   }
+
+   buffer[slength] = 0x00; /* null terminate the last string */
+
+   ep = buffer + 1;        /* skip unit byte */
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   width = strtod(ep, &vp);
+   if (*vp)
+   {
+       png_warning(png_ptr, "malformed width string in sCAL chunk");
+       return;
+   }
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   swidth = (png_charp)png_malloc(png_ptr, png_strlen(ep) + 1);
+   png_memcpy(swidth, ep, (png_size_t)png_strlen(ep));
+#endif
+#endif
+
+   for (ep = buffer; *ep; ep++)
+      /* empty loop */ ;
+   ep++;
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   height = strtod(ep, &vp);
+   if (*vp)
+   {
+       png_warning(png_ptr, "malformed height string in sCAL chunk");
+       return;
+   }
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   sheight = (png_charp)png_malloc(png_ptr, png_strlen(ep) + 1);
+   png_memcpy(sheight, ep, (png_size_t)png_strlen(ep));
+#endif
+#endif
+
+   if (buffer + slength < ep
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+      || width <= 0. || height <= 0.
+#endif
+      )
+   {
+      png_warning(png_ptr, "Invalid sCAL data");
+      png_free(png_ptr, buffer);
+#if defined(PNG_FIXED_POINT_SUPPORTED) && !defined(PNG_FLOATING_POINT_SUPPORTED)
+      png_free(png_ptr, swidth);
+      png_free(png_ptr, sheight);
+#endif
+      return;
+   }
+
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   png_set_sCAL(png_ptr, info_ptr, buffer[0], width, height);
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_set_sCAL_s(png_ptr, info_ptr, buffer[0], swidth, sheight);
+#endif
+#endif
+
+   png_free(png_ptr, buffer);
+#if defined(PNG_FIXED_POINT_SUPPORTED) && !defined(PNG_FLOATING_POINT_SUPPORTED)
+   png_free(png_ptr, swidth);
+   png_free(png_ptr, sheight);
+#endif
+}
+#endif
+
+#if defined(PNG_READ_tIME_SUPPORTED)
+void /* PRIVATE */
+png_handle_tIME(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_byte buf[7];
+   png_time mod_time;
+
+   png_debug(1, "in png_handle_tIME\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Out of place tIME chunk");
+   else if (info_ptr != NULL && (info_ptr->valid & PNG_INFO_tIME))
+   {
+      png_warning(png_ptr, "Duplicate tIME chunk");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   if (png_ptr->mode & PNG_HAVE_IDAT)
+      png_ptr->mode |= PNG_AFTER_IDAT;
+
+   if (length != 7)
+   {
+      png_warning(png_ptr, "Incorrect tIME chunk length");
+      png_crc_finish(png_ptr, length);
+      return;
+   }
+
+   png_crc_read(png_ptr, buf, 7);
+   if (png_crc_finish(png_ptr, 0))
+      return;
+
+   mod_time.second = buf[6];
+   mod_time.minute = buf[5];
+   mod_time.hour = buf[4];
+   mod_time.day = buf[3];
+   mod_time.month = buf[2];
+   mod_time.year = png_get_uint_16(buf);
+
+   png_set_tIME(png_ptr, info_ptr, &mod_time);
+}
+#endif
+
+#if defined(PNG_READ_tEXt_SUPPORTED)
+/* Note: this does not properly handle chunks that are > 64K under DOS */
+void /* PRIVATE */
+png_handle_tEXt(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_textp text_ptr;
+   png_charp key;
+   png_charp text;
+   png_uint_32 skip = 0;
+   png_size_t slength;
+
+   png_debug(1, "in png_handle_tEXt\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before tEXt");
+
+   if (png_ptr->mode & PNG_HAVE_IDAT)
+      png_ptr->mode |= PNG_AFTER_IDAT;
+
+#ifdef PNG_MAX_MALLOC_64K
+   if (length > (png_uint_32)65535L)
+   {
+      png_warning(png_ptr, "tEXt chunk too large to fit in memory");
+      skip = length - (png_uint_32)65535L;
+      length = (png_uint_32)65535L;
+   }
+#endif
+
+   key = (png_charp)png_malloc(png_ptr, length + 1);
+   slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)key, slength);
+
+   if (png_crc_finish(png_ptr, skip))
+   {
+      png_free(png_ptr, key);
+      return;
+   }
+
+   key[slength] = 0x00;
+
+   for (text = key; *text; text++)
+      /* empty loop to find end of key */ ;
+
+   if (text != key + slength)
+      text++;
+
+   text_ptr = (png_textp)png_malloc(png_ptr, (png_uint_32)sizeof(png_text));
+   text_ptr->compression = PNG_TEXT_COMPRESSION_NONE;
+   text_ptr->key = key;
+#ifdef PNG_iTXt_SUPPORTED
+   text_ptr->lang = NULL;
+   text_ptr->lang_key = NULL;
+   text_ptr->itxt_length = 0;
+#endif
+   text_ptr->text = text;
+   text_ptr->text_length = png_strlen(text);
+
+   png_set_text(png_ptr, info_ptr, text_ptr, 1);
+
+   png_free(png_ptr, key);
+   png_free(png_ptr, text_ptr);
+}
+#endif
+
+#if defined(PNG_READ_zTXt_SUPPORTED)
+/* note: this does not correctly handle chunks that are > 64K under DOS */
+void /* PRIVATE */
+png_handle_zTXt(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_textp text_ptr;
+   png_charp chunkdata;
+   png_charp text;
+   int comp_type;
+   png_size_t slength, prefix_len, data_len;
+
+   png_debug(1, "in png_handle_zTXt\n");
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before zTXt");
+
+   if (png_ptr->mode & PNG_HAVE_IDAT)
+      png_ptr->mode |= PNG_AFTER_IDAT;
+
+#ifdef PNG_MAX_MALLOC_64K
+   /* We will no doubt have problems with chunks even half this size, but
+      there is no hard and fast rule to tell us where to stop. */
+   if (length > (png_uint_32)65535L)
+   {
+     png_warning(png_ptr,"zTXt chunk too large to fit in memory");
+     png_crc_finish(png_ptr, length);
+     return;
+   }
+#endif
+
+   chunkdata = (png_charp)png_malloc(png_ptr, length + 1);
+     slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)chunkdata, slength);
+   if (png_crc_finish(png_ptr, 0))
+   {
+      png_free(png_ptr, chunkdata);
+      return;
+   }
+
+   chunkdata[slength] = 0x00;
+
+   for (text = chunkdata; *text; text++)
+      /* empty loop */ ;
+
+   /* zTXt must have some text after the chunkdataword */
+   if (text == chunkdata + slength)
+   {
+      comp_type = PNG_TEXT_COMPRESSION_NONE;
+      png_warning(png_ptr, "Zero length zTXt chunk");
+   }
+   else
+   {
+       comp_type = *(++text);
+       if (comp_type != PNG_TEXT_COMPRESSION_zTXt)
+       {
+          png_warning(png_ptr, "Unknown compression type in zTXt chunk");
+          comp_type = PNG_TEXT_COMPRESSION_zTXt;
+       }
+       text++;        /* skip the compression_method byte */
+   }
+   prefix_len = text - chunkdata;
+
+   chunkdata = (png_charp)png_decompress_chunk(png_ptr, comp_type, chunkdata,
+                                    (png_size_t)length, prefix_len, &data_len);
+
+   text_ptr = (png_textp)png_malloc(png_ptr, (png_uint_32)sizeof(png_text));
+   text_ptr->compression = comp_type;
+   text_ptr->key = chunkdata;
+#ifdef PNG_iTXt_SUPPORTED
+   text_ptr->lang = NULL;
+   text_ptr->lang_key = NULL;
+   text_ptr->itxt_length = 0;
+#endif
+   text_ptr->text = chunkdata + prefix_len;
+   text_ptr->text_length = data_len;
+
+   png_set_text(png_ptr, info_ptr, text_ptr, 1);
+
+   png_free(png_ptr, text_ptr);
+   png_free(png_ptr, chunkdata);
+}
+#endif
+
+#if defined(PNG_READ_iTXt_SUPPORTED)
+/* note: this does not correctly handle chunks that are > 64K under DOS */
+void /* PRIVATE */
+png_handle_iTXt(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_textp text_ptr;
+   png_charp chunkdata;
+   png_charp key, lang, text, lang_key;
+   int comp_flag;
+   int comp_type = 0;
+   png_size_t slength, prefix_len, data_len;
+
+   png_debug(1, "in png_handle_iTXt\n");
+
+   if (!(png_ptr->mode & PNG_HAVE_IHDR))
+      png_error(png_ptr, "Missing IHDR before iTXt");
+
+   if (png_ptr->mode & PNG_HAVE_IDAT)
+      png_ptr->mode |= PNG_AFTER_IDAT;
+
+#ifdef PNG_MAX_MALLOC_64K
+   /* We will no doubt have problems with chunks even half this size, but
+      there is no hard and fast rule to tell us where to stop. */
+   if (length > (png_uint_32)65535L)
+   {
+     png_warning(png_ptr,"iTXt chunk too large to fit in memory");
+     png_crc_finish(png_ptr, length);
+     return;
+   }
+#endif
+
+   chunkdata = (png_charp)png_malloc(png_ptr, length + 1);
+   slength = (png_size_t)length;
+   png_crc_read(png_ptr, (png_bytep)chunkdata, slength);
+   if (png_crc_finish(png_ptr, 0))
+   {
+      png_free(png_ptr, chunkdata);
+      return;
+   }
+
+   chunkdata[slength] = 0x00;
+
+   for (lang = chunkdata; *lang; lang++)
+      /* empty loop */ ;
+   lang++;        /* skip NUL separator */
+
+   /* iTXt must have a language tag (possibly empty), two compression bytes,
+      translated keyword (possibly empty), and possibly some text after the
+      keyword */
+
+   if (lang >= chunkdata + slength)
+   {
+      comp_flag = PNG_TEXT_COMPRESSION_NONE;
+      png_warning(png_ptr, "Zero length iTXt chunk");
+   }
+   else
+   {
+       comp_flag = *lang++;
+       comp_type = *lang++;
+   }
+
+   for (lang_key = lang; *lang_key; lang_key++)
+      /* empty loop */ ;
+   lang_key++;        /* skip NUL separator */
+
+   for (text = lang_key; *text; text++)
+      /* empty loop */ ;
+   text++;        /* skip NUL separator */
+
+   prefix_len = text - chunkdata;
+
+   key=chunkdata;
+   if (comp_flag)
+       chunkdata = png_decompress_chunk(png_ptr, comp_type, chunkdata,
+          (size_t)length, prefix_len, &data_len);
+   else
+       data_len=png_strlen(chunkdata + prefix_len);
+   text_ptr = (png_textp)png_malloc(png_ptr, (png_uint_32)sizeof(png_text));
+   text_ptr->compression = (int)comp_flag + 1;
+   text_ptr->lang_key = chunkdata+(lang_key-key);
+   text_ptr->lang = chunkdata+(lang-key);
+   text_ptr->itxt_length = data_len;
+   text_ptr->text_length = 0;
+   text_ptr->key = chunkdata;
+   text_ptr->text = chunkdata + prefix_len;
+
+   png_set_text(png_ptr, info_ptr, text_ptr, 1);
+
+   png_free(png_ptr, text_ptr);
+   png_free(png_ptr, chunkdata);
+}
+#endif
+
+/* This function is called when we haven't found a handler for a
+   chunk.  If there isn't a problem with the chunk itself (ie bad
+   chunk name, CRC, or a critical chunk), the chunk is silently ignored
+   -- unless the PNG_FLAG_UNKNOWN_CHUNKS_SUPPORTED flag is on in which
+   case it will be saved away to be written out later. */
+void /* PRIVATE */
+png_handle_unknown(png_structp png_ptr, png_infop info_ptr, png_uint_32 length)
+{
+   png_uint_32 skip = 0;
+
+   png_debug(1, "in png_handle_unknown\n");
+
+   if (png_ptr->mode & PNG_HAVE_IDAT)
+   {
+#ifdef PNG_USE_LOCAL_ARRAYS
+      PNG_IDAT;
+#endif
+      if (png_memcmp(png_ptr->chunk_name, png_IDAT, 4))  /* not an IDAT */
+         png_ptr->mode |= PNG_AFTER_IDAT;
+   }
+
+   png_check_chunk_name(png_ptr, png_ptr->chunk_name);
+
+   if (!(png_ptr->chunk_name[0] & 0x20))
+   {
+#if defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED)
+      if(png_handle_as_unknown(png_ptr, png_ptr->chunk_name) !=
+           HANDLE_CHUNK_ALWAYS
+#if defined(PNG_READ_USER_CHUNKS_SUPPORTED)
+           && png_ptr->read_user_chunk_fn == NULL
+#endif
+        )
+#endif
+          png_chunk_error(png_ptr, "unknown critical chunk");
+   }
+
+#if defined(PNG_READ_UNKNOWN_CHUNKS_SUPPORTED)
+   if (png_ptr->flags & PNG_FLAG_KEEP_UNKNOWN_CHUNKS)
+   {
+       png_unknown_chunk chunk;
+
+#ifdef PNG_MAX_MALLOC_64K
+       if (length > (png_uint_32)65535L)
+       {
+           png_warning(png_ptr, "unknown chunk too large to fit in memory");
+           skip = length - (png_uint_32)65535L;
+           length = (png_uint_32)65535L;
+       }
+#endif
+       png_strcpy((png_charp)chunk.name, (png_charp)png_ptr->chunk_name);
+       chunk.data = (png_bytep)png_malloc(png_ptr, length);
+       chunk.size = (png_size_t)length;
+       png_crc_read(png_ptr, (png_bytep)chunk.data, length);
+#if defined(PNG_READ_USER_CHUNKS_SUPPORTED)
+       if(png_ptr->read_user_chunk_fn != NULL)
+       {
+          /* callback to user unknown chunk handler */
+          if ((*(png_ptr->read_user_chunk_fn)) (png_ptr, &chunk) <= 0)
+          {
+             if (!(png_ptr->chunk_name[0] & 0x20))
+                if(png_handle_as_unknown(png_ptr, png_ptr->chunk_name) !=
+                     HANDLE_CHUNK_ALWAYS)
+                   png_chunk_error(png_ptr, "unknown critical chunk");
+             png_set_unknown_chunks(png_ptr, info_ptr, &chunk, 1);
+          }
+       }
+       else
+#endif
+          png_set_unknown_chunks(png_ptr, info_ptr, &chunk, 1);
+       png_free(png_ptr, chunk.data);
+   }
+   else
+#endif
+      skip = length;
+
+   png_crc_finish(png_ptr, skip);
+
+#if !defined(PNG_READ_USER_CHUNKS_SUPPORTED)
+   info_ptr = info_ptr; /* quiet compiler warnings about unused info_ptr */
+#endif
+}
+
+/* This function is called to verify that a chunk name is valid.
+   This function can't have the "critical chunk check" incorporated
+   into it, since in the future we will need to be able to call user
+   functions to handle unknown critical chunks after we check that
+   the chunk name itself is valid. */
+
+#define isnonalpha(c) ((c) < 41 || (c) > 122 || ((c) > 90 && (c) < 97))
+
+void /* PRIVATE */
+png_check_chunk_name(png_structp png_ptr, png_bytep chunk_name)
+{
+   png_debug(1, "in png_check_chunk_name\n");
+   if (isnonalpha(chunk_name[0]) || isnonalpha(chunk_name[1]) ||
+       isnonalpha(chunk_name[2]) || isnonalpha(chunk_name[3]))
+   {
+      png_chunk_error(png_ptr, "invalid chunk type");
+   }
+}
+
+/* Combines the row recently read in with the existing pixels in the
+   row.  This routine takes care of alpha and transparency if requested.
+   This routine also handles the two methods of progressive display
+   of interlaced images, depending on the mask value.
+   The mask value describes which pixels are to be combined with
+   the row.  The pattern always repeats every 8 pixels, so just 8
+   bits are needed.  A one indicates the pixel is to be combined,
+   a zero indicates the pixel is to be skipped.  This is in addition
+   to any alpha or transparency value associated with the pixel.  If
+   you want all pixels to be combined, pass 0xff (255) in mask.  */
+#ifndef PNG_HAVE_ASSEMBLER_COMBINE_ROW
+void /* PRIVATE */
+png_combine_row(png_structp png_ptr, png_bytep row, int mask)
+{
+   png_debug(1,"in png_combine_row\n");
+   if (mask == 0xff)
+   {
+      png_memcpy(row, png_ptr->row_buf + 1,
+         (png_size_t)((png_ptr->width *
+         png_ptr->row_info.pixel_depth + 7) >> 3));
+   }
+   else
+   {
+      switch (png_ptr->row_info.pixel_depth)
+      {
+         case 1:
+         {
+            png_bytep sp = png_ptr->row_buf + 1;
+            png_bytep dp = row;
+            int s_inc, s_start, s_end;
+            int m = 0x80;
+            int shift;
+            png_uint_32 i;
+            png_uint_32 row_width = png_ptr->width;
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+            if (png_ptr->transformations & PNG_PACKSWAP)
+            {
+                s_start = 0;
+                s_end = 7;
+                s_inc = 1;
+            }
+            else
+#endif
+            {
+                s_start = 7;
+                s_end = 0;
+                s_inc = -1;
+            }
+
+            shift = s_start;
+
+            for (i = 0; i < row_width; i++)
+            {
+               if (m & mask)
+               {
+                  int value;
+
+                  value = (*sp >> shift) & 0x01;
+                  *dp &= (png_byte)((0x7f7f >> (7 - shift)) & 0xff);
+                  *dp |= (png_byte)(value << shift);
+               }
+
+               if (shift == s_end)
+               {
+                  shift = s_start;
+                  sp++;
+                  dp++;
+               }
+               else
+                  shift += s_inc;
+
+               if (m == 1)
+                  m = 0x80;
+               else
+                  m >>= 1;
+            }
+            break;
+         }
+         case 2:
+         {
+            png_bytep sp = png_ptr->row_buf + 1;
+            png_bytep dp = row;
+            int s_start, s_end, s_inc;
+            int m = 0x80;
+            int shift;
+            png_uint_32 i;
+            png_uint_32 row_width = png_ptr->width;
+            int value;
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+            if (png_ptr->transformations & PNG_PACKSWAP)
+            {
+               s_start = 0;
+               s_end = 6;
+               s_inc = 2;
+            }
+            else
+#endif
+            {
+               s_start = 6;
+               s_end = 0;
+               s_inc = -2;
+            }
+
+            shift = s_start;
+
+            for (i = 0; i < row_width; i++)
+            {
+               if (m & mask)
+               {
+                  value = (*sp >> shift) & 0x03;
+                  *dp &= (png_byte)((0x3f3f >> (6 - shift)) & 0xff);
+                  *dp |= (png_byte)(value << shift);
+               }
+
+               if (shift == s_end)
+               {
+                  shift = s_start;
+                  sp++;
+                  dp++;
+               }
+               else
+                  shift += s_inc;
+               if (m == 1)
+                  m = 0x80;
+               else
+                  m >>= 1;
+            }
+            break;
+         }
+         case 4:
+         {
+            png_bytep sp = png_ptr->row_buf + 1;
+            png_bytep dp = row;
+            int s_start, s_end, s_inc;
+            int m = 0x80;
+            int shift;
+            png_uint_32 i;
+            png_uint_32 row_width = png_ptr->width;
+            int value;
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+            if (png_ptr->transformations & PNG_PACKSWAP)
+            {
+               s_start = 0;
+               s_end = 4;
+               s_inc = 4;
+            }
+            else
+#endif
+            {
+               s_start = 4;
+               s_end = 0;
+               s_inc = -4;
+            }
+            shift = s_start;
+
+            for (i = 0; i < row_width; i++)
+            {
+               if (m & mask)
+               {
+                  value = (*sp >> shift) & 0xf;
+                  *dp &= (png_byte)((0xf0f >> (4 - shift)) & 0xff);
+                  *dp |= (png_byte)(value << shift);
+               }
+
+               if (shift == s_end)
+               {
+                  shift = s_start;
+                  sp++;
+                  dp++;
+               }
+               else
+                  shift += s_inc;
+               if (m == 1)
+                  m = 0x80;
+               else
+                  m >>= 1;
+            }
+            break;
+         }
+         default:
+         {
+            png_bytep sp = png_ptr->row_buf + 1;
+            png_bytep dp = row;
+            png_size_t pixel_bytes = (png_ptr->row_info.pixel_depth >> 3);
+            png_uint_32 i;
+            png_uint_32 row_width = png_ptr->width;
+            png_byte m = 0x80;
+
+
+            for (i = 0; i < row_width; i++)
+            {
+               if (m & mask)
+               {
+                  png_memcpy(dp, sp, pixel_bytes);
+               }
+
+               sp += pixel_bytes;
+               dp += pixel_bytes;
+
+               if (m == 1)
+                  m = 0x80;
+               else
+                  m >>= 1;
+            }
+            break;
+         }
+      }
+   }
+}
+#endif /* !PNG_HAVE_ASSEMBLER_COMBINE_ROW */
+
+#ifdef PNG_READ_INTERLACING_SUPPORTED
+#ifndef PNG_HAVE_ASSEMBLER_READ_INTERLACE   /* else in pngvcrd.c, pnggccrd.c */
+/* OLD pre-1.0.9 interface:
+void png_do_read_interlace(png_row_infop row_info, png_bytep row, int pass,
+   png_uint_32 transformations)
+ */
+void /* PRIVATE */
+png_do_read_interlace(png_structp png_ptr)
+{
+   png_row_infop row_info = &(png_ptr->row_info);
+   png_bytep row = png_ptr->row_buf + 1;
+   int pass = png_ptr->pass;
+   png_uint_32 transformations = png_ptr->transformations;
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+   /* offset to next interlace block */
+   const int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1};
+#endif
+
+   png_debug(1,"in png_do_read_interlace (stock C version)\n");
+   if (row != NULL && row_info != NULL)
+   {
+      png_uint_32 final_width;
+
+      final_width = row_info->width * png_pass_inc[pass];
+
+      switch (row_info->pixel_depth)
+      {
+         case 1:
+         {
+            png_bytep sp = row + (png_size_t)((row_info->width - 1) >> 3);
+            png_bytep dp = row + (png_size_t)((final_width - 1) >> 3);
+            int sshift, dshift;
+            int s_start, s_end, s_inc;
+            int jstop = png_pass_inc[pass];
+            png_byte v;
+            png_uint_32 i;
+            int j;
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+            if (transformations & PNG_PACKSWAP)
+            {
+                sshift = (int)((row_info->width + 7) & 0x07);
+                dshift = (int)((final_width + 7) & 0x07);
+                s_start = 7;
+                s_end = 0;
+                s_inc = -1;
+            }
+            else
+#endif
+            {
+                sshift = 7 - (int)((row_info->width + 7) & 0x07);
+                dshift = 7 - (int)((final_width + 7) & 0x07);
+                s_start = 0;
+                s_end = 7;
+                s_inc = 1;
+            }
+
+            for (i = 0; i < row_info->width; i++)
+            {
+               v = (png_byte)((*sp >> sshift) & 0x01);
+               for (j = 0; j < jstop; j++)
+               {
+                  *dp &= (png_byte)((0x7f7f >> (7 - dshift)) & 0xff);
+                  *dp |= (png_byte)(v << dshift);
+                  if (dshift == s_end)
+                  {
+                     dshift = s_start;
+                     dp--;
+                  }
+                  else
+                     dshift += s_inc;
+               }
+               if (sshift == s_end)
+               {
+                  sshift = s_start;
+                  sp--;
+               }
+               else
+                  sshift += s_inc;
+            }
+            break;
+         }
+         case 2:
+         {
+            png_bytep sp = row + (png_uint_32)((row_info->width - 1) >> 2);
+            png_bytep dp = row + (png_uint_32)((final_width - 1) >> 2);
+            int sshift, dshift;
+            int s_start, s_end, s_inc;
+            int jstop = png_pass_inc[pass];
+            png_uint_32 i;
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+            if (transformations & PNG_PACKSWAP)
+            {
+               sshift = (int)(((row_info->width + 3) & 0x03) << 1);
+               dshift = (int)(((final_width + 3) & 0x03) << 1);
+               s_start = 6;
+               s_end = 0;
+               s_inc = -2;
+            }
+            else
+#endif
+            {
+               sshift = (int)((3 - ((row_info->width + 3) & 0x03)) << 1);
+               dshift = (int)((3 - ((final_width + 3) & 0x03)) << 1);
+               s_start = 0;
+               s_end = 6;
+               s_inc = 2;
+            }
+
+            for (i = 0; i < row_info->width; i++)
+            {
+               png_byte v;
+               int j;
+
+               v = (png_byte)((*sp >> sshift) & 0x03);
+               for (j = 0; j < jstop; j++)
+               {
+                  *dp &= (png_byte)((0x3f3f >> (6 - dshift)) & 0xff);
+                  *dp |= (png_byte)(v << dshift);
+                  if (dshift == s_end)
+                  {
+                     dshift = s_start;
+                     dp--;
+                  }
+                  else
+                     dshift += s_inc;
+               }
+               if (sshift == s_end)
+               {
+                  sshift = s_start;
+                  sp--;
+               }
+               else
+                  sshift += s_inc;
+            }
+            break;
+         }
+         case 4:
+         {
+            png_bytep sp = row + (png_size_t)((row_info->width - 1) >> 1);
+            png_bytep dp = row + (png_size_t)((final_width - 1) >> 1);
+            int sshift, dshift;
+            int s_start, s_end, s_inc;
+            png_uint_32 i;
+            int jstop = png_pass_inc[pass];
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)
+            if (transformations & PNG_PACKSWAP)
+            {
+               sshift = (int)(((row_info->width + 1) & 0x01) << 2);
+               dshift = (int)(((final_width + 1) & 0x01) << 2);
+               s_start = 4;
+               s_end = 0;
+               s_inc = -4;
+            }
+            else
+#endif
+            {
+               sshift = (int)((1 - ((row_info->width + 1) & 0x01)) << 2);
+               dshift = (int)((1 - ((final_width + 1) & 0x01)) << 2);
+               s_start = 0;
+               s_end = 4;
+               s_inc = 4;
+            }
+
+            for (i = 0; i < row_info->width; i++)
+            {
+               png_byte v = (png_byte)((*sp >> sshift) & 0xf);
+               int j;
+
+               for (j = 0; j < jstop; j++)
+               {
+                  *dp &= (png_byte)((0xf0f >> (4 - dshift)) & 0xff);
+                  *dp |= (png_byte)(v << dshift);
+                  if (dshift == s_end)
+                  {
+                     dshift = s_start;
+                     dp--;
+                  }
+                  else
+                     dshift += s_inc;
+               }
+               if (sshift == s_end)
+               {
+                  sshift = s_start;
+                  sp--;
+               }
+               else
+                  sshift += s_inc;
+            }
+            break;
+         }
+         default:
+         {
+            png_size_t pixel_bytes = (row_info->pixel_depth >> 3);
+            png_bytep sp = row + (png_size_t)(row_info->width - 1) * pixel_bytes;
+            png_bytep dp = row + (png_size_t)(final_width - 1) * pixel_bytes;
+
+            int jstop = png_pass_inc[pass];
+            png_uint_32 i;
+
+            for (i = 0; i < row_info->width; i++)
+            {
+               png_byte v[8];
+               int j;
+
+               png_memcpy(v, sp, pixel_bytes);
+               for (j = 0; j < jstop; j++)
+               {
+                  png_memcpy(dp, v, pixel_bytes);
+                  dp -= pixel_bytes;
+               }
+               sp -= pixel_bytes;
+            }
+            break;
+         }
+      }
+      row_info->width = final_width;
+      row_info->rowbytes = ((final_width *
+         (png_uint_32)row_info->pixel_depth + 7) >> 3);
+   }
+#if !defined(PNG_READ_PACKSWAP_SUPPORTED)
+   transformations = transformations; /* silence compiler warning */
+#endif
+}
+#endif /* !PNG_HAVE_ASSEMBLER_READ_INTERLACE */
+#endif /* PNG_READ_INTERLACING_SUPPORTED */
+
+#ifndef PNG_HAVE_ASSEMBLER_READ_FILTER_ROW
+void /* PRIVATE */
+png_read_filter_row(png_structp png_ptr, png_row_infop row_info, png_bytep row,
+   png_bytep prev_row, int filter)
+{
+   png_debug(1, "in png_read_filter_row\n");
+   png_debug2(2,"row = %lu, filter = %d\n", png_ptr->row_number, filter);
+   switch (filter)
+   {
+      case PNG_FILTER_VALUE_NONE:
+         break;
+      case PNG_FILTER_VALUE_SUB:
+      {
+         png_uint_32 i;
+         png_uint_32 istop = row_info->rowbytes;
+         png_uint_32 bpp = (row_info->pixel_depth + 7) >> 3;
+         png_bytep rp = row + bpp;
+         png_bytep lp = row;
+
+         for (i = bpp; i < istop; i++)
+         {
+            *rp = (png_byte)(((int)(*rp) + (int)(*lp++)) & 0xff);
+            rp++;
+         }
+         break;
+      }
+      case PNG_FILTER_VALUE_UP:
+      {
+         png_uint_32 i;
+         png_uint_32 istop = row_info->rowbytes;
+         png_bytep rp = row;
+         png_bytep pp = prev_row;
+
+         for (i = 0; i < istop; i++)
+         {
+            *rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff);
+            rp++;
+         }
+         break;
+      }
+      case PNG_FILTER_VALUE_AVG:
+      {
+         png_uint_32 i;
+         png_bytep rp = row;
+         png_bytep pp = prev_row;
+         png_bytep lp = row;
+         png_uint_32 bpp = (row_info->pixel_depth + 7) >> 3;
+         png_uint_32 istop = row_info->rowbytes - bpp;
+
+         for (i = 0; i < bpp; i++)
+         {
+            *rp = (png_byte)(((int)(*rp) +
+               ((int)(*pp++) / 2 )) & 0xff);
+            rp++;
+         }
+
+         for (i = 0; i < istop; i++)
+         {
+            *rp = (png_byte)(((int)(*rp) +
+               (int)(*pp++ + *lp++) / 2 ) & 0xff);
+            rp++;
+         }
+         break;
+      }
+      case PNG_FILTER_VALUE_PAETH:
+      {
+         png_uint_32 i;
+         png_bytep rp = row;
+         png_bytep pp = prev_row;
+         png_bytep lp = row;
+         png_bytep cp = prev_row;
+         png_uint_32 bpp = (row_info->pixel_depth + 7) >> 3;
+         png_uint_32 istop=row_info->rowbytes - bpp;
+
+         for (i = 0; i < bpp; i++)
+         {
+            *rp = (png_byte)(((int)(*rp) + (int)(*pp++)) & 0xff);
+            rp++;
+         }
+
+         for (i = 0; i < istop; i++)   /* use leftover rp,pp */
+         {
+            int a, b, c, pa, pb, pc, p;
+
+            a = *lp++;
+            b = *pp++;
+            c = *cp++;
+
+            p = b - c;
+            pc = a - c;
+
+#ifdef PNG_USE_ABS
+            pa = abs(p);
+            pb = abs(pc);
+            pc = abs(p + pc);
+#else
+            pa = p < 0 ? -p : p;
+            pb = pc < 0 ? -pc : pc;
+            pc = (p + pc) < 0 ? -(p + pc) : p + pc;
+#endif
+
+            /*
+               if (pa <= pb && pa <= pc)
+                  p = a;
+               else if (pb <= pc)
+                  p = b;
+               else
+                  p = c;
+             */
+
+            p = (pa <= pb && pa <=pc) ? a : (pb <= pc) ? b : c;
+
+            *rp = (png_byte)(((int)(*rp) + p) & 0xff);
+            rp++;
+         }
+         break;
+      }
+      default:
+         png_warning(png_ptr, "Ignoring bad adaptive filter type");
+         *row=0;
+         break;
+   }
+}
+#endif /* !PNG_HAVE_ASSEMBLER_READ_FILTER_ROW */
+
+void /* PRIVATE */
+png_read_finish_row(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+   /* start of interlace block */
+   const int png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0};
+
+   /* offset to next interlace block */
+   const int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1};
+
+   /* start of interlace block in the y direction */
+   const int png_pass_ystart[7] = {0, 0, 4, 0, 2, 0, 1};
+
+   /* offset to next interlace block in the y direction */
+   const int png_pass_yinc[7] = {8, 8, 8, 4, 4, 2, 2};
+#endif
+
+   png_debug(1, "in png_read_finish_row\n");
+   png_ptr->row_number++;
+   if (png_ptr->row_number < png_ptr->num_rows)
+      return;
+
+   if (png_ptr->interlaced)
+   {
+      png_ptr->row_number = 0;
+      png_memset_check(png_ptr, png_ptr->prev_row, 0, png_ptr->rowbytes + 1);
+      do
+      {
+         png_ptr->pass++;
+         if (png_ptr->pass >= 7)
+            break;
+         png_ptr->iwidth = (png_ptr->width +
+            png_pass_inc[png_ptr->pass] - 1 -
+            png_pass_start[png_ptr->pass]) /
+            png_pass_inc[png_ptr->pass];
+            png_ptr->irowbytes = ((png_ptr->iwidth *
+               (png_uint_32)png_ptr->pixel_depth + 7) >> 3) +1;
+
+         if (!(png_ptr->transformations & PNG_INTERLACE))
+         {
+            png_ptr->num_rows = (png_ptr->height +
+               png_pass_yinc[png_ptr->pass] - 1 -
+               png_pass_ystart[png_ptr->pass]) /
+               png_pass_yinc[png_ptr->pass];
+            if (!(png_ptr->num_rows))
+               continue;
+         }
+         else  /* if (png_ptr->transformations & PNG_INTERLACE) */
+            break;
+      } while (png_ptr->iwidth == 0);
+
+      if (png_ptr->pass < 7)
+         return;
+   }
+
+   if (!(png_ptr->flags & PNG_FLAG_ZLIB_FINISHED))
+   {
+#ifdef PNG_USE_LOCAL_ARRAYS
+      PNG_IDAT;
+#endif
+      char extra;
+      int ret;
+
+      png_ptr->zstream.next_out = (Byte *)&extra;
+      png_ptr->zstream.avail_out = (uInt)1;
+      for(;;)
+      {
+         if (!(png_ptr->zstream.avail_in))
+         {
+            while (!png_ptr->idat_size)
+            {
+               png_byte chunk_length[4];
+
+               png_crc_finish(png_ptr, 0);
+
+               png_read_data(png_ptr, chunk_length, 4);
+               png_ptr->idat_size = png_get_uint_32(chunk_length);
+
+               png_reset_crc(png_ptr);
+               png_crc_read(png_ptr, png_ptr->chunk_name, 4);
+               if (png_memcmp(png_ptr->chunk_name, (png_bytep)png_IDAT, 4))
+                  png_error(png_ptr, "Not enough image data");
+
+            }
+            png_ptr->zstream.avail_in = (uInt)png_ptr->zbuf_size;
+            png_ptr->zstream.next_in = png_ptr->zbuf;
+            if (png_ptr->zbuf_size > png_ptr->idat_size)
+               png_ptr->zstream.avail_in = (uInt)png_ptr->idat_size;
+            png_crc_read(png_ptr, png_ptr->zbuf, png_ptr->zstream.avail_in);
+            png_ptr->idat_size -= png_ptr->zstream.avail_in;
+         }
+         ret = inflate(&png_ptr->zstream, Z_PARTIAL_FLUSH);
+         if (ret == Z_STREAM_END)
+         {
+            if (!(png_ptr->zstream.avail_out) || png_ptr->zstream.avail_in ||
+               png_ptr->idat_size)
+               png_error(png_ptr, "Extra compressed data");
+            png_ptr->mode |= PNG_AFTER_IDAT;
+            png_ptr->flags |= PNG_FLAG_ZLIB_FINISHED;
+            break;
+         }
+         if (ret != Z_OK)
+            png_error(png_ptr, png_ptr->zstream.msg ? png_ptr->zstream.msg :
+                      "Decompression Error");
+
+         if (!(png_ptr->zstream.avail_out))
+            png_error(png_ptr, "Extra compressed data");
+
+      }
+      png_ptr->zstream.avail_out = 0;
+   }
+
+   if (png_ptr->idat_size || png_ptr->zstream.avail_in)
+      png_error(png_ptr, "Extra compression data");
+
+   inflateReset(&png_ptr->zstream);
+
+   png_ptr->mode |= PNG_AFTER_IDAT;
+}
+
+void /* PRIVATE */
+png_read_start_row(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+   /* start of interlace block */
+   const int png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0};
+
+   /* offset to next interlace block */
+   const int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1};
+
+   /* start of interlace block in the y direction */
+   const int png_pass_ystart[7] = {0, 0, 4, 0, 2, 0, 1};
+
+   /* offset to next interlace block in the y direction */
+   const int png_pass_yinc[7] = {8, 8, 8, 4, 4, 2, 2};
+#endif
+
+   int max_pixel_depth;
+   png_uint_32 row_bytes;
+
+   png_debug(1, "in png_read_start_row\n");
+   png_ptr->zstream.avail_in = 0;
+   png_init_read_transformations(png_ptr);
+   if (png_ptr->interlaced)
+   {
+      if (!(png_ptr->transformations & PNG_INTERLACE))
+         png_ptr->num_rows = (png_ptr->height + png_pass_yinc[0] - 1 -
+            png_pass_ystart[0]) / png_pass_yinc[0];
+      else
+         png_ptr->num_rows = png_ptr->height;
+
+      png_ptr->iwidth = (png_ptr->width +
+         png_pass_inc[png_ptr->pass] - 1 -
+         png_pass_start[png_ptr->pass]) /
+         png_pass_inc[png_ptr->pass];
+
+         row_bytes = ((png_ptr->iwidth *
+            (png_uint_32)png_ptr->pixel_depth + 7) >> 3) +1;
+         png_ptr->irowbytes = (png_size_t)row_bytes;
+         if((png_uint_32)png_ptr->irowbytes != row_bytes)
+            png_error(png_ptr, "Rowbytes overflow in png_read_start_row");
+   }
+   else
+   {
+      png_ptr->num_rows = png_ptr->height;
+      png_ptr->iwidth = png_ptr->width;
+      png_ptr->irowbytes = png_ptr->rowbytes + 1;
+   }
+   max_pixel_depth = png_ptr->pixel_depth;
+
+#if defined(PNG_READ_PACK_SUPPORTED)
+   if ((png_ptr->transformations & PNG_PACK) && png_ptr->bit_depth < 8)
+      max_pixel_depth = 8;
+#endif
+
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+   if (png_ptr->transformations & PNG_EXPAND)
+   {
+      if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      {
+         if (png_ptr->num_trans)
+            max_pixel_depth = 32;
+         else
+            max_pixel_depth = 24;
+      }
+      else if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY)
+      {
+         if (max_pixel_depth < 8)
+            max_pixel_depth = 8;
+         if (png_ptr->num_trans)
+            max_pixel_depth *= 2;
+      }
+      else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB)
+      {
+         if (png_ptr->num_trans)
+         {
+            max_pixel_depth *= 4;
+            max_pixel_depth /= 3;
+         }
+      }
+   }
+#endif
+
+#if defined(PNG_READ_FILLER_SUPPORTED)
+   if (png_ptr->transformations & (PNG_FILLER))
+   {
+      if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+         max_pixel_depth = 32;
+      else if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY)
+      {
+         if (max_pixel_depth <= 8)
+            max_pixel_depth = 16;
+         else
+            max_pixel_depth = 32;
+      }
+      else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB)
+      {
+         if (max_pixel_depth <= 32)
+            max_pixel_depth = 32;
+         else
+            max_pixel_depth = 64;
+      }
+   }
+#endif
+
+#if defined(PNG_READ_GRAY_TO_RGB_SUPPORTED)
+   if (png_ptr->transformations & PNG_GRAY_TO_RGB)
+   {
+      if (
+#if defined(PNG_READ_EXPAND_SUPPORTED)
+        (png_ptr->num_trans && (png_ptr->transformations & PNG_EXPAND)) ||
+#endif
+#if defined(PNG_READ_FILLER_SUPPORTED)
+        (png_ptr->transformations & (PNG_FILLER)) ||
+#endif
+        png_ptr->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      {
+         if (max_pixel_depth <= 16)
+            max_pixel_depth = 32;
+         else
+            max_pixel_depth = 64;
+      }
+      else
+      {
+         if (max_pixel_depth <= 8)
+           {
+             if (png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+               max_pixel_depth = 32;
+             else
+               max_pixel_depth = 24;
+           }
+         else if (png_ptr->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+            max_pixel_depth = 64;
+         else
+            max_pixel_depth = 48;
+      }
+   }
+#endif
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) && \
+defined(PNG_USER_TRANSFORM_PTR_SUPPORTED)
+   if(png_ptr->transformations & PNG_USER_TRANSFORM)
+     {
+       int user_pixel_depth=png_ptr->user_transform_depth*
+         png_ptr->user_transform_channels;
+       if(user_pixel_depth > max_pixel_depth)
+         max_pixel_depth=user_pixel_depth;
+     }
+#endif
+
+   /* align the width on the next larger 8 pixels.  Mainly used
+      for interlacing */
+   row_bytes = ((png_ptr->width + 7) & ~((png_uint_32)7));
+   /* calculate the maximum bytes needed, adding a byte and a pixel
+      for safety's sake */
+   row_bytes = ((row_bytes * (png_uint_32)max_pixel_depth + 7) >> 3) +
+      1 + ((max_pixel_depth + 7) >> 3);
+#ifdef PNG_MAX_MALLOC_64K
+   if (row_bytes > (png_uint_32)65536L)
+      png_error(png_ptr, "This image requires a row greater than 64KB");
+#endif
+   png_ptr->big_row_buf = (png_bytep)png_malloc(png_ptr, row_bytes+64);
+   png_ptr->row_buf = png_ptr->big_row_buf+32;
+#if defined(PNG_DEBUG) && defined(PNG_USE_PNGGCCRD)
+   png_ptr->row_buf_size = row_bytes;
+#endif
+
+#ifdef PNG_MAX_MALLOC_64K
+   if ((png_uint_32)png_ptr->rowbytes + 1 > (png_uint_32)65536L)
+      png_error(png_ptr, "This image requires a row greater than 64KB");
+#endif
+   png_ptr->prev_row = (png_bytep)png_malloc(png_ptr, (png_uint_32)(
+      png_ptr->rowbytes + 1));
+
+   png_memset_check(png_ptr, png_ptr->prev_row, 0, png_ptr->rowbytes + 1);
+
+   png_debug1(3, "width = %lu,\n", png_ptr->width);
+   png_debug1(3, "height = %lu,\n", png_ptr->height);
+   png_debug1(3, "iwidth = %lu,\n", png_ptr->iwidth);
+   png_debug1(3, "num_rows = %lu\n", png_ptr->num_rows);
+   png_debug1(3, "rowbytes = %lu,\n", png_ptr->rowbytes);
+   png_debug1(3, "irowbytes = %lu,\n", png_ptr->irowbytes);
+
+   png_ptr->flags |= PNG_FLAG_ROW_INIT;
+}
diff --git a/png/pngset.c b/png/pngset.c
new file mode 100644
index 000000000..7878f3708
--- /dev/null
+++ b/png/pngset.c
@@ -0,0 +1,1033 @@
+
+/* pngset.c - storage of image information into info struct
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * The functions here are used during reads to store data from the file
+ * into the info struct, and during writes to store application data
+ * into the info struct for writing into the file.  This abstracts the
+ * info struct and allows us to change the structure in the future.
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+#if defined(PNG_bKGD_SUPPORTED)
+void PNGAPI
+png_set_bKGD(png_structp png_ptr, png_infop info_ptr, png_color_16p background)
+{
+   png_debug1(1, "in %s storage function\n", "bKGD");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   png_memcpy(&(info_ptr->background), background, sizeof(png_color_16));
+   info_ptr->valid |= PNG_INFO_bKGD;
+}
+#endif
+
+#if defined(PNG_cHRM_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+void PNGAPI
+png_set_cHRM(png_structp png_ptr, png_infop info_ptr,
+   double white_x, double white_y, double red_x, double red_y,
+   double green_x, double green_y, double blue_x, double blue_y)
+{
+   png_debug1(1, "in %s storage function\n", "cHRM");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->x_white = (float)white_x;
+   info_ptr->y_white = (float)white_y;
+   info_ptr->x_red   = (float)red_x;
+   info_ptr->y_red   = (float)red_y;
+   info_ptr->x_green = (float)green_x;
+   info_ptr->y_green = (float)green_y;
+   info_ptr->x_blue  = (float)blue_x;
+   info_ptr->y_blue  = (float)blue_y;
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   info_ptr->int_x_white = (png_fixed_point)(white_x*100000.+0.5);
+   info_ptr->int_y_white = (png_fixed_point)(white_y*100000.+0.5);
+   info_ptr->int_x_red   = (png_fixed_point)(red_x*100000.+0.5);
+   info_ptr->int_y_red   = (png_fixed_point)(red_y*100000.+0.5);
+   info_ptr->int_x_green = (png_fixed_point)(green_x*100000.+0.5);
+   info_ptr->int_y_green = (png_fixed_point)(green_y*100000.+0.5);
+   info_ptr->int_x_blue  = (png_fixed_point)(blue_x*100000.+0.5);
+   info_ptr->int_y_blue  = (png_fixed_point)(blue_y*100000.+0.5);
+#endif
+   info_ptr->valid |= PNG_INFO_cHRM;
+}
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+void PNGAPI
+png_set_cHRM_fixed(png_structp png_ptr, png_infop info_ptr,
+   png_fixed_point white_x, png_fixed_point white_y, png_fixed_point red_x,
+   png_fixed_point red_y, png_fixed_point green_x, png_fixed_point green_y,
+   png_fixed_point blue_x, png_fixed_point blue_y)
+{
+   png_debug1(1, "in %s storage function\n", "cHRM");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->int_x_white = white_x;
+   info_ptr->int_y_white = white_y;
+   info_ptr->int_x_red   = red_x;
+   info_ptr->int_y_red   = red_y;
+   info_ptr->int_x_green = green_x;
+   info_ptr->int_y_green = green_y;
+   info_ptr->int_x_blue  = blue_x;
+   info_ptr->int_y_blue  = blue_y;
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   info_ptr->x_white = (float)(white_x/100000.);
+   info_ptr->y_white = (float)(white_y/100000.);
+   info_ptr->x_red   = (float)(red_x/100000.);
+   info_ptr->y_red   = (float)(red_y/100000.);
+   info_ptr->x_green = (float)(green_x/100000.);
+   info_ptr->y_green = (float)(green_y/100000.);
+   info_ptr->x_blue  = (float)(blue_x/100000.);
+   info_ptr->y_blue  = (float)(blue_y/100000.);
+#endif
+   info_ptr->valid |= PNG_INFO_cHRM;
+}
+#endif
+#endif
+
+#if defined(PNG_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+void PNGAPI
+png_set_gAMA(png_structp png_ptr, png_infop info_ptr, double file_gamma)
+{
+   png_debug1(1, "in %s storage function\n", "gAMA");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->gamma = (float)file_gamma;
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   info_ptr->int_gamma = (int)(file_gamma*100000.+.5);
+#endif
+   info_ptr->valid |= PNG_INFO_gAMA;
+   if(file_gamma == 0.0)
+      png_warning(png_ptr, "Setting gamma=0");
+}
+#endif
+void PNGAPI
+png_set_gAMA_fixed(png_structp png_ptr, png_infop info_ptr, png_fixed_point
+   int_gamma)
+{
+   png_debug1(1, "in %s storage function\n", "gAMA");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   info_ptr->gamma = (float)(int_gamma/100000.);
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   info_ptr->int_gamma = int_gamma;
+#endif
+   info_ptr->valid |= PNG_INFO_gAMA;
+   if(int_gamma == 0)
+      png_warning(png_ptr, "Setting gamma=0");
+}
+#endif
+
+#if defined(PNG_hIST_SUPPORTED)
+void PNGAPI
+png_set_hIST(png_structp png_ptr, png_infop info_ptr, png_uint_16p hist)
+{
+   int	i;
+
+   png_debug1(1, "in %s storage function\n", "hIST");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+   if (info_ptr->num_palette == 0)
+   {
+       png_warning(png_ptr,
+          "Palette size 0, hIST allocation skipped.");
+       return;
+   }
+
+#ifdef PNG_FREE_ME_SUPPORTED
+   png_free_data(png_ptr, info_ptr, PNG_FREE_HIST, 0);
+#endif
+   /* Changed from info->num_palette to 256 in version 1.2.1 */
+   png_ptr->hist = (png_uint_16p)png_malloc(png_ptr,
+      (png_uint_32)(256 * sizeof (png_uint_16)));
+
+   for (i = 0; i < info_ptr->num_palette; i++)
+       png_ptr->hist[i] = hist[i];
+   info_ptr->hist = png_ptr->hist;
+   info_ptr->valid |= PNG_INFO_hIST;
+
+#ifdef PNG_FREE_ME_SUPPORTED
+   info_ptr->free_me |= PNG_FREE_HIST;
+#else
+   png_ptr->flags |= PNG_FLAG_FREE_HIST;
+#endif
+}
+#endif
+
+void PNGAPI
+png_set_IHDR(png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 width, png_uint_32 height, int bit_depth,
+   int color_type, int interlace_type, int compression_type,
+   int filter_type)
+{
+   int rowbytes_per_pixel;
+   png_debug1(1, "in %s storage function\n", "IHDR");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   /* check for width and height valid values */
+   if (width == 0 || height == 0)
+      png_error(png_ptr, "Image width or height is zero in IHDR");
+   if (width > PNG_MAX_UINT || height > PNG_MAX_UINT)
+      png_error(png_ptr, "Invalid image size in IHDR");
+
+   /* check other values */
+   if (bit_depth != 1 && bit_depth != 2 && bit_depth != 4 &&
+      bit_depth != 8 && bit_depth != 16)
+      png_error(png_ptr, "Invalid bit depth in IHDR");
+
+   if (color_type < 0 || color_type == 1 ||
+      color_type == 5 || color_type > 6)
+      png_error(png_ptr, "Invalid color type in IHDR");
+
+   if (((color_type == PNG_COLOR_TYPE_PALETTE) && bit_depth > 8) ||
+       ((color_type == PNG_COLOR_TYPE_RGB ||
+         color_type == PNG_COLOR_TYPE_GRAY_ALPHA ||
+         color_type == PNG_COLOR_TYPE_RGB_ALPHA) && bit_depth < 8))
+      png_error(png_ptr, "Invalid color type/bit depth combination in IHDR");
+
+   if (interlace_type >= PNG_INTERLACE_LAST)
+      png_error(png_ptr, "Unknown interlace method in IHDR");
+
+   if (compression_type != PNG_COMPRESSION_TYPE_BASE)
+      png_error(png_ptr, "Unknown compression method in IHDR");
+
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   /* Accept filter_method 64 (intrapixel differencing) only if
+    * 1. Libpng was compiled with PNG_MNG_FEATURES_SUPPORTED and
+    * 2. Libpng did not read a PNG signature (this filter_method is only
+    *    used in PNG datastreams that are embedded in MNG datastreams) and
+    * 3. The application called png_permit_mng_features with a mask that
+    *    included PNG_FLAG_MNG_FILTER_64 and
+    * 4. The filter_method is 64 and
+    * 5. The color_type is RGB or RGBA
+    */
+   if((png_ptr->mode&PNG_HAVE_PNG_SIGNATURE)&&png_ptr->mng_features_permitted)
+      png_warning(png_ptr,"MNG features are not allowed in a PNG datastream\n");
+   if(filter_type != PNG_FILTER_TYPE_BASE)
+   {
+     if(!((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) &&
+        (filter_type == PNG_INTRAPIXEL_DIFFERENCING) &&
+        ((png_ptr->mode&PNG_HAVE_PNG_SIGNATURE) == 0) &&
+        (color_type == PNG_COLOR_TYPE_RGB || 
+         color_type == PNG_COLOR_TYPE_RGB_ALPHA)))
+        png_error(png_ptr, "Unknown filter method in IHDR");
+     if(png_ptr->mode&PNG_HAVE_PNG_SIGNATURE)
+        png_warning(png_ptr, "Invalid filter method in IHDR");
+   }
+#else
+   if(filter_type != PNG_FILTER_TYPE_BASE)
+      png_error(png_ptr, "Unknown filter method in IHDR");
+#endif
+
+   info_ptr->width = width;
+   info_ptr->height = height;
+   info_ptr->bit_depth = (png_byte)bit_depth;
+   info_ptr->color_type =(png_byte) color_type;
+   info_ptr->compression_type = (png_byte)compression_type;
+   info_ptr->filter_type = (png_byte)filter_type;
+   info_ptr->interlace_type = (png_byte)interlace_type;
+   if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      info_ptr->channels = 1;
+   else if (info_ptr->color_type & PNG_COLOR_MASK_COLOR)
+      info_ptr->channels = 3;
+   else
+      info_ptr->channels = 1;
+   if (info_ptr->color_type & PNG_COLOR_MASK_ALPHA)
+      info_ptr->channels++;
+   info_ptr->pixel_depth = (png_byte)(info_ptr->channels * info_ptr->bit_depth);
+
+   /* check for overflow */
+   rowbytes_per_pixel = (info_ptr->pixel_depth + 7) >> 3;
+   if (( width > PNG_MAX_UINT/rowbytes_per_pixel))
+   {
+      png_warning(png_ptr,
+         "Width too large to process image data; rowbytes will overflow.");
+      info_ptr->rowbytes = (png_size_t)0;
+   }
+   else
+      info_ptr->rowbytes = (info_ptr->width * info_ptr->pixel_depth + 7) >> 3;
+}
+
+#if defined(PNG_oFFs_SUPPORTED)
+void PNGAPI
+png_set_oFFs(png_structp png_ptr, png_infop info_ptr,
+   png_int_32 offset_x, png_int_32 offset_y, int unit_type)
+{
+   png_debug1(1, "in %s storage function\n", "oFFs");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->x_offset = offset_x;
+   info_ptr->y_offset = offset_y;
+   info_ptr->offset_unit_type = (png_byte)unit_type;
+   info_ptr->valid |= PNG_INFO_oFFs;
+}
+#endif
+
+#if defined(PNG_pCAL_SUPPORTED)
+void PNGAPI
+png_set_pCAL(png_structp png_ptr, png_infop info_ptr,
+   png_charp purpose, png_int_32 X0, png_int_32 X1, int type, int nparams,
+   png_charp units, png_charpp params)
+{
+   png_uint_32 length;
+   int i;
+
+   png_debug1(1, "in %s storage function\n", "pCAL");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   length = png_strlen(purpose) + 1;
+   png_debug1(3, "allocating purpose for info (%lu bytes)\n", length);
+   info_ptr->pcal_purpose = (png_charp)png_malloc(png_ptr, length);
+   png_memcpy(info_ptr->pcal_purpose, purpose, (png_size_t)length);
+
+   png_debug(3, "storing X0, X1, type, and nparams in info\n");
+   info_ptr->pcal_X0 = X0;
+   info_ptr->pcal_X1 = X1;
+   info_ptr->pcal_type = (png_byte)type;
+   info_ptr->pcal_nparams = (png_byte)nparams;
+
+   length = png_strlen(units) + 1;
+   png_debug1(3, "allocating units for info (%lu bytes)\n", length);
+   info_ptr->pcal_units = (png_charp)png_malloc(png_ptr, length);
+   png_memcpy(info_ptr->pcal_units, units, (png_size_t)length);
+
+   info_ptr->pcal_params = (png_charpp)png_malloc(png_ptr,
+      (png_uint_32)((nparams + 1) * sizeof(png_charp)));
+
+   info_ptr->pcal_params[nparams] = NULL;
+
+   for (i = 0; i < nparams; i++)
+   {
+      length = png_strlen(params[i]) + 1;
+      png_debug2(3, "allocating parameter %d for info (%lu bytes)\n", i, length);
+      info_ptr->pcal_params[i] = (png_charp)png_malloc(png_ptr, length);
+      png_memcpy(info_ptr->pcal_params[i], params[i], (png_size_t)length);
+   }
+
+   info_ptr->valid |= PNG_INFO_pCAL;
+#ifdef PNG_FREE_ME_SUPPORTED
+   info_ptr->free_me |= PNG_FREE_PCAL;
+#endif
+}
+#endif
+
+#if defined(PNG_READ_sCAL_SUPPORTED) || defined(PNG_WRITE_sCAL_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+void PNGAPI
+png_set_sCAL(png_structp png_ptr, png_infop info_ptr,
+             int unit, double width, double height)
+{
+   png_debug1(1, "in %s storage function\n", "sCAL");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->scal_unit = (png_byte)unit;
+   info_ptr->scal_pixel_width = width;
+   info_ptr->scal_pixel_height = height;
+
+   info_ptr->valid |= PNG_INFO_sCAL;
+}
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+void PNGAPI
+png_set_sCAL_s(png_structp png_ptr, png_infop info_ptr,
+             int unit, png_charp swidth, png_charp sheight)
+{
+   png_uint_32 length;
+
+   png_debug1(1, "in %s storage function\n", "sCAL");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->scal_unit = (png_byte)unit;
+
+   length = png_strlen(swidth) + 1;
+   png_debug1(3, "allocating unit for info (%d bytes)\n", length);
+   info_ptr->scal_s_width = (png_charp)png_malloc(png_ptr, length);
+   png_memcpy(info_ptr->scal_s_width, swidth, (png_size_t)length);
+
+   length = png_strlen(sheight) + 1;
+   png_debug1(3, "allocating unit for info (%d bytes)\n", length);
+   info_ptr->scal_s_height = (png_charp)png_malloc(png_ptr, length);
+   png_memcpy(info_ptr->scal_s_height, sheight, (png_size_t)length);
+
+   info_ptr->valid |= PNG_INFO_sCAL;
+#ifdef PNG_FREE_ME_SUPPORTED
+   info_ptr->free_me |= PNG_FREE_SCAL;
+#endif
+}
+#endif
+#endif
+#endif
+
+#if defined(PNG_pHYs_SUPPORTED)
+void PNGAPI
+png_set_pHYs(png_structp png_ptr, png_infop info_ptr,
+   png_uint_32 res_x, png_uint_32 res_y, int unit_type)
+{
+   png_debug1(1, "in %s storage function\n", "pHYs");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->x_pixels_per_unit = res_x;
+   info_ptr->y_pixels_per_unit = res_y;
+   info_ptr->phys_unit_type = (png_byte)unit_type;
+   info_ptr->valid |= PNG_INFO_pHYs;
+}
+#endif
+
+void PNGAPI
+png_set_PLTE(png_structp png_ptr, png_infop info_ptr,
+   png_colorp palette, int num_palette)
+{
+
+   png_debug1(1, "in %s storage function\n", "PLTE");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   /*
+    * It may not actually be necessary to set png_ptr->palette here;
+    * we do it for backward compatibility with the way the png_handle_tRNS
+    * function used to do the allocation.
+    */
+#ifdef PNG_FREE_ME_SUPPORTED
+   png_free_data(png_ptr, info_ptr, PNG_FREE_PLTE, 0);
+#endif
+   /* Changed in libpng-1.2.1 to allocate 256 instead of num_palette entries,
+      in case of an invalid PNG file that has too-large sample values. */
+   png_ptr->palette = (png_colorp)png_zalloc(png_ptr, (uInt)256,
+      sizeof (png_color));
+   png_memcpy(png_ptr->palette, palette, num_palette * sizeof (png_color));
+   info_ptr->palette = png_ptr->palette;
+   info_ptr->num_palette = png_ptr->num_palette = (png_uint_16)num_palette;
+
+#ifdef PNG_FREE_ME_SUPPORTED
+   info_ptr->free_me |= PNG_FREE_PLTE;
+#else
+   png_ptr->flags |= PNG_FLAG_FREE_PLTE;
+#endif
+
+   info_ptr->valid |= PNG_INFO_PLTE;
+}
+
+#if defined(PNG_sBIT_SUPPORTED)
+void PNGAPI
+png_set_sBIT(png_structp png_ptr, png_infop info_ptr,
+   png_color_8p sig_bit)
+{
+   png_debug1(1, "in %s storage function\n", "sBIT");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   png_memcpy(&(info_ptr->sig_bit), sig_bit, sizeof (png_color_8));
+   info_ptr->valid |= PNG_INFO_sBIT;
+}
+#endif
+
+#if defined(PNG_sRGB_SUPPORTED)
+void PNGAPI
+png_set_sRGB(png_structp png_ptr, png_infop info_ptr, int intent)
+{
+   png_debug1(1, "in %s storage function\n", "sRGB");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   info_ptr->srgb_intent = (png_byte)intent;
+   info_ptr->valid |= PNG_INFO_sRGB;
+}
+
+void PNGAPI
+png_set_sRGB_gAMA_and_cHRM(png_structp png_ptr, png_infop info_ptr,
+   int intent)
+{
+#if defined(PNG_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   float file_gamma;
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_fixed_point int_file_gamma;
+#endif
+#endif
+#if defined(PNG_cHRM_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   float white_x, white_y, red_x, red_y, green_x, green_y, blue_x, blue_y;
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   png_fixed_point int_white_x, int_white_y, int_red_x, int_red_y, int_green_x,
+      int_green_y, int_blue_x, int_blue_y;
+#endif
+#endif
+   png_debug1(1, "in %s storage function\n", "sRGB_gAMA_and_cHRM");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   png_set_sRGB(png_ptr, info_ptr, intent);
+
+#if defined(PNG_gAMA_SUPPORTED)
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   file_gamma = (float).45455;
+   png_set_gAMA(png_ptr, info_ptr, file_gamma);
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   int_file_gamma = 45455L;
+   png_set_gAMA_fixed(png_ptr, info_ptr, int_file_gamma);
+#endif
+#endif
+
+#if defined(PNG_cHRM_SUPPORTED)
+#ifdef PNG_FIXED_POINT_SUPPORTED
+   int_white_x = 31270L;
+   int_white_y = 32900L;
+   int_red_x   = 64000L;
+   int_red_y   = 33000L;
+   int_green_x = 30000L;
+   int_green_y = 60000L;
+   int_blue_x  = 15000L;
+   int_blue_y  =  6000L;
+
+   png_set_cHRM_fixed(png_ptr, info_ptr,
+      int_white_x, int_white_y, int_red_x, int_red_y, int_green_x, int_green_y,
+      int_blue_x, int_blue_y);
+#endif
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+   white_x = (float).3127;
+   white_y = (float).3290;
+   red_x   = (float).64;
+   red_y   = (float).33;
+   green_x = (float).30;
+   green_y = (float).60;
+   blue_x  = (float).15;
+   blue_y  = (float).06;
+
+   png_set_cHRM(png_ptr, info_ptr,
+      white_x, white_y, red_x, red_y, green_x, green_y, blue_x, blue_y);
+#endif
+#endif
+}
+#endif
+
+
+#if defined(PNG_iCCP_SUPPORTED)
+void PNGAPI
+png_set_iCCP(png_structp png_ptr, png_infop info_ptr,
+             png_charp name, int compression_type,
+             png_charp profile, png_uint_32 proflen)
+{
+   png_charp new_iccp_name;
+   png_charp new_iccp_profile;
+
+   png_debug1(1, "in %s storage function\n", "iCCP");
+   if (png_ptr == NULL || info_ptr == NULL || name == NULL || profile == NULL)
+      return;
+
+   new_iccp_name = (png_charp)png_malloc(png_ptr, png_strlen(name)+1);
+   png_strcpy(new_iccp_name, name);
+   new_iccp_profile = (png_charp)png_malloc(png_ptr, proflen);
+   png_memcpy(new_iccp_profile, profile, (png_size_t)proflen);
+
+   png_free_data(png_ptr, info_ptr, PNG_FREE_ICCP, 0);
+
+   info_ptr->iccp_proflen = proflen;
+   info_ptr->iccp_name = new_iccp_name;
+   info_ptr->iccp_profile = new_iccp_profile;
+   /* Compression is always zero but is here so the API and info structure
+    * does not have to change if we introduce multiple compression types */
+   info_ptr->iccp_compression = (png_byte)compression_type;
+#ifdef PNG_FREE_ME_SUPPORTED
+   info_ptr->free_me |= PNG_FREE_ICCP;
+#endif
+   info_ptr->valid |= PNG_INFO_iCCP;
+}
+#endif
+
+#if defined(PNG_TEXT_SUPPORTED)
+void PNGAPI
+png_set_text(png_structp png_ptr, png_infop info_ptr, png_textp text_ptr,
+   int num_text)
+{
+   int i;
+
+   png_debug1(1, "in %s storage function\n", (png_ptr->chunk_name[0] == '\0' ?
+      "text" : (png_const_charp)png_ptr->chunk_name));
+
+   if (png_ptr == NULL || info_ptr == NULL || num_text == 0)
+      return;
+
+   /* Make sure we have enough space in the "text" array in info_struct
+    * to hold all of the incoming text_ptr objects.
+    */
+   if (info_ptr->num_text + num_text > info_ptr->max_text)
+   {
+      if (info_ptr->text != NULL)
+      {
+         png_textp old_text;
+         int old_max;
+
+         old_max = info_ptr->max_text;
+         info_ptr->max_text = info_ptr->num_text + num_text + 8;
+         old_text = info_ptr->text;
+         info_ptr->text = (png_textp)png_malloc(png_ptr,
+            (png_uint_32)(info_ptr->max_text * sizeof (png_text)));
+         png_memcpy(info_ptr->text, old_text, (png_size_t)(old_max *
+            sizeof(png_text)));
+         png_free(png_ptr, old_text);
+      }
+      else
+      {
+         info_ptr->max_text = num_text + 8;
+         info_ptr->num_text = 0;
+         info_ptr->text = (png_textp)png_malloc(png_ptr,
+            (png_uint_32)(info_ptr->max_text * sizeof (png_text)));
+#ifdef PNG_FREE_ME_SUPPORTED
+         info_ptr->free_me |= PNG_FREE_TEXT;
+#endif
+      }
+      png_debug1(3, "allocated %d entries for info_ptr->text\n",
+         info_ptr->max_text);
+   }
+   for (i = 0; i < num_text; i++)
+   {
+      png_size_t text_length,key_len;
+      png_size_t lang_len,lang_key_len;
+      png_textp textp = &(info_ptr->text[info_ptr->num_text]);
+
+      if (text_ptr[i].key == NULL)
+          continue;
+
+      key_len = png_strlen(text_ptr[i].key);
+
+      if(text_ptr[i].compression <= 0)
+      {
+        lang_len = 0;
+        lang_key_len = 0;
+      }
+      else
+#ifdef PNG_iTXt_SUPPORTED
+      {
+        /* set iTXt data */
+        if (text_ptr[i].key != NULL)
+          lang_len = png_strlen(text_ptr[i].lang);
+        else
+          lang_len = 0;
+        if (text_ptr[i].lang_key != NULL)
+          lang_key_len = png_strlen(text_ptr[i].lang_key);
+        else
+          lang_key_len = 0;
+      }
+#else
+      {
+        png_warning(png_ptr, "iTXt chunk not supported.");
+        continue;
+      }
+#endif
+
+      if (text_ptr[i].text == NULL || text_ptr[i].text[0] == '\0')
+      {
+         text_length = 0;
+#ifdef PNG_iTXt_SUPPORTED
+         if(text_ptr[i].compression > 0)
+            textp->compression = PNG_ITXT_COMPRESSION_NONE;
+         else
+#endif
+            textp->compression = PNG_TEXT_COMPRESSION_NONE;
+      }
+      else
+      {
+         text_length = png_strlen(text_ptr[i].text);
+         textp->compression = text_ptr[i].compression;
+      }
+
+      textp->key = (png_charp)png_malloc(png_ptr,
+         (png_uint_32)(key_len + text_length + lang_len + lang_key_len + 4));
+      png_debug2(2, "Allocated %lu bytes at %x in png_set_text\n",
+         (png_uint_32)(key_len + lang_len + lang_key_len + text_length + 4),
+         (int)textp->key);
+
+      png_memcpy(textp->key, text_ptr[i].key,
+         (png_size_t)(key_len));
+      *(textp->key+key_len) = '\0';
+#ifdef PNG_iTXt_SUPPORTED
+      if (text_ptr[i].compression > 0)
+      {
+         textp->lang=textp->key + key_len + 1;
+         png_memcpy(textp->lang, text_ptr[i].lang, lang_len);
+         *(textp->lang+lang_len) = '\0';
+         textp->lang_key=textp->lang + lang_len + 1;
+         png_memcpy(textp->lang_key, text_ptr[i].lang_key, lang_key_len);
+         *(textp->lang_key+lang_key_len) = '\0';
+         textp->text=textp->lang_key + lang_key_len + 1;
+      }
+      else
+#endif
+      {
+#ifdef PNG_iTXt_SUPPORTED
+         textp->lang=NULL;
+         textp->lang_key=NULL;
+#endif
+         textp->text=textp->key + key_len + 1;
+      }
+      if(text_length)
+         png_memcpy(textp->text, text_ptr[i].text,
+            (png_size_t)(text_length));
+      *(textp->text+text_length) = '\0';
+
+#ifdef PNG_iTXt_SUPPORTED
+      if(textp->compression > 0)
+      {
+         textp->text_length = 0;
+         textp->itxt_length = text_length;
+      }
+      else
+#endif
+      {
+         textp->text_length = text_length;
+#ifdef PNG_iTXt_SUPPORTED
+         textp->itxt_length = 0;
+#endif
+      }
+      info_ptr->text[info_ptr->num_text]= *textp;
+      info_ptr->num_text++;
+      png_debug1(3, "transferred text chunk %d\n", info_ptr->num_text);
+   }
+}
+#endif
+
+#if defined(PNG_tIME_SUPPORTED)
+void PNGAPI
+png_set_tIME(png_structp png_ptr, png_infop info_ptr, png_timep mod_time)
+{
+   png_debug1(1, "in %s storage function\n", "tIME");
+   if (png_ptr == NULL || info_ptr == NULL ||
+       (png_ptr->mode & PNG_WROTE_tIME))
+      return;
+
+   png_memcpy(&(info_ptr->mod_time), mod_time, sizeof (png_time));
+   info_ptr->valid |= PNG_INFO_tIME;
+}
+#endif
+
+#if defined(PNG_tRNS_SUPPORTED)
+void PNGAPI
+png_set_tRNS(png_structp png_ptr, png_infop info_ptr,
+   png_bytep trans, int num_trans, png_color_16p trans_values)
+{
+   png_debug1(1, "in %s storage function\n", "tRNS");
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   if (trans != NULL)
+   {
+       /*
+	* It may not actually be necessary to set png_ptr->trans here;
+	* we do it for backward compatibility with the way the png_handle_tRNS
+	* function used to do the allocation.
+	*/
+#ifdef PNG_FREE_ME_SUPPORTED
+       png_free_data(png_ptr, info_ptr, PNG_FREE_TRNS, 0);
+#endif
+       /* Changed from num_trans to 256 in version 1.2.1 */
+       png_ptr->trans = info_ptr->trans = (png_bytep)png_malloc(png_ptr,
+           (png_uint_32)256);
+       png_memcpy(info_ptr->trans, trans, (png_size_t)num_trans);
+#ifdef PNG_FREE_ME_SUPPORTED
+       info_ptr->free_me |= PNG_FREE_TRNS;
+#else
+       png_ptr->flags |= PNG_FLAG_FREE_TRNS;
+#endif
+   }
+
+   if (trans_values != NULL)
+   {
+      png_memcpy(&(info_ptr->trans_values), trans_values,
+         sizeof(png_color_16));
+      if (num_trans == 0)
+        num_trans = 1;
+   }
+   info_ptr->num_trans = (png_uint_16)num_trans;
+   info_ptr->valid |= PNG_INFO_tRNS;
+}
+#endif
+
+#if defined(PNG_sPLT_SUPPORTED)
+void PNGAPI
+png_set_sPLT(png_structp png_ptr,
+             png_infop info_ptr, png_sPLT_tp entries, int nentries)
+{
+    png_sPLT_tp np;
+    int i;
+
+    np = (png_sPLT_tp)png_malloc(png_ptr,
+        (info_ptr->splt_palettes_num + nentries) * sizeof(png_sPLT_t));
+
+    png_memcpy(np, info_ptr->splt_palettes,
+           info_ptr->splt_palettes_num * sizeof(png_sPLT_t));
+    png_free(png_ptr, info_ptr->splt_palettes);
+    info_ptr->splt_palettes=NULL;
+
+    for (i = 0; i < nentries; i++)
+    {
+        png_sPLT_tp to = np + info_ptr->splt_palettes_num + i;
+        png_sPLT_tp from = entries + i;
+
+        to->name = (png_charp)png_malloc(png_ptr,
+            png_strlen(from->name) + 1);
+        png_strcpy(to->name, from->name);
+        to->entries = (png_sPLT_entryp)png_malloc(png_ptr,
+            from->nentries * sizeof(png_sPLT_t));
+        png_memcpy(to->entries, from->entries,
+            from->nentries * sizeof(png_sPLT_t));
+        to->nentries = from->nentries;
+        to->depth = from->depth;
+    }
+
+    info_ptr->splt_palettes = np;
+    info_ptr->splt_palettes_num += nentries;
+    info_ptr->valid |= PNG_INFO_sPLT;
+#ifdef PNG_FREE_ME_SUPPORTED
+    info_ptr->free_me |= PNG_FREE_SPLT;
+#endif
+}
+#endif /* PNG_sPLT_SUPPORTED */
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+void PNGAPI
+png_set_unknown_chunks(png_structp png_ptr,
+   png_infop info_ptr, png_unknown_chunkp unknowns, int num_unknowns)
+{
+    png_unknown_chunkp np;
+    int i;
+
+    if (png_ptr == NULL || info_ptr == NULL || num_unknowns == 0)
+        return;
+
+    np = (png_unknown_chunkp)png_malloc(png_ptr,
+        (info_ptr->unknown_chunks_num + num_unknowns) *
+        sizeof(png_unknown_chunk));
+
+    png_memcpy(np, info_ptr->unknown_chunks,
+           info_ptr->unknown_chunks_num * sizeof(png_unknown_chunk));
+    png_free(png_ptr, info_ptr->unknown_chunks);
+    info_ptr->unknown_chunks=NULL;
+
+    for (i = 0; i < num_unknowns; i++)
+    {
+        png_unknown_chunkp to = np + info_ptr->unknown_chunks_num + i;
+        png_unknown_chunkp from = unknowns + i;
+
+        png_strcpy((png_charp)to->name, (png_charp)from->name);
+        to->data = (png_bytep)png_malloc(png_ptr, from->size);
+        png_memcpy(to->data, from->data, from->size);
+        to->size = from->size;
+
+        /* note our location in the read or write sequence */
+        to->location = (png_byte)(png_ptr->mode & 0xff);
+    }
+
+    info_ptr->unknown_chunks = np;
+    info_ptr->unknown_chunks_num += num_unknowns;
+#ifdef PNG_FREE_ME_SUPPORTED
+    info_ptr->free_me |= PNG_FREE_UNKN;
+#endif
+}
+void PNGAPI
+png_set_unknown_chunk_location(png_structp png_ptr, png_infop info_ptr,
+   int chunk, int location)
+{
+   if(png_ptr != NULL && info_ptr != NULL && chunk >= 0 && chunk <
+         (int)info_ptr->unknown_chunks_num)
+      info_ptr->unknown_chunks[chunk].location = (png_byte)location;
+}
+#endif
+
+#if defined(PNG_READ_EMPTY_PLTE_SUPPORTED) || \
+    defined(PNG_WRITE_EMPTY_PLTE_SUPPORTED)
+void PNGAPI
+png_permit_empty_plte (png_structp png_ptr, int empty_plte_permitted)
+{
+   /* This function is deprecated in favor of png_permit_mng_features()
+      and will be removed from libpng-2.0.0 */
+   png_debug(1, "in png_permit_empty_plte, DEPRECATED.\n");
+   if (png_ptr == NULL)
+      return;
+   png_ptr->mng_features_permitted = (png_byte)
+     ((png_ptr->mng_features_permitted & (~(PNG_FLAG_MNG_EMPTY_PLTE))) |
+     ((empty_plte_permitted & PNG_FLAG_MNG_EMPTY_PLTE)));
+}
+#endif
+
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+png_uint_32 PNGAPI
+png_permit_mng_features (png_structp png_ptr, png_uint_32 mng_features)
+{
+   png_debug(1, "in png_permit_mng_features\n");
+   if (png_ptr == NULL)
+      return (png_uint_32)0;
+   png_ptr->mng_features_permitted =
+     (png_byte)(mng_features & PNG_ALL_MNG_FEATURES);
+   return (png_uint_32)png_ptr->mng_features_permitted;
+}
+#endif
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+void PNGAPI
+png_set_keep_unknown_chunks(png_structp png_ptr, int keep, png_bytep
+   chunk_list, int num_chunks)
+{
+    png_bytep new_list, p;
+    int i, old_num_chunks;
+    if (num_chunks == 0)
+    {
+      if(keep == HANDLE_CHUNK_ALWAYS || keep == HANDLE_CHUNK_IF_SAFE)
+        png_ptr->flags |= PNG_FLAG_KEEP_UNKNOWN_CHUNKS;
+      else
+        png_ptr->flags &= ~PNG_FLAG_KEEP_UNKNOWN_CHUNKS;
+
+      if(keep == HANDLE_CHUNK_ALWAYS)
+        png_ptr->flags |= PNG_FLAG_KEEP_UNSAFE_CHUNKS;
+      else
+        png_ptr->flags &= ~PNG_FLAG_KEEP_UNSAFE_CHUNKS;
+      return;
+    }
+    if (chunk_list == NULL)
+      return;
+    old_num_chunks=png_ptr->num_chunk_list;
+    new_list=(png_bytep)png_malloc(png_ptr,
+       (png_uint_32)(5*(num_chunks+old_num_chunks)));
+    if(png_ptr->chunk_list != NULL)
+    {
+       png_memcpy(new_list, png_ptr->chunk_list,
+          (png_size_t)(5*old_num_chunks));
+       png_free(png_ptr, png_ptr->chunk_list);
+       png_ptr->chunk_list=NULL;
+    }
+    png_memcpy(new_list+5*old_num_chunks, chunk_list,
+       (png_size_t)(5*num_chunks));
+    for (p=new_list+5*old_num_chunks+4, i=0; i<num_chunks; i++, p+=5)
+       *p=(png_byte)keep;
+    png_ptr->num_chunk_list=old_num_chunks+num_chunks;
+    png_ptr->chunk_list=new_list;
+#ifdef PNG_FREE_ME_SUPPORTED
+    png_ptr->free_me |= PNG_FREE_LIST;
+#endif
+}
+#endif
+
+#if defined(PNG_READ_USER_CHUNKS_SUPPORTED)
+void PNGAPI
+png_set_read_user_chunk_fn(png_structp png_ptr, png_voidp user_chunk_ptr,
+   png_user_chunk_ptr read_user_chunk_fn)
+{
+   png_debug(1, "in png_set_read_user_chunk_fn\n");
+   png_ptr->read_user_chunk_fn = read_user_chunk_fn;
+   png_ptr->user_chunk_ptr = user_chunk_ptr;
+}
+#endif
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+void PNGAPI
+png_set_rows(png_structp png_ptr, png_infop info_ptr, png_bytepp row_pointers)
+{
+   png_debug1(1, "in %s storage function\n", "rows");
+
+   if (png_ptr == NULL || info_ptr == NULL)
+      return;
+
+   if(info_ptr->row_pointers && (info_ptr->row_pointers != row_pointers))
+      png_free_data(png_ptr, info_ptr, PNG_FREE_ROWS, 0);
+   info_ptr->row_pointers = row_pointers;
+   if(row_pointers)
+      info_ptr->valid |= PNG_INFO_IDAT;
+}
+#endif
+
+void PNGAPI
+png_set_compression_buffer_size(png_structp png_ptr, png_uint_32 size)
+{
+    if(png_ptr->zbuf)
+       png_free(png_ptr, png_ptr->zbuf);
+    png_ptr->zbuf_size = (png_size_t)size;
+    png_ptr->zbuf = (png_bytep)png_malloc(png_ptr, size);
+    png_ptr->zstream.next_out = png_ptr->zbuf;
+    png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+}
+
+void PNGAPI
+png_set_invalid(png_structp png_ptr, png_infop info_ptr, int mask)
+{
+   if (png_ptr && info_ptr)
+      info_ptr->valid &= ~(mask);
+}
+
+
+#ifdef PNG_ASSEMBLER_CODE_SUPPORTED
+/* this function was added to libpng 1.2.0 and should always exist by default */
+void PNGAPI
+png_set_asm_flags (png_structp png_ptr, png_uint_32 asm_flags)
+{
+    png_uint_32 settable_asm_flags;
+    png_uint_32 settable_mmx_flags;
+
+    settable_mmx_flags =
+#ifdef PNG_HAVE_ASSEMBLER_COMBINE_ROW
+                         PNG_ASM_FLAG_MMX_READ_COMBINE_ROW  |
+#endif
+#ifdef PNG_HAVE_ASSEMBLER_READ_INTERLACE
+                         PNG_ASM_FLAG_MMX_READ_INTERLACE    |
+#endif
+#ifdef PNG_HAVE_ASSEMBLER_READ_FILTER_ROW
+                         PNG_ASM_FLAG_MMX_READ_FILTER_SUB   |
+                         PNG_ASM_FLAG_MMX_READ_FILTER_UP    |
+                         PNG_ASM_FLAG_MMX_READ_FILTER_AVG   |
+                         PNG_ASM_FLAG_MMX_READ_FILTER_PAETH |
+#endif
+                         0;
+
+    /* could be some non-MMX ones in the future, but not currently: */
+    settable_asm_flags = settable_mmx_flags;
+
+    if (!(png_ptr->asm_flags & PNG_ASM_FLAG_MMX_SUPPORT_COMPILED) ||
+        !(png_ptr->asm_flags & PNG_ASM_FLAG_MMX_SUPPORT_IN_CPU))
+    {
+        /* clear all MMX flags if MMX isn't supported */
+        settable_asm_flags &= ~settable_mmx_flags;
+        png_ptr->asm_flags &= ~settable_mmx_flags;
+    }
+
+    /* we're replacing the settable bits with those passed in by the user,
+     * so first zero them out of the master copy, then logical-OR in the
+     * allowed subset that was requested */
+
+    png_ptr->asm_flags &= ~settable_asm_flags;			/* zero them */
+    png_ptr->asm_flags |= (asm_flags & settable_asm_flags);	/* set them */
+}
+#endif /* ?PNG_ASSEMBLER_CODE_SUPPORTED */
+
+#ifdef PNG_ASSEMBLER_CODE_SUPPORTED
+/* this function was added to libpng 1.2.0 */
+void PNGAPI
+png_set_mmx_thresholds (png_structp png_ptr,
+                        png_byte mmx_bitdepth_threshold,
+                        png_uint_32 mmx_rowbytes_threshold)
+{
+    png_ptr->mmx_bitdepth_threshold = mmx_bitdepth_threshold;
+    png_ptr->mmx_rowbytes_threshold = mmx_rowbytes_threshold;
+}
+#endif /* ?PNG_ASSEMBLER_CODE_SUPPORTED */
diff --git a/png/pngtrans.c b/png/pngtrans.c
new file mode 100644
index 000000000..6eee9fb1e
--- /dev/null
+++ b/png/pngtrans.c
@@ -0,0 +1,640 @@
+
+/* pngtrans.c - transforms the data in a row (used by both readers and writers)
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+
+#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED)
+/* turn on BGR-to-RGB mapping */
+void PNGAPI
+png_set_bgr(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_bgr\n");
+   png_ptr->transformations |= PNG_BGR;
+}
+#endif
+
+#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED)
+/* turn on 16 bit byte swapping */
+void PNGAPI
+png_set_swap(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_swap\n");
+   if (png_ptr->bit_depth == 16)
+      png_ptr->transformations |= PNG_SWAP_BYTES;
+}
+#endif
+
+#if defined(PNG_READ_PACK_SUPPORTED) || defined(PNG_WRITE_PACK_SUPPORTED)
+/* turn on pixel packing */
+void PNGAPI
+png_set_packing(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_packing\n");
+   if (png_ptr->bit_depth < 8)
+   {
+      png_ptr->transformations |= PNG_PACK;
+      png_ptr->usr_bit_depth = 8;
+   }
+}
+#endif
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)||defined(PNG_WRITE_PACKSWAP_SUPPORTED)
+/* turn on packed pixel swapping */
+void PNGAPI
+png_set_packswap(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_packswap\n");
+   if (png_ptr->bit_depth < 8)
+      png_ptr->transformations |= PNG_PACKSWAP;
+}
+#endif
+
+#if defined(PNG_READ_SHIFT_SUPPORTED) || defined(PNG_WRITE_SHIFT_SUPPORTED)
+void PNGAPI
+png_set_shift(png_structp png_ptr, png_color_8p true_bits)
+{
+   png_debug(1, "in png_set_shift\n");
+   png_ptr->transformations |= PNG_SHIFT;
+   png_ptr->shift = *true_bits;
+}
+#endif
+
+#if defined(PNG_READ_INTERLACING_SUPPORTED) || \
+    defined(PNG_WRITE_INTERLACING_SUPPORTED)
+int PNGAPI
+png_set_interlace_handling(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_interlace handling\n");
+   if (png_ptr->interlaced)
+   {
+      png_ptr->transformations |= PNG_INTERLACE;
+      return (7);
+   }
+
+   return (1);
+}
+#endif
+
+#if defined(PNG_READ_FILLER_SUPPORTED) || defined(PNG_WRITE_FILLER_SUPPORTED)
+/* Add a filler byte on read, or remove a filler or alpha byte on write.
+ * The filler type has changed in v0.95 to allow future 2-byte fillers
+ * for 48-bit input data, as well as to avoid problems with some compilers
+ * that don't like bytes as parameters.
+ */
+void PNGAPI
+png_set_filler(png_structp png_ptr, png_uint_32 filler, int filler_loc)
+{
+   png_debug(1, "in png_set_filler\n");
+   png_ptr->transformations |= PNG_FILLER;
+   png_ptr->filler = (png_byte)filler;
+   if (filler_loc == PNG_FILLER_AFTER)
+      png_ptr->flags |= PNG_FLAG_FILLER_AFTER;
+   else
+      png_ptr->flags &= ~PNG_FLAG_FILLER_AFTER;
+
+   /* This should probably go in the "do_filler" routine.
+    * I attempted to do that in libpng-1.0.1a but that caused problems
+    * so I restored it in libpng-1.0.2a
+   */
+
+   if (png_ptr->color_type == PNG_COLOR_TYPE_RGB)
+   {
+      png_ptr->usr_channels = 4;
+   }
+
+   /* Also I added this in libpng-1.0.2a (what happens when we expand
+    * a less-than-8-bit grayscale to GA? */
+
+   if (png_ptr->color_type == PNG_COLOR_TYPE_GRAY && png_ptr->bit_depth >= 8)
+   {
+      png_ptr->usr_channels = 2;
+   }
+}
+#endif
+
+#if defined(PNG_READ_SWAP_ALPHA_SUPPORTED) || \
+    defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED)
+void PNGAPI
+png_set_swap_alpha(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_swap_alpha\n");
+   png_ptr->transformations |= PNG_SWAP_ALPHA;
+}
+#endif
+
+#if defined(PNG_READ_INVERT_ALPHA_SUPPORTED) || \
+    defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+void PNGAPI
+png_set_invert_alpha(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_invert_alpha\n");
+   png_ptr->transformations |= PNG_INVERT_ALPHA;
+}
+#endif
+
+#if defined(PNG_READ_INVERT_SUPPORTED) || defined(PNG_WRITE_INVERT_SUPPORTED)
+void PNGAPI
+png_set_invert_mono(png_structp png_ptr)
+{
+   png_debug(1, "in png_set_invert_mono\n");
+   png_ptr->transformations |= PNG_INVERT_MONO;
+}
+
+/* invert monochrome grayscale data */
+void /* PRIVATE */
+png_do_invert(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_invert\n");
+  /* This test removed from libpng version 1.0.13 and 1.2.0:
+   *   if (row_info->bit_depth == 1 &&
+   */
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row == NULL || row_info == NULL)
+     return;
+#endif
+   if (row_info->color_type == PNG_COLOR_TYPE_GRAY)
+   {
+      png_bytep rp = row;
+      png_uint_32 i;
+      png_uint_32 istop = row_info->rowbytes;
+
+      for (i = 0; i < istop; i++)
+      {
+         *rp = (png_byte)(~(*rp));
+         rp++;
+      }
+   }
+   else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA &&
+      row_info->bit_depth == 8)
+   {
+      png_bytep rp = row;
+      png_uint_32 i;
+      png_uint_32 istop = row_info->rowbytes;
+
+      for (i = 0; i < istop; i+=2)
+      {
+         *rp = (png_byte)(~(*rp));
+         rp+=2;
+      }
+   }
+   else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA &&
+      row_info->bit_depth == 16)
+   {
+      png_bytep rp = row;
+      png_uint_32 i;
+      png_uint_32 istop = row_info->rowbytes;
+
+      for (i = 0; i < istop; i+=4)
+      {
+         *rp = (png_byte)(~(*rp));
+         *(rp+1) = (png_byte)(~(*(rp+1)));
+         rp+=4;
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_SWAP_SUPPORTED) || defined(PNG_WRITE_SWAP_SUPPORTED)
+/* swaps byte order on 16 bit depth images */
+void /* PRIVATE */
+png_do_swap(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_swap\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       row_info->bit_depth == 16)
+   {
+      png_bytep rp = row;
+      png_uint_32 i;
+      png_uint_32 istop= row_info->width * row_info->channels;
+
+      for (i = 0; i < istop; i++, rp += 2)
+      {
+         png_byte t = *rp;
+         *rp = *(rp + 1);
+         *(rp + 1) = t;
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_PACKSWAP_SUPPORTED)||defined(PNG_WRITE_PACKSWAP_SUPPORTED)
+static png_byte onebppswaptable[256] = {
+   0x00, 0x80, 0x40, 0xC0, 0x20, 0xA0, 0x60, 0xE0,
+   0x10, 0x90, 0x50, 0xD0, 0x30, 0xB0, 0x70, 0xF0,
+   0x08, 0x88, 0x48, 0xC8, 0x28, 0xA8, 0x68, 0xE8,
+   0x18, 0x98, 0x58, 0xD8, 0x38, 0xB8, 0x78, 0xF8,
+   0x04, 0x84, 0x44, 0xC4, 0x24, 0xA4, 0x64, 0xE4,
+   0x14, 0x94, 0x54, 0xD4, 0x34, 0xB4, 0x74, 0xF4,
+   0x0C, 0x8C, 0x4C, 0xCC, 0x2C, 0xAC, 0x6C, 0xEC,
+   0x1C, 0x9C, 0x5C, 0xDC, 0x3C, 0xBC, 0x7C, 0xFC,
+   0x02, 0x82, 0x42, 0xC2, 0x22, 0xA2, 0x62, 0xE2,
+   0x12, 0x92, 0x52, 0xD2, 0x32, 0xB2, 0x72, 0xF2,
+   0x0A, 0x8A, 0x4A, 0xCA, 0x2A, 0xAA, 0x6A, 0xEA,
+   0x1A, 0x9A, 0x5A, 0xDA, 0x3A, 0xBA, 0x7A, 0xFA,
+   0x06, 0x86, 0x46, 0xC6, 0x26, 0xA6, 0x66, 0xE6,
+   0x16, 0x96, 0x56, 0xD6, 0x36, 0xB6, 0x76, 0xF6,
+   0x0E, 0x8E, 0x4E, 0xCE, 0x2E, 0xAE, 0x6E, 0xEE,
+   0x1E, 0x9E, 0x5E, 0xDE, 0x3E, 0xBE, 0x7E, 0xFE,
+   0x01, 0x81, 0x41, 0xC1, 0x21, 0xA1, 0x61, 0xE1,
+   0x11, 0x91, 0x51, 0xD1, 0x31, 0xB1, 0x71, 0xF1,
+   0x09, 0x89, 0x49, 0xC9, 0x29, 0xA9, 0x69, 0xE9,
+   0x19, 0x99, 0x59, 0xD9, 0x39, 0xB9, 0x79, 0xF9,
+   0x05, 0x85, 0x45, 0xC5, 0x25, 0xA5, 0x65, 0xE5,
+   0x15, 0x95, 0x55, 0xD5, 0x35, 0xB5, 0x75, 0xF5,
+   0x0D, 0x8D, 0x4D, 0xCD, 0x2D, 0xAD, 0x6D, 0xED,
+   0x1D, 0x9D, 0x5D, 0xDD, 0x3D, 0xBD, 0x7D, 0xFD,
+   0x03, 0x83, 0x43, 0xC3, 0x23, 0xA3, 0x63, 0xE3,
+   0x13, 0x93, 0x53, 0xD3, 0x33, 0xB3, 0x73, 0xF3,
+   0x0B, 0x8B, 0x4B, 0xCB, 0x2B, 0xAB, 0x6B, 0xEB,
+   0x1B, 0x9B, 0x5B, 0xDB, 0x3B, 0xBB, 0x7B, 0xFB,
+   0x07, 0x87, 0x47, 0xC7, 0x27, 0xA7, 0x67, 0xE7,
+   0x17, 0x97, 0x57, 0xD7, 0x37, 0xB7, 0x77, 0xF7,
+   0x0F, 0x8F, 0x4F, 0xCF, 0x2F, 0xAF, 0x6F, 0xEF,
+   0x1F, 0x9F, 0x5F, 0xDF, 0x3F, 0xBF, 0x7F, 0xFF
+};
+
+static png_byte twobppswaptable[256] = {
+   0x00, 0x40, 0x80, 0xC0, 0x10, 0x50, 0x90, 0xD0,
+   0x20, 0x60, 0xA0, 0xE0, 0x30, 0x70, 0xB0, 0xF0,
+   0x04, 0x44, 0x84, 0xC4, 0x14, 0x54, 0x94, 0xD4,
+   0x24, 0x64, 0xA4, 0xE4, 0x34, 0x74, 0xB4, 0xF4,
+   0x08, 0x48, 0x88, 0xC8, 0x18, 0x58, 0x98, 0xD8,
+   0x28, 0x68, 0xA8, 0xE8, 0x38, 0x78, 0xB8, 0xF8,
+   0x0C, 0x4C, 0x8C, 0xCC, 0x1C, 0x5C, 0x9C, 0xDC,
+   0x2C, 0x6C, 0xAC, 0xEC, 0x3C, 0x7C, 0xBC, 0xFC,
+   0x01, 0x41, 0x81, 0xC1, 0x11, 0x51, 0x91, 0xD1,
+   0x21, 0x61, 0xA1, 0xE1, 0x31, 0x71, 0xB1, 0xF1,
+   0x05, 0x45, 0x85, 0xC5, 0x15, 0x55, 0x95, 0xD5,
+   0x25, 0x65, 0xA5, 0xE5, 0x35, 0x75, 0xB5, 0xF5,
+   0x09, 0x49, 0x89, 0xC9, 0x19, 0x59, 0x99, 0xD9,
+   0x29, 0x69, 0xA9, 0xE9, 0x39, 0x79, 0xB9, 0xF9,
+   0x0D, 0x4D, 0x8D, 0xCD, 0x1D, 0x5D, 0x9D, 0xDD,
+   0x2D, 0x6D, 0xAD, 0xED, 0x3D, 0x7D, 0xBD, 0xFD,
+   0x02, 0x42, 0x82, 0xC2, 0x12, 0x52, 0x92, 0xD2,
+   0x22, 0x62, 0xA2, 0xE2, 0x32, 0x72, 0xB2, 0xF2,
+   0x06, 0x46, 0x86, 0xC6, 0x16, 0x56, 0x96, 0xD6,
+   0x26, 0x66, 0xA6, 0xE6, 0x36, 0x76, 0xB6, 0xF6,
+   0x0A, 0x4A, 0x8A, 0xCA, 0x1A, 0x5A, 0x9A, 0xDA,
+   0x2A, 0x6A, 0xAA, 0xEA, 0x3A, 0x7A, 0xBA, 0xFA,
+   0x0E, 0x4E, 0x8E, 0xCE, 0x1E, 0x5E, 0x9E, 0xDE,
+   0x2E, 0x6E, 0xAE, 0xEE, 0x3E, 0x7E, 0xBE, 0xFE,
+   0x03, 0x43, 0x83, 0xC3, 0x13, 0x53, 0x93, 0xD3,
+   0x23, 0x63, 0xA3, 0xE3, 0x33, 0x73, 0xB3, 0xF3,
+   0x07, 0x47, 0x87, 0xC7, 0x17, 0x57, 0x97, 0xD7,
+   0x27, 0x67, 0xA7, 0xE7, 0x37, 0x77, 0xB7, 0xF7,
+   0x0B, 0x4B, 0x8B, 0xCB, 0x1B, 0x5B, 0x9B, 0xDB,
+   0x2B, 0x6B, 0xAB, 0xEB, 0x3B, 0x7B, 0xBB, 0xFB,
+   0x0F, 0x4F, 0x8F, 0xCF, 0x1F, 0x5F, 0x9F, 0xDF,
+   0x2F, 0x6F, 0xAF, 0xEF, 0x3F, 0x7F, 0xBF, 0xFF
+};
+
+static png_byte fourbppswaptable[256] = {
+   0x00, 0x10, 0x20, 0x30, 0x40, 0x50, 0x60, 0x70,
+   0x80, 0x90, 0xA0, 0xB0, 0xC0, 0xD0, 0xE0, 0xF0,
+   0x01, 0x11, 0x21, 0x31, 0x41, 0x51, 0x61, 0x71,
+   0x81, 0x91, 0xA1, 0xB1, 0xC1, 0xD1, 0xE1, 0xF1,
+   0x02, 0x12, 0x22, 0x32, 0x42, 0x52, 0x62, 0x72,
+   0x82, 0x92, 0xA2, 0xB2, 0xC2, 0xD2, 0xE2, 0xF2,
+   0x03, 0x13, 0x23, 0x33, 0x43, 0x53, 0x63, 0x73,
+   0x83, 0x93, 0xA3, 0xB3, 0xC3, 0xD3, 0xE3, 0xF3,
+   0x04, 0x14, 0x24, 0x34, 0x44, 0x54, 0x64, 0x74,
+   0x84, 0x94, 0xA4, 0xB4, 0xC4, 0xD4, 0xE4, 0xF4,
+   0x05, 0x15, 0x25, 0x35, 0x45, 0x55, 0x65, 0x75,
+   0x85, 0x95, 0xA5, 0xB5, 0xC5, 0xD5, 0xE5, 0xF5,
+   0x06, 0x16, 0x26, 0x36, 0x46, 0x56, 0x66, 0x76,
+   0x86, 0x96, 0xA6, 0xB6, 0xC6, 0xD6, 0xE6, 0xF6,
+   0x07, 0x17, 0x27, 0x37, 0x47, 0x57, 0x67, 0x77,
+   0x87, 0x97, 0xA7, 0xB7, 0xC7, 0xD7, 0xE7, 0xF7,
+   0x08, 0x18, 0x28, 0x38, 0x48, 0x58, 0x68, 0x78,
+   0x88, 0x98, 0xA8, 0xB8, 0xC8, 0xD8, 0xE8, 0xF8,
+   0x09, 0x19, 0x29, 0x39, 0x49, 0x59, 0x69, 0x79,
+   0x89, 0x99, 0xA9, 0xB9, 0xC9, 0xD9, 0xE9, 0xF9,
+   0x0A, 0x1A, 0x2A, 0x3A, 0x4A, 0x5A, 0x6A, 0x7A,
+   0x8A, 0x9A, 0xAA, 0xBA, 0xCA, 0xDA, 0xEA, 0xFA,
+   0x0B, 0x1B, 0x2B, 0x3B, 0x4B, 0x5B, 0x6B, 0x7B,
+   0x8B, 0x9B, 0xAB, 0xBB, 0xCB, 0xDB, 0xEB, 0xFB,
+   0x0C, 0x1C, 0x2C, 0x3C, 0x4C, 0x5C, 0x6C, 0x7C,
+   0x8C, 0x9C, 0xAC, 0xBC, 0xCC, 0xDC, 0xEC, 0xFC,
+   0x0D, 0x1D, 0x2D, 0x3D, 0x4D, 0x5D, 0x6D, 0x7D,
+   0x8D, 0x9D, 0xAD, 0xBD, 0xCD, 0xDD, 0xED, 0xFD,
+   0x0E, 0x1E, 0x2E, 0x3E, 0x4E, 0x5E, 0x6E, 0x7E,
+   0x8E, 0x9E, 0xAE, 0xBE, 0xCE, 0xDE, 0xEE, 0xFE,
+   0x0F, 0x1F, 0x2F, 0x3F, 0x4F, 0x5F, 0x6F, 0x7F,
+   0x8F, 0x9F, 0xAF, 0xBF, 0xCF, 0xDF, 0xEF, 0xFF
+};
+
+/* swaps pixel packing order within bytes */
+void /* PRIVATE */
+png_do_packswap(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_packswap\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       row_info->bit_depth < 8)
+   {
+      png_bytep rp, end, table;
+
+      end = row + row_info->rowbytes;
+
+      if (row_info->bit_depth == 1)
+         table = onebppswaptable;
+      else if (row_info->bit_depth == 2)
+         table = twobppswaptable;
+      else if (row_info->bit_depth == 4)
+         table = fourbppswaptable;
+      else
+         return;
+
+      for (rp = row; rp < end; rp++)
+         *rp = table[*rp];
+   }
+}
+#endif /* PNG_READ_PACKSWAP_SUPPORTED or PNG_WRITE_PACKSWAP_SUPPORTED */
+
+#if defined(PNG_WRITE_FILLER_SUPPORTED) || \
+    defined(PNG_READ_STRIP_ALPHA_SUPPORTED)
+/* remove filler or alpha byte(s) */
+void /* PRIVATE */
+png_do_strip_filler(png_row_infop row_info, png_bytep row, png_uint_32 flags)
+{
+   png_debug(1, "in png_do_strip_filler\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+/*
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB ||
+          row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+*/
+      png_bytep sp=row;
+      png_bytep dp=row;
+      png_uint_32 row_width=row_info->width;
+      png_uint_32 i;
+
+      if (row_info->channels == 4)
+      {
+         if (row_info->bit_depth == 8)
+         {
+            /* This converts from RGBX or RGBA to RGB */
+            if (flags & PNG_FLAG_FILLER_AFTER)
+            {
+               dp+=3; sp+=4;
+               for (i = 1; i < row_width; i++)
+               {
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  sp++;
+               }
+            }
+            /* This converts from XRGB or ARGB to RGB */
+            else
+            {
+               for (i = 0; i < row_width; i++)
+               {
+                  sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+               }
+            }
+            row_info->pixel_depth = 24;
+            row_info->rowbytes = row_width * 3;
+         }
+         else /* if (row_info->bit_depth == 16) */
+         {
+            if (flags & PNG_FLAG_FILLER_AFTER)
+            {
+               /* This converts from RRGGBBXX or RRGGBBAA to RRGGBB */
+               sp += 8; dp += 6;
+               for (i = 1; i < row_width; i++)
+               {
+                  /* This could be (although png_memcpy is probably slower):
+                  png_memcpy(dp, sp, 6);
+                  sp += 8;
+                  dp += 6;
+                  */
+
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  sp += 2;
+               }
+            }
+            else
+            {
+               /* This converts from XXRRGGBB or AARRGGBB to RRGGBB */
+               for (i = 0; i < row_width; i++)
+               {
+                  /* This could be (although png_memcpy is probably slower):
+                  png_memcpy(dp, sp, 6);
+                  sp += 8;
+                  dp += 6;
+                  */
+
+                  sp+=2;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+               }
+            }
+            row_info->pixel_depth = 48;
+            row_info->rowbytes = row_width * 6;
+         }
+         row_info->channels = 3;
+         row_info->color_type &= ~PNG_COLOR_MASK_ALPHA;
+      }
+/*
+      else if (row_info->color_type == PNG_COLOR_TYPE_GRAY ||
+               row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+*/
+      else if (row_info->channels == 2)
+      {
+         if (row_info->bit_depth == 8)
+         {
+            /* This converts from GX or GA to G */
+            if (flags & PNG_FLAG_FILLER_AFTER)
+            {
+               for (i = 0; i < row_width; i++)
+               {
+                  *dp++ = *sp++;
+                  sp++;
+               }
+            }
+            /* This converts from XG or AG to G */
+            else
+            {
+               for (i = 0; i < row_width; i++)
+               {
+                  sp++;
+                  *dp++ = *sp++;
+               }
+            }
+            row_info->pixel_depth = 8;
+            row_info->rowbytes = row_width;
+         }
+         else /* if (row_info->bit_depth == 16) */
+         {
+            if (flags & PNG_FLAG_FILLER_AFTER)
+            {
+               /* This converts from GGXX or GGAA to GG */
+               sp += 4; dp += 2;
+               for (i = 1; i < row_width; i++)
+               {
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+                  sp += 2;
+               }
+            }
+            else
+            {
+               /* This converts from XXGG or AAGG to GG */
+               for (i = 0; i < row_width; i++)
+               {
+                  sp += 2;
+                  *dp++ = *sp++;
+                  *dp++ = *sp++;
+               }
+            }
+            row_info->pixel_depth = 16;
+            row_info->rowbytes = row_width * 2;
+         }
+         row_info->channels = 1;
+         row_info->color_type &= ~PNG_COLOR_MASK_ALPHA;
+      }
+   }
+}
+#endif
+
+#if defined(PNG_READ_BGR_SUPPORTED) || defined(PNG_WRITE_BGR_SUPPORTED)
+/* swaps red and blue bytes within a pixel */
+void /* PRIVATE */
+png_do_bgr(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_bgr\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       (row_info->color_type & PNG_COLOR_MASK_COLOR))
+   {
+      png_uint_32 row_width = row_info->width;
+      if (row_info->bit_depth == 8)
+      {
+         if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+         {
+            png_bytep rp;
+            png_uint_32 i;
+
+            for (i = 0, rp = row; i < row_width; i++, rp += 3)
+            {
+               png_byte save = *rp;
+               *rp = *(rp + 2);
+               *(rp + 2) = save;
+            }
+         }
+         else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+         {
+            png_bytep rp;
+            png_uint_32 i;
+
+            for (i = 0, rp = row; i < row_width; i++, rp += 4)
+            {
+               png_byte save = *rp;
+               *rp = *(rp + 2);
+               *(rp + 2) = save;
+            }
+         }
+      }
+      else if (row_info->bit_depth == 16)
+      {
+         if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+         {
+            png_bytep rp;
+            png_uint_32 i;
+
+            for (i = 0, rp = row; i < row_width; i++, rp += 6)
+            {
+               png_byte save = *rp;
+               *rp = *(rp + 4);
+               *(rp + 4) = save;
+               save = *(rp + 1);
+               *(rp + 1) = *(rp + 5);
+               *(rp + 5) = save;
+            }
+         }
+         else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+         {
+            png_bytep rp;
+            png_uint_32 i;
+
+            for (i = 0, rp = row; i < row_width; i++, rp += 8)
+            {
+               png_byte save = *rp;
+               *rp = *(rp + 4);
+               *(rp + 4) = save;
+               save = *(rp + 1);
+               *(rp + 1) = *(rp + 5);
+               *(rp + 5) = save;
+            }
+         }
+      }
+   }
+}
+#endif /* PNG_READ_BGR_SUPPORTED or PNG_WRITE_BGR_SUPPORTED */
+
+#if defined(PNG_READ_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED) || \
+    defined(PNG_LEGACY_SUPPORTED)
+void PNGAPI
+png_set_user_transform_info(png_structp png_ptr, png_voidp
+   user_transform_ptr, int user_transform_depth, int user_transform_channels)
+{
+   png_debug(1, "in png_set_user_transform_info\n");
+#if defined(PNG_USER_TRANSFORM_PTR_SUPPORTED)
+   png_ptr->user_transform_ptr = user_transform_ptr;
+   png_ptr->user_transform_depth = (png_byte)user_transform_depth;
+   png_ptr->user_transform_channels = (png_byte)user_transform_channels;
+#else
+   if(user_transform_ptr || user_transform_depth || user_transform_channels)
+      png_warning(png_ptr,
+        "This version of libpng does not support user transform info");
+#endif
+}
+#endif
+
+/* This function returns a pointer to the user_transform_ptr associated with
+ * the user transform functions.  The application should free any memory
+ * associated with this pointer before png_write_destroy and png_read_destroy
+ * are called.
+ */
+png_voidp PNGAPI
+png_get_user_transform_ptr(png_structp png_ptr)
+{
+#if defined(PNG_USER_TRANSFORM_PTR_SUPPORTED)
+   return ((png_voidp)png_ptr->user_transform_ptr);
+#else
+   if(png_ptr)
+     return (NULL);
+   return (NULL);
+#endif
+}
diff --git a/png/pngwio.c b/png/pngwio.c
new file mode 100644
index 000000000..01ebbcf01
--- /dev/null
+++ b/png/pngwio.c
@@ -0,0 +1,228 @@
+
+/* pngwio.c - functions for data output
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ *
+ * This file provides a location for all output.  Users who need
+ * special handling are expected to write functions that have the same
+ * arguments as these and perform similar functions, but that possibly
+ * use different output methods.  Note that you shouldn't change these
+ * functions, but rather write replacement functions and then change
+ * them at run time with png_set_write_fn(...).
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+#ifdef PNG_WRITE_SUPPORTED
+
+/* Write the data to whatever output you are using.  The default routine
+   writes to a file pointer.  Note that this routine sometimes gets called
+   with very small lengths, so you should implement some kind of simple
+   buffering if you are using unbuffered writes.  This should never be asked
+   to write more than 64K on a 16 bit machine.  */
+
+void /* PRIVATE */
+png_write_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   if (png_ptr->write_data_fn != NULL )
+      (*(png_ptr->write_data_fn))(png_ptr, data, length);
+   else
+      png_error(png_ptr, "Call to NULL write function");
+}
+
+#if !defined(PNG_NO_STDIO)
+/* This is the function that does the actual writing of data.  If you are
+   not writing to a standard C stream, you should create a replacement
+   write_data function and use it at run time with png_set_write_fn(), rather
+   than changing the library. */
+#ifndef USE_FAR_KEYWORD
+static void /* PRIVATE */
+png_default_write_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   png_uint_32 check;
+
+#if defined(_WIN32_WCE)
+   if ( !WriteFile((HANDLE)(png_ptr->io_ptr), data, length, &check, NULL) )
+      check = 0;
+#else
+   check = fwrite(data, 1, length, (png_FILE_p)(png_ptr->io_ptr));
+#endif
+   if (check != length)
+      png_error(png_ptr, "Write Error");
+}
+#else
+/* this is the model-independent version. Since the standard I/O library
+   can't handle far buffers in the medium and small models, we have to copy
+   the data.
+*/
+
+#define NEAR_BUF_SIZE 1024
+#define MIN(a,b) (a <= b ? a : b)
+
+static void /* PRIVATE */
+png_default_write_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   png_uint_32 check;
+   png_byte *near_data;  /* Needs to be "png_byte *" instead of "png_bytep" */
+   png_FILE_p io_ptr;
+
+   /* Check if data really is near. If so, use usual code. */
+   near_data = (png_byte *)CVT_PTR_NOCHECK(data);
+   io_ptr = (png_FILE_p)CVT_PTR(png_ptr->io_ptr);
+   if ((png_bytep)near_data == data)
+   {
+#if defined(_WIN32_WCE)
+      if ( !WriteFile(io_ptr, near_data, length, &check, NULL) )
+         check = 0;
+#else
+      check = fwrite(near_data, 1, length, io_ptr);
+#endif
+   }
+   else
+   {
+      png_byte buf[NEAR_BUF_SIZE];
+      png_size_t written, remaining, err;
+      check = 0;
+      remaining = length;
+      do
+      {
+         written = MIN(NEAR_BUF_SIZE, remaining);
+         png_memcpy(buf, data, written); /* copy far buffer to near buffer */
+#if defined(_WIN32_WCE)
+         if ( !WriteFile(io_ptr, buf, written, &err, NULL) )
+            err = 0;
+#else
+         err = fwrite(buf, 1, written, io_ptr);
+#endif
+         if (err != written)
+            break;
+         else
+            check += err;
+         data += written;
+         remaining -= written;
+      }
+      while (remaining != 0);
+   }
+   if (check != length)
+      png_error(png_ptr, "Write Error");
+}
+
+#endif
+#endif
+
+/* This function is called to output any data pending writing (normally
+   to disk).  After png_flush is called, there should be no data pending
+   writing in any buffers. */
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+void /* PRIVATE */
+png_flush(png_structp png_ptr)
+{
+   if (png_ptr->output_flush_fn != NULL)
+      (*(png_ptr->output_flush_fn))(png_ptr);
+}
+
+#if !defined(PNG_NO_STDIO)
+static void /* PRIVATE */
+png_default_flush(png_structp png_ptr)
+{
+#if !defined(_WIN32_WCE)
+   png_FILE_p io_ptr;
+   io_ptr = (png_FILE_p)CVT_PTR((png_ptr->io_ptr));
+   if (io_ptr != NULL)
+      fflush(io_ptr);
+#endif
+}
+#endif
+#endif
+
+/* This function allows the application to supply new output functions for
+   libpng if standard C streams aren't being used.
+
+   This function takes as its arguments:
+   png_ptr       - pointer to a png output data structure
+   io_ptr        - pointer to user supplied structure containing info about
+                   the output functions.  May be NULL.
+   write_data_fn - pointer to a new output function that takes as its
+                   arguments a pointer to a png_struct, a pointer to
+                   data to be written, and a 32-bit unsigned int that is
+                   the number of bytes to be written.  The new write
+                   function should call png_error(png_ptr, "Error msg")
+                   to exit and output any fatal error messages.
+   flush_data_fn - pointer to a new flush function that takes as its
+                   arguments a pointer to a png_struct.  After a call to
+                   the flush function, there should be no data in any buffers
+                   or pending transmission.  If the output method doesn't do
+                   any buffering of ouput, a function prototype must still be
+                   supplied although it doesn't have to do anything.  If
+                   PNG_WRITE_FLUSH_SUPPORTED is not defined at libpng compile
+                   time, output_flush_fn will be ignored, although it must be
+                   supplied for compatibility. */
+void PNGAPI
+png_set_write_fn(png_structp png_ptr, png_voidp io_ptr,
+   png_rw_ptr write_data_fn, png_flush_ptr output_flush_fn)
+{
+   png_ptr->io_ptr = io_ptr;
+
+#if !defined(PNG_NO_STDIO)
+   if (write_data_fn != NULL)
+      png_ptr->write_data_fn = write_data_fn;
+   else
+      png_ptr->write_data_fn = png_default_write_data;
+#else
+   png_ptr->write_data_fn = write_data_fn;
+#endif
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+#if !defined(PNG_NO_STDIO)
+   if (output_flush_fn != NULL)
+      png_ptr->output_flush_fn = output_flush_fn;
+   else
+      png_ptr->output_flush_fn = png_default_flush;
+#else
+   png_ptr->output_flush_fn = output_flush_fn;
+#endif
+#endif /* PNG_WRITE_FLUSH_SUPPORTED */
+
+   /* It is an error to read while writing a png file */
+   if (png_ptr->read_data_fn != NULL)
+   {
+      png_ptr->read_data_fn = NULL;
+      png_warning(png_ptr,
+         "Attempted to set both read_data_fn and write_data_fn in");
+      png_warning(png_ptr,
+         "the same structure.  Resetting read_data_fn to NULL.");
+   }
+}
+
+#if defined(USE_FAR_KEYWORD)
+#if defined(_MSC_VER)
+void *png_far_to_near(png_structp png_ptr,png_voidp ptr, int check)
+{
+   void *near_ptr;
+   void FAR *far_ptr;
+   FP_OFF(near_ptr) = FP_OFF(ptr);
+   far_ptr = (void FAR *)near_ptr;
+   if(check != 0)
+      if(FP_SEG(ptr) != FP_SEG(far_ptr))
+         png_error(png_ptr,"segment lost in conversion");
+   return(near_ptr);
+}
+#  else
+void *png_far_to_near(png_structp png_ptr,png_voidp ptr, int check)
+{
+   void *near_ptr;
+   void FAR *far_ptr;
+   near_ptr = (void FAR *)ptr;
+   far_ptr = (void FAR *)near_ptr;
+   if(check != 0)
+      if(far_ptr != ptr)
+         png_error(png_ptr,"segment lost in conversion");
+   return(near_ptr);
+}
+#   endif
+#   endif
+#endif /* PNG_WRITE_SUPPORTED */
diff --git a/png/pngwrite.c b/png/pngwrite.c
new file mode 100644
index 000000000..f46ce0f6b
--- /dev/null
+++ b/png/pngwrite.c
@@ -0,0 +1,1432 @@
+
+/* pngwrite.c - general routines to write a PNG file
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+/* get internal access to png.h */
+#define PNG_INTERNAL
+#include "png.h"
+#ifdef PNG_WRITE_SUPPORTED
+
+/* Writes all the PNG information.  This is the suggested way to use the
+ * library.  If you have a new chunk to add, make a function to write it,
+ * and put it in the correct location here.  If you want the chunk written
+ * after the image data, put it in png_write_end().  I strongly encourage
+ * you to supply a PNG_INFO_ flag, and check info_ptr->valid before writing
+ * the chunk, as that will keep the code from breaking if you want to just
+ * write a plain PNG file.  If you have long comments, I suggest writing
+ * them in png_write_end(), and compressing them.
+ */
+void PNGAPI
+png_write_info_before_PLTE(png_structp png_ptr, png_infop info_ptr)
+{
+   png_debug(1, "in png_write_info_before_PLTE\n");
+   if (!(png_ptr->mode & PNG_WROTE_INFO_BEFORE_PLTE))
+   {
+   png_write_sig(png_ptr); /* write PNG signature */
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   if((png_ptr->mode&PNG_HAVE_PNG_SIGNATURE)&&(png_ptr->mng_features_permitted))
+   {
+      png_warning(png_ptr,"MNG features are not allowed in a PNG datastream\n");
+      png_ptr->mng_features_permitted=0;
+   }
+#endif
+   /* write IHDR information. */
+   png_write_IHDR(png_ptr, info_ptr->width, info_ptr->height,
+      info_ptr->bit_depth, info_ptr->color_type, info_ptr->compression_type,
+      info_ptr->filter_type,
+#if defined(PNG_WRITE_INTERLACING_SUPPORTED)
+      info_ptr->interlace_type);
+#else
+      0);
+#endif
+   /* the rest of these check to see if the valid field has the appropriate
+      flag set, and if it does, writes the chunk. */
+#if defined(PNG_WRITE_gAMA_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_gAMA)
+   {
+#  ifdef PNG_FLOATING_POINT_SUPPORTED
+      png_write_gAMA(png_ptr, info_ptr->gamma);
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+      png_write_gAMA_fixed(png_ptr, info_ptr->int_gamma);
+#  endif
+#endif
+   }
+#endif
+#if defined(PNG_WRITE_sRGB_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_sRGB)
+      png_write_sRGB(png_ptr, (int)info_ptr->srgb_intent);
+#endif
+#if defined(PNG_WRITE_iCCP_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_iCCP)
+      png_write_iCCP(png_ptr, info_ptr->iccp_name, PNG_COMPRESSION_TYPE_BASE,
+                     info_ptr->iccp_profile, (int)info_ptr->iccp_proflen);
+#endif
+#if defined(PNG_WRITE_sBIT_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_sBIT)
+      png_write_sBIT(png_ptr, &(info_ptr->sig_bit), info_ptr->color_type);
+#endif
+#if defined(PNG_WRITE_cHRM_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_cHRM)
+   {
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+      png_write_cHRM(png_ptr,
+         info_ptr->x_white, info_ptr->y_white,
+         info_ptr->x_red, info_ptr->y_red,
+         info_ptr->x_green, info_ptr->y_green,
+         info_ptr->x_blue, info_ptr->y_blue);
+#else
+#  ifdef PNG_FIXED_POINT_SUPPORTED
+      png_write_cHRM_fixed(png_ptr,
+         info_ptr->int_x_white, info_ptr->int_y_white,
+         info_ptr->int_x_red, info_ptr->int_y_red,
+         info_ptr->int_x_green, info_ptr->int_y_green,
+         info_ptr->int_x_blue, info_ptr->int_y_blue);
+#  endif
+#endif
+   }
+#endif
+#if defined(PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED)
+   if (info_ptr->unknown_chunks_num)
+   {
+       png_unknown_chunk *up;
+
+       png_debug(5, "writing extra chunks\n");
+
+       for (up = info_ptr->unknown_chunks;
+            up < info_ptr->unknown_chunks + info_ptr->unknown_chunks_num;
+            up++)
+       {
+         int keep=png_handle_as_unknown(png_ptr, up->name);
+         if (keep != HANDLE_CHUNK_NEVER &&
+            up->location && (!(up->location & PNG_HAVE_PLTE)) &&
+            ((up->name[3] & 0x20) || keep == HANDLE_CHUNK_ALWAYS ||
+            (png_ptr->flags & PNG_FLAG_KEEP_UNSAFE_CHUNKS)))
+         {
+            png_write_chunk(png_ptr, up->name, up->data, up->size);
+         }
+       }
+   }
+#endif
+      png_ptr->mode |= PNG_WROTE_INFO_BEFORE_PLTE;
+   }
+}
+
+void PNGAPI
+png_write_info(png_structp png_ptr, png_infop info_ptr)
+{
+#if defined(PNG_WRITE_TEXT_SUPPORTED) || defined(PNG_WRITE_sPLT_SUPPORTED)
+   int i;
+#endif
+
+   png_debug(1, "in png_write_info\n");
+
+   png_write_info_before_PLTE(png_ptr, info_ptr);
+
+   if (info_ptr->valid & PNG_INFO_PLTE)
+      png_write_PLTE(png_ptr, info_ptr->palette,
+         (png_uint_32)info_ptr->num_palette);
+   else if (info_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+      png_error(png_ptr, "Valid palette required for paletted images\n");
+
+#if defined(PNG_WRITE_tRNS_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_tRNS)
+      {
+#if defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+         /* invert the alpha channel (in tRNS) */
+         if ((png_ptr->transformations & PNG_INVERT_ALPHA) &&
+            info_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+         {
+            int j;
+            for (j=0; j<(int)info_ptr->num_trans; j++)
+               info_ptr->trans[j] = (png_byte)(255 - info_ptr->trans[j]);
+         }
+#endif
+      png_write_tRNS(png_ptr, info_ptr->trans, &(info_ptr->trans_values),
+         info_ptr->num_trans, info_ptr->color_type);
+      }
+#endif
+#if defined(PNG_WRITE_bKGD_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_bKGD)
+      png_write_bKGD(png_ptr, &(info_ptr->background), info_ptr->color_type);
+#endif
+#if defined(PNG_WRITE_hIST_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_hIST)
+      png_write_hIST(png_ptr, info_ptr->hist, info_ptr->num_palette);
+#endif
+#if defined(PNG_WRITE_oFFs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_oFFs)
+      png_write_oFFs(png_ptr, info_ptr->x_offset, info_ptr->y_offset,
+         info_ptr->offset_unit_type);
+#endif
+#if defined(PNG_WRITE_pCAL_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_pCAL)
+      png_write_pCAL(png_ptr, info_ptr->pcal_purpose, info_ptr->pcal_X0,
+         info_ptr->pcal_X1, info_ptr->pcal_type, info_ptr->pcal_nparams,
+         info_ptr->pcal_units, info_ptr->pcal_params);
+#endif
+#if defined(PNG_WRITE_sCAL_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_sCAL)
+#if defined(PNG_FLOATING_POINT_SUPPORTED) && !defined(PNG_NO_STDIO)
+      png_write_sCAL(png_ptr, (int)info_ptr->scal_unit,
+          info_ptr->scal_pixel_width, info_ptr->scal_pixel_height);
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+      png_write_sCAL_s(png_ptr, (int)info_ptr->scal_unit,
+          info_ptr->scal_s_width, info_ptr->scal_s_height);
+#else
+      png_warning(png_ptr,
+          "png_write_sCAL not supported; sCAL chunk not written.\n");
+#endif
+#endif
+#endif
+#if defined(PNG_WRITE_pHYs_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_pHYs)
+      png_write_pHYs(png_ptr, info_ptr->x_pixels_per_unit,
+         info_ptr->y_pixels_per_unit, info_ptr->phys_unit_type);
+#endif
+#if defined(PNG_WRITE_tIME_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_tIME)
+   {
+      png_write_tIME(png_ptr, &(info_ptr->mod_time));
+      png_ptr->mode |= PNG_WROTE_tIME;
+   }
+#endif
+#if defined(PNG_WRITE_sPLT_SUPPORTED)
+   if (info_ptr->valid & PNG_INFO_sPLT)
+     for (i = 0; i < (int)info_ptr->splt_palettes_num; i++)
+       png_write_sPLT(png_ptr, info_ptr->splt_palettes + i);
+#endif
+#if defined(PNG_WRITE_TEXT_SUPPORTED)
+   /* Check to see if we need to write text chunks */
+   for (i = 0; i < info_ptr->num_text; i++)
+   {
+      png_debug2(2, "Writing header text chunk %d, type %d\n", i,
+         info_ptr->text[i].compression);
+      /* an internationalized chunk? */
+      if (info_ptr->text[i].compression > 0)
+      {
+#if defined(PNG_WRITE_iTXt_SUPPORTED)
+          /* write international chunk */
+          png_write_iTXt(png_ptr,
+                         info_ptr->text[i].compression,
+                         info_ptr->text[i].key,
+                         info_ptr->text[i].lang,
+                         info_ptr->text[i].lang_key,
+                         info_ptr->text[i].text);
+#else
+          png_warning(png_ptr, "Unable to write international text\n");
+#endif
+          /* Mark this chunk as written */
+          info_ptr->text[i].compression = PNG_TEXT_COMPRESSION_NONE_WR;
+      }
+      /* If we want a compressed text chunk */
+      else if (info_ptr->text[i].compression == PNG_TEXT_COMPRESSION_zTXt)
+      {
+#if defined(PNG_WRITE_zTXt_SUPPORTED)
+         /* write compressed chunk */
+         png_write_zTXt(png_ptr, info_ptr->text[i].key,
+            info_ptr->text[i].text, 0,
+            info_ptr->text[i].compression);
+#else
+         png_warning(png_ptr, "Unable to write compressed text\n");
+#endif
+         /* Mark this chunk as written */
+         info_ptr->text[i].compression = PNG_TEXT_COMPRESSION_zTXt_WR;
+      }
+      else if (info_ptr->text[i].compression == PNG_TEXT_COMPRESSION_NONE)
+      {
+#if defined(PNG_WRITE_tEXt_SUPPORTED)
+         /* write uncompressed chunk */
+         png_write_tEXt(png_ptr, info_ptr->text[i].key,
+                         info_ptr->text[i].text,
+                         0);
+#else
+         png_warning(png_ptr, "Unable to write uncompressed text\n");
+#endif
+         /* Mark this chunk as written */
+         info_ptr->text[i].compression = PNG_TEXT_COMPRESSION_NONE_WR;
+      }
+   }
+#endif
+#if defined(PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED)
+   if (info_ptr->unknown_chunks_num)
+   {
+       png_unknown_chunk *up;
+
+       png_debug(5, "writing extra chunks\n");
+
+       for (up = info_ptr->unknown_chunks;
+            up < info_ptr->unknown_chunks + info_ptr->unknown_chunks_num;
+            up++)
+       {
+         int keep=png_handle_as_unknown(png_ptr, up->name);
+         if (keep != HANDLE_CHUNK_NEVER &&
+            up->location && (up->location & PNG_HAVE_PLTE) &&
+            !(up->location & PNG_HAVE_IDAT) &&
+            ((up->name[3] & 0x20) || keep == HANDLE_CHUNK_ALWAYS ||
+            (png_ptr->flags & PNG_FLAG_KEEP_UNSAFE_CHUNKS)))
+         {
+            png_write_chunk(png_ptr, up->name, up->data, up->size);
+         }
+       }
+   }
+#endif
+}
+
+/* Writes the end of the PNG file.  If you don't want to write comments or
+ * time information, you can pass NULL for info.  If you already wrote these
+ * in png_write_info(), do not write them again here.  If you have long
+ * comments, I suggest writing them here, and compressing them.
+ */
+void PNGAPI
+png_write_end(png_structp png_ptr, png_infop info_ptr)
+{
+   png_debug(1, "in png_write_end\n");
+   if (!(png_ptr->mode & PNG_HAVE_IDAT))
+      png_error(png_ptr, "No IDATs written into file");
+
+   /* see if user wants us to write information chunks */
+   if (info_ptr != NULL)
+   {
+#if defined(PNG_WRITE_TEXT_SUPPORTED)
+      int i; /* local index variable */
+#endif
+#if defined(PNG_WRITE_tIME_SUPPORTED)
+      /* check to see if user has supplied a time chunk */
+      if ((info_ptr->valid & PNG_INFO_tIME) &&
+         !(png_ptr->mode & PNG_WROTE_tIME))
+         png_write_tIME(png_ptr, &(info_ptr->mod_time));
+#endif
+#if defined(PNG_WRITE_TEXT_SUPPORTED)
+      /* loop through comment chunks */
+      for (i = 0; i < info_ptr->num_text; i++)
+      {
+         png_debug2(2, "Writing trailer text chunk %d, type %d\n", i,
+            info_ptr->text[i].compression);
+         /* an internationalized chunk? */
+         if (info_ptr->text[i].compression > 0)
+         {
+#if defined(PNG_WRITE_iTXt_SUPPORTED)
+             /* write international chunk */
+             png_write_iTXt(png_ptr,
+                         info_ptr->text[i].compression,
+                         info_ptr->text[i].key,
+                         info_ptr->text[i].lang,
+                         info_ptr->text[i].lang_key,
+                         info_ptr->text[i].text);
+#else
+             png_warning(png_ptr, "Unable to write international text\n");
+#endif
+             /* Mark this chunk as written */
+             info_ptr->text[i].compression = PNG_TEXT_COMPRESSION_NONE_WR;
+         }
+         else if (info_ptr->text[i].compression >= PNG_TEXT_COMPRESSION_zTXt)
+         {
+#if defined(PNG_WRITE_zTXt_SUPPORTED)
+            /* write compressed chunk */
+            png_write_zTXt(png_ptr, info_ptr->text[i].key,
+               info_ptr->text[i].text, 0,
+               info_ptr->text[i].compression);
+#else
+            png_warning(png_ptr, "Unable to write compressed text\n");
+#endif
+            /* Mark this chunk as written */
+            info_ptr->text[i].compression = PNG_TEXT_COMPRESSION_zTXt_WR;
+         }
+         else if (info_ptr->text[i].compression == PNG_TEXT_COMPRESSION_NONE)
+         {
+#if defined(PNG_WRITE_tEXt_SUPPORTED)
+            /* write uncompressed chunk */
+            png_write_tEXt(png_ptr, info_ptr->text[i].key,
+               info_ptr->text[i].text, 0);
+#else
+            png_warning(png_ptr, "Unable to write uncompressed text\n");
+#endif
+
+            /* Mark this chunk as written */
+            info_ptr->text[i].compression = PNG_TEXT_COMPRESSION_NONE_WR;
+         }
+      }
+#endif
+#if defined(PNG_WRITE_UNKNOWN_CHUNKS_SUPPORTED)
+   if (info_ptr->unknown_chunks_num)
+   {
+       png_unknown_chunk *up;
+
+       png_debug(5, "writing extra chunks\n");
+
+       for (up = info_ptr->unknown_chunks;
+            up < info_ptr->unknown_chunks + info_ptr->unknown_chunks_num;
+            up++)
+       {
+         int keep=png_handle_as_unknown(png_ptr, up->name);
+         if (keep != HANDLE_CHUNK_NEVER &&
+            up->location && (up->location & PNG_AFTER_IDAT) &&
+            ((up->name[3] & 0x20) || keep == HANDLE_CHUNK_ALWAYS ||
+            (png_ptr->flags & PNG_FLAG_KEEP_UNSAFE_CHUNKS)))
+         {
+            png_write_chunk(png_ptr, up->name, up->data, up->size);
+         }
+       }
+   }
+#endif
+   }
+
+   png_ptr->mode |= PNG_AFTER_IDAT;
+
+   /* write end of PNG file */
+   png_write_IEND(png_ptr);
+#if 0
+/* This flush, added in libpng-1.0.8,  causes some applications to crash
+   because they do not set png_ptr->output_flush_fn */
+   png_flush(png_ptr);
+#endif
+}
+
+#if defined(PNG_WRITE_tIME_SUPPORTED)
+#if !defined(_WIN32_WCE)
+/* "time.h" functions are not supported on WindowsCE */
+void PNGAPI
+png_convert_from_struct_tm(png_timep ptime, struct tm FAR * ttime)
+{
+   png_debug(1, "in png_convert_from_struct_tm\n");
+   ptime->year = (png_uint_16)(1900 + ttime->tm_year);
+   ptime->month = (png_byte)(ttime->tm_mon + 1);
+   ptime->day = (png_byte)ttime->tm_mday;
+   ptime->hour = (png_byte)ttime->tm_hour;
+   ptime->minute = (png_byte)ttime->tm_min;
+   ptime->second = (png_byte)ttime->tm_sec;
+}
+
+void PNGAPI
+png_convert_from_time_t(png_timep ptime, time_t ttime)
+{
+   struct tm *tbuf;
+
+   png_debug(1, "in png_convert_from_time_t\n");
+   tbuf = gmtime(&ttime);
+   png_convert_from_struct_tm(ptime, tbuf);
+}
+#endif
+#endif
+
+/* Initialize png_ptr structure, and allocate any memory needed */
+png_structp PNGAPI
+png_create_write_struct(png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn)
+{
+#ifdef PNG_USER_MEM_SUPPORTED
+   return (png_create_write_struct_2(user_png_ver, error_ptr, error_fn,
+      warn_fn, png_voidp_NULL, png_malloc_ptr_NULL, png_free_ptr_NULL));
+}
+
+/* Alternate initialize png_ptr structure, and allocate any memory needed */
+png_structp PNGAPI
+png_create_write_struct_2(png_const_charp user_png_ver, png_voidp error_ptr,
+   png_error_ptr error_fn, png_error_ptr warn_fn, png_voidp mem_ptr,
+   png_malloc_ptr malloc_fn, png_free_ptr free_fn)
+{
+#endif /* PNG_USER_MEM_SUPPORTED */
+   png_structp png_ptr;
+#ifdef PNG_SETJMP_SUPPORTED
+#ifdef USE_FAR_KEYWORD
+   jmp_buf jmpbuf;
+#endif
+#endif
+   int i;
+   png_debug(1, "in png_create_write_struct\n");
+#ifdef PNG_USER_MEM_SUPPORTED
+   if ((png_ptr = (png_structp)png_create_struct_2(PNG_STRUCT_PNG,
+      (png_malloc_ptr)malloc_fn, (png_voidp)mem_ptr)) == NULL)
+#else
+   if ((png_ptr = (png_structp)png_create_struct(PNG_STRUCT_PNG)) == NULL)
+#endif /* PNG_USER_MEM_SUPPORTED */
+      return (NULL);
+
+#ifdef PNG_ASSEMBLER_CODE_SUPPORTED
+   png_init_mmx_flags(png_ptr);   /* 1.2.0 addition */
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+#ifdef USE_FAR_KEYWORD
+   if (setjmp(jmpbuf))
+#else
+   if (setjmp(png_ptr->jmpbuf))
+#endif
+   {
+      png_free(png_ptr, png_ptr->zbuf);
+      png_ptr->zbuf=NULL;
+      png_destroy_struct(png_ptr);
+      return (NULL);
+   }
+#ifdef USE_FAR_KEYWORD
+   png_memcpy(png_ptr->jmpbuf,jmpbuf,sizeof(jmp_buf));
+#endif
+#endif
+
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_set_mem_fn(png_ptr, mem_ptr, malloc_fn, free_fn);
+#endif /* PNG_USER_MEM_SUPPORTED */
+   png_set_error_fn(png_ptr, error_ptr, error_fn, warn_fn);
+
+   i=0;
+   do
+   {
+     if(user_png_ver[i] != png_libpng_ver[i])
+        png_ptr->flags |= PNG_FLAG_LIBRARY_MISMATCH;
+   } while (png_libpng_ver[i++]);
+
+   if (png_ptr->flags & PNG_FLAG_LIBRARY_MISMATCH)
+   {
+     /* Libpng 0.90 and later are binary incompatible with libpng 0.89, so
+      * we must recompile any applications that use any older library version.
+      * For versions after libpng 1.0, we will be compatible, so we need
+      * only check the first digit.
+      */
+     if (user_png_ver == NULL || user_png_ver[0] != png_libpng_ver[0] ||
+         (user_png_ver[0] == '1' && user_png_ver[2] != png_libpng_ver[2]) ||
+         (user_png_ver[0] == '0' && user_png_ver[2] < '9'))
+     {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+        char msg[80];
+        if (user_png_ver)
+        {
+          sprintf(msg, "Application was compiled with png.h from libpng-%.20s",
+             user_png_ver);
+          png_warning(png_ptr, msg);
+        }
+        sprintf(msg, "Application  is  running with png.c from libpng-%.20s",
+           png_libpng_ver);
+        png_warning(png_ptr, msg);
+#endif
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+        png_ptr->flags=0;
+#endif
+        png_error(png_ptr,
+           "Incompatible libpng version in application and library");
+     }
+   }
+
+   /* initialize zbuf - compression buffer */
+   png_ptr->zbuf_size = PNG_ZBUF_SIZE;
+   png_ptr->zbuf = (png_bytep)png_malloc(png_ptr,
+      (png_uint_32)png_ptr->zbuf_size);
+
+   png_set_write_fn(png_ptr, png_voidp_NULL, png_rw_ptr_NULL,
+      png_flush_ptr_NULL);
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+   png_set_filter_heuristics(png_ptr, PNG_FILTER_HEURISTIC_DEFAULT,
+      1, png_doublep_NULL, png_doublep_NULL);
+#endif
+
+   return ((png_structp)png_ptr);
+}
+
+/* Initialize png_ptr structure, and allocate any memory needed */
+#undef png_write_init
+void PNGAPI
+png_write_init(png_structp png_ptr)
+{
+   /* We only come here via pre-1.0.7-compiled applications */
+   png_write_init_2(png_ptr, "1.0.6 or earlier", 0, 0);
+}
+
+#undef png_write_init_2
+void PNGAPI
+png_write_init_2(png_structp png_ptr, png_const_charp user_png_ver,
+   png_size_t png_struct_size, png_size_t png_info_size)
+{
+   /* We only come here via pre-1.0.12-compiled applications */
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+   if(sizeof(png_struct) > png_struct_size || sizeof(png_info) > png_info_size)
+   {
+      char msg[80];
+      png_ptr->warning_fn=NULL;
+      if (user_png_ver)
+      {
+        sprintf(msg, "Application was compiled with png.h from libpng-%.20s",
+           user_png_ver);
+        png_warning(png_ptr, msg);
+      }
+      sprintf(msg, "Application  is  running with png.c from libpng-%.20s",
+         png_libpng_ver);
+      png_warning(png_ptr, msg);
+   }
+#endif
+   if(sizeof(png_struct) > png_struct_size)
+     {
+       png_ptr->error_fn=NULL;
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+       png_ptr->flags=0;
+#endif
+       png_error(png_ptr,
+       "The png struct allocated by the application for writing is too small.");
+     }
+   if(sizeof(png_info) > png_info_size)
+     {
+       png_ptr->error_fn=NULL;
+#ifdef PNG_ERROR_NUMBERS_SUPPORTED
+       png_ptr->flags=0;
+#endif
+       png_error(png_ptr,
+       "The info struct allocated by the application for writing is too small.");
+     }
+   png_write_init_3(&png_ptr, user_png_ver, png_struct_size);
+}
+
+
+void PNGAPI
+png_write_init_3(png_structpp ptr_ptr, png_const_charp user_png_ver,
+   png_size_t png_struct_size)
+{
+   png_structp png_ptr=*ptr_ptr;
+#ifdef PNG_SETJMP_SUPPORTED
+   jmp_buf tmp_jmp; /* to save current jump buffer */
+#endif
+   int i = 0;
+   do
+   {
+     if (user_png_ver[i] != png_libpng_ver[i])
+     {
+#ifdef PNG_LEGACY_SUPPORTED
+       png_ptr->flags |= PNG_FLAG_LIBRARY_MISMATCH;
+#else
+       png_ptr->warning_fn=NULL;
+       png_warning(png_ptr,
+     "Application uses deprecated png_write_init() and should be recompiled.");
+       break;
+#endif
+     }
+   } while (png_libpng_ver[i++]);
+
+   png_debug(1, "in png_write_init_3\n");
+
+#ifdef PNG_SETJMP_SUPPORTED
+   /* save jump buffer and error functions */
+   png_memcpy(tmp_jmp, png_ptr->jmpbuf, sizeof (jmp_buf));
+#endif
+
+   if (sizeof(png_struct) > png_struct_size)
+     {
+       png_destroy_struct(png_ptr);
+       png_ptr = (png_structp)png_create_struct(PNG_STRUCT_PNG);
+       *ptr_ptr = png_ptr;
+     }
+
+   /* reset all variables to 0 */
+   png_memset(png_ptr, 0, sizeof (png_struct));
+
+#ifdef PNG_ASSEMBLER_CODE_SUPPORTED
+   png_init_mmx_flags(png_ptr);   /* 1.2.0 addition */
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+   /* restore jump buffer */
+   png_memcpy(png_ptr->jmpbuf, tmp_jmp, sizeof (jmp_buf));
+#endif
+
+   png_set_write_fn(png_ptr, png_voidp_NULL, png_rw_ptr_NULL,
+      png_flush_ptr_NULL);
+
+   /* initialize zbuf - compression buffer */
+   png_ptr->zbuf_size = PNG_ZBUF_SIZE;
+   png_ptr->zbuf = (png_bytep)png_malloc(png_ptr,
+      (png_uint_32)png_ptr->zbuf_size);
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+   png_set_filter_heuristics(png_ptr, PNG_FILTER_HEURISTIC_DEFAULT,
+      1, png_doublep_NULL, png_doublep_NULL);
+#endif
+}
+
+/* Write a few rows of image data.  If the image is interlaced,
+ * either you will have to write the 7 sub images, or, if you
+ * have called png_set_interlace_handling(), you will have to
+ * "write" the image seven times.
+ */
+void PNGAPI
+png_write_rows(png_structp png_ptr, png_bytepp row,
+   png_uint_32 num_rows)
+{
+   png_uint_32 i; /* row counter */
+   png_bytepp rp; /* row pointer */
+
+   png_debug(1, "in png_write_rows\n");
+   /* loop through the rows */
+   for (i = 0, rp = row; i < num_rows; i++, rp++)
+   {
+      png_write_row(png_ptr, *rp);
+   }
+}
+
+/* Write the image.  You only need to call this function once, even
+ * if you are writing an interlaced image.
+ */
+void PNGAPI
+png_write_image(png_structp png_ptr, png_bytepp image)
+{
+   png_uint_32 i; /* row index */
+   int pass, num_pass; /* pass variables */
+   png_bytepp rp; /* points to current row */
+
+   png_debug(1, "in png_write_image\n");
+#if defined(PNG_WRITE_INTERLACING_SUPPORTED)
+   /* intialize interlace handling.  If image is not interlaced,
+      this will set pass to 1 */
+   num_pass = png_set_interlace_handling(png_ptr);
+#else
+   num_pass = 1;
+#endif
+   /* loop through passes */
+   for (pass = 0; pass < num_pass; pass++)
+   {
+      /* loop through image */
+      for (i = 0, rp = image; i < png_ptr->height; i++, rp++)
+      {
+         png_write_row(png_ptr, *rp);
+      }
+   }
+}
+
+/* called by user to write a row of image data */
+void PNGAPI
+png_write_row(png_structp png_ptr, png_bytep row)
+{
+   png_debug2(1, "in png_write_row (row %ld, pass %d)\n",
+      png_ptr->row_number, png_ptr->pass);
+   /* initialize transformations and other stuff if first time */
+   if (png_ptr->row_number == 0 && png_ptr->pass == 0)
+   {
+   /* make sure we wrote the header info */
+   if (!(png_ptr->mode & PNG_WROTE_INFO_BEFORE_PLTE))
+      png_error(png_ptr,
+         "png_write_info was never called before png_write_row.");
+
+   /* check for transforms that have been set but were defined out */
+#if !defined(PNG_WRITE_INVERT_SUPPORTED) && defined(PNG_READ_INVERT_SUPPORTED)
+   if (png_ptr->transformations & PNG_INVERT_MONO)
+      png_warning(png_ptr, "PNG_WRITE_INVERT_SUPPORTED is not defined.");
+#endif
+#if !defined(PNG_WRITE_FILLER_SUPPORTED) && defined(PNG_READ_FILLER_SUPPORTED)
+   if (png_ptr->transformations & PNG_FILLER)
+      png_warning(png_ptr, "PNG_WRITE_FILLER_SUPPORTED is not defined.");
+#endif
+#if !defined(PNG_WRITE_PACKSWAP_SUPPORTED) && defined(PNG_READ_PACKSWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACKSWAP)
+      png_warning(png_ptr, "PNG_WRITE_PACKSWAP_SUPPORTED is not defined.");
+#endif
+#if !defined(PNG_WRITE_PACK_SUPPORTED) && defined(PNG_READ_PACK_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACK)
+      png_warning(png_ptr, "PNG_WRITE_PACK_SUPPORTED is not defined.");
+#endif
+#if !defined(PNG_WRITE_SHIFT_SUPPORTED) && defined(PNG_READ_SHIFT_SUPPORTED)
+   if (png_ptr->transformations & PNG_SHIFT)
+      png_warning(png_ptr, "PNG_WRITE_SHIFT_SUPPORTED is not defined.");
+#endif
+#if !defined(PNG_WRITE_BGR_SUPPORTED) && defined(PNG_READ_BGR_SUPPORTED)
+   if (png_ptr->transformations & PNG_BGR)
+      png_warning(png_ptr, "PNG_WRITE_BGR_SUPPORTED is not defined.");
+#endif
+#if !defined(PNG_WRITE_SWAP_SUPPORTED) && defined(PNG_READ_SWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_SWAP_BYTES)
+      png_warning(png_ptr, "PNG_WRITE_SWAP_SUPPORTED is not defined.");
+#endif
+
+      png_write_start_row(png_ptr);
+   }
+
+#if defined(PNG_WRITE_INTERLACING_SUPPORTED)
+   /* if interlaced and not interested in row, return */
+   if (png_ptr->interlaced && (png_ptr->transformations & PNG_INTERLACE))
+   {
+      switch (png_ptr->pass)
+      {
+         case 0:
+            if (png_ptr->row_number & 0x07)
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 1:
+            if ((png_ptr->row_number & 0x07) || png_ptr->width < 5)
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 2:
+            if ((png_ptr->row_number & 0x07) != 4)
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 3:
+            if ((png_ptr->row_number & 0x03) || png_ptr->width < 3)
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 4:
+            if ((png_ptr->row_number & 0x03) != 2)
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 5:
+            if ((png_ptr->row_number & 0x01) || png_ptr->width < 2)
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+         case 6:
+            if (!(png_ptr->row_number & 0x01))
+            {
+               png_write_finish_row(png_ptr);
+               return;
+            }
+            break;
+      }
+   }
+#endif
+
+   /* set up row info for transformations */
+   png_ptr->row_info.color_type = png_ptr->color_type;
+   png_ptr->row_info.width = png_ptr->usr_width;
+   png_ptr->row_info.channels = png_ptr->usr_channels;
+   png_ptr->row_info.bit_depth = png_ptr->usr_bit_depth;
+   png_ptr->row_info.pixel_depth = (png_byte)(png_ptr->row_info.bit_depth *
+      png_ptr->row_info.channels);
+
+   png_ptr->row_info.rowbytes = ((png_ptr->row_info.width *
+      (png_uint_32)png_ptr->row_info.pixel_depth + 7) >> 3);
+
+   png_debug1(3, "row_info->color_type = %d\n", png_ptr->row_info.color_type);
+   png_debug1(3, "row_info->width = %lu\n", png_ptr->row_info.width);
+   png_debug1(3, "row_info->channels = %d\n", png_ptr->row_info.channels);
+   png_debug1(3, "row_info->bit_depth = %d\n", png_ptr->row_info.bit_depth);
+   png_debug1(3, "row_info->pixel_depth = %d\n", png_ptr->row_info.pixel_depth);
+   png_debug1(3, "row_info->rowbytes = %lu\n", png_ptr->row_info.rowbytes);
+
+   /* Copy user's row into buffer, leaving room for filter byte. */
+   png_memcpy_check(png_ptr, png_ptr->row_buf + 1, row,
+      png_ptr->row_info.rowbytes);
+
+#if defined(PNG_WRITE_INTERLACING_SUPPORTED)
+   /* handle interlacing */
+   if (png_ptr->interlaced && png_ptr->pass < 6 &&
+      (png_ptr->transformations & PNG_INTERLACE))
+   {
+      png_do_write_interlace(&(png_ptr->row_info),
+         png_ptr->row_buf + 1, png_ptr->pass);
+      /* this should always get caught above, but still ... */
+      if (!(png_ptr->row_info.width))
+      {
+         png_write_finish_row(png_ptr);
+         return;
+      }
+   }
+#endif
+
+   /* handle other transformations */
+   if (png_ptr->transformations)
+      png_do_write_transformations(png_ptr);
+
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   /* Write filter_method 64 (intrapixel differencing) only if
+    * 1. Libpng was compiled with PNG_MNG_FEATURES_SUPPORTED and
+    * 2. Libpng did not write a PNG signature (this filter_method is only
+    *    used in PNG datastreams that are embedded in MNG datastreams) and
+    * 3. The application called png_permit_mng_features with a mask that
+    *    included PNG_FLAG_MNG_FILTER_64 and
+    * 4. The filter_method is 64 and
+    * 5. The color_type is RGB or RGBA
+    */
+   if((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) &&
+      (png_ptr->filter_type == PNG_INTRAPIXEL_DIFFERENCING))
+   {
+      /* Intrapixel differencing */
+      png_do_write_intrapixel(&(png_ptr->row_info), png_ptr->row_buf + 1);
+   }
+#endif
+
+   /* Find a filter if necessary, filter the row and write it out. */
+   png_write_find_filter(png_ptr, &(png_ptr->row_info));
+
+   if (png_ptr->write_row_fn != NULL)
+      (*(png_ptr->write_row_fn))(png_ptr, png_ptr->row_number, png_ptr->pass);
+}
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+/* Set the automatic flush interval or 0 to turn flushing off */
+void PNGAPI
+png_set_flush(png_structp png_ptr, int nrows)
+{
+   png_debug(1, "in png_set_flush\n");
+   png_ptr->flush_dist = (nrows < 0 ? 0 : nrows);
+}
+
+/* flush the current output buffers now */
+void PNGAPI
+png_write_flush(png_structp png_ptr)
+{
+   int wrote_IDAT;
+
+   png_debug(1, "in png_write_flush\n");
+   /* We have already written out all of the data */
+   if (png_ptr->row_number >= png_ptr->num_rows)
+     return;
+
+   do
+   {
+      int ret;
+
+      /* compress the data */
+      ret = deflate(&png_ptr->zstream, Z_SYNC_FLUSH);
+      wrote_IDAT = 0;
+
+      /* check for compression errors */
+      if (ret != Z_OK)
+      {
+         if (png_ptr->zstream.msg != NULL)
+            png_error(png_ptr, png_ptr->zstream.msg);
+         else
+            png_error(png_ptr, "zlib error");
+      }
+
+      if (!(png_ptr->zstream.avail_out))
+      {
+         /* write the IDAT and reset the zlib output buffer */
+         png_write_IDAT(png_ptr, png_ptr->zbuf,
+                        png_ptr->zbuf_size);
+         png_ptr->zstream.next_out = png_ptr->zbuf;
+         png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+         wrote_IDAT = 1;
+      }
+   } while(wrote_IDAT == 1);
+
+   /* If there is any data left to be output, write it into a new IDAT */
+   if (png_ptr->zbuf_size != png_ptr->zstream.avail_out)
+   {
+      /* write the IDAT and reset the zlib output buffer */
+      png_write_IDAT(png_ptr, png_ptr->zbuf,
+                     png_ptr->zbuf_size - png_ptr->zstream.avail_out);
+      png_ptr->zstream.next_out = png_ptr->zbuf;
+      png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+   }
+   png_ptr->flush_rows = 0;
+   png_flush(png_ptr);
+}
+#endif /* PNG_WRITE_FLUSH_SUPPORTED */
+
+/* free all memory used by the write */
+void PNGAPI
+png_destroy_write_struct(png_structpp png_ptr_ptr, png_infopp info_ptr_ptr)
+{
+   png_structp png_ptr = NULL;
+   png_infop info_ptr = NULL;
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_free_ptr free_fn = NULL;
+   png_voidp mem_ptr = NULL;
+#endif
+
+   png_debug(1, "in png_destroy_write_struct\n");
+   if (png_ptr_ptr != NULL)
+   {
+      png_ptr = *png_ptr_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+      free_fn = png_ptr->free_fn;
+#endif
+   }
+
+   if (info_ptr_ptr != NULL)
+      info_ptr = *info_ptr_ptr;
+
+   if (info_ptr != NULL)
+   {
+      png_free_data(png_ptr, info_ptr, PNG_FREE_ALL, -1);
+
+#if defined(PNG_UNKNOWN_CHUNKS_SUPPORTED)
+      if (png_ptr->num_chunk_list)
+      {
+         png_free(png_ptr, png_ptr->chunk_list);
+         png_ptr->chunk_list=NULL;
+         png_ptr->num_chunk_list=0;
+      }
+#endif
+
+#ifdef PNG_USER_MEM_SUPPORTED
+      png_destroy_struct_2((png_voidp)info_ptr, (png_free_ptr)free_fn,
+         (png_voidp)mem_ptr);
+#else
+      png_destroy_struct((png_voidp)info_ptr);
+#endif
+      *info_ptr_ptr = NULL;
+   }
+
+   if (png_ptr != NULL)
+   {
+      png_write_destroy(png_ptr);
+#ifdef PNG_USER_MEM_SUPPORTED
+      png_destroy_struct_2((png_voidp)png_ptr, (png_free_ptr)free_fn,
+         (png_voidp)mem_ptr);
+#else
+      png_destroy_struct((png_voidp)png_ptr);
+#endif
+      *png_ptr_ptr = NULL;
+   }
+}
+
+
+/* Free any memory used in png_ptr struct (old method) */
+void /* PRIVATE */
+png_write_destroy(png_structp png_ptr)
+{
+#ifdef PNG_SETJMP_SUPPORTED
+   jmp_buf tmp_jmp; /* save jump buffer */
+#endif
+   png_error_ptr error_fn;
+   png_error_ptr warning_fn;
+   png_voidp error_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_free_ptr free_fn;
+#endif
+
+   png_debug(1, "in png_write_destroy\n");
+   /* free any memory zlib uses */
+   deflateEnd(&png_ptr->zstream);
+
+   /* free our memory.  png_free checks NULL for us. */
+   png_free(png_ptr, png_ptr->zbuf);
+   png_free(png_ptr, png_ptr->row_buf);
+   png_free(png_ptr, png_ptr->prev_row);
+   png_free(png_ptr, png_ptr->sub_row);
+   png_free(png_ptr, png_ptr->up_row);
+   png_free(png_ptr, png_ptr->avg_row);
+   png_free(png_ptr, png_ptr->paeth_row);
+
+#if defined(PNG_TIME_RFC1123_SUPPORTED)
+   png_free(png_ptr, png_ptr->time_buffer);
+#endif
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+   png_free(png_ptr, png_ptr->prev_filters);
+   png_free(png_ptr, png_ptr->filter_weights);
+   png_free(png_ptr, png_ptr->inv_filter_weights);
+   png_free(png_ptr, png_ptr->filter_costs);
+   png_free(png_ptr, png_ptr->inv_filter_costs);
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+   /* reset structure */
+   png_memcpy(tmp_jmp, png_ptr->jmpbuf, sizeof (jmp_buf));
+#endif
+
+   error_fn = png_ptr->error_fn;
+   warning_fn = png_ptr->warning_fn;
+   error_ptr = png_ptr->error_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+   free_fn = png_ptr->free_fn;
+#endif
+
+   png_memset(png_ptr, 0, sizeof (png_struct));
+
+   png_ptr->error_fn = error_fn;
+   png_ptr->warning_fn = warning_fn;
+   png_ptr->error_ptr = error_ptr;
+#ifdef PNG_USER_MEM_SUPPORTED
+   png_ptr->free_fn = free_fn;
+#endif
+
+#ifdef PNG_SETJMP_SUPPORTED
+   png_memcpy(png_ptr->jmpbuf, tmp_jmp, sizeof (jmp_buf));
+#endif
+}
+
+/* Allow the application to select one or more row filters to use. */
+void PNGAPI
+png_set_filter(png_structp png_ptr, int method, int filters)
+{
+   png_debug(1, "in png_set_filter\n");
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   if((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) &&
+      (method == PNG_INTRAPIXEL_DIFFERENCING))
+         method = PNG_FILTER_TYPE_BASE;
+#endif
+   if (method == PNG_FILTER_TYPE_BASE)
+   {
+      switch (filters & (PNG_ALL_FILTERS | 0x07))
+      {
+         case 5:
+         case 6:
+         case 7: png_warning(png_ptr, "Unknown row filter for method 0");
+         case PNG_FILTER_VALUE_NONE:  png_ptr->do_filter=PNG_FILTER_NONE; break;
+         case PNG_FILTER_VALUE_SUB:   png_ptr->do_filter=PNG_FILTER_SUB;  break;
+         case PNG_FILTER_VALUE_UP:    png_ptr->do_filter=PNG_FILTER_UP;   break;
+         case PNG_FILTER_VALUE_AVG:   png_ptr->do_filter=PNG_FILTER_AVG;  break;
+         case PNG_FILTER_VALUE_PAETH: png_ptr->do_filter=PNG_FILTER_PAETH;break;
+         default: png_ptr->do_filter = (png_byte)filters; break;
+      }
+
+      /* If we have allocated the row_buf, this means we have already started
+       * with the image and we should have allocated all of the filter buffers
+       * that have been selected.  If prev_row isn't already allocated, then
+       * it is too late to start using the filters that need it, since we
+       * will be missing the data in the previous row.  If an application
+       * wants to start and stop using particular filters during compression,
+       * it should start out with all of the filters, and then add and
+       * remove them after the start of compression.
+       */
+      if (png_ptr->row_buf != NULL)
+      {
+         if ((png_ptr->do_filter & PNG_FILTER_SUB) && png_ptr->sub_row == NULL)
+         {
+            png_ptr->sub_row = (png_bytep)png_malloc(png_ptr,
+              (png_ptr->rowbytes + 1));
+            png_ptr->sub_row[0] = PNG_FILTER_VALUE_SUB;
+         }
+
+         if ((png_ptr->do_filter & PNG_FILTER_UP) && png_ptr->up_row == NULL)
+         {
+            if (png_ptr->prev_row == NULL)
+            {
+               png_warning(png_ptr, "Can't add Up filter after starting");
+               png_ptr->do_filter &= ~PNG_FILTER_UP;
+            }
+            else
+            {
+               png_ptr->up_row = (png_bytep)png_malloc(png_ptr,
+                  (png_ptr->rowbytes + 1));
+               png_ptr->up_row[0] = PNG_FILTER_VALUE_UP;
+            }
+         }
+
+         if ((png_ptr->do_filter & PNG_FILTER_AVG) && png_ptr->avg_row == NULL)
+         {
+            if (png_ptr->prev_row == NULL)
+            {
+               png_warning(png_ptr, "Can't add Average filter after starting");
+               png_ptr->do_filter &= ~PNG_FILTER_AVG;
+            }
+            else
+            {
+               png_ptr->avg_row = (png_bytep)png_malloc(png_ptr,
+                  (png_ptr->rowbytes + 1));
+               png_ptr->avg_row[0] = PNG_FILTER_VALUE_AVG;
+            }
+         }
+
+         if ((png_ptr->do_filter & PNG_FILTER_PAETH) &&
+             png_ptr->paeth_row == NULL)
+         {
+            if (png_ptr->prev_row == NULL)
+            {
+               png_warning(png_ptr, "Can't add Paeth filter after starting");
+               png_ptr->do_filter &= (png_byte)(~PNG_FILTER_PAETH);
+            }
+            else
+            {
+               png_ptr->paeth_row = (png_bytep)png_malloc(png_ptr,
+                  (png_ptr->rowbytes + 1));
+               png_ptr->paeth_row[0] = PNG_FILTER_VALUE_PAETH;
+            }
+         }
+
+         if (png_ptr->do_filter == PNG_NO_FILTERS)
+            png_ptr->do_filter = PNG_FILTER_NONE;
+      }
+   }
+   else
+      png_error(png_ptr, "Unknown custom filter method");
+}
+
+/* This allows us to influence the way in which libpng chooses the "best"
+ * filter for the current scanline.  While the "minimum-sum-of-absolute-
+ * differences metric is relatively fast and effective, there is some
+ * question as to whether it can be improved upon by trying to keep the
+ * filtered data going to zlib more consistent, hopefully resulting in
+ * better compression.
+ */
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)      /* GRR 970116 */
+void PNGAPI
+png_set_filter_heuristics(png_structp png_ptr, int heuristic_method,
+   int num_weights, png_doublep filter_weights,
+   png_doublep filter_costs)
+{
+   int i;
+
+   png_debug(1, "in png_set_filter_heuristics\n");
+   if (heuristic_method >= PNG_FILTER_HEURISTIC_LAST)
+   {
+      png_warning(png_ptr, "Unknown filter heuristic method");
+      return;
+   }
+
+   if (heuristic_method == PNG_FILTER_HEURISTIC_DEFAULT)
+   {
+      heuristic_method = PNG_FILTER_HEURISTIC_UNWEIGHTED;
+   }
+
+   if (num_weights < 0 || filter_weights == NULL ||
+      heuristic_method == PNG_FILTER_HEURISTIC_UNWEIGHTED)
+   {
+      num_weights = 0;
+   }
+
+   png_ptr->num_prev_filters = (png_byte)num_weights;
+   png_ptr->heuristic_method = (png_byte)heuristic_method;
+
+   if (num_weights > 0)
+   {
+      if (png_ptr->prev_filters == NULL)
+      {
+         png_ptr->prev_filters = (png_bytep)png_malloc(png_ptr,
+            (png_uint_32)(sizeof(png_byte) * num_weights));
+
+         /* To make sure that the weighting starts out fairly */
+         for (i = 0; i < num_weights; i++)
+         {
+            png_ptr->prev_filters[i] = 255;
+         }
+      }
+
+      if (png_ptr->filter_weights == NULL)
+      {
+         png_ptr->filter_weights = (png_uint_16p)png_malloc(png_ptr,
+            (png_uint_32)(sizeof(png_uint_16) * num_weights));
+
+         png_ptr->inv_filter_weights = (png_uint_16p)png_malloc(png_ptr,
+            (png_uint_32)(sizeof(png_uint_16) * num_weights));
+         for (i = 0; i < num_weights; i++)
+         {
+            png_ptr->inv_filter_weights[i] =
+            png_ptr->filter_weights[i] = PNG_WEIGHT_FACTOR;
+         }
+      }
+
+      for (i = 0; i < num_weights; i++)
+      {
+         if (filter_weights[i] < 0.0)
+         {
+            png_ptr->inv_filter_weights[i] =
+            png_ptr->filter_weights[i] = PNG_WEIGHT_FACTOR;
+         }
+         else
+         {
+            png_ptr->inv_filter_weights[i] =
+               (png_uint_16)((double)PNG_WEIGHT_FACTOR*filter_weights[i]+0.5);
+            png_ptr->filter_weights[i] =
+               (png_uint_16)((double)PNG_WEIGHT_FACTOR/filter_weights[i]+0.5);
+         }
+      }
+   }
+
+   /* If, in the future, there are other filter methods, this would
+    * need to be based on png_ptr->filter.
+    */
+   if (png_ptr->filter_costs == NULL)
+   {
+      png_ptr->filter_costs = (png_uint_16p)png_malloc(png_ptr,
+         (png_uint_32)(sizeof(png_uint_16) * PNG_FILTER_VALUE_LAST));
+
+      png_ptr->inv_filter_costs = (png_uint_16p)png_malloc(png_ptr,
+         (png_uint_32)(sizeof(png_uint_16) * PNG_FILTER_VALUE_LAST));
+
+      for (i = 0; i < PNG_FILTER_VALUE_LAST; i++)
+      {
+         png_ptr->inv_filter_costs[i] =
+         png_ptr->filter_costs[i] = PNG_COST_FACTOR;
+      }
+   }
+
+   /* Here is where we set the relative costs of the different filters.  We
+    * should take the desired compression level into account when setting
+    * the costs, so that Paeth, for instance, has a high relative cost at low
+    * compression levels, while it has a lower relative cost at higher
+    * compression settings.  The filter types are in order of increasing
+    * relative cost, so it would be possible to do this with an algorithm.
+    */
+   for (i = 0; i < PNG_FILTER_VALUE_LAST; i++)
+   {
+      if (filter_costs == NULL || filter_costs[i] < 0.0)
+      {
+         png_ptr->inv_filter_costs[i] =
+         png_ptr->filter_costs[i] = PNG_COST_FACTOR;
+      }
+      else if (filter_costs[i] >= 1.0)
+      {
+         png_ptr->inv_filter_costs[i] =
+            (png_uint_16)((double)PNG_COST_FACTOR / filter_costs[i] + 0.5);
+         png_ptr->filter_costs[i] =
+            (png_uint_16)((double)PNG_COST_FACTOR * filter_costs[i] + 0.5);
+      }
+   }
+}
+#endif /* PNG_WRITE_WEIGHTED_FILTER_SUPPORTED */
+
+void PNGAPI
+png_set_compression_level(png_structp png_ptr, int level)
+{
+   png_debug(1, "in png_set_compression_level\n");
+   png_ptr->flags |= PNG_FLAG_ZLIB_CUSTOM_LEVEL;
+   png_ptr->zlib_level = level;
+}
+
+void PNGAPI
+png_set_compression_mem_level(png_structp png_ptr, int mem_level)
+{
+   png_debug(1, "in png_set_compression_mem_level\n");
+   png_ptr->flags |= PNG_FLAG_ZLIB_CUSTOM_MEM_LEVEL;
+   png_ptr->zlib_mem_level = mem_level;
+}
+
+void PNGAPI
+png_set_compression_strategy(png_structp png_ptr, int strategy)
+{
+   png_debug(1, "in png_set_compression_strategy\n");
+   png_ptr->flags |= PNG_FLAG_ZLIB_CUSTOM_STRATEGY;
+   png_ptr->zlib_strategy = strategy;
+}
+
+void PNGAPI
+png_set_compression_window_bits(png_structp png_ptr, int window_bits)
+{
+   if (window_bits > 15)
+      png_warning(png_ptr, "Only compression windows <= 32k supported by PNG");
+   else if (window_bits < 8)
+      png_warning(png_ptr, "Only compression windows >= 256 supported by PNG");
+#ifndef WBITS_8_OK
+   /* avoid libpng bug with 256-byte windows */
+   if (window_bits == 8)
+     {
+       png_warning(png_ptr, "Compression window is being reset to 512");
+       window_bits=9;
+     }
+#endif
+   png_ptr->flags |= PNG_FLAG_ZLIB_CUSTOM_WINDOW_BITS;
+   png_ptr->zlib_window_bits = window_bits;
+}
+
+void PNGAPI
+png_set_compression_method(png_structp png_ptr, int method)
+{
+   png_debug(1, "in png_set_compression_method\n");
+   if (method != 8)
+      png_warning(png_ptr, "Only compression method 8 is supported by PNG");
+   png_ptr->flags |= PNG_FLAG_ZLIB_CUSTOM_METHOD;
+   png_ptr->zlib_method = method;
+}
+
+void PNGAPI
+png_set_write_status_fn(png_structp png_ptr, png_write_status_ptr write_row_fn)
+{
+   png_ptr->write_row_fn = write_row_fn;
+}
+
+#if defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED)
+void PNGAPI
+png_set_write_user_transform_fn(png_structp png_ptr, png_user_transform_ptr
+   write_user_transform_fn)
+{
+   png_debug(1, "in png_set_write_user_transform_fn\n");
+   png_ptr->transformations |= PNG_USER_TRANSFORM;
+   png_ptr->write_user_transform_fn = write_user_transform_fn;
+}
+#endif
+
+
+#if defined(PNG_INFO_IMAGE_SUPPORTED)
+void PNGAPI
+png_write_png(png_structp png_ptr, png_infop info_ptr,
+              int transforms, voidp params)
+{
+#if defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+   /* invert the alpha channel from opacity to transparency */
+   if (transforms & PNG_TRANSFORM_INVERT_ALPHA)
+       png_set_invert_alpha(png_ptr);
+#endif
+
+   /* Write the file header information. */
+   png_write_info(png_ptr, info_ptr);
+
+   /* ------ these transformations don't touch the info structure ------- */
+
+#if defined(PNG_WRITE_INVERT_SUPPORTED)
+   /* invert monochrome pixels */
+   if (transforms & PNG_TRANSFORM_INVERT_MONO)
+       png_set_invert_mono(png_ptr);
+#endif
+
+#if defined(PNG_WRITE_SHIFT_SUPPORTED)
+   /* Shift the pixels up to a legal bit depth and fill in
+    * as appropriate to correctly scale the image.
+    */
+   if ((transforms & PNG_TRANSFORM_SHIFT)
+               && (info_ptr->valid & PNG_INFO_sBIT))
+       png_set_shift(png_ptr, &info_ptr->sig_bit);
+#endif
+
+#if defined(PNG_WRITE_PACK_SUPPORTED)
+   /* pack pixels into bytes */
+   if (transforms & PNG_TRANSFORM_PACKING)
+       png_set_packing(png_ptr);
+#endif
+
+#if defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED)
+   /* swap location of alpha bytes from ARGB to RGBA */
+   if (transforms & PNG_TRANSFORM_SWAP_ALPHA)
+       png_set_swap_alpha(png_ptr);
+#endif
+
+#if defined(PNG_WRITE_FILLER_SUPPORTED)
+   /* Get rid of filler (OR ALPHA) bytes, pack XRGB/RGBX/ARGB/RGBA into
+    * RGB (4 channels -> 3 channels). The second parameter is not used.
+    */
+   if (transforms & PNG_TRANSFORM_STRIP_FILLER)
+       png_set_filler(png_ptr, 0, PNG_FILLER_BEFORE);
+#endif
+
+#if defined(PNG_WRITE_BGR_SUPPORTED)
+   /* flip BGR pixels to RGB */
+   if (transforms & PNG_TRANSFORM_BGR)
+       png_set_bgr(png_ptr);
+#endif
+
+#if defined(PNG_WRITE_SWAP_SUPPORTED)
+   /* swap bytes of 16-bit files to most significant byte first */
+   if (transforms & PNG_TRANSFORM_SWAP_ENDIAN)
+       png_set_swap(png_ptr);
+#endif
+
+#if defined(PNG_WRITE_PACKSWAP_SUPPORTED)
+   /* swap bits of 1, 2, 4 bit packed pixel formats */
+   if (transforms & PNG_TRANSFORM_PACKSWAP)
+       png_set_packswap(png_ptr);
+#endif
+
+   /* ----------------------- end of transformations ------------------- */
+
+   /* write the bits */
+   if (info_ptr->valid & PNG_INFO_IDAT)
+       png_write_image(png_ptr, info_ptr->row_pointers);
+
+   /* It is REQUIRED to call this to finish writing the rest of the file */
+   png_write_end(png_ptr, info_ptr);
+
+   if(transforms == 0 || params == NULL)
+      /* quiet compiler warnings */ return;
+}
+#endif
+#endif /* PNG_WRITE_SUPPORTED */
diff --git a/png/pngwtran.c b/png/pngwtran.c
new file mode 100644
index 000000000..55e691d28
--- /dev/null
+++ b/png/pngwtran.c
@@ -0,0 +1,563 @@
+
+/* pngwtran.c - transforms the data in a row for PNG writers
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+#ifdef PNG_WRITE_SUPPORTED
+
+/* Transform the data according to the user's wishes.  The order of
+ * transformations is significant.
+ */
+void /* PRIVATE */
+png_do_write_transformations(png_structp png_ptr)
+{
+   png_debug(1, "in png_do_write_transformations\n");
+
+   if (png_ptr == NULL)
+      return;
+
+#if defined(PNG_WRITE_USER_TRANSFORM_SUPPORTED)
+   if (png_ptr->transformations & PNG_USER_TRANSFORM)
+      if(png_ptr->write_user_transform_fn != NULL)
+        (*(png_ptr->write_user_transform_fn)) /* user write transform function */
+          (png_ptr,                    /* png_ptr */
+           &(png_ptr->row_info),       /* row_info:     */
+             /*  png_uint_32 width;          width of row */
+             /*  png_uint_32 rowbytes;       number of bytes in row */
+             /*  png_byte color_type;        color type of pixels */
+             /*  png_byte bit_depth;         bit depth of samples */
+             /*  png_byte channels;          number of channels (1-4) */
+             /*  png_byte pixel_depth;       bits per pixel (depth*channels) */
+           png_ptr->row_buf + 1);      /* start of pixel data for row */
+#endif
+#if defined(PNG_WRITE_FILLER_SUPPORTED)
+   if (png_ptr->transformations & PNG_FILLER)
+      png_do_strip_filler(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         png_ptr->flags);
+#endif
+#if defined(PNG_WRITE_PACKSWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACKSWAP)
+      png_do_packswap(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+#if defined(PNG_WRITE_PACK_SUPPORTED)
+   if (png_ptr->transformations & PNG_PACK)
+      png_do_pack(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         (png_uint_32)png_ptr->bit_depth);
+#endif
+#if defined(PNG_WRITE_SWAP_SUPPORTED)
+   if (png_ptr->transformations & PNG_SWAP_BYTES)
+      png_do_swap(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+#if defined(PNG_WRITE_SHIFT_SUPPORTED)
+   if (png_ptr->transformations & PNG_SHIFT)
+      png_do_shift(&(png_ptr->row_info), png_ptr->row_buf + 1,
+         &(png_ptr->shift));
+#endif
+#if defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+   if (png_ptr->transformations & PNG_INVERT_ALPHA)
+      png_do_write_invert_alpha(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+#if defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED)
+   if (png_ptr->transformations & PNG_SWAP_ALPHA)
+      png_do_write_swap_alpha(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+#if defined(PNG_WRITE_BGR_SUPPORTED)
+   if (png_ptr->transformations & PNG_BGR)
+      png_do_bgr(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+#if defined(PNG_WRITE_INVERT_SUPPORTED)
+   if (png_ptr->transformations & PNG_INVERT_MONO)
+      png_do_invert(&(png_ptr->row_info), png_ptr->row_buf + 1);
+#endif
+}
+
+#if defined(PNG_WRITE_PACK_SUPPORTED)
+/* Pack pixels into bytes.  Pass the true bit depth in bit_depth.  The
+ * row_info bit depth should be 8 (one pixel per byte).  The channels
+ * should be 1 (this only happens on grayscale and paletted images).
+ */
+void /* PRIVATE */
+png_do_pack(png_row_infop row_info, png_bytep row, png_uint_32 bit_depth)
+{
+   png_debug(1, "in png_do_pack\n");
+   if (row_info->bit_depth == 8 &&
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+      row_info->channels == 1)
+   {
+      switch ((int)bit_depth)
+      {
+         case 1:
+         {
+            png_bytep sp, dp;
+            int mask, v;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            sp = row;
+            dp = row;
+            mask = 0x80;
+            v = 0;
+
+            for (i = 0; i < row_width; i++)
+            {
+               if (*sp != 0)
+                  v |= mask;
+               sp++;
+               if (mask > 1)
+                  mask >>= 1;
+               else
+               {
+                  mask = 0x80;
+                  *dp = (png_byte)v;
+                  dp++;
+                  v = 0;
+               }
+            }
+            if (mask != 0x80)
+               *dp = (png_byte)v;
+            break;
+         }
+         case 2:
+         {
+            png_bytep sp, dp;
+            int shift, v;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            sp = row;
+            dp = row;
+            shift = 6;
+            v = 0;
+            for (i = 0; i < row_width; i++)
+            {
+               png_byte value;
+
+               value = (png_byte)(*sp & 0x03);
+               v |= (value << shift);
+               if (shift == 0)
+               {
+                  shift = 6;
+                  *dp = (png_byte)v;
+                  dp++;
+                  v = 0;
+               }
+               else
+                  shift -= 2;
+               sp++;
+            }
+            if (shift != 6)
+               *dp = (png_byte)v;
+            break;
+         }
+         case 4:
+         {
+            png_bytep sp, dp;
+            int shift, v;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            sp = row;
+            dp = row;
+            shift = 4;
+            v = 0;
+            for (i = 0; i < row_width; i++)
+            {
+               png_byte value;
+
+               value = (png_byte)(*sp & 0x0f);
+               v |= (value << shift);
+
+               if (shift == 0)
+               {
+                  shift = 4;
+                  *dp = (png_byte)v;
+                  dp++;
+                  v = 0;
+               }
+               else
+                  shift -= 4;
+
+               sp++;
+            }
+            if (shift != 4)
+               *dp = (png_byte)v;
+            break;
+         }
+      }
+      row_info->bit_depth = (png_byte)bit_depth;
+      row_info->pixel_depth = (png_byte)(bit_depth * row_info->channels);
+      row_info->rowbytes =
+         ((row_info->width * row_info->pixel_depth + 7) >> 3);
+   }
+}
+#endif
+
+#if defined(PNG_WRITE_SHIFT_SUPPORTED)
+/* Shift pixel values to take advantage of whole range.  Pass the
+ * true number of bits in bit_depth.  The row should be packed
+ * according to row_info->bit_depth.  Thus, if you had a row of
+ * bit depth 4, but the pixels only had values from 0 to 7, you
+ * would pass 3 as bit_depth, and this routine would translate the
+ * data to 0 to 15.
+ */
+void /* PRIVATE */
+png_do_shift(png_row_infop row_info, png_bytep row, png_color_8p bit_depth)
+{
+   png_debug(1, "in png_do_shift\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL &&
+#else
+   if (
+#endif
+      row_info->color_type != PNG_COLOR_TYPE_PALETTE)
+   {
+      int shift_start[4], shift_dec[4];
+      int channels = 0;
+
+      if (row_info->color_type & PNG_COLOR_MASK_COLOR)
+      {
+         shift_start[channels] = row_info->bit_depth - bit_depth->red;
+         shift_dec[channels] = bit_depth->red;
+         channels++;
+         shift_start[channels] = row_info->bit_depth - bit_depth->green;
+         shift_dec[channels] = bit_depth->green;
+         channels++;
+         shift_start[channels] = row_info->bit_depth - bit_depth->blue;
+         shift_dec[channels] = bit_depth->blue;
+         channels++;
+      }
+      else
+      {
+         shift_start[channels] = row_info->bit_depth - bit_depth->gray;
+         shift_dec[channels] = bit_depth->gray;
+         channels++;
+      }
+      if (row_info->color_type & PNG_COLOR_MASK_ALPHA)
+      {
+         shift_start[channels] = row_info->bit_depth - bit_depth->alpha;
+         shift_dec[channels] = bit_depth->alpha;
+         channels++;
+      }
+
+      /* with low row depths, could only be grayscale, so one channel */
+      if (row_info->bit_depth < 8)
+      {
+         png_bytep bp = row;
+         png_uint_32 i;
+         png_byte mask;
+         png_uint_32 row_bytes = row_info->rowbytes;
+
+         if (bit_depth->gray == 1 && row_info->bit_depth == 2)
+            mask = 0x55;
+         else if (row_info->bit_depth == 4 && bit_depth->gray == 3)
+            mask = 0x11;
+         else
+            mask = 0xff;
+
+         for (i = 0; i < row_bytes; i++, bp++)
+         {
+            png_uint_16 v;
+            int j;
+
+            v = *bp;
+            *bp = 0;
+            for (j = shift_start[0]; j > -shift_dec[0]; j -= shift_dec[0])
+            {
+               if (j > 0)
+                  *bp |= (png_byte)((v << j) & 0xff);
+               else
+                  *bp |= (png_byte)((v >> (-j)) & mask);
+            }
+         }
+      }
+      else if (row_info->bit_depth == 8)
+      {
+         png_bytep bp = row;
+         png_uint_32 i;
+         png_uint_32 istop = channels * row_info->width;
+
+         for (i = 0; i < istop; i++, bp++)
+         {
+
+            png_uint_16 v;
+            int j;
+            int c = (int)(i%channels);
+
+            v = *bp;
+            *bp = 0;
+            for (j = shift_start[c]; j > -shift_dec[c]; j -= shift_dec[c])
+            {
+               if (j > 0)
+                  *bp |= (png_byte)((v << j) & 0xff);
+               else
+                  *bp |= (png_byte)((v >> (-j)) & 0xff);
+            }
+         }
+      }
+      else
+      {
+         png_bytep bp;
+         png_uint_32 i;
+         png_uint_32 istop = channels * row_info->width;
+
+         for (bp = row, i = 0; i < istop; i++)
+         {
+            int c = (int)(i%channels);
+            png_uint_16 value, v;
+            int j;
+
+            v = (png_uint_16)(((png_uint_16)(*bp) << 8) + *(bp + 1));
+            value = 0;
+            for (j = shift_start[c]; j > -shift_dec[c]; j -= shift_dec[c])
+            {
+               if (j > 0)
+                  value |= (png_uint_16)((v << j) & (png_uint_16)0xffff);
+               else
+                  value |= (png_uint_16)((v >> (-j)) & (png_uint_16)0xffff);
+            }
+            *bp++ = (png_byte)(value >> 8);
+            *bp++ = (png_byte)(value & 0xff);
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_WRITE_SWAP_ALPHA_SUPPORTED)
+void /* PRIVATE */
+png_do_write_swap_alpha(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_write_swap_alpha\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+      {
+         /* This converts from ARGB to RGBA */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               png_byte save = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = save;
+            }
+         }
+         /* This converts from AARRGGBB to RRGGBBAA */
+         else
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               png_byte save[2];
+               save[0] = *(sp++);
+               save[1] = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = save[0];
+               *(dp++) = save[1];
+            }
+         }
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      {
+         /* This converts from AG to GA */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               png_byte save = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = save;
+            }
+         }
+         /* This converts from AAGG to GGAA */
+         else
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               png_byte save[2];
+               save[0] = *(sp++);
+               save[1] = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = save[0];
+               *(dp++) = save[1];
+            }
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_WRITE_INVERT_ALPHA_SUPPORTED)
+void /* PRIVATE */
+png_do_write_invert_alpha(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_write_invert_alpha\n");
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL)
+#endif
+   {
+      if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+      {
+         /* This inverts the alpha channel in RGBA */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = (png_byte)(255 - *(sp++));
+            }
+         }
+         /* This inverts the alpha channel in RRGGBBAA */
+         else
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = (png_byte)(255 - *(sp++));
+               *(dp++) = (png_byte)(255 - *(sp++));
+            }
+         }
+      }
+      else if (row_info->color_type == PNG_COLOR_TYPE_GRAY_ALPHA)
+      {
+         /* This inverts the alpha channel in GA */
+         if (row_info->bit_depth == 8)
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               *(dp++) = *(sp++);
+               *(dp++) = (png_byte)(255 - *(sp++));
+            }
+         }
+         /* This inverts the alpha channel in GGAA */
+         else
+         {
+            png_bytep sp, dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            for (i = 0, sp = dp = row; i < row_width; i++)
+            {
+               *(dp++) = *(sp++);
+               *(dp++) = *(sp++);
+               *(dp++) = (png_byte)(255 - *(sp++));
+               *(dp++) = (png_byte)(255 - *(sp++));
+            }
+         }
+      }
+   }
+}
+#endif
+
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+/* undoes intrapixel differencing  */
+void /* PRIVATE */
+png_do_write_intrapixel(png_row_infop row_info, png_bytep row)
+{
+   png_debug(1, "in png_do_write_intrapixel\n");
+   if (
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+       row != NULL && row_info != NULL &&
+#endif
+       (row_info->color_type & PNG_COLOR_MASK_COLOR))
+   {
+      int bytes_per_pixel;
+      png_uint_32 row_width = row_info->width;
+      if (row_info->bit_depth == 8)
+      {
+         png_bytep rp;
+         png_uint_32 i;
+
+         if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+            bytes_per_pixel = 3;
+         else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+            bytes_per_pixel = 4;
+         else
+            return;
+
+         for (i = 0, rp = row; i < row_width; i++, rp += bytes_per_pixel)
+         {
+            *(rp)   = (png_byte)((*rp     - *(rp+1))&0xff);
+            *(rp+2) = (png_byte)((*(rp+2) - *(rp+1))&0xff);
+         }
+      }
+      else if (row_info->bit_depth == 16)
+      {
+         png_bytep rp;
+         png_uint_32 i;
+
+         if (row_info->color_type == PNG_COLOR_TYPE_RGB)
+            bytes_per_pixel = 6;
+         else if (row_info->color_type == PNG_COLOR_TYPE_RGB_ALPHA)
+            bytes_per_pixel = 8;
+         else
+            return;
+
+         for (i = 0, rp = row; i < row_width; i++, rp += bytes_per_pixel)
+         {
+            png_uint_32 s0=*(rp  )<<8 | *(rp+1);
+            png_uint_32 s1=*(rp+2)<<8 | *(rp+3);
+            png_uint_32 s2=*(rp+4)<<8 | *(rp+5);
+            png_uint_32 red=(s0-s1)&0xffff;
+            png_uint_32 blue=(s2-s1)&0xffff;
+            *(rp  ) = (png_byte)((red>>8)&0xff);
+            *(rp+1) = (png_byte)(red&0xff);
+            *(rp+4) = (png_byte)((blue>>8)&0xff);
+            *(rp+5) = (png_byte)(blue&0xff);
+         }
+      }
+   }
+}
+#endif /* PNG_MNG_FEATURES_SUPPORTED */
+#endif /* PNG_WRITE_SUPPORTED */
diff --git a/png/pngwutil.c b/png/pngwutil.c
new file mode 100644
index 000000000..1f2d72442
--- /dev/null
+++ b/png/pngwutil.c
@@ -0,0 +1,2669 @@
+
+/* pngwutil.c - utilities to write a PNG file
+ *
+ * libpng 1.2.1 - December 12, 2001
+ * For conditions of distribution and use, see copyright notice in png.h
+ * Copyright (c) 1998-2001 Glenn Randers-Pehrson
+ * (Version 0.96 Copyright (c) 1996, 1997 Andreas Dilger)
+ * (Version 0.88 Copyright (c) 1995, 1996 Guy Eric Schalnat, Group 42, Inc.)
+ */
+
+#define PNG_INTERNAL
+#include "png.h"
+#ifdef PNG_WRITE_SUPPORTED
+
+/* Place a 32-bit number into a buffer in PNG byte order.  We work
+ * with unsigned numbers for convenience, although one supported
+ * ancillary chunk uses signed (two's complement) numbers.
+ */
+void /* PRIVATE */
+png_save_uint_32(png_bytep buf, png_uint_32 i)
+{
+   buf[0] = (png_byte)((i >> 24) & 0xff);
+   buf[1] = (png_byte)((i >> 16) & 0xff);
+   buf[2] = (png_byte)((i >> 8) & 0xff);
+   buf[3] = (png_byte)(i & 0xff);
+}
+
+#if defined(PNG_WRITE_pCAL_SUPPORTED) || defined(PNG_WRITE_oFFs_SUPPORTED)
+/* The png_save_int_32 function assumes integers are stored in two's
+ * complement format.  If this isn't the case, then this routine needs to
+ * be modified to write data in two's complement format.
+ */
+void /* PRIVATE */
+png_save_int_32(png_bytep buf, png_int_32 i)
+{
+   buf[0] = (png_byte)((i >> 24) & 0xff);
+   buf[1] = (png_byte)((i >> 16) & 0xff);
+   buf[2] = (png_byte)((i >> 8) & 0xff);
+   buf[3] = (png_byte)(i & 0xff);
+}
+#endif
+
+/* Place a 16-bit number into a buffer in PNG byte order.
+ * The parameter is declared unsigned int, not png_uint_16,
+ * just to avoid potential problems on pre-ANSI C compilers.
+ */
+void /* PRIVATE */
+png_save_uint_16(png_bytep buf, unsigned int i)
+{
+   buf[0] = (png_byte)((i >> 8) & 0xff);
+   buf[1] = (png_byte)(i & 0xff);
+}
+
+/* Write a PNG chunk all at once.  The type is an array of ASCII characters
+ * representing the chunk name.  The array must be at least 4 bytes in
+ * length, and does not need to be null terminated.  To be safe, pass the
+ * pre-defined chunk names here, and if you need a new one, define it
+ * where the others are defined.  The length is the length of the data.
+ * All the data must be present.  If that is not possible, use the
+ * png_write_chunk_start(), png_write_chunk_data(), and png_write_chunk_end()
+ * functions instead.
+ */
+void PNGAPI
+png_write_chunk(png_structp png_ptr, png_bytep chunk_name,
+   png_bytep data, png_size_t length)
+{
+   png_write_chunk_start(png_ptr, chunk_name, (png_uint_32)length);
+   png_write_chunk_data(png_ptr, data, length);
+   png_write_chunk_end(png_ptr);
+}
+
+/* Write the start of a PNG chunk.  The type is the chunk type.
+ * The total_length is the sum of the lengths of all the data you will be
+ * passing in png_write_chunk_data().
+ */
+void PNGAPI
+png_write_chunk_start(png_structp png_ptr, png_bytep chunk_name,
+   png_uint_32 length)
+{
+   png_byte buf[4];
+   png_debug2(0, "Writing %s chunk (%lu bytes)\n", chunk_name, length);
+
+   /* write the length */
+   png_save_uint_32(buf, length);
+   png_write_data(png_ptr, buf, (png_size_t)4);
+
+   /* write the chunk name */
+   png_write_data(png_ptr, chunk_name, (png_size_t)4);
+   /* reset the crc and run it over the chunk name */
+   png_reset_crc(png_ptr);
+   png_calculate_crc(png_ptr, chunk_name, (png_size_t)4);
+}
+
+/* Write the data of a PNG chunk started with png_write_chunk_start().
+ * Note that multiple calls to this function are allowed, and that the
+ * sum of the lengths from these calls *must* add up to the total_length
+ * given to png_write_chunk_start().
+ */
+void PNGAPI
+png_write_chunk_data(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+   /* write the data, and run the CRC over it */
+   if (data != NULL && length > 0)
+   {
+      png_calculate_crc(png_ptr, data, length);
+      png_write_data(png_ptr, data, length);
+   }
+}
+
+/* Finish a chunk started with png_write_chunk_start(). */
+void PNGAPI
+png_write_chunk_end(png_structp png_ptr)
+{
+   png_byte buf[4];
+
+   /* write the crc */
+   png_save_uint_32(buf, png_ptr->crc);
+
+   png_write_data(png_ptr, buf, (png_size_t)4);
+}
+
+/* Simple function to write the signature.  If we have already written
+ * the magic bytes of the signature, or more likely, the PNG stream is
+ * being embedded into another stream and doesn't need its own signature,
+ * we should call png_set_sig_bytes() to tell libpng how many of the
+ * bytes have already been written.
+ */
+void /* PRIVATE */
+png_write_sig(png_structp png_ptr)
+{
+   png_byte png_signature[8] = {137, 80, 78, 71, 13, 10, 26, 10};
+   /* write the rest of the 8 byte signature */
+   png_write_data(png_ptr, &png_signature[png_ptr->sig_bytes],
+      (png_size_t)8 - png_ptr->sig_bytes);
+   if(png_ptr->sig_bytes < 3)
+      png_ptr->mode |= PNG_HAVE_PNG_SIGNATURE;
+}
+
+#if defined(PNG_WRITE_TEXT_SUPPORTED) || defined(PNG_WRITE_iCCP_SUPPORTED)
+/*
+ * This pair of functions encapsulates the operation of (a) compressing a
+ * text string, and (b) issuing it later as a series of chunk data writes.
+ * The compression_state structure is shared context for these functions
+ * set up by the caller in order to make the whole mess thread-safe.
+ */
+
+typedef struct
+{
+    char *input;   /* the uncompressed input data */
+    int input_len;   /* its length */
+    int num_output_ptr; /* number of output pointers used */
+    int max_output_ptr; /* size of output_ptr */
+    png_charpp output_ptr; /* array of pointers to output */
+} compression_state;
+
+/* compress given text into storage in the png_ptr structure */
+static int /* PRIVATE */
+png_text_compress(png_structp png_ptr,
+        png_charp text, png_size_t text_len, int compression,
+        compression_state *comp)
+{
+   int ret;
+
+   comp->num_output_ptr = comp->max_output_ptr = 0;
+   comp->output_ptr = NULL;
+   comp->input = NULL;
+
+   /* we may just want to pass the text right through */
+   if (compression == PNG_TEXT_COMPRESSION_NONE)
+   {
+       comp->input = text;
+       comp->input_len = text_len;
+       return((int)text_len);
+   }
+
+   if (compression >= PNG_TEXT_COMPRESSION_LAST)
+   {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+      char msg[50];
+      sprintf(msg, "Unknown compression type %d", compression);
+      png_warning(png_ptr, msg);
+#else
+      png_warning(png_ptr, "Unknown compression type");
+#endif
+   }
+
+   /* We can't write the chunk until we find out how much data we have,
+    * which means we need to run the compressor first and save the
+    * output.  This shouldn't be a problem, as the vast majority of
+    * comments should be reasonable, but we will set up an array of
+    * malloc'd pointers to be sure.
+    *
+    * If we knew the application was well behaved, we could simplify this
+    * greatly by assuming we can always malloc an output buffer large
+    * enough to hold the compressed text ((1001 * text_len / 1000) + 12)
+    * and malloc this directly.  The only time this would be a bad idea is
+    * if we can't malloc more than 64K and we have 64K of random input
+    * data, or if the input string is incredibly large (although this
+    * wouldn't cause a failure, just a slowdown due to swapping).
+    */
+
+   /* set up the compression buffers */
+   png_ptr->zstream.avail_in = (uInt)text_len;
+   png_ptr->zstream.next_in = (Bytef *)text;
+   png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+   png_ptr->zstream.next_out = (Bytef *)png_ptr->zbuf;
+
+   /* this is the same compression loop as in png_write_row() */
+   do
+   {
+      /* compress the data */
+      ret = deflate(&png_ptr->zstream, Z_NO_FLUSH);
+      if (ret != Z_OK)
+      {
+         /* error */
+         if (png_ptr->zstream.msg != NULL)
+            png_error(png_ptr, png_ptr->zstream.msg);
+         else
+            png_error(png_ptr, "zlib error");
+      }
+      /* check to see if we need more room */
+      if (!png_ptr->zstream.avail_out && png_ptr->zstream.avail_in)
+      {
+         /* make sure the output array has room */
+         if (comp->num_output_ptr >= comp->max_output_ptr)
+         {
+            int old_max;
+
+            old_max = comp->max_output_ptr;
+            comp->max_output_ptr = comp->num_output_ptr + 4;
+            if (comp->output_ptr != NULL)
+            {
+               png_charpp old_ptr;
+
+               old_ptr = comp->output_ptr;
+               comp->output_ptr = (png_charpp)png_malloc(png_ptr,
+                  (png_uint_32)(comp->max_output_ptr * sizeof (png_charpp)));
+               png_memcpy(comp->output_ptr, old_ptr, old_max
+                  * sizeof (png_charp));
+               png_free(png_ptr, old_ptr);
+            }
+            else
+               comp->output_ptr = (png_charpp)png_malloc(png_ptr,
+                  (png_uint_32)(comp->max_output_ptr * sizeof (png_charp)));
+         }
+
+         /* save the data */
+         comp->output_ptr[comp->num_output_ptr] = (png_charp)png_malloc(png_ptr,
+            (png_uint_32)png_ptr->zbuf_size);
+         png_memcpy(comp->output_ptr[comp->num_output_ptr], png_ptr->zbuf,
+            png_ptr->zbuf_size);
+         comp->num_output_ptr++;
+
+         /* and reset the buffer */
+         png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+         png_ptr->zstream.next_out = png_ptr->zbuf;
+      }
+   /* continue until we don't have any more to compress */
+   } while (png_ptr->zstream.avail_in);
+
+   /* finish the compression */
+   do
+   {
+      /* tell zlib we are finished */
+      ret = deflate(&png_ptr->zstream, Z_FINISH);
+
+      if (ret == Z_OK)
+      {
+         /* check to see if we need more room */
+         if (!(png_ptr->zstream.avail_out))
+         {
+            /* check to make sure our output array has room */
+            if (comp->num_output_ptr >= comp->max_output_ptr)
+            {
+               int old_max;
+
+               old_max = comp->max_output_ptr;
+               comp->max_output_ptr = comp->num_output_ptr + 4;
+               if (comp->output_ptr != NULL)
+               {
+                  png_charpp old_ptr;
+
+                  old_ptr = comp->output_ptr;
+                  /* This could be optimized to realloc() */
+                  comp->output_ptr = (png_charpp)png_malloc(png_ptr,
+                     (png_uint_32)(comp->max_output_ptr * sizeof (png_charpp)));
+                  png_memcpy(comp->output_ptr, old_ptr,
+                     old_max * sizeof (png_charp));
+                  png_free(png_ptr, old_ptr);
+               }
+               else
+                  comp->output_ptr = (png_charpp)png_malloc(png_ptr,
+                     (png_uint_32)(comp->max_output_ptr * sizeof (png_charp)));
+            }
+
+            /* save off the data */
+            comp->output_ptr[comp->num_output_ptr] =
+               (png_charp)png_malloc(png_ptr, (png_uint_32)png_ptr->zbuf_size);
+            png_memcpy(comp->output_ptr[comp->num_output_ptr], png_ptr->zbuf,
+               png_ptr->zbuf_size);
+            comp->num_output_ptr++;
+
+            /* and reset the buffer pointers */
+            png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+            png_ptr->zstream.next_out = png_ptr->zbuf;
+         }
+      }
+      else if (ret != Z_STREAM_END)
+      {
+         /* we got an error */
+         if (png_ptr->zstream.msg != NULL)
+            png_error(png_ptr, png_ptr->zstream.msg);
+         else
+            png_error(png_ptr, "zlib error");
+      }
+   } while (ret != Z_STREAM_END);
+
+   /* text length is number of buffers plus last buffer */
+   text_len = png_ptr->zbuf_size * comp->num_output_ptr;
+   if (png_ptr->zstream.avail_out < png_ptr->zbuf_size)
+      text_len += png_ptr->zbuf_size - (png_size_t)png_ptr->zstream.avail_out;
+
+   return((int)text_len);
+}
+
+/* ship the compressed text out via chunk writes */
+static void /* PRIVATE */
+png_write_compressed_data_out(png_structp png_ptr, compression_state *comp)
+{
+   int i;
+
+   /* handle the no-compression case */
+   if (comp->input)
+   {
+       png_write_chunk_data(png_ptr, (png_bytep)comp->input,
+                            (png_size_t)comp->input_len);
+       return;
+   }
+
+   /* write saved output buffers, if any */
+   for (i = 0; i < comp->num_output_ptr; i++)
+   {
+      png_write_chunk_data(png_ptr,(png_bytep)comp->output_ptr[i],
+         png_ptr->zbuf_size);
+      png_free(png_ptr, comp->output_ptr[i]);
+      comp->output_ptr[i]=NULL;
+   }
+   if (comp->max_output_ptr != 0)
+      png_free(png_ptr, comp->output_ptr);
+      comp->output_ptr=NULL;
+   /* write anything left in zbuf */
+   if (png_ptr->zstream.avail_out < (png_uint_32)png_ptr->zbuf_size)
+      png_write_chunk_data(png_ptr, png_ptr->zbuf,
+         png_ptr->zbuf_size - png_ptr->zstream.avail_out);
+
+   /* reset zlib for another zTXt/iTXt or the image data */
+   deflateReset(&png_ptr->zstream);
+
+}
+#endif
+
+/* Write the IHDR chunk, and update the png_struct with the necessary
+ * information.  Note that the rest of this code depends upon this
+ * information being correct.
+ */
+void /* PRIVATE */
+png_write_IHDR(png_structp png_ptr, png_uint_32 width, png_uint_32 height,
+   int bit_depth, int color_type, int compression_type, int filter_type,
+   int interlace_type)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_IHDR;
+#endif
+   png_byte buf[13]; /* buffer to store the IHDR info */
+
+   png_debug(1, "in png_write_IHDR\n");
+   /* Check that we have valid input data from the application info */
+   switch (color_type)
+   {
+      case PNG_COLOR_TYPE_GRAY:
+         switch (bit_depth)
+         {
+            case 1:
+            case 2:
+            case 4:
+            case 8:
+            case 16: png_ptr->channels = 1; break;
+            default: png_error(png_ptr,"Invalid bit depth for grayscale image");
+         }
+         break;
+      case PNG_COLOR_TYPE_RGB:
+         if (bit_depth != 8 && bit_depth != 16)
+            png_error(png_ptr, "Invalid bit depth for RGB image");
+         png_ptr->channels = 3;
+         break;
+      case PNG_COLOR_TYPE_PALETTE:
+         switch (bit_depth)
+         {
+            case 1:
+            case 2:
+            case 4:
+            case 8: png_ptr->channels = 1; break;
+            default: png_error(png_ptr, "Invalid bit depth for paletted image");
+         }
+         break;
+      case PNG_COLOR_TYPE_GRAY_ALPHA:
+         if (bit_depth != 8 && bit_depth != 16)
+            png_error(png_ptr, "Invalid bit depth for grayscale+alpha image");
+         png_ptr->channels = 2;
+         break;
+      case PNG_COLOR_TYPE_RGB_ALPHA:
+         if (bit_depth != 8 && bit_depth != 16)
+            png_error(png_ptr, "Invalid bit depth for RGBA image");
+         png_ptr->channels = 4;
+         break;
+      default:
+         png_error(png_ptr, "Invalid image color type specified");
+   }
+
+   if (compression_type != PNG_COMPRESSION_TYPE_BASE)
+   {
+      png_warning(png_ptr, "Invalid compression type specified");
+      compression_type = PNG_COMPRESSION_TYPE_BASE;
+   }
+
+   /* Write filter_method 64 (intrapixel differencing) only if
+    * 1. Libpng was compiled with PNG_MNG_FEATURES_SUPPORTED and
+    * 2. Libpng did not write a PNG signature (this filter_method is only
+    *    used in PNG datastreams that are embedded in MNG datastreams) and
+    * 3. The application called png_permit_mng_features with a mask that
+    *    included PNG_FLAG_MNG_FILTER_64 and
+    * 4. The filter_method is 64 and
+    * 5. The color_type is RGB or RGBA
+    */
+   if (
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+      !((png_ptr->mng_features_permitted & PNG_FLAG_MNG_FILTER_64) &&
+      ((png_ptr->mode&PNG_HAVE_PNG_SIGNATURE) == 0) &&
+      (color_type == PNG_COLOR_TYPE_RGB ||
+       color_type == PNG_COLOR_TYPE_RGB_ALPHA) &&
+      (filter_type == PNG_INTRAPIXEL_DIFFERENCING)) &&
+#endif
+      filter_type != PNG_FILTER_TYPE_BASE)
+   {
+      png_warning(png_ptr, "Invalid filter type specified");
+      filter_type = PNG_FILTER_TYPE_BASE;
+   }
+
+#ifdef PNG_WRITE_INTERLACING_SUPPORTED
+   if (interlace_type != PNG_INTERLACE_NONE &&
+      interlace_type != PNG_INTERLACE_ADAM7)
+   {
+      png_warning(png_ptr, "Invalid interlace type specified");
+      interlace_type = PNG_INTERLACE_ADAM7;
+   }
+#else
+   interlace_type=PNG_INTERLACE_NONE;
+#endif
+
+   /* save off the relevent information */
+   png_ptr->bit_depth = (png_byte)bit_depth;
+   png_ptr->color_type = (png_byte)color_type;
+   png_ptr->interlaced = (png_byte)interlace_type;
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+   png_ptr->filter_type = (png_byte)filter_type;
+#endif
+   png_ptr->width = width;
+   png_ptr->height = height;
+
+   png_ptr->pixel_depth = (png_byte)(bit_depth * png_ptr->channels);
+   png_ptr->rowbytes = ((width * (png_size_t)png_ptr->pixel_depth + 7) >> 3);
+   /* set the usr info, so any transformations can modify it */
+   png_ptr->usr_width = png_ptr->width;
+   png_ptr->usr_bit_depth = png_ptr->bit_depth;
+   png_ptr->usr_channels = png_ptr->channels;
+
+   /* pack the header information into the buffer */
+   png_save_uint_32(buf, width);
+   png_save_uint_32(buf + 4, height);
+   buf[8] = (png_byte)bit_depth;
+   buf[9] = (png_byte)color_type;
+   buf[10] = (png_byte)compression_type;
+   buf[11] = (png_byte)filter_type;
+   buf[12] = (png_byte)interlace_type;
+
+   /* write the chunk */
+   png_write_chunk(png_ptr, (png_bytep)png_IHDR, buf, (png_size_t)13);
+
+   /* initialize zlib with PNG info */
+   png_ptr->zstream.zalloc = png_zalloc;
+   png_ptr->zstream.zfree = png_zfree;
+   png_ptr->zstream.opaque = (voidpf)png_ptr;
+   if (!(png_ptr->do_filter))
+   {
+      if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE ||
+         png_ptr->bit_depth < 8)
+         png_ptr->do_filter = PNG_FILTER_NONE;
+      else
+         png_ptr->do_filter = PNG_ALL_FILTERS;
+   }
+   if (!(png_ptr->flags & PNG_FLAG_ZLIB_CUSTOM_STRATEGY))
+   {
+      if (png_ptr->do_filter != PNG_FILTER_NONE)
+         png_ptr->zlib_strategy = Z_FILTERED;
+      else
+         png_ptr->zlib_strategy = Z_DEFAULT_STRATEGY;
+   }
+   if (!(png_ptr->flags & PNG_FLAG_ZLIB_CUSTOM_LEVEL))
+      png_ptr->zlib_level = Z_DEFAULT_COMPRESSION;
+   if (!(png_ptr->flags & PNG_FLAG_ZLIB_CUSTOM_MEM_LEVEL))
+      png_ptr->zlib_mem_level = 8;
+   if (!(png_ptr->flags & PNG_FLAG_ZLIB_CUSTOM_WINDOW_BITS))
+      png_ptr->zlib_window_bits = 15;
+   if (!(png_ptr->flags & PNG_FLAG_ZLIB_CUSTOM_METHOD))
+      png_ptr->zlib_method = 8;
+   deflateInit2(&png_ptr->zstream, png_ptr->zlib_level,
+      png_ptr->zlib_method, png_ptr->zlib_window_bits,
+      png_ptr->zlib_mem_level, png_ptr->zlib_strategy);
+   png_ptr->zstream.next_out = png_ptr->zbuf;
+   png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+
+   png_ptr->mode = PNG_HAVE_IHDR;
+}
+
+/* write the palette.  We are careful not to trust png_color to be in the
+ * correct order for PNG, so people can redefine it to any convenient
+ * structure.
+ */
+void /* PRIVATE */
+png_write_PLTE(png_structp png_ptr, png_colorp palette, png_uint_32 num_pal)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_PLTE;
+#endif
+   png_uint_32 i;
+   png_colorp pal_ptr;
+   png_byte buf[3];
+
+   png_debug(1, "in png_write_PLTE\n");
+   if ((
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+        !(png_ptr->mng_features_permitted & PNG_FLAG_MNG_EMPTY_PLTE) &&
+#endif
+        num_pal == 0) || num_pal > 256)
+   {
+     if (png_ptr->color_type == PNG_COLOR_TYPE_PALETTE)
+     {
+        png_error(png_ptr, "Invalid number of colors in palette");
+     }
+     else
+     {
+        png_warning(png_ptr, "Invalid number of colors in palette");
+        return;
+     }
+   }
+
+   if (!(png_ptr->color_type&PNG_COLOR_MASK_COLOR))
+   {
+      png_warning(png_ptr,
+        "Ignoring request to write a PLTE chunk in grayscale PNG");
+      return;
+   }
+
+   png_ptr->num_palette = (png_uint_16)num_pal;
+   png_debug1(3, "num_palette = %d\n", png_ptr->num_palette);
+
+   png_write_chunk_start(png_ptr, (png_bytep)png_PLTE, num_pal * 3);
+#ifndef PNG_NO_POINTER_INDEXING
+   for (i = 0, pal_ptr = palette; i < num_pal; i++, pal_ptr++)
+   {
+      buf[0] = pal_ptr->red;
+      buf[1] = pal_ptr->green;
+      buf[2] = pal_ptr->blue;
+      png_write_chunk_data(png_ptr, buf, (png_size_t)3);
+   }
+#else
+   /* This is a little slower but some buggy compilers need to do this instead */
+   pal_ptr=palette;
+   for (i = 0; i < num_pal; i++)
+   {
+      buf[0] = pal_ptr[i].red;
+      buf[1] = pal_ptr[i].green;
+      buf[2] = pal_ptr[i].blue;
+      png_write_chunk_data(png_ptr, buf, (png_size_t)3);
+   }
+#endif
+   png_write_chunk_end(png_ptr);
+   png_ptr->mode |= PNG_HAVE_PLTE;
+}
+
+/* write an IDAT chunk */
+void /* PRIVATE */
+png_write_IDAT(png_structp png_ptr, png_bytep data, png_size_t length)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_IDAT;
+#endif
+   png_debug(1, "in png_write_IDAT\n");
+   png_write_chunk(png_ptr, (png_bytep)png_IDAT, data, length);
+   png_ptr->mode |= PNG_HAVE_IDAT;
+}
+
+/* write an IEND chunk */
+void /* PRIVATE */
+png_write_IEND(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_IEND;
+#endif
+   png_debug(1, "in png_write_IEND\n");
+   png_write_chunk(png_ptr, (png_bytep)png_IEND, png_bytep_NULL,
+     (png_size_t)0);
+   png_ptr->mode |= PNG_HAVE_IEND;
+}
+
+#if defined(PNG_WRITE_gAMA_SUPPORTED)
+/* write a gAMA chunk */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+void /* PRIVATE */
+png_write_gAMA(png_structp png_ptr, double file_gamma)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_gAMA;
+#endif
+   png_uint_32 igamma;
+   png_byte buf[4];
+
+   png_debug(1, "in png_write_gAMA\n");
+   /* file_gamma is saved in 1/100,000ths */
+   igamma = (png_uint_32)(file_gamma * 100000.0 + 0.5);
+   png_save_uint_32(buf, igamma);
+   png_write_chunk(png_ptr, (png_bytep)png_gAMA, buf, (png_size_t)4);
+}
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+void /* PRIVATE */
+png_write_gAMA_fixed(png_structp png_ptr, png_fixed_point file_gamma)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_gAMA;
+#endif
+   png_byte buf[4];
+
+   png_debug(1, "in png_write_gAMA\n");
+   /* file_gamma is saved in 1/100,000ths */
+   png_save_uint_32(buf, (png_uint_32)file_gamma);
+   png_write_chunk(png_ptr, (png_bytep)png_gAMA, buf, (png_size_t)4);
+}
+#endif
+#endif
+
+#if defined(PNG_WRITE_sRGB_SUPPORTED)
+/* write a sRGB chunk */
+void /* PRIVATE */
+png_write_sRGB(png_structp png_ptr, int srgb_intent)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_sRGB;
+#endif
+   png_byte buf[1];
+
+   png_debug(1, "in png_write_sRGB\n");
+   if(srgb_intent >= PNG_sRGB_INTENT_LAST)
+         png_warning(png_ptr,
+            "Invalid sRGB rendering intent specified");
+   buf[0]=(png_byte)srgb_intent;
+   png_write_chunk(png_ptr, (png_bytep)png_sRGB, buf, (png_size_t)1);
+}
+#endif
+
+#if defined(PNG_WRITE_iCCP_SUPPORTED)
+/* write an iCCP chunk */
+void /* PRIVATE */
+png_write_iCCP(png_structp png_ptr, png_charp name, int compression_type,
+   png_charp profile, int profile_len)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_iCCP;
+#endif
+   png_size_t name_len;
+   png_charp new_name;
+   compression_state comp;
+
+   png_debug(1, "in png_write_iCCP\n");
+   if (name == NULL || (name_len = png_check_keyword(png_ptr, name,
+      &new_name)) == 0)
+   {
+      png_warning(png_ptr, "Empty keyword in iCCP chunk");
+      return;
+   }
+
+   if (compression_type != PNG_COMPRESSION_TYPE_BASE)
+      png_warning(png_ptr, "Unknown compression type in iCCP chunk");
+
+   if (profile == NULL)
+      profile_len = 0;
+
+   if (profile_len)
+       profile_len = png_text_compress(png_ptr, profile, (png_size_t)profile_len,
+          PNG_COMPRESSION_TYPE_BASE, &comp);
+
+   /* make sure we include the NULL after the name and the compression type */
+   png_write_chunk_start(png_ptr, (png_bytep)png_iCCP,
+          (png_uint_32)name_len+profile_len+2);
+   new_name[name_len+1]=0x00;
+   png_write_chunk_data(png_ptr, (png_bytep)new_name, name_len + 2);
+
+   if (profile_len)
+      png_write_compressed_data_out(png_ptr, &comp);
+
+   png_write_chunk_end(png_ptr);
+   png_free(png_ptr, new_name);
+}
+#endif
+
+#if defined(PNG_WRITE_sPLT_SUPPORTED)
+/* write a sPLT chunk */
+void /* PRIVATE */
+png_write_sPLT(png_structp png_ptr, png_sPLT_tp spalette)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_sPLT;
+#endif
+   png_size_t name_len;
+   png_charp new_name;
+   png_byte entrybuf[10];
+   int entry_size = (spalette->depth == 8 ? 6 : 10);
+   int palette_size = entry_size * spalette->nentries;
+   png_sPLT_entryp ep;
+#ifdef PNG_NO_POINTER_INDEXING
+   int i;
+#endif
+
+   png_debug(1, "in png_write_sPLT\n");
+   if (spalette->name == NULL || (name_len = png_check_keyword(png_ptr,
+      spalette->name, &new_name))==0)
+   {
+      png_warning(png_ptr, "Empty keyword in sPLT chunk");
+      return;
+   }
+
+   /* make sure we include the NULL after the name */
+   png_write_chunk_start(png_ptr, (png_bytep)png_sPLT,
+          (png_uint_32)(name_len + 2 + palette_size));
+   png_write_chunk_data(png_ptr, (png_bytep)new_name, name_len + 1);
+   png_write_chunk_data(png_ptr, (png_bytep)&spalette->depth, 1);
+
+   /* loop through each palette entry, writing appropriately */
+#ifndef PNG_NO_POINTER_INDEXING
+   for (ep = spalette->entries; ep<spalette->entries+spalette->nentries; ep++)
+   {
+       if (spalette->depth == 8)
+       {
+           entrybuf[0] = (png_byte)ep->red;
+           entrybuf[1] = (png_byte)ep->green;
+           entrybuf[2] = (png_byte)ep->blue;
+           entrybuf[3] = (png_byte)ep->alpha;
+           png_save_uint_16(entrybuf + 4, ep->frequency);
+       }
+       else
+       {
+           png_save_uint_16(entrybuf + 0, ep->red);
+           png_save_uint_16(entrybuf + 2, ep->green);
+           png_save_uint_16(entrybuf + 4, ep->blue);
+           png_save_uint_16(entrybuf + 6, ep->alpha);
+           png_save_uint_16(entrybuf + 8, ep->frequency);
+       }
+       png_write_chunk_data(png_ptr, entrybuf, (png_size_t)entry_size);
+   }
+#else
+   ep=spalette->entries;
+   for (i=0; i>spalette->nentries; i++)
+   {
+       if (spalette->depth == 8)
+       {
+           entrybuf[0] = (png_byte)ep[i].red;
+           entrybuf[1] = (png_byte)ep[i].green;
+           entrybuf[2] = (png_byte)ep[i].blue;
+           entrybuf[3] = (png_byte)ep[i].alpha;
+           png_save_uint_16(entrybuf + 4, ep[i].frequency);
+       }
+       else
+       {
+           png_save_uint_16(entrybuf + 0, ep[i].red);
+           png_save_uint_16(entrybuf + 2, ep[i].green);
+           png_save_uint_16(entrybuf + 4, ep[i].blue);
+           png_save_uint_16(entrybuf + 6, ep[i].alpha);
+           png_save_uint_16(entrybuf + 8, ep[i].frequency);
+       }
+       png_write_chunk_data(png_ptr, entrybuf, entry_size);
+   }
+#endif
+
+   png_write_chunk_end(png_ptr);
+   png_free(png_ptr, new_name);
+}
+#endif
+
+#if defined(PNG_WRITE_sBIT_SUPPORTED)
+/* write the sBIT chunk */
+void /* PRIVATE */
+png_write_sBIT(png_structp png_ptr, png_color_8p sbit, int color_type)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_sBIT;
+#endif
+   png_byte buf[4];
+   png_size_t size;
+
+   png_debug(1, "in png_write_sBIT\n");
+   /* make sure we don't depend upon the order of PNG_COLOR_8 */
+   if (color_type & PNG_COLOR_MASK_COLOR)
+   {
+      png_byte maxbits;
+
+      maxbits = (png_byte)(color_type==PNG_COLOR_TYPE_PALETTE ? 8 :
+                png_ptr->usr_bit_depth);
+      if (sbit->red == 0 || sbit->red > maxbits ||
+          sbit->green == 0 || sbit->green > maxbits ||
+          sbit->blue == 0 || sbit->blue > maxbits)
+      {
+         png_warning(png_ptr, "Invalid sBIT depth specified");
+         return;
+      }
+      buf[0] = sbit->red;
+      buf[1] = sbit->green;
+      buf[2] = sbit->blue;
+      size = 3;
+   }
+   else
+   {
+      if (sbit->gray == 0 || sbit->gray > png_ptr->usr_bit_depth)
+      {
+         png_warning(png_ptr, "Invalid sBIT depth specified");
+         return;
+      }
+      buf[0] = sbit->gray;
+      size = 1;
+   }
+
+   if (color_type & PNG_COLOR_MASK_ALPHA)
+   {
+      if (sbit->alpha == 0 || sbit->alpha > png_ptr->usr_bit_depth)
+      {
+         png_warning(png_ptr, "Invalid sBIT depth specified");
+         return;
+      }
+      buf[size++] = sbit->alpha;
+   }
+
+   png_write_chunk(png_ptr, (png_bytep)png_sBIT, buf, size);
+}
+#endif
+
+#if defined(PNG_WRITE_cHRM_SUPPORTED)
+/* write the cHRM chunk */
+#ifdef PNG_FLOATING_POINT_SUPPORTED
+void /* PRIVATE */
+png_write_cHRM(png_structp png_ptr, double white_x, double white_y,
+   double red_x, double red_y, double green_x, double green_y,
+   double blue_x, double blue_y)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_cHRM;
+#endif
+   png_byte buf[32];
+   png_uint_32 itemp;
+
+   png_debug(1, "in png_write_cHRM\n");
+   /* each value is saved in 1/100,000ths */
+   if (white_x < 0 || white_x > 0.8 || white_y < 0 || white_y > 0.8 ||
+       white_x + white_y > 1.0)
+   {
+      png_warning(png_ptr, "Invalid cHRM white point specified");
+#if !defined(PNG_NO_CONSOLE_IO)
+      fprintf(stderr,"white_x=%f, white_y=%f\n",white_x, white_y);
+#endif
+      return;
+   }
+   itemp = (png_uint_32)(white_x * 100000.0 + 0.5);
+   png_save_uint_32(buf, itemp);
+   itemp = (png_uint_32)(white_y * 100000.0 + 0.5);
+   png_save_uint_32(buf + 4, itemp);
+
+   if (red_x < 0 || red_x > 0.8 || red_y < 0 || red_y > 0.8 ||
+       red_x + red_y > 1.0)
+   {
+      png_warning(png_ptr, "Invalid cHRM red point specified");
+      return;
+   }
+   itemp = (png_uint_32)(red_x * 100000.0 + 0.5);
+   png_save_uint_32(buf + 8, itemp);
+   itemp = (png_uint_32)(red_y * 100000.0 + 0.5);
+   png_save_uint_32(buf + 12, itemp);
+
+   if (green_x < 0 || green_x > 0.8 || green_y < 0 || green_y > 0.8 ||
+       green_x + green_y > 1.0)
+   {
+      png_warning(png_ptr, "Invalid cHRM green point specified");
+      return;
+   }
+   itemp = (png_uint_32)(green_x * 100000.0 + 0.5);
+   png_save_uint_32(buf + 16, itemp);
+   itemp = (png_uint_32)(green_y * 100000.0 + 0.5);
+   png_save_uint_32(buf + 20, itemp);
+
+   if (blue_x < 0 || blue_x > 0.8 || blue_y < 0 || blue_y > 0.8 ||
+       blue_x + blue_y > 1.0)
+   {
+      png_warning(png_ptr, "Invalid cHRM blue point specified");
+      return;
+   }
+   itemp = (png_uint_32)(blue_x * 100000.0 + 0.5);
+   png_save_uint_32(buf + 24, itemp);
+   itemp = (png_uint_32)(blue_y * 100000.0 + 0.5);
+   png_save_uint_32(buf + 28, itemp);
+
+   png_write_chunk(png_ptr, (png_bytep)png_cHRM, buf, (png_size_t)32);
+}
+#endif
+#ifdef PNG_FIXED_POINT_SUPPORTED
+void /* PRIVATE */
+png_write_cHRM_fixed(png_structp png_ptr, png_fixed_point white_x,
+   png_fixed_point white_y, png_fixed_point red_x, png_fixed_point red_y,
+   png_fixed_point green_x, png_fixed_point green_y, png_fixed_point blue_x,
+   png_fixed_point blue_y)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_cHRM;
+#endif
+   png_byte buf[32];
+
+   png_debug(1, "in png_write_cHRM\n");
+   /* each value is saved in 1/100,000ths */
+   if (white_x > 80000L || white_y > 80000L || white_x + white_y > 100000L)
+   {
+      png_warning(png_ptr, "Invalid fixed cHRM white point specified");
+#if !defined(PNG_NO_CONSOLE_IO)
+      fprintf(stderr,"white_x=%ld, white_y=%ld\n",white_x, white_y);
+#endif
+      return;
+   }
+   png_save_uint_32(buf, (png_uint_32)white_x);
+   png_save_uint_32(buf + 4, (png_uint_32)white_y);
+
+   if (red_x > 80000L || red_y > 80000L || red_x + red_y > 100000L)
+   {
+      png_warning(png_ptr, "Invalid cHRM fixed red point specified");
+      return;
+   }
+   png_save_uint_32(buf + 8, (png_uint_32)red_x);
+   png_save_uint_32(buf + 12, (png_uint_32)red_y);
+
+   if (green_x > 80000L || green_y > 80000L || green_x + green_y > 100000L)
+   {
+      png_warning(png_ptr, "Invalid fixed cHRM green point specified");
+      return;
+   }
+   png_save_uint_32(buf + 16, (png_uint_32)green_x);
+   png_save_uint_32(buf + 20, (png_uint_32)green_y);
+
+   if (blue_x > 80000L || blue_y > 80000L || blue_x + blue_y > 100000L)
+   {
+      png_warning(png_ptr, "Invalid fixed cHRM blue point specified");
+      return;
+   }
+   png_save_uint_32(buf + 24, (png_uint_32)blue_x);
+   png_save_uint_32(buf + 28, (png_uint_32)blue_y);
+
+   png_write_chunk(png_ptr, (png_bytep)png_cHRM, buf, (png_size_t)32);
+}
+#endif
+#endif
+
+#if defined(PNG_WRITE_tRNS_SUPPORTED)
+/* write the tRNS chunk */
+void /* PRIVATE */
+png_write_tRNS(png_structp png_ptr, png_bytep trans, png_color_16p tran,
+   int num_trans, int color_type)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_tRNS;
+#endif
+   png_byte buf[6];
+
+   png_debug(1, "in png_write_tRNS\n");
+   if (color_type == PNG_COLOR_TYPE_PALETTE)
+   {
+      if (num_trans <= 0 || num_trans > (int)png_ptr->num_palette)
+      {
+         png_warning(png_ptr,"Invalid number of transparent colors specified");
+         return;
+      }
+      /* write the chunk out as it is */
+      png_write_chunk(png_ptr, (png_bytep)png_tRNS, trans, (png_size_t)num_trans);
+   }
+   else if (color_type == PNG_COLOR_TYPE_GRAY)
+   {
+      /* one 16 bit value */
+      if(tran->gray >= (1 << png_ptr->bit_depth))
+      {
+         png_warning(png_ptr,
+           "Ignoring attempt to write tRNS chunk out-of-range for bit_depth");
+         return;
+      }
+      png_save_uint_16(buf, tran->gray);
+      png_write_chunk(png_ptr, (png_bytep)png_tRNS, buf, (png_size_t)2);
+   }
+   else if (color_type == PNG_COLOR_TYPE_RGB)
+   {
+      /* three 16 bit values */
+      png_save_uint_16(buf, tran->red);
+      png_save_uint_16(buf + 2, tran->green);
+      png_save_uint_16(buf + 4, tran->blue);
+      if(png_ptr->bit_depth == 8 && (buf[0] | buf[2] | buf[4]))
+         {
+            png_warning(png_ptr,
+              "Ignoring attempt to write 16-bit tRNS chunk when bit_depth is 8");
+            return;
+         }
+      png_write_chunk(png_ptr, (png_bytep)png_tRNS, buf, (png_size_t)6);
+   }
+   else
+   {
+      png_warning(png_ptr, "Can't write tRNS with an alpha channel");
+   }
+}
+#endif
+
+#if defined(PNG_WRITE_bKGD_SUPPORTED)
+/* write the background chunk */
+void /* PRIVATE */
+png_write_bKGD(png_structp png_ptr, png_color_16p back, int color_type)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_bKGD;
+#endif
+   png_byte buf[6];
+
+   png_debug(1, "in png_write_bKGD\n");
+   if (color_type == PNG_COLOR_TYPE_PALETTE)
+   {
+      if (
+#if defined(PNG_MNG_FEATURES_SUPPORTED)
+          (png_ptr->num_palette ||
+          (!(png_ptr->mng_features_permitted & PNG_FLAG_MNG_EMPTY_PLTE))) &&
+#endif
+         back->index > png_ptr->num_palette)
+      {
+         png_warning(png_ptr, "Invalid background palette index");
+         return;
+      }
+      buf[0] = back->index;
+      png_write_chunk(png_ptr, (png_bytep)png_bKGD, buf, (png_size_t)1);
+   }
+   else if (color_type & PNG_COLOR_MASK_COLOR)
+   {
+      png_save_uint_16(buf, back->red);
+      png_save_uint_16(buf + 2, back->green);
+      png_save_uint_16(buf + 4, back->blue);
+      if(png_ptr->bit_depth == 8 && (buf[0] | buf[2] | buf[4]))
+         {
+            png_warning(png_ptr,
+              "Ignoring attempt to write 16-bit bKGD chunk when bit_depth is 8");
+            return;
+         }
+      png_write_chunk(png_ptr, (png_bytep)png_bKGD, buf, (png_size_t)6);
+   }
+   else
+   {
+      if(back->gray >= (1 << png_ptr->bit_depth))
+      {
+         png_warning(png_ptr,
+           "Ignoring attempt to write bKGD chunk out-of-range for bit_depth");
+         return;
+      }
+      png_save_uint_16(buf, back->gray);
+      png_write_chunk(png_ptr, (png_bytep)png_bKGD, buf, (png_size_t)2);
+   }
+}
+#endif
+
+#if defined(PNG_WRITE_hIST_SUPPORTED)
+/* write the histogram */
+void /* PRIVATE */
+png_write_hIST(png_structp png_ptr, png_uint_16p hist, int num_hist)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_hIST;
+#endif
+   int i;
+   png_byte buf[3];
+
+   png_debug(1, "in png_write_hIST\n");
+   if (num_hist > (int)png_ptr->num_palette)
+   {
+      png_debug2(3, "num_hist = %d, num_palette = %d\n", num_hist,
+         png_ptr->num_palette);
+      png_warning(png_ptr, "Invalid number of histogram entries specified");
+      return;
+   }
+
+   png_write_chunk_start(png_ptr, (png_bytep)png_hIST, (png_uint_32)(num_hist * 2));
+   for (i = 0; i < num_hist; i++)
+   {
+      png_save_uint_16(buf, hist[i]);
+      png_write_chunk_data(png_ptr, buf, (png_size_t)2);
+   }
+   png_write_chunk_end(png_ptr);
+}
+#endif
+
+#if defined(PNG_WRITE_TEXT_SUPPORTED) || defined(PNG_WRITE_pCAL_SUPPORTED) || \
+    defined(PNG_WRITE_iCCP_SUPPORTED) || defined(PNG_WRITE_sPLT_SUPPORTED)
+/* Check that the tEXt or zTXt keyword is valid per PNG 1.0 specification,
+ * and if invalid, correct the keyword rather than discarding the entire
+ * chunk.  The PNG 1.0 specification requires keywords 1-79 characters in
+ * length, forbids leading or trailing whitespace, multiple internal spaces,
+ * and the non-break space (0x80) from ISO 8859-1.  Returns keyword length.
+ *
+ * The new_key is allocated to hold the corrected keyword and must be freed
+ * by the calling routine.  This avoids problems with trying to write to
+ * static keywords without having to have duplicate copies of the strings.
+ */
+png_size_t /* PRIVATE */
+png_check_keyword(png_structp png_ptr, png_charp key, png_charpp new_key)
+{
+   png_size_t key_len;
+   png_charp kp, dp;
+   int kflag;
+   int kwarn=0;
+
+   png_debug(1, "in png_check_keyword\n");
+   *new_key = NULL;
+
+   if (key == NULL || (key_len = png_strlen(key)) == 0)
+   {
+      png_warning(png_ptr, "zero length keyword");
+      return ((png_size_t)0);
+   }
+
+   png_debug1(2, "Keyword to be checked is '%s'\n", key);
+
+   *new_key = (png_charp)png_malloc(png_ptr, (png_uint_32)(key_len + 2));
+
+   /* Replace non-printing characters with a blank and print a warning */
+   for (kp = key, dp = *new_key; *kp != '\0'; kp++, dp++)
+   {
+      if (*kp < 0x20 || (*kp > 0x7E && (png_byte)*kp < 0xA1))
+      {
+#if !defined(PNG_NO_STDIO) && !defined(_WIN32_WCE)
+         char msg[40];
+
+         sprintf(msg, "invalid keyword character 0x%02X", *kp);
+         png_warning(png_ptr, msg);
+#else
+         png_warning(png_ptr, "invalid character in keyword");
+#endif
+         *dp = ' ';
+      }
+      else
+      {
+         *dp = *kp;
+      }
+   }
+   *dp = '\0';
+
+   /* Remove any trailing white space. */
+   kp = *new_key + key_len - 1;
+   if (*kp == ' ')
+   {
+      png_warning(png_ptr, "trailing spaces removed from keyword");
+
+      while (*kp == ' ')
+      {
+        *(kp--) = '\0';
+        key_len--;
+      }
+   }
+
+   /* Remove any leading white space. */
+   kp = *new_key;
+   if (*kp == ' ')
+   {
+      png_warning(png_ptr, "leading spaces removed from keyword");
+
+      while (*kp == ' ')
+      {
+        kp++;
+        key_len--;
+      }
+   }
+
+   png_debug1(2, "Checking for multiple internal spaces in '%s'\n", kp);
+
+   /* Remove multiple internal spaces. */
+   for (kflag = 0, dp = *new_key; *kp != '\0'; kp++)
+   {
+      if (*kp == ' ' && kflag == 0)
+      {
+         *(dp++) = *kp;
+         kflag = 1;
+      }
+      else if (*kp == ' ')
+      {
+         key_len--;
+         kwarn=1;
+      }
+      else
+      {
+         *(dp++) = *kp;
+         kflag = 0;
+      }
+   }
+   *dp = '\0';
+   if(kwarn)
+      png_warning(png_ptr, "extra interior spaces removed from keyword");
+
+   if (key_len == 0)
+   {
+      png_free(png_ptr, *new_key);
+      *new_key=NULL;
+      png_warning(png_ptr, "Zero length keyword");
+   }
+
+   if (key_len > 79)
+   {
+      png_warning(png_ptr, "keyword length must be 1 - 79 characters");
+      new_key[79] = '\0';
+      key_len = 79;
+   }
+
+   return (key_len);
+}
+#endif
+
+#if defined(PNG_WRITE_tEXt_SUPPORTED)
+/* write a tEXt chunk */
+void /* PRIVATE */
+png_write_tEXt(png_structp png_ptr, png_charp key, png_charp text,
+   png_size_t text_len)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_tEXt;
+#endif
+   png_size_t key_len;
+   png_charp new_key;
+
+   png_debug(1, "in png_write_tEXt\n");
+   if (key == NULL || (key_len = png_check_keyword(png_ptr, key, &new_key))==0)
+   {
+      png_warning(png_ptr, "Empty keyword in tEXt chunk");
+      return;
+   }
+
+   if (text == NULL || *text == '\0')
+      text_len = 0;
+   else
+      text_len = png_strlen(text);
+
+   /* make sure we include the 0 after the key */
+   png_write_chunk_start(png_ptr, (png_bytep)png_tEXt, (png_uint_32)key_len+text_len+1);
+   /*
+    * We leave it to the application to meet PNG-1.0 requirements on the
+    * contents of the text.  PNG-1.0 through PNG-1.2 discourage the use of
+    * any non-Latin-1 characters except for NEWLINE.  ISO PNG will forbid them.
+    * The NUL character is forbidden by PNG-1.0 through PNG-1.2 and ISO PNG.
+    */
+   png_write_chunk_data(png_ptr, (png_bytep)new_key, key_len + 1);
+   if (text_len)
+      png_write_chunk_data(png_ptr, (png_bytep)text, text_len);
+
+   png_write_chunk_end(png_ptr);
+   png_free(png_ptr, new_key);
+}
+#endif
+
+#if defined(PNG_WRITE_zTXt_SUPPORTED)
+/* write a compressed text chunk */
+void /* PRIVATE */
+png_write_zTXt(png_structp png_ptr, png_charp key, png_charp text,
+   png_size_t text_len, int compression)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_zTXt;
+#endif
+   png_size_t key_len;
+   char buf[1];
+   png_charp new_key;
+   compression_state comp;
+
+   png_debug(1, "in png_write_zTXt\n");
+
+   if (key == NULL || (key_len = png_check_keyword(png_ptr, key, &new_key))==0)
+   {
+      png_warning(png_ptr, "Empty keyword in zTXt chunk");
+      return;
+   }
+
+   if (text == NULL || *text == '\0' || compression==PNG_TEXT_COMPRESSION_NONE)
+   {
+      png_write_tEXt(png_ptr, new_key, text, (png_size_t)0);
+      png_free(png_ptr, new_key);
+      return;
+   }
+
+   text_len = png_strlen(text);
+
+   png_free(png_ptr, new_key);
+
+   /* compute the compressed data; do it now for the length */
+   text_len = png_text_compress(png_ptr, text, text_len, compression,
+       &comp);
+
+   /* write start of chunk */
+   png_write_chunk_start(png_ptr, (png_bytep)png_zTXt, (png_uint_32)
+      (key_len+text_len+2));
+   /* write key */
+   png_write_chunk_data(png_ptr, (png_bytep)key, key_len + 1);
+   buf[0] = (png_byte)compression;
+   /* write compression */
+   png_write_chunk_data(png_ptr, (png_bytep)buf, (png_size_t)1);
+   /* write the compressed data */
+   png_write_compressed_data_out(png_ptr, &comp);
+
+   /* close the chunk */
+   png_write_chunk_end(png_ptr);
+}
+#endif
+
+#if defined(PNG_WRITE_iTXt_SUPPORTED)
+/* write an iTXt chunk */
+void /* PRIVATE */
+png_write_iTXt(png_structp png_ptr, int compression, png_charp key,
+    png_charp lang, png_charp lang_key, png_charp text)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_iTXt;
+#endif
+   png_size_t lang_len, key_len, lang_key_len, text_len;
+   png_charp new_lang, new_key;
+   png_byte cbuf[2];
+   compression_state comp;
+
+   png_debug(1, "in png_write_iTXt\n");
+
+   if (key == NULL || (key_len = png_check_keyword(png_ptr, key, &new_key))==0)
+   {
+      png_warning(png_ptr, "Empty keyword in iTXt chunk");
+      return;
+   }
+   if (lang == NULL || (lang_len = png_check_keyword(png_ptr, lang,
+      &new_lang))==0)
+   {
+      png_warning(png_ptr, "Empty language field in iTXt chunk");
+      return;
+   }
+   lang_key_len = png_strlen(lang_key);
+   text_len = png_strlen(text);
+
+   if (text == NULL || *text == '\0')
+      text_len = 0;
+
+   /* compute the compressed data; do it now for the length */
+   text_len = png_text_compress(png_ptr, text, text_len, compression-2,
+      &comp);
+
+   /* make sure we include the compression flag, the compression byte,
+    * and the NULs after the key, lang, and lang_key parts */
+
+   png_write_chunk_start(png_ptr, (png_bytep)png_iTXt,
+          (png_uint_32)(
+        5 /* comp byte, comp flag, terminators for key, lang and lang_key */
+        + key_len
+        + lang_len
+        + lang_key_len
+        + text_len));
+
+   /*
+    * We leave it to the application to meet PNG-1.0 requirements on the
+    * contents of the text.  PNG-1.0 through PNG-1.2 discourage the use of
+    * any non-Latin-1 characters except for NEWLINE.  ISO PNG will forbid them.
+    * The NUL character is forbidden by PNG-1.0 through PNG-1.2 and ISO PNG.
+    */
+   png_write_chunk_data(png_ptr, (png_bytep)new_key, key_len + 1);
+
+   /* set the compression flag */
+   if (compression == PNG_ITXT_COMPRESSION_NONE || \
+       compression == PNG_TEXT_COMPRESSION_NONE)
+       cbuf[0] = 0;
+   else /* compression == PNG_ITXT_COMPRESSION_zTXt */
+       cbuf[0] = 1;
+   /* set the compression method */
+   cbuf[1] = 0;
+   png_write_chunk_data(png_ptr, cbuf, 2);
+
+   png_write_chunk_data(png_ptr, (png_bytep)new_lang, lang_len + 1);
+   png_write_chunk_data(png_ptr, (png_bytep)lang_key, lang_key_len+1);
+   png_write_chunk_data(png_ptr, '\0', 1);
+
+   png_write_compressed_data_out(png_ptr, &comp);
+
+   png_write_chunk_end(png_ptr);
+   png_free(png_ptr, new_key);
+   png_free(png_ptr, new_lang);
+}
+#endif
+
+#if defined(PNG_WRITE_oFFs_SUPPORTED)
+/* write the oFFs chunk */
+void /* PRIVATE */
+png_write_oFFs(png_structp png_ptr, png_int_32 x_offset, png_int_32 y_offset,
+   int unit_type)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_oFFs;
+#endif
+   png_byte buf[9];
+
+   png_debug(1, "in png_write_oFFs\n");
+   if (unit_type >= PNG_OFFSET_LAST)
+      png_warning(png_ptr, "Unrecognized unit type for oFFs chunk");
+
+   png_save_int_32(buf, x_offset);
+   png_save_int_32(buf + 4, y_offset);
+   buf[8] = (png_byte)unit_type;
+
+   png_write_chunk(png_ptr, (png_bytep)png_oFFs, buf, (png_size_t)9);
+}
+#endif
+
+#if defined(PNG_WRITE_pCAL_SUPPORTED)
+/* write the pCAL chunk (described in the PNG extensions document) */
+void /* PRIVATE */
+png_write_pCAL(png_structp png_ptr, png_charp purpose, png_int_32 X0,
+   png_int_32 X1, int type, int nparams, png_charp units, png_charpp params)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_pCAL;
+#endif
+   png_size_t purpose_len, units_len, total_len;
+   png_uint_32p params_len;
+   png_byte buf[10];
+   png_charp new_purpose;
+   int i;
+
+   png_debug1(1, "in png_write_pCAL (%d parameters)\n", nparams);
+   if (type >= PNG_EQUATION_LAST)
+      png_warning(png_ptr, "Unrecognized equation type for pCAL chunk");
+
+   purpose_len = png_check_keyword(png_ptr, purpose, &new_purpose) + 1;
+   png_debug1(3, "pCAL purpose length = %d\n", (int)purpose_len);
+   units_len = png_strlen(units) + (nparams == 0 ? 0 : 1);
+   png_debug1(3, "pCAL units length = %d\n", (int)units_len);
+   total_len = purpose_len + units_len + 10;
+
+   params_len = (png_uint_32p)png_malloc(png_ptr, (png_uint_32)(nparams
+      *sizeof(png_uint_32)));
+
+   /* Find the length of each parameter, making sure we don't count the
+      null terminator for the last parameter. */
+   for (i = 0; i < nparams; i++)
+   {
+      params_len[i] = png_strlen(params[i]) + (i == nparams - 1 ? 0 : 1);
+      png_debug2(3, "pCAL parameter %d length = %lu\n", i, params_len[i]);
+      total_len += (png_size_t)params_len[i];
+   }
+
+   png_debug1(3, "pCAL total length = %d\n", (int)total_len);
+   png_write_chunk_start(png_ptr, (png_bytep)png_pCAL, (png_uint_32)total_len);
+   png_write_chunk_data(png_ptr, (png_bytep)new_purpose, purpose_len);
+   png_save_int_32(buf, X0);
+   png_save_int_32(buf + 4, X1);
+   buf[8] = (png_byte)type;
+   buf[9] = (png_byte)nparams;
+   png_write_chunk_data(png_ptr, buf, (png_size_t)10);
+   png_write_chunk_data(png_ptr, (png_bytep)units, (png_size_t)units_len);
+
+   png_free(png_ptr, new_purpose);
+
+   for (i = 0; i < nparams; i++)
+   {
+      png_write_chunk_data(png_ptr, (png_bytep)params[i],
+         (png_size_t)params_len[i]);
+   }
+
+   png_free(png_ptr, params_len);
+   png_write_chunk_end(png_ptr);
+}
+#endif
+
+#if defined(PNG_WRITE_sCAL_SUPPORTED)
+/* write the sCAL chunk */
+#if defined(PNG_FLOATING_POINT_SUPPORTED) && !defined(PNG_NO_STDIO)
+void /* PRIVATE */
+png_write_sCAL(png_structp png_ptr, int unit, double width,double height)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_sCAL;
+#endif
+   png_size_t total_len;
+   char wbuf[32], hbuf[32];
+
+   png_debug(1, "in png_write_sCAL\n");
+
+#if defined(_WIN32_WCE)
+/* sprintf() function is not supported on WindowsCE */
+   {
+      wchar_t wc_buf[32];
+      swprintf(wc_buf, TEXT("%12.12e"), width);
+      WideCharToMultiByte(CP_ACP, 0, wc_buf, -1, wbuf, 32, NULL, NULL);
+      swprintf(wc_buf, TEXT("%12.12e"), height);
+      WideCharToMultiByte(CP_ACP, 0, wc_buf, -1, hbuf, 32, NULL, NULL);
+   }
+#else
+   sprintf(wbuf, "%12.12e", width);
+   sprintf(hbuf, "%12.12e", height);
+#endif
+   total_len = 1 + png_strlen(wbuf)+1 + png_strlen(hbuf);
+
+   png_debug1(3, "sCAL total length = %d\n", (int)total_len);
+   png_write_chunk_start(png_ptr, (png_bytep)png_sCAL, (png_uint_32)total_len);
+   png_write_chunk_data(png_ptr, (png_bytep)&unit, 1);
+   png_write_chunk_data(png_ptr, (png_bytep)wbuf, png_strlen(wbuf)+1);
+   png_write_chunk_data(png_ptr, (png_bytep)hbuf, png_strlen(hbuf));
+
+   png_write_chunk_end(png_ptr);
+}
+#else
+#ifdef PNG_FIXED_POINT_SUPPORTED
+void /* PRIVATE */
+png_write_sCAL_s(png_structp png_ptr, int unit, png_charp width,
+   png_charp height)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_sCAL;
+#endif
+   png_size_t total_len;
+   char wbuf[32], hbuf[32];
+
+   png_debug(1, "in png_write_sCAL_s\n");
+
+   png_strcpy(wbuf,(const char *)width);
+   png_strcpy(hbuf,(const char *)height);
+   total_len = 1 + png_strlen(wbuf)+1 + png_strlen(hbuf);
+
+   png_debug1(3, "sCAL total length = %d\n", total_len);
+   png_write_chunk_start(png_ptr, (png_bytep)png_sCAL, (png_uint_32)total_len);
+   png_write_chunk_data(png_ptr, (png_bytep)&unit, 1);
+   png_write_chunk_data(png_ptr, (png_bytep)wbuf, png_strlen(wbuf)+1);
+   png_write_chunk_data(png_ptr, (png_bytep)hbuf, png_strlen(hbuf));
+
+   png_write_chunk_end(png_ptr);
+}
+#endif
+#endif
+#endif
+
+#if defined(PNG_WRITE_pHYs_SUPPORTED)
+/* write the pHYs chunk */
+void /* PRIVATE */
+png_write_pHYs(png_structp png_ptr, png_uint_32 x_pixels_per_unit,
+   png_uint_32 y_pixels_per_unit,
+   int unit_type)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_pHYs;
+#endif
+   png_byte buf[9];
+
+   png_debug(1, "in png_write_pHYs\n");
+   if (unit_type >= PNG_RESOLUTION_LAST)
+      png_warning(png_ptr, "Unrecognized unit type for pHYs chunk");
+
+   png_save_uint_32(buf, x_pixels_per_unit);
+   png_save_uint_32(buf + 4, y_pixels_per_unit);
+   buf[8] = (png_byte)unit_type;
+
+   png_write_chunk(png_ptr, (png_bytep)png_pHYs, buf, (png_size_t)9);
+}
+#endif
+
+#if defined(PNG_WRITE_tIME_SUPPORTED)
+/* Write the tIME chunk.  Use either png_convert_from_struct_tm()
+ * or png_convert_from_time_t(), or fill in the structure yourself.
+ */
+void /* PRIVATE */
+png_write_tIME(png_structp png_ptr, png_timep mod_time)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   PNG_tIME;
+#endif
+   png_byte buf[7];
+
+   png_debug(1, "in png_write_tIME\n");
+   if (mod_time->month  > 12 || mod_time->month  < 1 ||
+       mod_time->day    > 31 || mod_time->day    < 1 ||
+       mod_time->hour   > 23 || mod_time->second > 60)
+   {
+      png_warning(png_ptr, "Invalid time specified for tIME chunk");
+      return;
+   }
+
+   png_save_uint_16(buf, mod_time->year);
+   buf[2] = mod_time->month;
+   buf[3] = mod_time->day;
+   buf[4] = mod_time->hour;
+   buf[5] = mod_time->minute;
+   buf[6] = mod_time->second;
+
+   png_write_chunk(png_ptr, (png_bytep)png_tIME, buf, (png_size_t)7);
+}
+#endif
+
+/* initializes the row writing capability of libpng */
+void /* PRIVATE */
+png_write_start_row(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+   /* start of interlace block */
+   int png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0};
+
+   /* offset to next interlace block */
+   int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1};
+
+   /* start of interlace block in the y direction */
+   int png_pass_ystart[7] = {0, 0, 4, 0, 2, 0, 1};
+
+   /* offset to next interlace block in the y direction */
+   int png_pass_yinc[7] = {8, 8, 8, 4, 4, 2, 2};
+#endif
+
+   png_size_t buf_size;
+
+   png_debug(1, "in png_write_start_row\n");
+   buf_size = (png_size_t)(((png_ptr->width * png_ptr->usr_channels *
+                            png_ptr->usr_bit_depth + 7) >> 3) + 1);
+
+   /* set up row buffer */
+   png_ptr->row_buf = (png_bytep)png_malloc(png_ptr, (png_uint_32)buf_size);
+   png_ptr->row_buf[0] = PNG_FILTER_VALUE_NONE;
+
+   /* set up filtering buffer, if using this filter */
+   if (png_ptr->do_filter & PNG_FILTER_SUB)
+   {
+      png_ptr->sub_row = (png_bytep)png_malloc(png_ptr,
+         (png_ptr->rowbytes + 1));
+      png_ptr->sub_row[0] = PNG_FILTER_VALUE_SUB;
+   }
+
+   /* We only need to keep the previous row if we are using one of these. */
+   if (png_ptr->do_filter & (PNG_FILTER_AVG | PNG_FILTER_UP | PNG_FILTER_PAETH))
+   {
+     /* set up previous row buffer */
+      png_ptr->prev_row = (png_bytep)png_malloc(png_ptr, (png_uint_32)buf_size);
+      png_memset(png_ptr->prev_row, 0, buf_size);
+
+      if (png_ptr->do_filter & PNG_FILTER_UP)
+      {
+         png_ptr->up_row = (png_bytep )png_malloc(png_ptr,
+            (png_ptr->rowbytes + 1));
+         png_ptr->up_row[0] = PNG_FILTER_VALUE_UP;
+      }
+
+      if (png_ptr->do_filter & PNG_FILTER_AVG)
+      {
+         png_ptr->avg_row = (png_bytep)png_malloc(png_ptr,
+            (png_ptr->rowbytes + 1));
+         png_ptr->avg_row[0] = PNG_FILTER_VALUE_AVG;
+      }
+
+      if (png_ptr->do_filter & PNG_FILTER_PAETH)
+      {
+         png_ptr->paeth_row = (png_bytep )png_malloc(png_ptr,
+            (png_ptr->rowbytes + 1));
+         png_ptr->paeth_row[0] = PNG_FILTER_VALUE_PAETH;
+      }
+   }
+
+#ifdef PNG_WRITE_INTERLACING_SUPPORTED
+   /* if interlaced, we need to set up width and height of pass */
+   if (png_ptr->interlaced)
+   {
+      if (!(png_ptr->transformations & PNG_INTERLACE))
+      {
+         png_ptr->num_rows = (png_ptr->height + png_pass_yinc[0] - 1 -
+            png_pass_ystart[0]) / png_pass_yinc[0];
+         png_ptr->usr_width = (png_ptr->width + png_pass_inc[0] - 1 -
+            png_pass_start[0]) / png_pass_inc[0];
+      }
+      else
+      {
+         png_ptr->num_rows = png_ptr->height;
+         png_ptr->usr_width = png_ptr->width;
+      }
+   }
+   else
+#endif
+   {
+      png_ptr->num_rows = png_ptr->height;
+      png_ptr->usr_width = png_ptr->width;
+   }
+   png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+   png_ptr->zstream.next_out = png_ptr->zbuf;
+}
+
+/* Internal use only.  Called when finished processing a row of data. */
+void /* PRIVATE */
+png_write_finish_row(png_structp png_ptr)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+   /* start of interlace block */
+   int png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0};
+
+   /* offset to next interlace block */
+   int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1};
+
+   /* start of interlace block in the y direction */
+   int png_pass_ystart[7] = {0, 0, 4, 0, 2, 0, 1};
+
+   /* offset to next interlace block in the y direction */
+   int png_pass_yinc[7] = {8, 8, 8, 4, 4, 2, 2};
+#endif
+
+   int ret;
+
+   png_debug(1, "in png_write_finish_row\n");
+   /* next row */
+   png_ptr->row_number++;
+
+   /* see if we are done */
+   if (png_ptr->row_number < png_ptr->num_rows)
+      return;
+
+#ifdef PNG_WRITE_INTERLACING_SUPPORTED
+   /* if interlaced, go to next pass */
+   if (png_ptr->interlaced)
+   {
+      png_ptr->row_number = 0;
+      if (png_ptr->transformations & PNG_INTERLACE)
+      {
+         png_ptr->pass++;
+      }
+      else
+      {
+         /* loop until we find a non-zero width or height pass */
+         do
+         {
+            png_ptr->pass++;
+            if (png_ptr->pass >= 7)
+               break;
+            png_ptr->usr_width = (png_ptr->width +
+               png_pass_inc[png_ptr->pass] - 1 -
+               png_pass_start[png_ptr->pass]) /
+               png_pass_inc[png_ptr->pass];
+            png_ptr->num_rows = (png_ptr->height +
+               png_pass_yinc[png_ptr->pass] - 1 -
+               png_pass_ystart[png_ptr->pass]) /
+               png_pass_yinc[png_ptr->pass];
+            if (png_ptr->transformations & PNG_INTERLACE)
+               break;
+         } while (png_ptr->usr_width == 0 || png_ptr->num_rows == 0);
+
+      }
+
+      /* reset the row above the image for the next pass */
+      if (png_ptr->pass < 7)
+      {
+         if (png_ptr->prev_row != NULL)
+            png_memset(png_ptr->prev_row, 0,
+               (png_size_t) (((png_uint_32)png_ptr->usr_channels *
+               (png_uint_32)png_ptr->usr_bit_depth *
+               png_ptr->width + 7) >> 3) + 1);
+         return;
+      }
+   }
+#endif
+
+   /* if we get here, we've just written the last row, so we need
+      to flush the compressor */
+   do
+   {
+      /* tell the compressor we are done */
+      ret = deflate(&png_ptr->zstream, Z_FINISH);
+      /* check for an error */
+      if (ret == Z_OK)
+      {
+         /* check to see if we need more room */
+         if (!(png_ptr->zstream.avail_out))
+         {
+            png_write_IDAT(png_ptr, png_ptr->zbuf, png_ptr->zbuf_size);
+            png_ptr->zstream.next_out = png_ptr->zbuf;
+            png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+         }
+      }
+      else if (ret != Z_STREAM_END)
+      {
+         if (png_ptr->zstream.msg != NULL)
+            png_error(png_ptr, png_ptr->zstream.msg);
+         else
+            png_error(png_ptr, "zlib error");
+      }
+   } while (ret != Z_STREAM_END);
+
+   /* write any extra space */
+   if (png_ptr->zstream.avail_out < png_ptr->zbuf_size)
+   {
+      png_write_IDAT(png_ptr, png_ptr->zbuf, png_ptr->zbuf_size -
+         png_ptr->zstream.avail_out);
+   }
+
+   deflateReset(&png_ptr->zstream);
+}
+
+#if defined(PNG_WRITE_INTERLACING_SUPPORTED)
+/* Pick out the correct pixels for the interlace pass.
+ * The basic idea here is to go through the row with a source
+ * pointer and a destination pointer (sp and dp), and copy the
+ * correct pixels for the pass.  As the row gets compacted,
+ * sp will always be >= dp, so we should never overwrite anything.
+ * See the default: case for the easiest code to understand.
+ */
+void /* PRIVATE */
+png_do_write_interlace(png_row_infop row_info, png_bytep row, int pass)
+{
+#ifdef PNG_USE_LOCAL_ARRAYS
+   /* arrays to facilitate easy interlacing - use pass (0 - 6) as index */
+
+   /* start of interlace block */
+   int png_pass_start[7] = {0, 4, 0, 2, 0, 1, 0};
+
+   /* offset to next interlace block */
+   int png_pass_inc[7] = {8, 8, 4, 4, 2, 2, 1};
+#endif
+
+   png_debug(1, "in png_do_write_interlace\n");
+   /* we don't have to do anything on the last pass (6) */
+#if defined(PNG_USELESS_TESTS_SUPPORTED)
+   if (row != NULL && row_info != NULL && pass < 6)
+#else
+   if (pass < 6)
+#endif
+   {
+      /* each pixel depth is handled separately */
+      switch (row_info->pixel_depth)
+      {
+         case 1:
+         {
+            png_bytep sp;
+            png_bytep dp;
+            int shift;
+            int d;
+            int value;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            dp = row;
+            d = 0;
+            shift = 7;
+            for (i = png_pass_start[pass]; i < row_width;
+               i += png_pass_inc[pass])
+            {
+               sp = row + (png_size_t)(i >> 3);
+               value = (int)(*sp >> (7 - (int)(i & 0x07))) & 0x01;
+               d |= (value << shift);
+
+               if (shift == 0)
+               {
+                  shift = 7;
+                  *dp++ = (png_byte)d;
+                  d = 0;
+               }
+               else
+                  shift--;
+
+            }
+            if (shift != 7)
+               *dp = (png_byte)d;
+            break;
+         }
+         case 2:
+         {
+            png_bytep sp;
+            png_bytep dp;
+            int shift;
+            int d;
+            int value;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            dp = row;
+            shift = 6;
+            d = 0;
+            for (i = png_pass_start[pass]; i < row_width;
+               i += png_pass_inc[pass])
+            {
+               sp = row + (png_size_t)(i >> 2);
+               value = (*sp >> ((3 - (int)(i & 0x03)) << 1)) & 0x03;
+               d |= (value << shift);
+
+               if (shift == 0)
+               {
+                  shift = 6;
+                  *dp++ = (png_byte)d;
+                  d = 0;
+               }
+               else
+                  shift -= 2;
+            }
+            if (shift != 6)
+                   *dp = (png_byte)d;
+            break;
+         }
+         case 4:
+         {
+            png_bytep sp;
+            png_bytep dp;
+            int shift;
+            int d;
+            int value;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+
+            dp = row;
+            shift = 4;
+            d = 0;
+            for (i = png_pass_start[pass]; i < row_width;
+               i += png_pass_inc[pass])
+            {
+               sp = row + (png_size_t)(i >> 1);
+               value = (*sp >> ((1 - (int)(i & 0x01)) << 2)) & 0x0f;
+               d |= (value << shift);
+
+               if (shift == 0)
+               {
+                  shift = 4;
+                  *dp++ = (png_byte)d;
+                  d = 0;
+               }
+               else
+                  shift -= 4;
+            }
+            if (shift != 4)
+               *dp = (png_byte)d;
+            break;
+         }
+         default:
+         {
+            png_bytep sp;
+            png_bytep dp;
+            png_uint_32 i;
+            png_uint_32 row_width = row_info->width;
+            png_size_t pixel_bytes;
+
+            /* start at the beginning */
+            dp = row;
+            /* find out how many bytes each pixel takes up */
+            pixel_bytes = (row_info->pixel_depth >> 3);
+            /* loop through the row, only looking at the pixels that
+               matter */
+            for (i = png_pass_start[pass]; i < row_width;
+               i += png_pass_inc[pass])
+            {
+               /* find out where the original pixel is */
+               sp = row + (png_size_t)i * pixel_bytes;
+               /* move the pixel */
+               if (dp != sp)
+                  png_memcpy(dp, sp, pixel_bytes);
+               /* next pixel */
+               dp += pixel_bytes;
+            }
+            break;
+         }
+      }
+      /* set new row width */
+      row_info->width = (row_info->width +
+         png_pass_inc[pass] - 1 -
+         png_pass_start[pass]) /
+         png_pass_inc[pass];
+         row_info->rowbytes = ((row_info->width *
+            row_info->pixel_depth + 7) >> 3);
+   }
+}
+#endif
+
+/* This filters the row, chooses which filter to use, if it has not already
+ * been specified by the application, and then writes the row out with the
+ * chosen filter.
+ */
+#define PNG_MAXSUM (~((png_uint_32)0) >> 1)
+#define PNG_HISHIFT 10
+#define PNG_LOMASK ((png_uint_32)0xffffL)
+#define PNG_HIMASK ((png_uint_32)(~PNG_LOMASK >> PNG_HISHIFT))
+void /* PRIVATE */
+png_write_find_filter(png_structp png_ptr, png_row_infop row_info)
+{
+   png_bytep prev_row, best_row, row_buf;
+   png_uint_32 mins, bpp;
+   png_byte filter_to_do = png_ptr->do_filter;
+   png_uint_32 row_bytes = row_info->rowbytes;
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+   int num_p_filters = (int)png_ptr->num_prev_filters;
+#endif
+
+   png_debug(1, "in png_write_find_filter\n");
+   /* find out how many bytes offset each pixel is */
+   bpp = (row_info->pixel_depth + 7) / 8;
+
+   prev_row = png_ptr->prev_row;
+   best_row = row_buf = png_ptr->row_buf;
+   mins = PNG_MAXSUM;
+
+   /* The prediction method we use is to find which method provides the
+    * smallest value when summing the absolute values of the distances
+    * from zero, using anything >= 128 as negative numbers.  This is known
+    * as the "minimum sum of absolute differences" heuristic.  Other
+    * heuristics are the "weighted minimum sum of absolute differences"
+    * (experimental and can in theory improve compression), and the "zlib
+    * predictive" method (not implemented yet), which does test compressions
+    * of lines using different filter methods, and then chooses the
+    * (series of) filter(s) that give minimum compressed data size (VERY
+    * computationally expensive).
+    *
+    * GRR 980525:  consider also
+    *   (1) minimum sum of absolute differences from running average (i.e.,
+    *       keep running sum of non-absolute differences & count of bytes)
+    *       [track dispersion, too?  restart average if dispersion too large?]
+    *  (1b) minimum sum of absolute differences from sliding average, probably
+    *       with window size <= deflate window (usually 32K)
+    *   (2) minimum sum of squared differences from zero or running average
+    *       (i.e., ~ root-mean-square approach)
+    */
+
+
+   /* We don't need to test the 'no filter' case if this is the only filter
+    * that has been chosen, as it doesn't actually do anything to the data.
+    */
+   if ((filter_to_do & PNG_FILTER_NONE) &&
+       filter_to_do != PNG_FILTER_NONE)
+   {
+      png_bytep rp;
+      png_uint_32 sum = 0;
+      png_uint_32 i;
+      int v;
+
+      for (i = 0, rp = row_buf + 1; i < row_bytes; i++, rp++)
+      {
+         v = *rp;
+         sum += (v < 128) ? v : 256 - v;
+      }
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         png_uint_32 sumhi, sumlo;
+         int j;
+         sumlo = sum & PNG_LOMASK;
+         sumhi = (sum >> PNG_HISHIFT) & PNG_HIMASK; /* Gives us some footroom */
+
+         /* Reduce the sum if we match any of the previous rows */
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_NONE)
+            {
+               sumlo = (sumlo * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               sumhi = (sumhi * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         /* Factor in the cost of this filter (this is here for completeness,
+          * but it makes no sense to have a "cost" for the NONE filter, as
+          * it has the minimum possible computational cost - none).
+          */
+         sumlo = (sumlo * png_ptr->filter_costs[PNG_FILTER_VALUE_NONE]) >>
+            PNG_COST_SHIFT;
+         sumhi = (sumhi * png_ptr->filter_costs[PNG_FILTER_VALUE_NONE]) >>
+            PNG_COST_SHIFT;
+
+         if (sumhi > PNG_HIMASK)
+            sum = PNG_MAXSUM;
+         else
+            sum = (sumhi << PNG_HISHIFT) + sumlo;
+      }
+#endif
+      mins = sum;
+   }
+
+   /* sub filter */
+   if (filter_to_do == PNG_FILTER_SUB)
+   /* it's the only filter so no testing is needed */
+   {
+      png_bytep rp, lp, dp;
+      png_uint_32 i;
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->sub_row + 1; i < bpp;
+           i++, rp++, dp++)
+      {
+         *dp = *rp;
+      }
+      for (lp = row_buf + 1; i < row_bytes;
+         i++, rp++, lp++, dp++)
+      {
+         *dp = (png_byte)(((int)*rp - (int)*lp) & 0xff);
+      }
+      best_row = png_ptr->sub_row;
+   }
+
+   else if (filter_to_do & PNG_FILTER_SUB)
+   {
+      png_bytep rp, dp, lp;
+      png_uint_32 sum = 0, lmins = mins;
+      png_uint_32 i;
+      int v;
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      /* We temporarily increase the "minimum sum" by the factor we
+       * would reduce the sum of this filter, so that we can do the
+       * early exit comparison without scaling the sum each time.
+       */
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 lmhi, lmlo;
+         lmlo = lmins & PNG_LOMASK;
+         lmhi = (lmins >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_SUB)
+            {
+               lmlo = (lmlo * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               lmhi = (lmhi * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         lmlo = (lmlo * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_SUB]) >>
+            PNG_COST_SHIFT;
+         lmhi = (lmhi * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_SUB]) >>
+            PNG_COST_SHIFT;
+
+         if (lmhi > PNG_HIMASK)
+            lmins = PNG_MAXSUM;
+         else
+            lmins = (lmhi << PNG_HISHIFT) + lmlo;
+      }
+#endif
+
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->sub_row + 1; i < bpp;
+           i++, rp++, dp++)
+      {
+         v = *dp = *rp;
+
+         sum += (v < 128) ? v : 256 - v;
+      }
+      for (lp = row_buf + 1; i < row_info->rowbytes;
+         i++, rp++, lp++, dp++)
+      {
+         v = *dp = (png_byte)(((int)*rp - (int)*lp) & 0xff);
+
+         sum += (v < 128) ? v : 256 - v;
+
+         if (sum > lmins)  /* We are already worse, don't continue. */
+            break;
+      }
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 sumhi, sumlo;
+         sumlo = sum & PNG_LOMASK;
+         sumhi = (sum >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_SUB)
+            {
+               sumlo = (sumlo * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               sumhi = (sumhi * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         sumlo = (sumlo * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_SUB]) >>
+            PNG_COST_SHIFT;
+         sumhi = (sumhi * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_SUB]) >>
+            PNG_COST_SHIFT;
+
+         if (sumhi > PNG_HIMASK)
+            sum = PNG_MAXSUM;
+         else
+            sum = (sumhi << PNG_HISHIFT) + sumlo;
+      }
+#endif
+
+      if (sum < mins)
+      {
+         mins = sum;
+         best_row = png_ptr->sub_row;
+      }
+   }
+
+   /* up filter */
+   if (filter_to_do == PNG_FILTER_UP)
+   {
+      png_bytep rp, dp, pp;
+      png_uint_32 i;
+
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->up_row + 1,
+           pp = prev_row + 1; i < row_bytes;
+           i++, rp++, pp++, dp++)
+      {
+         *dp = (png_byte)(((int)*rp - (int)*pp) & 0xff);
+      }
+      best_row = png_ptr->up_row;
+   }
+
+   else if (filter_to_do & PNG_FILTER_UP)
+   {
+      png_bytep rp, dp, pp;
+      png_uint_32 sum = 0, lmins = mins;
+      png_uint_32 i;
+      int v;
+
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 lmhi, lmlo;
+         lmlo = lmins & PNG_LOMASK;
+         lmhi = (lmins >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_UP)
+            {
+               lmlo = (lmlo * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               lmhi = (lmhi * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         lmlo = (lmlo * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_UP]) >>
+            PNG_COST_SHIFT;
+         lmhi = (lmhi * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_UP]) >>
+            PNG_COST_SHIFT;
+
+         if (lmhi > PNG_HIMASK)
+            lmins = PNG_MAXSUM;
+         else
+            lmins = (lmhi << PNG_HISHIFT) + lmlo;
+      }
+#endif
+
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->up_row + 1,
+           pp = prev_row + 1; i < row_bytes; i++)
+      {
+         v = *dp++ = (png_byte)(((int)*rp++ - (int)*pp++) & 0xff);
+
+         sum += (v < 128) ? v : 256 - v;
+
+         if (sum > lmins)  /* We are already worse, don't continue. */
+            break;
+      }
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 sumhi, sumlo;
+         sumlo = sum & PNG_LOMASK;
+         sumhi = (sum >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_UP)
+            {
+               sumlo = (sumlo * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               sumhi = (sumhi * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         sumlo = (sumlo * png_ptr->filter_costs[PNG_FILTER_VALUE_UP]) >>
+            PNG_COST_SHIFT;
+         sumhi = (sumhi * png_ptr->filter_costs[PNG_FILTER_VALUE_UP]) >>
+            PNG_COST_SHIFT;
+
+         if (sumhi > PNG_HIMASK)
+            sum = PNG_MAXSUM;
+         else
+            sum = (sumhi << PNG_HISHIFT) + sumlo;
+      }
+#endif
+
+      if (sum < mins)
+      {
+         mins = sum;
+         best_row = png_ptr->up_row;
+      }
+   }
+
+   /* avg filter */
+   if (filter_to_do == PNG_FILTER_AVG)
+   {
+      png_bytep rp, dp, pp, lp;
+      png_uint_32 i;
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->avg_row + 1,
+           pp = prev_row + 1; i < bpp; i++)
+      {
+         *dp++ = (png_byte)(((int)*rp++ - ((int)*pp++ / 2)) & 0xff);
+      }
+      for (lp = row_buf + 1; i < row_bytes; i++)
+      {
+         *dp++ = (png_byte)(((int)*rp++ - (((int)*pp++ + (int)*lp++) / 2))
+                 & 0xff);
+      }
+      best_row = png_ptr->avg_row;
+   }
+
+   else if (filter_to_do & PNG_FILTER_AVG)
+   {
+      png_bytep rp, dp, pp, lp;
+      png_uint_32 sum = 0, lmins = mins;
+      png_uint_32 i;
+      int v;
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 lmhi, lmlo;
+         lmlo = lmins & PNG_LOMASK;
+         lmhi = (lmins >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_AVG)
+            {
+               lmlo = (lmlo * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               lmhi = (lmhi * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         lmlo = (lmlo * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_AVG]) >>
+            PNG_COST_SHIFT;
+         lmhi = (lmhi * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_AVG]) >>
+            PNG_COST_SHIFT;
+
+         if (lmhi > PNG_HIMASK)
+            lmins = PNG_MAXSUM;
+         else
+            lmins = (lmhi << PNG_HISHIFT) + lmlo;
+      }
+#endif
+
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->avg_row + 1,
+           pp = prev_row + 1; i < bpp; i++)
+      {
+         v = *dp++ = (png_byte)(((int)*rp++ - ((int)*pp++ / 2)) & 0xff);
+
+         sum += (v < 128) ? v : 256 - v;
+      }
+      for (lp = row_buf + 1; i < row_bytes; i++)
+      {
+         v = *dp++ =
+          (png_byte)(((int)*rp++ - (((int)*pp++ + (int)*lp++) / 2)) & 0xff);
+
+         sum += (v < 128) ? v : 256 - v;
+
+         if (sum > lmins)  /* We are already worse, don't continue. */
+            break;
+      }
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 sumhi, sumlo;
+         sumlo = sum & PNG_LOMASK;
+         sumhi = (sum >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_NONE)
+            {
+               sumlo = (sumlo * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               sumhi = (sumhi * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         sumlo = (sumlo * png_ptr->filter_costs[PNG_FILTER_VALUE_AVG]) >>
+            PNG_COST_SHIFT;
+         sumhi = (sumhi * png_ptr->filter_costs[PNG_FILTER_VALUE_AVG]) >>
+            PNG_COST_SHIFT;
+
+         if (sumhi > PNG_HIMASK)
+            sum = PNG_MAXSUM;
+         else
+            sum = (sumhi << PNG_HISHIFT) + sumlo;
+      }
+#endif
+
+      if (sum < mins)
+      {
+         mins = sum;
+         best_row = png_ptr->avg_row;
+      }
+   }
+
+   /* Paeth filter */
+   if (filter_to_do == PNG_FILTER_PAETH)
+   {
+      png_bytep rp, dp, pp, cp, lp;
+      png_uint_32 i;
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->paeth_row + 1,
+           pp = prev_row + 1; i < bpp; i++)
+      {
+         *dp++ = (png_byte)(((int)*rp++ - (int)*pp++) & 0xff);
+      }
+
+      for (lp = row_buf + 1, cp = prev_row + 1; i < row_bytes; i++)
+      {
+         int a, b, c, pa, pb, pc, p;
+
+         b = *pp++;
+         c = *cp++;
+         a = *lp++;
+
+         p = b - c;
+         pc = a - c;
+
+#ifdef PNG_USE_ABS
+         pa = abs(p);
+         pb = abs(pc);
+         pc = abs(p + pc);
+#else
+         pa = p < 0 ? -p : p;
+         pb = pc < 0 ? -pc : pc;
+         pc = (p + pc) < 0 ? -(p + pc) : p + pc;
+#endif
+
+         p = (pa <= pb && pa <=pc) ? a : (pb <= pc) ? b : c;
+
+         *dp++ = (png_byte)(((int)*rp++ - p) & 0xff);
+      }
+      best_row = png_ptr->paeth_row;
+   }
+
+   else if (filter_to_do & PNG_FILTER_PAETH)
+   {
+      png_bytep rp, dp, pp, cp, lp;
+      png_uint_32 sum = 0, lmins = mins;
+      png_uint_32 i;
+      int v;
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 lmhi, lmlo;
+         lmlo = lmins & PNG_LOMASK;
+         lmhi = (lmins >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_PAETH)
+            {
+               lmlo = (lmlo * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               lmhi = (lmhi * png_ptr->inv_filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         lmlo = (lmlo * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_PAETH]) >>
+            PNG_COST_SHIFT;
+         lmhi = (lmhi * png_ptr->inv_filter_costs[PNG_FILTER_VALUE_PAETH]) >>
+            PNG_COST_SHIFT;
+
+         if (lmhi > PNG_HIMASK)
+            lmins = PNG_MAXSUM;
+         else
+            lmins = (lmhi << PNG_HISHIFT) + lmlo;
+      }
+#endif
+
+      for (i = 0, rp = row_buf + 1, dp = png_ptr->paeth_row + 1,
+           pp = prev_row + 1; i < bpp; i++)
+      {
+         v = *dp++ = (png_byte)(((int)*rp++ - (int)*pp++) & 0xff);
+
+         sum += (v < 128) ? v : 256 - v;
+      }
+
+      for (lp = row_buf + 1, cp = prev_row + 1; i < row_bytes; i++)
+      {
+         int a, b, c, pa, pb, pc, p;
+
+         b = *pp++;
+         c = *cp++;
+         a = *lp++;
+
+#ifndef PNG_SLOW_PAETH
+         p = b - c;
+         pc = a - c;
+#ifdef PNG_USE_ABS
+         pa = abs(p);
+         pb = abs(pc);
+         pc = abs(p + pc);
+#else
+         pa = p < 0 ? -p : p;
+         pb = pc < 0 ? -pc : pc;
+         pc = (p + pc) < 0 ? -(p + pc) : p + pc;
+#endif
+         p = (pa <= pb && pa <=pc) ? a : (pb <= pc) ? b : c;
+#else /* PNG_SLOW_PAETH */
+         p = a + b - c;
+         pa = abs(p - a);
+         pb = abs(p - b);
+         pc = abs(p - c);
+         if (pa <= pb && pa <= pc)
+            p = a;
+         else if (pb <= pc)
+            p = b;
+         else
+            p = c;
+#endif /* PNG_SLOW_PAETH */
+
+         v = *dp++ = (png_byte)(((int)*rp++ - p) & 0xff);
+
+         sum += (v < 128) ? v : 256 - v;
+
+         if (sum > lmins)  /* We are already worse, don't continue. */
+            break;
+      }
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+      if (png_ptr->heuristic_method == PNG_FILTER_HEURISTIC_WEIGHTED)
+      {
+         int j;
+         png_uint_32 sumhi, sumlo;
+         sumlo = sum & PNG_LOMASK;
+         sumhi = (sum >> PNG_HISHIFT) & PNG_HIMASK;
+
+         for (j = 0; j < num_p_filters; j++)
+         {
+            if (png_ptr->prev_filters[j] == PNG_FILTER_VALUE_PAETH)
+            {
+               sumlo = (sumlo * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+               sumhi = (sumhi * png_ptr->filter_weights[j]) >>
+                  PNG_WEIGHT_SHIFT;
+            }
+         }
+
+         sumlo = (sumlo * png_ptr->filter_costs[PNG_FILTER_VALUE_PAETH]) >>
+            PNG_COST_SHIFT;
+         sumhi = (sumhi * png_ptr->filter_costs[PNG_FILTER_VALUE_PAETH]) >>
+            PNG_COST_SHIFT;
+
+         if (sumhi > PNG_HIMASK)
+            sum = PNG_MAXSUM;
+         else
+            sum = (sumhi << PNG_HISHIFT) + sumlo;
+      }
+#endif
+
+      if (sum < mins)
+      {
+         best_row = png_ptr->paeth_row;
+      }
+   }
+
+   /* Do the actual writing of the filtered row data from the chosen filter. */
+
+   png_write_filtered_row(png_ptr, best_row);
+
+#if defined(PNG_WRITE_WEIGHTED_FILTER_SUPPORTED)
+   /* Save the type of filter we picked this time for future calculations */
+   if (png_ptr->num_prev_filters > 0)
+   {
+      int j;
+      for (j = 1; j < num_p_filters; j++)
+      {
+         png_ptr->prev_filters[j] = png_ptr->prev_filters[j - 1];
+      }
+      png_ptr->prev_filters[j] = best_row[0];
+   }
+#endif
+}
+
+
+/* Do the actual writing of a previously filtered row. */
+void /* PRIVATE */
+png_write_filtered_row(png_structp png_ptr, png_bytep filtered_row)
+{
+   png_debug(1, "in png_write_filtered_row\n");
+   png_debug1(2, "filter = %d\n", filtered_row[0]);
+   /* set up the zlib input buffer */
+
+   png_ptr->zstream.next_in = filtered_row;
+   png_ptr->zstream.avail_in = (uInt)png_ptr->row_info.rowbytes + 1;
+   /* repeat until we have compressed all the data */
+   do
+   {
+      int ret; /* return of zlib */
+
+      /* compress the data */
+      ret = deflate(&png_ptr->zstream, Z_NO_FLUSH);
+      /* check for compression errors */
+      if (ret != Z_OK)
+      {
+         if (png_ptr->zstream.msg != NULL)
+            png_error(png_ptr, png_ptr->zstream.msg);
+         else
+            png_error(png_ptr, "zlib error");
+      }
+
+      /* see if it is time to write another IDAT */
+      if (!(png_ptr->zstream.avail_out))
+      {
+         /* write the IDAT and reset the zlib output buffer */
+         png_write_IDAT(png_ptr, png_ptr->zbuf, png_ptr->zbuf_size);
+         png_ptr->zstream.next_out = png_ptr->zbuf;
+         png_ptr->zstream.avail_out = (uInt)png_ptr->zbuf_size;
+      }
+   /* repeat until all data has been compressed */
+   } while (png_ptr->zstream.avail_in);
+
+   /* swap the current and previous rows */
+   if (png_ptr->prev_row != NULL)
+   {
+      png_bytep tptr;
+
+      tptr = png_ptr->prev_row;
+      png_ptr->prev_row = png_ptr->row_buf;
+      png_ptr->row_buf = tptr;
+   }
+
+   /* finish row - updates counters and flushes zlib if last row */
+   png_write_finish_row(png_ptr);
+
+#if defined(PNG_WRITE_FLUSH_SUPPORTED)
+   png_ptr->flush_rows++;
+
+   if (png_ptr->flush_dist > 0 &&
+       png_ptr->flush_rows >= png_ptr->flush_dist)
+   {
+      png_write_flush(png_ptr);
+   }
+#endif
+}
+#endif /* PNG_WRITE_SUPPORTED */
diff --git a/zlib/ChangeLog b/zlib/ChangeLog
new file mode 100644
index 000000000..0ccf603c0
--- /dev/null
+++ b/zlib/ChangeLog
@@ -0,0 +1,481 @@
+
+		ChangeLog file for zlib
+
+Changes in 1.1.4 (11 March 2002)
+- ZFREE was repeated on same allocation on some error conditions.
+  This creates a security problem described in
+  http://www.zlib.org/advisory-2002-03-11.txt
+- Returned incorrect error (Z_MEM_ERROR) on some invalid data
+- Avoid accesses before window for invalid distances with inflate window
+  less than 32K.
+- force windowBits > 8 to avoid a bug in the encoder for a window size
+  of 256 bytes. (A complete fix will be available in 1.1.5).
+	
+Changes in 1.1.3 (9 July 1998)
+- fix "an inflate input buffer bug that shows up on rare but persistent
+  occasions" (Mark)
+- fix gzread and gztell for concatenated .gz files (Didier Le Botlan)
+- fix gzseek(..., SEEK_SET) in write mode
+- fix crc check after a gzeek (Frank Faubert)
+- fix miniunzip when the last entry in a zip file is itself a zip file
+  (J Lillge)
+- add contrib/asm586 and contrib/asm686 (Brian Raiter)
+  See http://www.muppetlabs.com/~breadbox/software/assembly.html
+- add support for Delphi 3 in contrib/delphi (Bob Dellaca)
+- add support for C++Builder 3 and Delphi 3 in contrib/delphi2 (Davide Moretti)
+- do not exit prematurely in untgz if 0 at start of block (Magnus Holmgren)
+- use macro EXTERN instead of extern to support DLL for BeOS (Sander Stoks)
+- added a FAQ file
+
+- Support gzdopen on Mac with Metrowerks (Jason Linhart)
+- Do not redefine Byte on Mac (Brad Pettit & Jason Linhart)
+- define SEEK_END too if SEEK_SET is not defined (Albert Chin-A-Young)
+- avoid some warnings with Borland C (Tom Tanner)
+- fix a problem in contrib/minizip/zip.c for 16-bit MSDOS (Gilles Vollant)
+- emulate utime() for WIN32 in contrib/untgz  (Gilles Vollant)
+- allow several arguments to configure (Tim Mooney, Frodo Looijaard)
+- use libdir and includedir in Makefile.in (Tim Mooney)
+- support shared libraries on OSF1 V4 (Tim Mooney)
+- remove so_locations in "make clean"  (Tim Mooney)
+- fix maketree.c compilation error (Glenn, Mark)
+- Python interface to zlib now in Python 1.5 (Jeremy Hylton)
+- new Makefile.riscos (Rich Walker)
+- initialize static descriptors in trees.c for embedded targets (Nick Smith)
+- use "foo-gz" in example.c for RISCOS and VMS (Nick Smith)
+- add the OS/2 files in Makefile.in too (Andrew Zabolotny)
+- fix fdopen and halloc macros for Microsoft C 6.0 (Tom Lane)
+- fix maketree.c to allow clean compilation of inffixed.h (Mark)
+- fix parameter check in deflateCopy (Gunther Nikl)
+- cleanup trees.c, use compressed_len only in debug mode (Christian Spieler)
+- Many portability patches by Christian Spieler:
+  . zutil.c, zutil.h: added "const" for zmem*
+  . Make_vms.com: fixed some typos
+  . Make_vms.com: msdos/Makefile.*: removed zutil.h from some dependency lists
+  . msdos/Makefile.msc: remove "default rtl link library" info from obj files
+  . msdos/Makefile.*: use model-dependent name for the built zlib library
+  . msdos/Makefile.emx, nt/Makefile.emx, nt/Makefile.gcc:
+     new makefiles, for emx (DOS/OS2), emx&rsxnt and mingw32 (Windows 9x / NT)
+- use define instead of typedef for Bytef also for MSC small/medium (Tom Lane)
+- replace __far with _far for better portability (Christian Spieler, Tom Lane)
+- fix test for errno.h in configure (Tim Newsham)
+
+Changes in 1.1.2 (19 March 98)
+- added contrib/minzip, mini zip and unzip based on zlib (Gilles Vollant)
+  See http://www.winimage.com/zLibDll/unzip.html
+- preinitialize the inflate tables for fixed codes, to make the code
+  completely thread safe (Mark)
+- some simplifications and slight speed-up to the inflate code (Mark)
+- fix gzeof on non-compressed files (Allan Schrum)
+- add -std1 option in configure for OSF1 to fix gzprintf (Martin Mokrejs)
+- use default value of 4K for Z_BUFSIZE for 16-bit MSDOS (Tim Wegner + Glenn)
+- added os2/Makefile.def and os2/zlib.def (Andrew Zabolotny)
+- add shared lib support for UNIX_SV4.2MP (MATSUURA Takanori)
+- do not wrap extern "C" around system includes (Tom Lane)
+- mention zlib binding for TCL in README (Andreas Kupries)
+- added amiga/Makefile.pup for Amiga powerUP SAS/C PPC (Andreas Kleinert)
+- allow "make install prefix=..." even after configure (Glenn Randers-Pehrson)
+- allow "configure --prefix $HOME" (Tim Mooney)
+- remove warnings in example.c and gzio.c (Glenn Randers-Pehrson)
+- move Makefile.sas to amiga/Makefile.sas
+
+Changes in 1.1.1 (27 Feb 98)
+- fix macros _tr_tally_* in deflate.h for debug mode  (Glenn Randers-Pehrson)
+- remove block truncation heuristic which had very marginal effect for zlib
+  (smaller lit_bufsize than in gzip 1.2.4) and degraded a little the
+  compression ratio on some files. This also allows inlining _tr_tally for
+  matches in deflate_slow.
+- added msdos/Makefile.w32 for WIN32 Microsoft Visual C++ (Bob Frazier)
+
+Changes in 1.1.0 (24 Feb 98)
+- do not return STREAM_END prematurely in inflate (John Bowler)
+- revert to the zlib 1.0.8 inflate to avoid the gcc 2.8.0 bug (Jeremy Buhler)
+- compile with -DFASTEST to get compression code optimized for speed only
+- in minigzip, try mmap'ing the input file first (Miguel Albrecht)
+- increase size of I/O buffers in minigzip.c and gzio.c (not a big gain
+  on Sun but significant on HP)
+
+- add a pointer to experimental unzip library in README (Gilles Vollant)
+- initialize variable gcc in configure (Chris Herborth)
+
+Changes in 1.0.9 (17 Feb 1998)
+- added gzputs and gzgets functions
+- do not clear eof flag in gzseek (Mark Diekhans)
+- fix gzseek for files in transparent mode (Mark Diekhans)
+- do not assume that vsprintf returns the number of bytes written (Jens Krinke)
+- replace EXPORT with ZEXPORT to avoid conflict with other programs
+- added compress2 in zconf.h, zlib.def, zlib.dnt
+- new asm code from Gilles Vollant in contrib/asm386
+- simplify the inflate code (Mark):
+ . Replace ZALLOC's in huft_build() with single ZALLOC in inflate_blocks_new()
+ . ZALLOC the length list in inflate_trees_fixed() instead of using stack
+ . ZALLOC the value area for huft_build() instead of using stack
+ . Simplify Z_FINISH check in inflate()
+
+- Avoid gcc 2.8.0 comparison bug a little differently than zlib 1.0.8
+- in inftrees.c, avoid cc -O bug on HP (Farshid Elahi)
+- in zconf.h move the ZLIB_DLL stuff earlier to avoid problems with
+  the declaration of FAR (Gilles VOllant)
+- install libz.so* with mode 755 (executable) instead of 644 (Marc Lehmann)
+- read_buf buf parameter of type Bytef* instead of charf*
+- zmemcpy parameters are of type Bytef*, not charf* (Joseph Strout)
+- do not redeclare unlink in minigzip.c for WIN32 (John Bowler)
+- fix check for presence of directories in "make install" (Ian Willis)
+
+Changes in 1.0.8 (27 Jan 1998)
+- fixed offsets in contrib/asm386/gvmat32.asm (Gilles Vollant)
+- fix gzgetc and gzputc for big endian systems (Markus Oberhumer)
+- added compress2() to allow setting the compression level
+- include sys/types.h to get off_t on some systems (Marc Lehmann & QingLong)
+- use constant arrays for the static trees in trees.c instead of computing
+  them at run time (thanks to Ken Raeburn for this suggestion). To create
+  trees.h, compile with GEN_TREES_H and run "make test".
+- check return code of example in "make test" and display result
+- pass minigzip command line options to file_compress
+- simplifying code of inflateSync to avoid gcc 2.8 bug
+
+- support CC="gcc -Wall" in configure -s (QingLong)
+- avoid a flush caused by ftell in gzopen for write mode (Ken Raeburn)
+- fix test for shared library support to avoid compiler warnings
+- zlib.lib -> zlib.dll in msdos/zlib.rc (Gilles Vollant)
+- check for TARGET_OS_MAC in addition to MACOS (Brad Pettit)
+- do not use fdopen for Metrowerks on Mac (Brad Pettit))
+- add checks for gzputc and gzputc in example.c
+- avoid warnings in gzio.c and deflate.c (Andreas Kleinert)
+- use const for the CRC table (Ken Raeburn)
+- fixed "make uninstall" for shared libraries
+- use Tracev instead of Trace in infblock.c
+- in example.c use correct compressed length for test_sync
+- suppress +vnocompatwarnings in configure for HPUX (not always supported)
+
+Changes in 1.0.7 (20 Jan 1998)
+- fix gzseek which was broken in write mode
+- return error for gzseek to negative absolute position
+- fix configure for Linux (Chun-Chung Chen)
+- increase stack space for MSC (Tim Wegner)
+- get_crc_table and inflateSyncPoint are EXPORTed (Gilles Vollant)
+- define EXPORTVA for gzprintf (Gilles Vollant)
+- added man page zlib.3 (Rick Rodgers)
+- for contrib/untgz, fix makedir() and improve Makefile
+
+- check gzseek in write mode in example.c
+- allocate extra buffer for seeks only if gzseek is actually called
+- avoid signed/unsigned comparisons (Tim Wegner, Gilles Vollant)
+- add inflateSyncPoint in zconf.h
+- fix list of exported functions in nt/zlib.dnt and mdsos/zlib.def
+
+Changes in 1.0.6 (19 Jan 1998)
+- add functions gzprintf, gzputc, gzgetc, gztell, gzeof, gzseek, gzrewind and
+  gzsetparams (thanks to Roland Giersig and Kevin Ruland for some of this code)
+- Fix a deflate bug occuring only with compression level 0 (thanks to
+  Andy Buckler for finding this one).
+- In minigzip, pass transparently also the first byte for .Z files.
+- return Z_BUF_ERROR instead of Z_OK if output buffer full in uncompress()
+- check Z_FINISH in inflate (thanks to Marc Schluper)
+- Implement deflateCopy (thanks to Adam Costello)
+- make static libraries by default in configure, add --shared option.
+- move MSDOS or Windows specific files to directory msdos
+- suppress the notion of partial flush to simplify the interface
+  (but the symbol Z_PARTIAL_FLUSH is kept for compatibility with 1.0.4)
+- suppress history buffer provided by application to simplify the interface
+  (this feature was not implemented anyway in 1.0.4)
+- next_in and avail_in must be initialized before calling inflateInit or
+  inflateInit2
+- add EXPORT in all exported functions (for Windows DLL)
+- added Makefile.nt (thanks to Stephen Williams)
+- added the unsupported "contrib" directory:
+   contrib/asm386/ by Gilles Vollant <info@winimage.com>
+	386 asm code replacing longest_match().
+   contrib/iostream/ by Kevin Ruland <kevin@rodin.wustl.edu>
+        A C++ I/O streams interface to the zlib gz* functions
+   contrib/iostream2/  by Tyge Løvset <Tyge.Lovset@cmr.no>
+	Another C++ I/O streams interface
+   contrib/untgz/  by "Pedro A. Aranda Guti\irrez" <paag@tid.es>
+	A very simple tar.gz file extractor using zlib
+   contrib/visual-basic.txt by Carlos Rios <c_rios@sonda.cl>
+        How to use compress(), uncompress() and the gz* functions from VB.
+- pass params -f (filtered data), -h (huffman only), -1 to -9 (compression
+  level) in minigzip (thanks to Tom Lane)
+
+- use const for rommable constants in deflate
+- added test for gzseek and gztell in example.c
+- add undocumented function inflateSyncPoint() (hack for Paul Mackerras)
+- add undocumented function zError to convert error code to string
+  (for Tim Smithers)
+- Allow compilation of gzio with -DNO_DEFLATE to avoid the compression code.
+- Use default memcpy for Symantec MSDOS compiler.
+- Add EXPORT keyword for check_func (needed for Windows DLL)
+- add current directory to LD_LIBRARY_PATH for "make test"
+- create also a link for libz.so.1
+- added support for FUJITSU UXP/DS (thanks to Toshiaki Nomura)
+- use $(SHAREDLIB) instead of libz.so in Makefile.in (for HPUX)
+- added -soname for Linux in configure (Chun-Chung Chen,
+- assign numbers to the exported functions in zlib.def (for Windows DLL)
+- add advice in zlib.h for best usage of deflateSetDictionary
+- work around compiler bug on Atari (cast Z_NULL in call of s->checkfn)
+- allow compilation with ANSI keywords only enabled for TurboC in large model
+- avoid "versionString"[0] (Borland bug)
+- add NEED_DUMMY_RETURN for Borland
+- use variable z_verbose for tracing in debug mode (L. Peter Deutsch).
+- allow compilation with CC
+- defined STDC for OS/2 (David Charlap)	
+- limit external names to 8 chars for MVS (Thomas Lund)
+- in minigzip.c, use static buffers only for 16-bit systems
+- fix suffix check for "minigzip -d foo.gz"
+- do not return an error for the 2nd of two consecutive gzflush() (Felix Lee)
+- use _fdopen instead of fdopen for MSC >= 6.0 (Thomas Fanslau)
+- added makelcc.bat for lcc-win32 (Tom St Denis)
+- in Makefile.dj2, use copy and del instead of install and rm (Frank Donahoe)
+- Avoid expanded $Id: ChangeLog,v 1.1.2.1 2004/07/05 14:12:37 easysw Exp $. Use "rcs -kb" or "cvs admin -kb" to avoid Id expansion.
+- check for unistd.h in configure (for off_t)
+- remove useless check parameter in inflate_blocks_free
+- avoid useless assignment of s->check to itself in inflate_blocks_new
+- do not flush twice in gzclose (thanks to Ken Raeburn)
+- rename FOPEN as F_OPEN to avoid clash with /usr/include/sys/file.h
+- use NO_ERRNO_H instead of enumeration of operating systems with errno.h
+- work around buggy fclose on pipes for HP/UX
+- support zlib DLL with BORLAND C++ 5.0 (thanks to Glenn Randers-Pehrson)
+- fix configure if CC is already equal to gcc
+
+Changes in 1.0.5 (3 Jan 98)
+- Fix inflate to terminate gracefully when fed corrupted or invalid data
+- Use const for rommable constants in inflate
+- Eliminate memory leaks on error conditions in inflate
+- Removed some vestigial code in inflate
+- Update web address in README
+  
+Changes in 1.0.4 (24 Jul 96)
+- In very rare conditions, deflate(s, Z_FINISH) could fail to produce an EOF
+  bit, so the decompressor could decompress all the correct data but went
+  on to attempt decompressing extra garbage data. This affected minigzip too.
+- zlibVersion and gzerror return const char* (needed for DLL)
+- port to RISCOS (no fdopen, no multiple dots, no unlink, no fileno)
+- use z_error only for DEBUG (avoid problem with DLLs)
+
+Changes in 1.0.3 (2 Jul 96)
+- use z_streamp instead of z_stream *, which is now a far pointer in MSDOS
+  small and medium models; this makes the library incompatible with previous
+  versions for these models. (No effect in large model or on other systems.)
+- return OK instead of BUF_ERROR if previous deflate call returned with
+  avail_out as zero but there is nothing to do
+- added memcmp for non STDC compilers
+- define NO_DUMMY_DECL for more Mac compilers (.h files merged incorrectly)
+- define __32BIT__ if __386__ or i386 is defined (pb. with Watcom and SCO)
+- better check for 16-bit mode MSC (avoids problem with Symantec)
+
+Changes in 1.0.2 (23 May 96)
+- added Windows DLL support
+- added a function zlibVersion (for the DLL support)
+- fixed declarations using Bytef in infutil.c (pb with MSDOS medium model)
+- Bytef is define's instead of typedef'd only for Borland C
+- avoid reading uninitialized memory in example.c
+- mention in README that the zlib format is now RFC1950
+- updated Makefile.dj2
+- added algorithm.doc
+
+Changes in 1.0.1 (20 May 96) [1.0 skipped to avoid confusion]
+- fix array overlay in deflate.c which sometimes caused bad compressed data
+- fix inflate bug with empty stored block
+- fix MSDOS medium model which was broken in 0.99
+- fix deflateParams() which could generated bad compressed data.
+- Bytef is define'd instead of typedef'ed (work around Borland bug)
+- added an INDEX file
+- new makefiles for DJGPP (Makefile.dj2), 32-bit Borland (Makefile.b32),
+  Watcom (Makefile.wat), Amiga SAS/C (Makefile.sas)
+- speed up adler32 for modern machines without auto-increment
+- added -ansi for IRIX in configure
+- static_init_done in trees.c is an int
+- define unlink as delete for VMS
+- fix configure for QNX
+- add configure branch for SCO and HPUX
+- avoid many warnings (unused variables, dead assignments, etc...)
+- no fdopen for BeOS
+- fix the Watcom fix for 32 bit mode (define FAR as empty)
+- removed redefinition of Byte for MKWERKS
+- work around an MWKERKS bug (incorrect merge of all .h files)
+
+Changes in 0.99 (27 Jan 96)
+- allow preset dictionary shared between compressor and decompressor
+- allow compression level 0 (no compression)
+- add deflateParams in zlib.h: allow dynamic change of compression level
+  and compression strategy.
+- test large buffers and deflateParams in example.c
+- add optional "configure" to build zlib as a shared library
+- suppress Makefile.qnx, use configure instead
+- fixed deflate for 64-bit systems (detected on Cray)
+- fixed inflate_blocks for 64-bit systems (detected on Alpha)
+- declare Z_DEFLATED in zlib.h (possible parameter for deflateInit2)
+- always return Z_BUF_ERROR when deflate() has nothing to do
+- deflateInit and inflateInit are now macros to allow version checking
+- prefix all global functions and types with z_ with -DZ_PREFIX
+- make falloc completely reentrant (inftrees.c)
+- fixed very unlikely race condition in ct_static_init
+- free in reverse order of allocation to help memory manager
+- use zlib-1.0/* instead of zlib/* inside the tar.gz
+- make zlib warning-free with "gcc -O3 -Wall -Wwrite-strings -Wpointer-arith
+  -Wconversion -Wstrict-prototypes -Wmissing-prototypes"
+- allow gzread on concatenated .gz files
+- deflateEnd now returns Z_DATA_ERROR if it was premature
+- deflate is finally (?) fully deterministic (no matches beyond end of input)
+- Document Z_SYNC_FLUSH
+- add uninstall in Makefile
+- Check for __cpluplus in zlib.h
+- Better test in ct_align for partial flush
+- avoid harmless warnings for Borland C++
+- initialize hash_head in deflate.c
+- avoid warning on fdopen (gzio.c) for HP cc -Aa
+- include stdlib.h for STDC compilers
+- include errno.h for Cray
+- ignore error if ranlib doesn't exist
+- call ranlib twice for NeXTSTEP
+- use exec_prefix instead of prefix for libz.a
+- renamed ct_* as _tr_* to avoid conflict with applications
+- clear z->msg in inflateInit2 before any error return
+- initialize opaque in example.c, gzio.c, deflate.c and inflate.c
+- fixed typo in zconf.h (_GNUC__ => __GNUC__)
+- check for WIN32 in zconf.h and zutil.c (avoid farmalloc in 32-bit mode)
+- fix typo in Make_vms.com (f$trnlnm -> f$getsyi)
+- in fcalloc, normalize pointer if size > 65520 bytes
+- don't use special fcalloc for 32 bit Borland C++
+- use STDC instead of __GO32__ to avoid redeclaring exit, calloc, etc...
+- use Z_BINARY instead of BINARY
+- document that gzclose after gzdopen will close the file
+- allow "a" as mode in gzopen.
+- fix error checking in gzread
+- allow skipping .gz extra-field on pipes
+- added reference to Perl interface in README
+- put the crc table in FAR data (I dislike more and more the medium model :)
+- added get_crc_table
+- added a dimension to all arrays (Borland C can't count).
+- workaround Borland C bug in declaration of inflate_codes_new & inflate_fast
+- guard against multiple inclusion of *.h (for precompiled header on Mac)
+- Watcom C pretends to be Microsoft C small model even in 32 bit mode.
+- don't use unsized arrays to avoid silly warnings by Visual C++:
+     warning C4746: 'inflate_mask' : unsized array treated as  '__far'
+     (what's wrong with far data in far model?).
+- define enum out of inflate_blocks_state to allow compilation with C++
+
+Changes in 0.95 (16 Aug 95)
+- fix MSDOS small and medium model (now easier to adapt to any compiler)
+- inlined send_bits
+- fix the final (:-) bug for deflate with flush (output was correct but
+  not completely flushed in rare occasions).
+- default window size is same for compression and decompression
+  (it's now sufficient to set MAX_WBITS in zconf.h).
+- voidp -> voidpf and voidnp -> voidp (for consistency with other
+  typedefs and because voidnp was not near in large model).
+
+Changes in 0.94 (13 Aug 95)
+- support MSDOS medium model
+- fix deflate with flush (could sometimes generate bad output)
+- fix deflateReset (zlib header was incorrectly suppressed)
+- added support for VMS
+- allow a compression level in gzopen()
+- gzflush now calls fflush
+- For deflate with flush, flush even if no more input is provided.
+- rename libgz.a as libz.a
+- avoid complex expression in infcodes.c triggering Turbo C bug
+- work around a problem with gcc on Alpha (in INSERT_STRING)
+- don't use inline functions (problem with some gcc versions)
+- allow renaming of Byte, uInt, etc... with #define.
+- avoid warning about (unused) pointer before start of array in deflate.c
+- avoid various warnings in gzio.c, example.c, infblock.c, adler32.c, zutil.c
+- avoid reserved word 'new' in trees.c
+
+Changes in 0.93 (25 June 95)
+- temporarily disable inline functions
+- make deflate deterministic
+- give enough lookahead for PARTIAL_FLUSH
+- Set binary mode for stdin/stdout in minigzip.c for OS/2
+- don't even use signed char in inflate (not portable enough)
+- fix inflate memory leak for segmented architectures
+
+Changes in 0.92 (3 May 95)
+- don't assume that char is signed (problem on SGI)
+- Clear bit buffer when starting a stored block
+- no memcpy on Pyramid
+- suppressed inftest.c
+- optimized fill_window, put longest_match inline for gcc
+- optimized inflate on stored blocks.
+- untabify all sources to simplify patches
+
+Changes in 0.91 (2 May 95)
+- Default MEM_LEVEL is 8 (not 9 for Unix) as documented in zlib.h
+- Document the memory requirements in zconf.h
+- added "make install"
+- fix sync search logic in inflateSync
+- deflate(Z_FULL_FLUSH) now works even if output buffer too short
+- after inflateSync, don't scare people with just "lo world"
+- added support for DJGPP
+
+Changes in 0.9 (1 May 95)
+- don't assume that zalloc clears the allocated memory (the TurboC bug
+  was Mark's bug after all :)
+- let again gzread copy uncompressed data unchanged (was working in 0.71)
+- deflate(Z_FULL_FLUSH), inflateReset and inflateSync are now fully implemented
+- added a test of inflateSync in example.c
+- moved MAX_WBITS to zconf.h because users might want to change that.
+- document explicitly that zalloc(64K) on MSDOS must return a normalized
+  pointer (zero offset)
+- added Makefiles for Microsoft C, Turbo C, Borland C++
+- faster crc32()
+
+Changes in 0.8 (29 April 95)
+- added fast inflate (inffast.c)
+- deflate(Z_FINISH) now returns Z_STREAM_END when done. Warning: this
+  is incompatible with previous versions of zlib which returned Z_OK.
+- work around a TurboC compiler bug (bad code for b << 0, see infutil.h)
+  (actually that was not a compiler bug, see 0.81 above)
+- gzread no longer reads one extra byte in certain cases
+- In gzio destroy(), don't reference a freed structure
+- avoid many warnings for MSDOS
+- avoid the ERROR symbol which is used by MS Windows
+
+Changes in 0.71 (14 April 95)
+- Fixed more MSDOS compilation problems :( There is still a bug with
+  TurboC large model.
+
+Changes in 0.7 (14 April 95)
+- Added full inflate support.
+- Simplified the crc32() interface. The pre- and post-conditioning
+  (one's complement) is now done inside crc32(). WARNING: this is
+  incompatible with previous versions; see zlib.h for the new usage.
+
+Changes in 0.61 (12 April 95)
+- workaround for a bug in TurboC. example and minigzip now work on MSDOS.
+
+Changes in 0.6 (11 April 95)
+- added minigzip.c
+- added gzdopen to reopen a file descriptor as gzFile
+- added transparent reading of non-gziped files in gzread.
+- fixed bug in gzread (don't read crc as data)
+- fixed bug in destroy (gzio.c) (don't return Z_STREAM_END for gzclose).
+- don't allocate big arrays in the stack (for MSDOS)
+- fix some MSDOS compilation problems
+
+Changes in 0.5:
+- do real compression in deflate.c. Z_PARTIAL_FLUSH is supported but
+  not yet Z_FULL_FLUSH.
+- support decompression but only in a single step (forced Z_FINISH)
+- added opaque object for zalloc and zfree.
+- added deflateReset and inflateReset
+- added a variable zlib_version for consistency checking.
+- renamed the 'filter' parameter of deflateInit2 as 'strategy'.
+  Added Z_FILTERED and Z_HUFFMAN_ONLY constants.
+
+Changes in 0.4:
+- avoid "zip" everywhere, use zlib instead of ziplib.
+- suppress Z_BLOCK_FLUSH, interpret Z_PARTIAL_FLUSH as block flush
+  if compression method == 8.
+- added adler32 and crc32
+- renamed deflateOptions as deflateInit2, call one or the other but not both
+- added the method parameter for deflateInit2.
+- added inflateInit2
+- simplied considerably deflateInit and inflateInit by not supporting
+  user-provided history buffer. This is supported only in deflateInit2
+  and inflateInit2.
+
+Changes in 0.3:
+- prefix all macro names with Z_
+- use Z_FINISH instead of deflateEnd to finish compression.
+- added Z_HUFFMAN_ONLY
+- added gzerror()
diff --git a/zlib/FAQ b/zlib/FAQ
new file mode 100644
index 000000000..47a7d60c6
--- /dev/null
+++ b/zlib/FAQ
@@ -0,0 +1,100 @@
+
+		Frequently Asked Questions about zlib
+
+
+If your question is not there, please check the zlib home page 
+http://www.zlib.org which may have more recent information.
+The lastest zlib FAQ is at http://www.gzip.org/zlib/zlib_faq.html
+
+
+ 1. Is zlib Y2K-compliant?
+
+    Yes. zlib doesn't handle dates.
+
+ 2. Where can I get a Windows DLL version?
+
+    The zlib sources can be compiled without change to produce a DLL. If you
+    want a precompiled DLL, see http://www.winimage.com/zLibDll/ . Questions
+    about the zlib DLL should be sent to Gilles Vollant (info@winimage.com).
+
+ 3. Where can I get a Visual Basic interface to zlib?
+
+    See
+        * http://www.winimage.com/zLibDll/cmp-z-it.zip
+        * http://www.dogma.net/markn/articles/zlibtool/zlibtool.htm
+        * contrib/visual-basic.txt in the zlib distribution
+
+ 4. compress() returns Z_BUF_ERROR
+
+    Make sure that before the call of compress, the length of the compressed
+    buffer is equal to the total size of the compressed buffer and not
+    zero. For Visual Basic, check that this parameter is passed by reference
+    ("as any"), not by value ("as long").
+
+ 5. deflate() or inflate() returns Z_BUF_ERROR
+
+    Before making the call, make sure that avail_in and avail_out are not
+    zero. When setting the parameter flush equal to Z_FINISH, also make sure
+    that avail_out is big enough to allow processing all pending input.
+
+ 6. Where's the zlib documentation (man pages, etc.)?
+
+    It's in zlib.h for the moment, and Francis S. Lin has converted it to a
+    web page zlib.html. Volunteers to transform this to Unix-style man pages,
+    please contact Jean-loup Gailly (jloup@gzip.org). Examples of zlib usage
+    are in the files example.c and minigzip.c.
+
+ 7. Why don't you use GNU autoconf or libtool or ...?
+
+    Because we would like to keep zlib as a very small and simple
+    package. zlib is rather portable and doesn't need much configuration.
+
+ 8. I found a bug in zlib.
+
+    Most of the time, such problems are due to an incorrect usage of
+    zlib. Please try to reproduce the problem with a small program and send
+    the corresponding source to us at zlib@gzip.org . Do not send
+    multi-megabyte data files without prior agreement.
+
+ 9. Why do I get "undefined reference to gzputc"?
+
+    If "make test" produces something like
+
+       example.o(.text+0x154): undefined reference to `gzputc'
+      
+    check that you don't have old files libz.* in /usr/lib, /usr/local/lib or
+    /usr/X11R6/lib. Remove any old versions, then do "make install".
+
+10. I need a Delphi interface to zlib.
+
+    See the directories contrib/delphi and contrib/delphi2 in the zlib
+    distribution.
+
+11. Can zlib handle .zip archives?
+
+    See the directory contrib/minizip in the zlib distribution.
+
+12. Can zlib handle .Z files?
+
+    No, sorry. You have to spawn an uncompress or gunzip subprocess, or adapt
+    the code of uncompress on your own.
+
+13. How can I make a Unix shared library?
+
+    make clean
+    ./configure -s
+    make
+
+14. Why does "make test" fail on Mac OS X?
+
+    Mac OS X already includes zlib as a shared library, and so -lz links the
+    shared library instead of the one that the "make" compiled. For zlib
+    1.1.3, the two are incompatible due to different compile-time
+    options. Simply change the -lz in the Makefile to libz.a, and it will use
+    the compiled library instead of the shared one and the "make test" will
+    succeed.
+
+15. I have a question about OttoPDF
+
+    We are not the authors of OttoPDF. The real author is on the OttoPDF web
+    site Joel Hainley jhainley@myndkryme.com.
diff --git a/zlib/INDEX b/zlib/INDEX
new file mode 100644
index 000000000..8a2457664
--- /dev/null
+++ b/zlib/INDEX
@@ -0,0 +1,86 @@
+ChangeLog		history of changes
+INDEX			this file
+FAQ			Frequently Asked Questions about zlib
+Make_vms.com		script for Vax/VMS
+Makefile		makefile for Unix (generated by configure)
+Makefile.in		makefile for Unix (template for configure)
+Makefile.riscos 	makefile for RISCOS
+README			guess what
+algorithm.txt		description of the (de)compression algorithm
+configure		configure script for Unix
+descrip.mms		makefile for Vax/VMS
+zlib.3			mini man page for zlib (volunteers to write full
+			man pages from zlib.h welcome. write to jloup@gzip.org)
+
+amiga/Makefile.sas	makefile for Amiga SAS/C
+amiga/Makefile.pup      makefile for Amiga powerUP SAS/C PPC
+
+msdos/Makefile.w32      makefile for Microsoft Visual C++ 32-bit
+msdos/Makefile.b32	makefile for Borland C++   32-bit
+msdos/Makefile.bor	makefile for Borland C/C++ 16-bit
+msdos/Makefile.dj2	makefile for DJGPP 2.x
+msdos/Makefile.emx	makefile for EMX 0.9c (32-bit DOS/OS2)
+msdos/Makefile.msc	makefile for Microsoft C 16-bit
+msdos/Makefile.tc	makefile for Turbo C
+msdos/Makefile.wat	makefile for Watcom C
+msdos/zlib.def        	definition file for Windows DLL
+msdos/zlib.rc         	definition file for Windows DLL
+
+nt/Makefile.nt		makefile for Windows NT
+nt/zlib.dnt		definition file for Windows NT DLL
+nt/Makefile.emx		makefile for EMX 0.9c/RSXNT 1.41 (Win32 Intel)
+nt/Makefile.gcc		makefile for Windows NT using GCC (mingw32)
+
+
+		zlib public header files (must be kept):
+zconf.h
+zlib.h
+
+		private source files used to build the zlib library:
+adler32.c
+compress.c
+crc32.c
+deflate.c
+deflate.h
+gzio.c
+infblock.c
+infblock.h
+infcodes.c
+infcodes.h
+inffast.c
+inffast.h
+inflate.c
+inftrees.c
+inftrees.h
+infutil.c
+infutil.h
+maketree.c
+trees.c
+uncompr.c
+zutil.c
+zutil.h
+
+		source files for sample programs:
+example.c
+minigzip.c
+
+		unsupported contribution by third parties
+
+contrib/asm386/ by Gilles Vollant <info@winimage.com>
+	386 asm code replacing longest_match().
+
+contrib/minizip/ by Gilles Vollant <info@winimage.com>
+	Mini zip and unzip based on zlib
+        See http://www.winimage.com/zLibDll/unzip.html
+
+contrib/iostream/ by Kevin Ruland <kevin@rodin.wustl.edu>
+        A C++ I/O streams interface to the zlib gz* functions
+
+contrib/iostream2/  by Tyge Løvset <Tyge.Lovset@cmr.no>
+	Another C++ I/O streams interface
+
+contrib/untgz/  by "Pedro A. Aranda Guti\irrez" <paag@tid.es>
+	A very simple tar.gz extractor using zlib
+
+contrib/visual-basic.txt by Carlos Rios <c_rios@sonda.cl>
+        How to use compress(), uncompress() and the gz* functions from VB.
diff --git a/zlib/Makefile b/zlib/Makefile
new file mode 100644
index 000000000..ff505e2fd
--- /dev/null
+++ b/zlib/Makefile
@@ -0,0 +1,78 @@
+#
+# "$Id: Makefile,v 1.1.2.1 2004/07/05 14:12:37 easysw Exp $"
+#
+# GNU ZIP library makefile for the Fast Light Toolkit (FLTK).
+#
+# Copyright 1997-2004 by Easy Software Products.
+#
+# This library is free software; you can redistribute it and/or
+# modify it under the terms of the GNU Library General Public
+# License as published by the Free Software Foundation; either
+# version 2 of the License, or (at your option) any later version.
+#
+# This library is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# Library General Public License for more details.
+#
+# You should have received a copy of the GNU Library General Public
+# License along with this library; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+# USA.
+#
+# Please report all bugs and problems to "fltk-bugs@fltk.org".
+#
+
+include ../makeinclude
+
+
+#
+# Object files...
+#
+
+OBJS	=	adler32.o compress.o crc32.o gzio.o uncompr.o deflate.o \
+		trees.o zutil.o inflate.o infblock.o inftrees.o infcodes.o \
+		infutil.o inffast.o
+
+
+#
+# Make all targets...
+#
+
+all:	libz.a
+
+
+#
+# Clean all targets and object files...
+#
+
+clean:
+	$(RM) $(OBJS)
+	$(RM) libz.a
+
+
+#
+# libz.a
+#
+
+libz.a:	$(OBJS)
+	echo Archiving $@...
+	$(RM) $@
+	$(AR) $(ARFLAGS) $@ $(OBJS)
+	$(RANLIB) $@
+
+#
+# Make dependencies...
+#
+
+depend:	$(OBJS:.o=.c)
+	makedepend -Y -I.. -f makedepend $(OBJS:.o=.c)
+
+include makedepend
+
+$(OBJS):	../makeinclude
+
+
+#
+# End of "$Id: Makefile,v 1.1.2.1 2004/07/05 14:12:37 easysw Exp $".
+#
diff --git a/zlib/README b/zlib/README
new file mode 100644
index 000000000..29d67146a
--- /dev/null
+++ b/zlib/README
@@ -0,0 +1,147 @@
+zlib 1.1.4 is a general purpose data compression library.  All the code
+is thread safe.  The data format used by the zlib library
+is described by RFCs (Request for Comments) 1950 to 1952 in the files 
+http://www.ietf.org/rfc/rfc1950.txt (zlib format), rfc1951.txt (deflate
+format) and rfc1952.txt (gzip format). These documents are also available in
+other formats from ftp://ftp.uu.net/graphics/png/documents/zlib/zdoc-index.html
+
+All functions of the compression library are documented in the file zlib.h
+(volunteer to write man pages welcome, contact jloup@gzip.org). A usage
+example of the library is given in the file example.c which also tests that
+the library is working correctly. Another example is given in the file
+minigzip.c. The compression library itself is composed of all source files
+except example.c and minigzip.c.
+
+To compile all files and run the test program, follow the instructions
+given at the top of Makefile. In short "make test; make install"
+should work for most machines. For Unix: "./configure; make test; make install"
+For MSDOS, use one of the special makefiles such as Makefile.msc.
+For VMS, use Make_vms.com or descrip.mms.
+
+Questions about zlib should be sent to <zlib@gzip.org>, or to
+Gilles Vollant <info@winimage.com> for the Windows DLL version.
+The zlib home page is http://www.zlib.org or http://www.gzip.org/zlib/
+Before reporting a problem, please check this site to verify that
+you have the latest version of zlib; otherwise get the latest version and
+check whether the problem still exists or not.
+
+PLEASE read the zlib FAQ http://www.gzip.org/zlib/zlib_faq.html
+before asking for help.
+
+Mark Nelson <markn@ieee.org> wrote an article about zlib for the Jan. 1997
+issue of  Dr. Dobb's Journal; a copy of the article is available in
+http://dogma.net/markn/articles/zlibtool/zlibtool.htm
+
+The changes made in version 1.1.4 are documented in the file ChangeLog.
+The only changes made since 1.1.3 are bug corrections:
+
+- ZFREE was repeated on same allocation on some error conditions.
+  This creates a security problem described in
+  http://www.zlib.org/advisory-2002-03-11.txt
+- Returned incorrect error (Z_MEM_ERROR) on some invalid data
+- Avoid accesses before window for invalid distances with inflate window
+  less than 32K.
+- force windowBits > 8 to avoid a bug in the encoder for a window size
+  of 256 bytes. (A complete fix will be available in 1.1.5).
+
+The beta version 1.1.5beta includes many more changes. A new official
+version 1.1.5 will be released as soon as extensive testing has been
+completed on it.
+
+
+Unsupported third party contributions are provided in directory "contrib".
+
+A Java implementation of zlib is available in the Java Development Kit
+http://www.javasoft.com/products/JDK/1.1/docs/api/Package-java.util.zip.html
+See the zlib home page http://www.zlib.org for details.
+
+A Perl interface to zlib written by Paul Marquess <pmarquess@bfsec.bt.co.uk>
+is in the CPAN (Comprehensive Perl Archive Network) sites
+http://www.cpan.org/modules/by-module/Compress/
+
+A Python interface to zlib written by A.M. Kuchling <amk@magnet.com>
+is available in Python 1.5 and later versions, see
+http://www.python.org/doc/lib/module-zlib.html
+
+A zlib binding for TCL written by Andreas Kupries <a.kupries@westend.com>
+is availlable at http://www.westend.com/~kupries/doc/trf/man/man.html
+
+An experimental package to read and write files in .zip format,
+written on top of zlib by Gilles Vollant <info@winimage.com>, is
+available at http://www.winimage.com/zLibDll/unzip.html
+and also in the contrib/minizip directory of zlib.
+
+
+Notes for some targets:
+
+- To build a Windows DLL version, include in a DLL project zlib.def, zlib.rc
+  and all .c files except example.c and minigzip.c; compile with -DZLIB_DLL
+  The zlib DLL support was initially done by Alessandro Iacopetti and is
+  now maintained by Gilles Vollant <info@winimage.com>. Check the zlib DLL
+  home page at http://www.winimage.com/zLibDll
+
+  From Visual Basic, you can call the DLL functions which do not take
+  a structure as argument: compress, uncompress and all gz* functions.
+  See contrib/visual-basic.txt for more information, or get
+  http://www.tcfb.com/dowseware/cmp-z-it.zip
+
+- For 64-bit Irix, deflate.c must be compiled without any optimization.
+  With -O, one libpng test fails. The test works in 32 bit mode (with
+  the -n32 compiler flag). The compiler bug has been reported to SGI.
+
+- zlib doesn't work with gcc 2.6.3 on a DEC 3000/300LX under OSF/1 2.1   
+  it works when compiled with cc.
+
+- on Digital Unix 4.0D (formely OSF/1) on AlphaServer, the cc option -std1
+  is necessary to get gzprintf working correctly. This is done by configure.
+
+- zlib doesn't work on HP-UX 9.05 with some versions of /bin/cc. It works
+  with other compilers. Use "make test" to check your compiler.
+
+- gzdopen is not supported on RISCOS, BEOS and by some Mac compilers.
+
+- For Turbo C the small model is supported only with reduced performance to
+  avoid any far allocation; it was tested with -DMAX_WBITS=11 -DMAX_MEM_LEVEL=3
+
+- For PalmOs, see http://www.cs.uit.no/~perm/PASTA/pilot/software.html
+  Per Harald Myrvang <perm@stud.cs.uit.no>
+
+
+Acknowledgments:
+
+  The deflate format used by zlib was defined by Phil Katz. The deflate
+  and zlib specifications were written by L. Peter Deutsch. Thanks to all the
+  people who reported problems and suggested various improvements in zlib;
+  they are too numerous to cite here.
+
+Copyright notice:
+
+ (C) 1995-2002 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu
+
+If you use the zlib library in a product, we would appreciate *not*
+receiving lengthy legal documents to sign. The sources are provided
+for free but without warranty of any kind.  The library has been
+entirely written by Jean-loup Gailly and Mark Adler; it does not
+include third-party code.
+
+If you redistribute modified sources, we would appreciate that you include
+in the file ChangeLog history information documenting your changes.
diff --git a/zlib/adler32.c b/zlib/adler32.c
new file mode 100644
index 000000000..2650c2c9b
--- /dev/null
+++ b/zlib/adler32.c
@@ -0,0 +1,48 @@
+/* adler32.c -- compute the Adler-32 checksum of a data stream
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* @(#) $Id: adler32.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#include "zlib.h"
+
+#define BASE 65521L /* largest prime smaller than 65536 */
+#define NMAX 5552
+/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
+
+#define DO1(buf,i)  {s1 += buf[i]; s2 += s1;}
+#define DO2(buf,i)  DO1(buf,i); DO1(buf,i+1);
+#define DO4(buf,i)  DO2(buf,i); DO2(buf,i+2);
+#define DO8(buf,i)  DO4(buf,i); DO4(buf,i+4);
+#define DO16(buf)   DO8(buf,0); DO8(buf,8);
+
+/* ========================================================================= */
+uLong ZEXPORT adler32(adler, buf, len)
+    uLong adler;
+    const Bytef *buf;
+    uInt len;
+{
+    unsigned long s1 = adler & 0xffff;
+    unsigned long s2 = (adler >> 16) & 0xffff;
+    int k;
+
+    if (buf == Z_NULL) return 1L;
+
+    while (len > 0) {
+        k = len < NMAX ? len : NMAX;
+        len -= k;
+        while (k >= 16) {
+            DO16(buf);
+	    buf += 16;
+            k -= 16;
+        }
+        if (k != 0) do {
+            s1 += *buf++;
+	    s2 += s1;
+        } while (--k);
+        s1 %= BASE;
+        s2 %= BASE;
+    }
+    return (s2 << 16) | s1;
+}
diff --git a/zlib/algorithm.txt b/zlib/algorithm.txt
new file mode 100644
index 000000000..cdc830b5d
--- /dev/null
+++ b/zlib/algorithm.txt
@@ -0,0 +1,213 @@
+1. Compression algorithm (deflate)
+
+The deflation algorithm used by gzip (also zip and zlib) is a variation of
+LZ77 (Lempel-Ziv 1977, see reference below). It finds duplicated strings in
+the input data.  The second occurrence of a string is replaced by a
+pointer to the previous string, in the form of a pair (distance,
+length).  Distances are limited to 32K bytes, and lengths are limited
+to 258 bytes. When a string does not occur anywhere in the previous
+32K bytes, it is emitted as a sequence of literal bytes.  (In this
+description, `string' must be taken as an arbitrary sequence of bytes,
+and is not restricted to printable characters.)
+
+Literals or match lengths are compressed with one Huffman tree, and
+match distances are compressed with another tree. The trees are stored
+in a compact form at the start of each block. The blocks can have any
+size (except that the compressed data for one block must fit in
+available memory). A block is terminated when deflate() determines that
+it would be useful to start another block with fresh trees. (This is
+somewhat similar to the behavior of LZW-based _compress_.)
+
+Duplicated strings are found using a hash table. All input strings of
+length 3 are inserted in the hash table. A hash index is computed for
+the next 3 bytes. If the hash chain for this index is not empty, all
+strings in the chain are compared with the current input string, and
+the longest match is selected.
+
+The hash chains are searched starting with the most recent strings, to
+favor small distances and thus take advantage of the Huffman encoding.
+The hash chains are singly linked. There are no deletions from the
+hash chains, the algorithm simply discards matches that are too old.
+
+To avoid a worst-case situation, very long hash chains are arbitrarily
+truncated at a certain length, determined by a runtime option (level
+parameter of deflateInit). So deflate() does not always find the longest
+possible match but generally finds a match which is long enough.
+
+deflate() also defers the selection of matches with a lazy evaluation
+mechanism. After a match of length N has been found, deflate() searches for
+a longer match at the next input byte. If a longer match is found, the
+previous match is truncated to a length of one (thus producing a single
+literal byte) and the process of lazy evaluation begins again. Otherwise,
+the original match is kept, and the next match search is attempted only N
+steps later.
+
+The lazy match evaluation is also subject to a runtime parameter. If
+the current match is long enough, deflate() reduces the search for a longer
+match, thus speeding up the whole process. If compression ratio is more
+important than speed, deflate() attempts a complete second search even if
+the first match is already long enough.
+
+The lazy match evaluation is not performed for the fastest compression
+modes (level parameter 1 to 3). For these fast modes, new strings
+are inserted in the hash table only when no match was found, or
+when the match is not too long. This degrades the compression ratio
+but saves time since there are both fewer insertions and fewer searches.
+
+
+2. Decompression algorithm (inflate)
+
+2.1 Introduction
+
+The real question is, given a Huffman tree, how to decode fast.  The most
+important realization is that shorter codes are much more common than
+longer codes, so pay attention to decoding the short codes fast, and let
+the long codes take longer to decode.
+
+inflate() sets up a first level table that covers some number of bits of
+input less than the length of longest code.  It gets that many bits from the
+stream, and looks it up in the table.  The table will tell if the next
+code is that many bits or less and how many, and if it is, it will tell
+the value, else it will point to the next level table for which inflate()
+grabs more bits and tries to decode a longer code.
+
+How many bits to make the first lookup is a tradeoff between the time it
+takes to decode and the time it takes to build the table.  If building the
+table took no time (and if you had infinite memory), then there would only
+be a first level table to cover all the way to the longest code.  However,
+building the table ends up taking a lot longer for more bits since short
+codes are replicated many times in such a table.  What inflate() does is
+simply to make the number of bits in the first table a variable, and set it
+for the maximum speed.
+
+inflate() sends new trees relatively often, so it is possibly set for a
+smaller first level table than an application that has only one tree for
+all the data.  For inflate, which has 286 possible codes for the
+literal/length tree, the size of the first table is nine bits.  Also the
+distance trees have 30 possible values, and the size of the first table is
+six bits.  Note that for each of those cases, the table ended up one bit
+longer than the ``average'' code length, i.e. the code length of an
+approximately flat code which would be a little more than eight bits for
+286 symbols and a little less than five bits for 30 symbols.  It would be
+interesting to see if optimizing the first level table for other
+applications gave values within a bit or two of the flat code size.
+
+
+2.2 More details on the inflate table lookup
+
+Ok, you want to know what this cleverly obfuscated inflate tree actually  
+looks like.  You are correct that it's not a Huffman tree.  It is simply a  
+lookup table for the first, let's say, nine bits of a Huffman symbol.  The  
+symbol could be as short as one bit or as long as 15 bits.  If a particular  
+symbol is shorter than nine bits, then that symbol's translation is duplicated
+in all those entries that start with that symbol's bits.  For example, if the  
+symbol is four bits, then it's duplicated 32 times in a nine-bit table.  If a  
+symbol is nine bits long, it appears in the table once.
+
+If the symbol is longer than nine bits, then that entry in the table points  
+to another similar table for the remaining bits.  Again, there are duplicated  
+entries as needed.  The idea is that most of the time the symbol will be short
+and there will only be one table look up.  (That's whole idea behind data  
+compression in the first place.)  For the less frequent long symbols, there  
+will be two lookups.  If you had a compression method with really long  
+symbols, you could have as many levels of lookups as is efficient.  For  
+inflate, two is enough.
+
+So a table entry either points to another table (in which case nine bits in  
+the above example are gobbled), or it contains the translation for the symbol  
+and the number of bits to gobble.  Then you start again with the next  
+ungobbled bit.
+
+You may wonder: why not just have one lookup table for how ever many bits the  
+longest symbol is?  The reason is that if you do that, you end up spending  
+more time filling in duplicate symbol entries than you do actually decoding.   
+At least for deflate's output that generates new trees every several 10's of  
+kbytes.  You can imagine that filling in a 2^15 entry table for a 15-bit code  
+would take too long if you're only decoding several thousand symbols.  At the  
+other extreme, you could make a new table for every bit in the code.  In fact,
+that's essentially a Huffman tree.  But then you spend two much time  
+traversing the tree while decoding, even for short symbols.
+
+So the number of bits for the first lookup table is a trade of the time to  
+fill out the table vs. the time spent looking at the second level and above of
+the table.
+
+Here is an example, scaled down:
+
+The code being decoded, with 10 symbols, from 1 to 6 bits long:
+
+A: 0
+B: 10
+C: 1100
+D: 11010
+E: 11011
+F: 11100
+G: 11101
+H: 11110
+I: 111110
+J: 111111
+
+Let's make the first table three bits long (eight entries):
+
+000: A,1
+001: A,1
+010: A,1
+011: A,1
+100: B,2
+101: B,2
+110: -> table X (gobble 3 bits)
+111: -> table Y (gobble 3 bits)
+
+Each entry is what the bits decode to and how many bits that is, i.e. how  
+many bits to gobble.  Or the entry points to another table, with the number of
+bits to gobble implicit in the size of the table.
+
+Table X is two bits long since the longest code starting with 110 is five bits
+long:
+
+00: C,1
+01: C,1
+10: D,2
+11: E,2
+
+Table Y is three bits long since the longest code starting with 111 is six  
+bits long:
+
+000: F,2
+001: F,2
+010: G,2
+011: G,2
+100: H,2
+101: H,2
+110: I,3
+111: J,3
+
+So what we have here are three tables with a total of 20 entries that had to  
+be constructed.  That's compared to 64 entries for a single table.  Or  
+compared to 16 entries for a Huffman tree (six two entry tables and one four  
+entry table).  Assuming that the code ideally represents the probability of  
+the symbols, it takes on the average 1.25 lookups per symbol.  That's compared
+to one lookup for the single table, or 1.66 lookups per symbol for the  
+Huffman tree.
+
+There, I think that gives you a picture of what's going on.  For inflate, the  
+meaning of a particular symbol is often more than just a letter.  It can be a  
+byte (a "literal"), or it can be either a length or a distance which  
+indicates a base value and a number of bits to fetch after the code that is  
+added to the base value.  Or it might be the special end-of-block code.  The  
+data structures created in inftrees.c try to encode all that information  
+compactly in the tables.
+
+
+Jean-loup Gailly        Mark Adler
+jloup@gzip.org          madler@alumni.caltech.edu
+
+
+References:
+
+[LZ77] Ziv J., Lempel A., ``A Universal Algorithm for Sequential Data
+Compression,'' IEEE Transactions on Information Theory, Vol. 23, No. 3,
+pp. 337-343.
+
+``DEFLATE Compressed Data Format Specification'' available in
+ftp://ds.internic.net/rfc/rfc1951.txt
diff --git a/zlib/compress.c b/zlib/compress.c
new file mode 100644
index 000000000..f30c60651
--- /dev/null
+++ b/zlib/compress.c
@@ -0,0 +1,68 @@
+/* compress.c -- compress a memory buffer
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* @(#) $Id: compress.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#include "zlib.h"
+
+/* ===========================================================================
+     Compresses the source buffer into the destination buffer. The level
+   parameter has the same meaning as in deflateInit.  sourceLen is the byte
+   length of the source buffer. Upon entry, destLen is the total size of the
+   destination buffer, which must be at least 0.1% larger than sourceLen plus
+   12 bytes. Upon exit, destLen is the actual size of the compressed buffer.
+
+     compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer,
+   Z_STREAM_ERROR if the level parameter is invalid.
+*/
+int ZEXPORT compress2 (dest, destLen, source, sourceLen, level)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+    int level;
+{
+    z_stream stream;
+    int err;
+
+    stream.next_in = (Bytef*)source;
+    stream.avail_in = (uInt)sourceLen;
+#ifdef MAXSEG_64K
+    /* Check for source > 64K on 16-bit machine: */
+    if ((uLong)stream.avail_in != sourceLen) return Z_BUF_ERROR;
+#endif
+    stream.next_out = dest;
+    stream.avail_out = (uInt)*destLen;
+    if ((uLong)stream.avail_out != *destLen) return Z_BUF_ERROR;
+
+    stream.zalloc = (alloc_func)0;
+    stream.zfree = (free_func)0;
+    stream.opaque = (voidpf)0;
+
+    err = deflateInit(&stream, level);
+    if (err != Z_OK) return err;
+
+    err = deflate(&stream, Z_FINISH);
+    if (err != Z_STREAM_END) {
+        deflateEnd(&stream);
+        return err == Z_OK ? Z_BUF_ERROR : err;
+    }
+    *destLen = stream.total_out;
+
+    err = deflateEnd(&stream);
+    return err;
+}
+
+/* ===========================================================================
+ */
+int ZEXPORT compress (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+{
+    return compress2(dest, destLen, source, sourceLen, Z_DEFAULT_COMPRESSION);
+}
diff --git a/zlib/crc32.c b/zlib/crc32.c
new file mode 100644
index 000000000..cf956b127
--- /dev/null
+++ b/zlib/crc32.c
@@ -0,0 +1,162 @@
+/* crc32.c -- compute the CRC-32 of a data stream
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* @(#) $Id: crc32.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#include "zlib.h"
+
+#define local static
+
+#ifdef DYNAMIC_CRC_TABLE
+
+local int crc_table_empty = 1;
+local uLongf crc_table[256];
+local void make_crc_table OF((void));
+
+/*
+  Generate a table for a byte-wise 32-bit CRC calculation on the polynomial:
+  x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x+1.
+
+  Polynomials over GF(2) are represented in binary, one bit per coefficient,
+  with the lowest powers in the most significant bit.  Then adding polynomials
+  is just exclusive-or, and multiplying a polynomial by x is a right shift by
+  one.  If we call the above polynomial p, and represent a byte as the
+  polynomial q, also with the lowest power in the most significant bit (so the
+  byte 0xb1 is the polynomial x^7+x^3+x+1), then the CRC is (q*x^32) mod p,
+  where a mod b means the remainder after dividing a by b.
+
+  This calculation is done using the shift-register method of multiplying and
+  taking the remainder.  The register is initialized to zero, and for each
+  incoming bit, x^32 is added mod p to the register if the bit is a one (where
+  x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by
+  x (which is shifting right by one and adding x^32 mod p if the bit shifted
+  out is a one).  We start with the highest power (least significant bit) of
+  q and repeat for all eight bits of q.
+
+  The table is simply the CRC of all possible eight bit values.  This is all
+  the information needed to generate CRC's on data a byte at a time for all
+  combinations of CRC register values and incoming bytes.
+*/
+local void make_crc_table()
+{
+  uLong c;
+  int n, k;
+  uLong poly;            /* polynomial exclusive-or pattern */
+  /* terms of polynomial defining this crc (except x^32): */
+  static const Byte p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
+
+  /* make exclusive-or pattern from polynomial (0xedb88320L) */
+  poly = 0L;
+  for (n = 0; n < sizeof(p)/sizeof(Byte); n++)
+    poly |= 1L << (31 - p[n]);
+ 
+  for (n = 0; n < 256; n++)
+  {
+    c = (uLong)n;
+    for (k = 0; k < 8; k++)
+      c = c & 1 ? poly ^ (c >> 1) : c >> 1;
+    crc_table[n] = c;
+  }
+  crc_table_empty = 0;
+}
+#else
+/* ========================================================================
+ * Table of CRC-32's of all single-byte values (made by make_crc_table)
+ */
+local const uLongf crc_table[256] = {
+  0x00000000L, 0x77073096L, 0xee0e612cL, 0x990951baL, 0x076dc419L,
+  0x706af48fL, 0xe963a535L, 0x9e6495a3L, 0x0edb8832L, 0x79dcb8a4L,
+  0xe0d5e91eL, 0x97d2d988L, 0x09b64c2bL, 0x7eb17cbdL, 0xe7b82d07L,
+  0x90bf1d91L, 0x1db71064L, 0x6ab020f2L, 0xf3b97148L, 0x84be41deL,
+  0x1adad47dL, 0x6ddde4ebL, 0xf4d4b551L, 0x83d385c7L, 0x136c9856L,
+  0x646ba8c0L, 0xfd62f97aL, 0x8a65c9ecL, 0x14015c4fL, 0x63066cd9L,
+  0xfa0f3d63L, 0x8d080df5L, 0x3b6e20c8L, 0x4c69105eL, 0xd56041e4L,
+  0xa2677172L, 0x3c03e4d1L, 0x4b04d447L, 0xd20d85fdL, 0xa50ab56bL,
+  0x35b5a8faL, 0x42b2986cL, 0xdbbbc9d6L, 0xacbcf940L, 0x32d86ce3L,
+  0x45df5c75L, 0xdcd60dcfL, 0xabd13d59L, 0x26d930acL, 0x51de003aL,
+  0xc8d75180L, 0xbfd06116L, 0x21b4f4b5L, 0x56b3c423L, 0xcfba9599L,
+  0xb8bda50fL, 0x2802b89eL, 0x5f058808L, 0xc60cd9b2L, 0xb10be924L,
+  0x2f6f7c87L, 0x58684c11L, 0xc1611dabL, 0xb6662d3dL, 0x76dc4190L,
+  0x01db7106L, 0x98d220bcL, 0xefd5102aL, 0x71b18589L, 0x06b6b51fL,
+  0x9fbfe4a5L, 0xe8b8d433L, 0x7807c9a2L, 0x0f00f934L, 0x9609a88eL,
+  0xe10e9818L, 0x7f6a0dbbL, 0x086d3d2dL, 0x91646c97L, 0xe6635c01L,
+  0x6b6b51f4L, 0x1c6c6162L, 0x856530d8L, 0xf262004eL, 0x6c0695edL,
+  0x1b01a57bL, 0x8208f4c1L, 0xf50fc457L, 0x65b0d9c6L, 0x12b7e950L,
+  0x8bbeb8eaL, 0xfcb9887cL, 0x62dd1ddfL, 0x15da2d49L, 0x8cd37cf3L,
+  0xfbd44c65L, 0x4db26158L, 0x3ab551ceL, 0xa3bc0074L, 0xd4bb30e2L,
+  0x4adfa541L, 0x3dd895d7L, 0xa4d1c46dL, 0xd3d6f4fbL, 0x4369e96aL,
+  0x346ed9fcL, 0xad678846L, 0xda60b8d0L, 0x44042d73L, 0x33031de5L,
+  0xaa0a4c5fL, 0xdd0d7cc9L, 0x5005713cL, 0x270241aaL, 0xbe0b1010L,
+  0xc90c2086L, 0x5768b525L, 0x206f85b3L, 0xb966d409L, 0xce61e49fL,
+  0x5edef90eL, 0x29d9c998L, 0xb0d09822L, 0xc7d7a8b4L, 0x59b33d17L,
+  0x2eb40d81L, 0xb7bd5c3bL, 0xc0ba6cadL, 0xedb88320L, 0x9abfb3b6L,
+  0x03b6e20cL, 0x74b1d29aL, 0xead54739L, 0x9dd277afL, 0x04db2615L,
+  0x73dc1683L, 0xe3630b12L, 0x94643b84L, 0x0d6d6a3eL, 0x7a6a5aa8L,
+  0xe40ecf0bL, 0x9309ff9dL, 0x0a00ae27L, 0x7d079eb1L, 0xf00f9344L,
+  0x8708a3d2L, 0x1e01f268L, 0x6906c2feL, 0xf762575dL, 0x806567cbL,
+  0x196c3671L, 0x6e6b06e7L, 0xfed41b76L, 0x89d32be0L, 0x10da7a5aL,
+  0x67dd4accL, 0xf9b9df6fL, 0x8ebeeff9L, 0x17b7be43L, 0x60b08ed5L,
+  0xd6d6a3e8L, 0xa1d1937eL, 0x38d8c2c4L, 0x4fdff252L, 0xd1bb67f1L,
+  0xa6bc5767L, 0x3fb506ddL, 0x48b2364bL, 0xd80d2bdaL, 0xaf0a1b4cL,
+  0x36034af6L, 0x41047a60L, 0xdf60efc3L, 0xa867df55L, 0x316e8eefL,
+  0x4669be79L, 0xcb61b38cL, 0xbc66831aL, 0x256fd2a0L, 0x5268e236L,
+  0xcc0c7795L, 0xbb0b4703L, 0x220216b9L, 0x5505262fL, 0xc5ba3bbeL,
+  0xb2bd0b28L, 0x2bb45a92L, 0x5cb36a04L, 0xc2d7ffa7L, 0xb5d0cf31L,
+  0x2cd99e8bL, 0x5bdeae1dL, 0x9b64c2b0L, 0xec63f226L, 0x756aa39cL,
+  0x026d930aL, 0x9c0906a9L, 0xeb0e363fL, 0x72076785L, 0x05005713L,
+  0x95bf4a82L, 0xe2b87a14L, 0x7bb12baeL, 0x0cb61b38L, 0x92d28e9bL,
+  0xe5d5be0dL, 0x7cdcefb7L, 0x0bdbdf21L, 0x86d3d2d4L, 0xf1d4e242L,
+  0x68ddb3f8L, 0x1fda836eL, 0x81be16cdL, 0xf6b9265bL, 0x6fb077e1L,
+  0x18b74777L, 0x88085ae6L, 0xff0f6a70L, 0x66063bcaL, 0x11010b5cL,
+  0x8f659effL, 0xf862ae69L, 0x616bffd3L, 0x166ccf45L, 0xa00ae278L,
+  0xd70dd2eeL, 0x4e048354L, 0x3903b3c2L, 0xa7672661L, 0xd06016f7L,
+  0x4969474dL, 0x3e6e77dbL, 0xaed16a4aL, 0xd9d65adcL, 0x40df0b66L,
+  0x37d83bf0L, 0xa9bcae53L, 0xdebb9ec5L, 0x47b2cf7fL, 0x30b5ffe9L,
+  0xbdbdf21cL, 0xcabac28aL, 0x53b39330L, 0x24b4a3a6L, 0xbad03605L,
+  0xcdd70693L, 0x54de5729L, 0x23d967bfL, 0xb3667a2eL, 0xc4614ab8L,
+  0x5d681b02L, 0x2a6f2b94L, 0xb40bbe37L, 0xc30c8ea1L, 0x5a05df1bL,
+  0x2d02ef8dL
+};
+#endif
+
+/* =========================================================================
+ * This function can be used by asm versions of crc32()
+ */
+const uLongf * ZEXPORT get_crc_table()
+{
+#ifdef DYNAMIC_CRC_TABLE
+  if (crc_table_empty) make_crc_table();
+#endif
+  return (const uLongf *)crc_table;
+}
+
+/* ========================================================================= */
+#define DO1(buf) crc = crc_table[((int)crc ^ (*buf++)) & 0xff] ^ (crc >> 8);
+#define DO2(buf)  DO1(buf); DO1(buf);
+#define DO4(buf)  DO2(buf); DO2(buf);
+#define DO8(buf)  DO4(buf); DO4(buf);
+
+/* ========================================================================= */
+uLong ZEXPORT crc32(crc, buf, len)
+    uLong crc;
+    const Bytef *buf;
+    uInt len;
+{
+    if (buf == Z_NULL) return 0L;
+#ifdef DYNAMIC_CRC_TABLE
+    if (crc_table_empty)
+      make_crc_table();
+#endif
+    crc = crc ^ 0xffffffffL;
+    while (len >= 8)
+    {
+      DO8(buf);
+      len -= 8;
+    }
+    if (len) do {
+      DO1(buf);
+    } while (--len);
+    return crc ^ 0xffffffffL;
+}
diff --git a/zlib/deflate.c b/zlib/deflate.c
new file mode 100644
index 000000000..cd58ae66f
--- /dev/null
+++ b/zlib/deflate.c
@@ -0,0 +1,1350 @@
+/* deflate.c -- compress data using the deflation algorithm
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/*
+ *  ALGORITHM
+ *
+ *      The "deflation" process depends on being able to identify portions
+ *      of the input text which are identical to earlier input (within a
+ *      sliding window trailing behind the input currently being processed).
+ *
+ *      The most straightforward technique turns out to be the fastest for
+ *      most input files: try all possible matches and select the longest.
+ *      The key feature of this algorithm is that insertions into the string
+ *      dictionary are very simple and thus fast, and deletions are avoided
+ *      completely. Insertions are performed at each input character, whereas
+ *      string matches are performed only when the previous match ends. So it
+ *      is preferable to spend more time in matches to allow very fast string
+ *      insertions and avoid deletions. The matching algorithm for small
+ *      strings is inspired from that of Rabin & Karp. A brute force approach
+ *      is used to find longer strings when a small match has been found.
+ *      A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
+ *      (by Leonid Broukhis).
+ *         A previous version of this file used a more sophisticated algorithm
+ *      (by Fiala and Greene) which is guaranteed to run in linear amortized
+ *      time, but has a larger average cost, uses more memory and is patented.
+ *      However the F&G algorithm may be faster for some highly redundant
+ *      files if the parameter max_chain_length (described below) is too large.
+ *
+ *  ACKNOWLEDGEMENTS
+ *
+ *      The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
+ *      I found it in 'freeze' written by Leonid Broukhis.
+ *      Thanks to many people for bug reports and testing.
+ *
+ *  REFERENCES
+ *
+ *      Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
+ *      Available in ftp://ds.internic.net/rfc/rfc1951.txt
+ *
+ *      A description of the Rabin and Karp algorithm is given in the book
+ *         "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
+ *
+ *      Fiala,E.R., and Greene,D.H.
+ *         Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
+ *
+ */
+
+/* @(#) $Id: deflate.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#include "deflate.h"
+
+const char deflate_copyright[] =
+   " deflate 1.1.4 Copyright 1995-2002 Jean-loup Gailly ";
+/*
+  If you use the zlib library in a product, an acknowledgment is welcome
+  in the documentation of your product. If for some reason you cannot
+  include such an acknowledgment, I would appreciate that you keep this
+  copyright string in the executable of your product.
+ */
+
+/* ===========================================================================
+ *  Function prototypes.
+ */
+typedef enum {
+    need_more,      /* block not completed, need more input or more output */
+    block_done,     /* block flush performed */
+    finish_started, /* finish started, need only more output at next deflate */
+    finish_done     /* finish done, accept no more input or output */
+} block_state;
+
+typedef block_state (*compress_func) OF((deflate_state *s, int flush));
+/* Compression function. Returns the block state after the call. */
+
+local void fill_window    OF((deflate_state *s));
+local block_state deflate_stored OF((deflate_state *s, int flush));
+local block_state deflate_fast   OF((deflate_state *s, int flush));
+local block_state deflate_slow   OF((deflate_state *s, int flush));
+local void lm_init        OF((deflate_state *s));
+local void putShortMSB    OF((deflate_state *s, uInt b));
+local void flush_pending  OF((z_streamp strm));
+local int read_buf        OF((z_streamp strm, Bytef *buf, unsigned size));
+#ifdef ASMV
+      void match_init OF((void)); /* asm code initialization */
+      uInt longest_match  OF((deflate_state *s, IPos cur_match));
+#else
+local uInt longest_match  OF((deflate_state *s, IPos cur_match));
+#endif
+
+#ifdef DEBUG
+local  void check_match OF((deflate_state *s, IPos start, IPos match,
+                            int length));
+#endif
+
+/* ===========================================================================
+ * Local data
+ */
+
+#define NIL 0
+/* Tail of hash chains */
+
+#ifndef TOO_FAR
+#  define TOO_FAR 4096
+#endif
+/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
+
+#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
+/* Minimum amount of lookahead, except at the end of the input file.
+ * See deflate.c for comments about the MIN_MATCH+1.
+ */
+
+/* Values for max_lazy_match, good_match and max_chain_length, depending on
+ * the desired pack level (0..9). The values given below have been tuned to
+ * exclude worst case performance for pathological files. Better values may be
+ * found for specific files.
+ */
+typedef struct config_s {
+   ush good_length; /* reduce lazy search above this match length */
+   ush max_lazy;    /* do not perform lazy search above this match length */
+   ush nice_length; /* quit search above this match length */
+   ush max_chain;
+   compress_func func;
+} config;
+
+local const config configuration_table[10] = {
+/*      good lazy nice chain */
+/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
+/* 1 */ {4,    4,  8,    4, deflate_fast}, /* maximum speed, no lazy matches */
+/* 2 */ {4,    5, 16,    8, deflate_fast},
+/* 3 */ {4,    6, 32,   32, deflate_fast},
+
+/* 4 */ {4,    4, 16,   16, deflate_slow},  /* lazy matches */
+/* 5 */ {8,   16, 32,   32, deflate_slow},
+/* 6 */ {8,   16, 128, 128, deflate_slow},
+/* 7 */ {8,   32, 128, 256, deflate_slow},
+/* 8 */ {32, 128, 258, 1024, deflate_slow},
+/* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* maximum compression */
+
+/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
+ * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
+ * meaning.
+ */
+
+#define EQUAL 0
+/* result of memcmp for equal strings */
+
+struct static_tree_desc_s {int dummy;}; /* for buggy compilers */
+
+/* ===========================================================================
+ * Update a hash value with the given input byte
+ * IN  assertion: all calls to to UPDATE_HASH are made with consecutive
+ *    input characters, so that a running hash key can be computed from the
+ *    previous key instead of complete recalculation each time.
+ */
+#define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
+
+
+/* ===========================================================================
+ * Insert string str in the dictionary and set match_head to the previous head
+ * of the hash chain (the most recent string with same hash key). Return
+ * the previous length of the hash chain.
+ * If this file is compiled with -DFASTEST, the compression level is forced
+ * to 1, and no hash chains are maintained.
+ * IN  assertion: all calls to to INSERT_STRING are made with consecutive
+ *    input characters and the first MIN_MATCH bytes of str are valid
+ *    (except for the last MIN_MATCH-1 bytes of the input file).
+ */
+#ifdef FASTEST
+#define INSERT_STRING(s, str, match_head) \
+   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
+    match_head = s->head[s->ins_h], \
+    s->head[s->ins_h] = (Pos)(str))
+#else
+#define INSERT_STRING(s, str, match_head) \
+   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
+    s->prev[(str) & s->w_mask] = match_head = s->head[s->ins_h], \
+    s->head[s->ins_h] = (Pos)(str))
+#endif
+
+/* ===========================================================================
+ * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
+ * prev[] will be initialized on the fly.
+ */
+#define CLEAR_HASH(s) \
+    s->head[s->hash_size-1] = NIL; \
+    zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
+
+/* ========================================================================= */
+int ZEXPORT deflateInit_(strm, level, version, stream_size)
+    z_streamp strm;
+    int level;
+    const char *version;
+    int stream_size;
+{
+    return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
+			 Z_DEFAULT_STRATEGY, version, stream_size);
+    /* To do: ignore strm->next_in if we use it as window */
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
+		  version, stream_size)
+    z_streamp strm;
+    int  level;
+    int  method;
+    int  windowBits;
+    int  memLevel;
+    int  strategy;
+    const char *version;
+    int stream_size;
+{
+    deflate_state *s;
+    int noheader = 0;
+    static const char* my_version = ZLIB_VERSION;
+
+    ushf *overlay;
+    /* We overlay pending_buf and d_buf+l_buf. This works since the average
+     * output size for (length,distance) codes is <= 24 bits.
+     */
+
+    if (version == Z_NULL || version[0] != my_version[0] ||
+        stream_size != sizeof(z_stream)) {
+	return Z_VERSION_ERROR;
+    }
+    if (strm == Z_NULL) return Z_STREAM_ERROR;
+
+    strm->msg = Z_NULL;
+    if (strm->zalloc == Z_NULL) {
+	strm->zalloc = zcalloc;
+	strm->opaque = (voidpf)0;
+    }
+    if (strm->zfree == Z_NULL) strm->zfree = zcfree;
+
+    if (level == Z_DEFAULT_COMPRESSION) level = 6;
+#ifdef FASTEST
+    level = 1;
+#endif
+
+    if (windowBits < 0) { /* undocumented feature: suppress zlib header */
+        noheader = 1;
+        windowBits = -windowBits;
+    }
+    if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
+        windowBits < 9 || windowBits > 15 || level < 0 || level > 9 ||
+	strategy < 0 || strategy > Z_HUFFMAN_ONLY) {
+        return Z_STREAM_ERROR;
+    }
+    s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
+    if (s == Z_NULL) return Z_MEM_ERROR;
+    strm->state = (struct internal_state FAR *)s;
+    s->strm = strm;
+
+    s->noheader = noheader;
+    s->w_bits = windowBits;
+    s->w_size = 1 << s->w_bits;
+    s->w_mask = s->w_size - 1;
+
+    s->hash_bits = memLevel + 7;
+    s->hash_size = 1 << s->hash_bits;
+    s->hash_mask = s->hash_size - 1;
+    s->hash_shift =  ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
+
+    s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
+    s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
+    s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
+
+    s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
+
+    overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
+    s->pending_buf = (uchf *) overlay;
+    s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
+
+    if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
+        s->pending_buf == Z_NULL) {
+        strm->msg = (char*)ERR_MSG(Z_MEM_ERROR);
+        deflateEnd (strm);
+        return Z_MEM_ERROR;
+    }
+    s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
+    s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
+
+    s->level = level;
+    s->strategy = strategy;
+    s->method = (Byte)method;
+
+    return deflateReset(strm);
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
+    z_streamp strm;
+    const Bytef *dictionary;
+    uInt  dictLength;
+{
+    deflate_state *s;
+    uInt length = dictLength;
+    uInt n;
+    IPos hash_head = 0;
+
+    if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL ||
+        strm->state->status != INIT_STATE) return Z_STREAM_ERROR;
+
+    s = strm->state;
+    strm->adler = adler32(strm->adler, dictionary, dictLength);
+
+    if (length < MIN_MATCH) return Z_OK;
+    if (length > MAX_DIST(s)) {
+	length = MAX_DIST(s);
+#ifndef USE_DICT_HEAD
+	dictionary += dictLength - length; /* use the tail of the dictionary */
+#endif
+    }
+    zmemcpy(s->window, dictionary, length);
+    s->strstart = length;
+    s->block_start = (long)length;
+
+    /* Insert all strings in the hash table (except for the last two bytes).
+     * s->lookahead stays null, so s->ins_h will be recomputed at the next
+     * call of fill_window.
+     */
+    s->ins_h = s->window[0];
+    UPDATE_HASH(s, s->ins_h, s->window[1]);
+    for (n = 0; n <= length - MIN_MATCH; n++) {
+	INSERT_STRING(s, n, hash_head);
+    }
+    if (hash_head) hash_head = 0;  /* to make compiler happy */
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateReset (strm)
+    z_streamp strm;
+{
+    deflate_state *s;
+    
+    if (strm == Z_NULL || strm->state == Z_NULL ||
+        strm->zalloc == Z_NULL || strm->zfree == Z_NULL) return Z_STREAM_ERROR;
+
+    strm->total_in = strm->total_out = 0;
+    strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
+    strm->data_type = Z_UNKNOWN;
+
+    s = (deflate_state *)strm->state;
+    s->pending = 0;
+    s->pending_out = s->pending_buf;
+
+    if (s->noheader < 0) {
+        s->noheader = 0; /* was set to -1 by deflate(..., Z_FINISH); */
+    }
+    s->status = s->noheader ? BUSY_STATE : INIT_STATE;
+    strm->adler = 1;
+    s->last_flush = Z_NO_FLUSH;
+
+    _tr_init(s);
+    lm_init(s);
+
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateParams(strm, level, strategy)
+    z_streamp strm;
+    int level;
+    int strategy;
+{
+    deflate_state *s;
+    compress_func func;
+    int err = Z_OK;
+
+    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
+    s = strm->state;
+
+    if (level == Z_DEFAULT_COMPRESSION) {
+	level = 6;
+    }
+    if (level < 0 || level > 9 || strategy < 0 || strategy > Z_HUFFMAN_ONLY) {
+	return Z_STREAM_ERROR;
+    }
+    func = configuration_table[s->level].func;
+
+    if (func != configuration_table[level].func && strm->total_in != 0) {
+	/* Flush the last buffer: */
+	err = deflate(strm, Z_PARTIAL_FLUSH);
+    }
+    if (s->level != level) {
+	s->level = level;
+	s->max_lazy_match   = configuration_table[level].max_lazy;
+	s->good_match       = configuration_table[level].good_length;
+	s->nice_match       = configuration_table[level].nice_length;
+	s->max_chain_length = configuration_table[level].max_chain;
+    }
+    s->strategy = strategy;
+    return err;
+}
+
+/* =========================================================================
+ * Put a short in the pending buffer. The 16-bit value is put in MSB order.
+ * IN assertion: the stream state is correct and there is enough room in
+ * pending_buf.
+ */
+local void putShortMSB (s, b)
+    deflate_state *s;
+    uInt b;
+{
+    put_byte(s, (Byte)(b >> 8));
+    put_byte(s, (Byte)(b & 0xff));
+}   
+
+/* =========================================================================
+ * Flush as much pending output as possible. All deflate() output goes
+ * through this function so some applications may wish to modify it
+ * to avoid allocating a large strm->next_out buffer and copying into it.
+ * (See also read_buf()).
+ */
+local void flush_pending(strm)
+    z_streamp strm;
+{
+    unsigned len = strm->state->pending;
+
+    if (len > strm->avail_out) len = strm->avail_out;
+    if (len == 0) return;
+
+    zmemcpy(strm->next_out, strm->state->pending_out, len);
+    strm->next_out  += len;
+    strm->state->pending_out  += len;
+    strm->total_out += len;
+    strm->avail_out  -= len;
+    strm->state->pending -= len;
+    if (strm->state->pending == 0) {
+        strm->state->pending_out = strm->state->pending_buf;
+    }
+}
+
+/* ========================================================================= */
+int ZEXPORT deflate (strm, flush)
+    z_streamp strm;
+    int flush;
+{
+    int old_flush; /* value of flush param for previous deflate call */
+    deflate_state *s;
+
+    if (strm == Z_NULL || strm->state == Z_NULL ||
+	flush > Z_FINISH || flush < 0) {
+        return Z_STREAM_ERROR;
+    }
+    s = strm->state;
+
+    if (strm->next_out == Z_NULL ||
+        (strm->next_in == Z_NULL && strm->avail_in != 0) ||
+	(s->status == FINISH_STATE && flush != Z_FINISH)) {
+        ERR_RETURN(strm, Z_STREAM_ERROR);
+    }
+    if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
+
+    s->strm = strm; /* just in case */
+    old_flush = s->last_flush;
+    s->last_flush = flush;
+
+    /* Write the zlib header */
+    if (s->status == INIT_STATE) {
+
+        uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
+        uInt level_flags = (s->level-1) >> 1;
+
+        if (level_flags > 3) level_flags = 3;
+        header |= (level_flags << 6);
+	if (s->strstart != 0) header |= PRESET_DICT;
+        header += 31 - (header % 31);
+
+        s->status = BUSY_STATE;
+        putShortMSB(s, header);
+
+	/* Save the adler32 of the preset dictionary: */
+	if (s->strstart != 0) {
+	    putShortMSB(s, (uInt)(strm->adler >> 16));
+	    putShortMSB(s, (uInt)(strm->adler & 0xffff));
+	}
+	strm->adler = 1L;
+    }
+
+    /* Flush as much pending output as possible */
+    if (s->pending != 0) {
+        flush_pending(strm);
+        if (strm->avail_out == 0) {
+	    /* Since avail_out is 0, deflate will be called again with
+	     * more output space, but possibly with both pending and
+	     * avail_in equal to zero. There won't be anything to do,
+	     * but this is not an error situation so make sure we
+	     * return OK instead of BUF_ERROR at next call of deflate:
+             */
+	    s->last_flush = -1;
+	    return Z_OK;
+	}
+
+    /* Make sure there is something to do and avoid duplicate consecutive
+     * flushes. For repeated and useless calls with Z_FINISH, we keep
+     * returning Z_STREAM_END instead of Z_BUFF_ERROR.
+     */
+    } else if (strm->avail_in == 0 && flush <= old_flush &&
+	       flush != Z_FINISH) {
+        ERR_RETURN(strm, Z_BUF_ERROR);
+    }
+
+    /* User must not provide more input after the first FINISH: */
+    if (s->status == FINISH_STATE && strm->avail_in != 0) {
+        ERR_RETURN(strm, Z_BUF_ERROR);
+    }
+
+    /* Start a new block or continue the current one.
+     */
+    if (strm->avail_in != 0 || s->lookahead != 0 ||
+        (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
+        block_state bstate;
+
+	bstate = (*(configuration_table[s->level].func))(s, flush);
+
+        if (bstate == finish_started || bstate == finish_done) {
+            s->status = FINISH_STATE;
+        }
+        if (bstate == need_more || bstate == finish_started) {
+	    if (strm->avail_out == 0) {
+	        s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
+	    }
+	    return Z_OK;
+	    /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
+	     * of deflate should use the same flush parameter to make sure
+	     * that the flush is complete. So we don't have to output an
+	     * empty block here, this will be done at next call. This also
+	     * ensures that for a very small output buffer, we emit at most
+	     * one empty block.
+	     */
+	}
+        if (bstate == block_done) {
+            if (flush == Z_PARTIAL_FLUSH) {
+                _tr_align(s);
+            } else { /* FULL_FLUSH or SYNC_FLUSH */
+                _tr_stored_block(s, (char*)0, 0L, 0);
+                /* For a full flush, this empty block will be recognized
+                 * as a special marker by inflate_sync().
+                 */
+                if (flush == Z_FULL_FLUSH) {
+                    CLEAR_HASH(s);             /* forget history */
+                }
+            }
+            flush_pending(strm);
+	    if (strm->avail_out == 0) {
+	      s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
+	      return Z_OK;
+	    }
+        }
+    }
+    Assert(strm->avail_out > 0, "bug2");
+
+    if (flush != Z_FINISH) return Z_OK;
+    if (s->noheader) return Z_STREAM_END;
+
+    /* Write the zlib trailer (adler32) */
+    putShortMSB(s, (uInt)(strm->adler >> 16));
+    putShortMSB(s, (uInt)(strm->adler & 0xffff));
+    flush_pending(strm);
+    /* If avail_out is zero, the application will call deflate again
+     * to flush the rest.
+     */
+    s->noheader = -1; /* write the trailer only once! */
+    return s->pending != 0 ? Z_OK : Z_STREAM_END;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateEnd (strm)
+    z_streamp strm;
+{
+    int status;
+
+    if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR;
+
+    status = strm->state->status;
+    if (status != INIT_STATE && status != BUSY_STATE &&
+	status != FINISH_STATE) {
+      return Z_STREAM_ERROR;
+    }
+
+    /* Deallocate in reverse order of allocations: */
+    TRY_FREE(strm, strm->state->pending_buf);
+    TRY_FREE(strm, strm->state->head);
+    TRY_FREE(strm, strm->state->prev);
+    TRY_FREE(strm, strm->state->window);
+
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+
+    return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
+}
+
+/* =========================================================================
+ * Copy the source state to the destination state.
+ * To simplify the source, this is not supported for 16-bit MSDOS (which
+ * doesn't have enough memory anyway to duplicate compression states).
+ */
+int ZEXPORT deflateCopy (dest, source)
+    z_streamp dest;
+    z_streamp source;
+{
+#ifdef MAXSEG_64K
+    return Z_STREAM_ERROR;
+#else
+    deflate_state *ds;
+    deflate_state *ss;
+    ushf *overlay;
+
+
+    if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) {
+        return Z_STREAM_ERROR;
+    }
+
+    ss = source->state;
+
+    *dest = *source;
+
+    ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
+    if (ds == Z_NULL) return Z_MEM_ERROR;
+    dest->state = (struct internal_state FAR *) ds;
+    *ds = *ss;
+    ds->strm = dest;
+
+    ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
+    ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
+    ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
+    overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
+    ds->pending_buf = (uchf *) overlay;
+
+    if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
+        ds->pending_buf == Z_NULL) {
+        deflateEnd (dest);
+        return Z_MEM_ERROR;
+    }
+    /* following zmemcpy do not work for 16-bit MSDOS */
+    zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
+    zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos));
+    zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos));
+    zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
+
+    ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
+    ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
+    ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
+
+    ds->l_desc.dyn_tree = ds->dyn_ltree;
+    ds->d_desc.dyn_tree = ds->dyn_dtree;
+    ds->bl_desc.dyn_tree = ds->bl_tree;
+
+    return Z_OK;
+#endif
+}
+
+/* ===========================================================================
+ * Read a new buffer from the current input stream, update the adler32
+ * and total number of bytes read.  All deflate() input goes through
+ * this function so some applications may wish to modify it to avoid
+ * allocating a large strm->next_in buffer and copying from it.
+ * (See also flush_pending()).
+ */
+local int read_buf(strm, buf, size)
+    z_streamp strm;
+    Bytef *buf;
+    unsigned size;
+{
+    unsigned len = strm->avail_in;
+
+    if (len > size) len = size;
+    if (len == 0) return 0;
+
+    strm->avail_in  -= len;
+
+    if (!strm->state->noheader) {
+        strm->adler = adler32(strm->adler, strm->next_in, len);
+    }
+    zmemcpy(buf, strm->next_in, len);
+    strm->next_in  += len;
+    strm->total_in += len;
+
+    return (int)len;
+}
+
+/* ===========================================================================
+ * Initialize the "longest match" routines for a new zlib stream
+ */
+local void lm_init (s)
+    deflate_state *s;
+{
+    s->window_size = (ulg)2L*s->w_size;
+
+    CLEAR_HASH(s);
+
+    /* Set the default configuration parameters:
+     */
+    s->max_lazy_match   = configuration_table[s->level].max_lazy;
+    s->good_match       = configuration_table[s->level].good_length;
+    s->nice_match       = configuration_table[s->level].nice_length;
+    s->max_chain_length = configuration_table[s->level].max_chain;
+
+    s->strstart = 0;
+    s->block_start = 0L;
+    s->lookahead = 0;
+    s->match_length = s->prev_length = MIN_MATCH-1;
+    s->match_available = 0;
+    s->ins_h = 0;
+#ifdef ASMV
+    match_init(); /* initialize the asm code */
+#endif
+}
+
+/* ===========================================================================
+ * Set match_start to the longest match starting at the given string and
+ * return its length. Matches shorter or equal to prev_length are discarded,
+ * in which case the result is equal to prev_length and match_start is
+ * garbage.
+ * IN assertions: cur_match is the head of the hash chain for the current
+ *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
+ * OUT assertion: the match length is not greater than s->lookahead.
+ */
+#ifndef ASMV
+/* For 80x86 and 680x0, an optimized version will be provided in match.asm or
+ * match.S. The code will be functionally equivalent.
+ */
+#ifndef FASTEST
+local uInt longest_match(s, cur_match)
+    deflate_state *s;
+    IPos cur_match;                             /* current match */
+{
+    unsigned chain_length = s->max_chain_length;/* max hash chain length */
+    register Bytef *scan = s->window + s->strstart; /* current string */
+    register Bytef *match;                       /* matched string */
+    register int len;                           /* length of current match */
+    int best_len = s->prev_length;              /* best match length so far */
+    int nice_match = s->nice_match;             /* stop if match long enough */
+    IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
+        s->strstart - (IPos)MAX_DIST(s) : NIL;
+    /* Stop when cur_match becomes <= limit. To simplify the code,
+     * we prevent matches with the string of window index 0.
+     */
+    Posf *prev = s->prev;
+    uInt wmask = s->w_mask;
+
+#ifdef UNALIGNED_OK
+    /* Compare two bytes at a time. Note: this is not always beneficial.
+     * Try with and without -DUNALIGNED_OK to check.
+     */
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
+    register ush scan_start = *(ushf*)scan;
+    register ush scan_end   = *(ushf*)(scan+best_len-1);
+#else
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+    register Byte scan_end1  = scan[best_len-1];
+    register Byte scan_end   = scan[best_len];
+#endif
+
+    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+     * It is easy to get rid of this optimization if necessary.
+     */
+    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+    /* Do not waste too much time if we already have a good match: */
+    if (s->prev_length >= s->good_match) {
+        chain_length >>= 2;
+    }
+    /* Do not look for matches beyond the end of the input. This is necessary
+     * to make deflate deterministic.
+     */
+    if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead;
+
+    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+    do {
+        Assert(cur_match < s->strstart, "no future");
+        match = s->window + cur_match;
+
+        /* Skip to next match if the match length cannot increase
+         * or if the match length is less than 2:
+         */
+#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
+        /* This code assumes sizeof(unsigned short) == 2. Do not use
+         * UNALIGNED_OK if your compiler uses a different size.
+         */
+        if (*(ushf*)(match+best_len-1) != scan_end ||
+            *(ushf*)match != scan_start) continue;
+
+        /* It is not necessary to compare scan[2] and match[2] since they are
+         * always equal when the other bytes match, given that the hash keys
+         * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
+         * strstart+3, +5, ... up to strstart+257. We check for insufficient
+         * lookahead only every 4th comparison; the 128th check will be made
+         * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
+         * necessary to put more guard bytes at the end of the window, or
+         * to check more often for insufficient lookahead.
+         */
+        Assert(scan[2] == match[2], "scan[2]?");
+        scan++, match++;
+        do {
+        } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 scan < strend);
+        /* The funny "do {}" generates better code on most compilers */
+
+        /* Here, scan <= window+strstart+257 */
+        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+        if (*scan == *match) scan++;
+
+        len = (MAX_MATCH - 1) - (int)(strend-scan);
+        scan = strend - (MAX_MATCH-1);
+
+#else /* UNALIGNED_OK */
+
+        if (match[best_len]   != scan_end  ||
+            match[best_len-1] != scan_end1 ||
+            *match            != *scan     ||
+            *++match          != scan[1])      continue;
+
+        /* The check at best_len-1 can be removed because it will be made
+         * again later. (This heuristic is not always a win.)
+         * It is not necessary to compare scan[2] and match[2] since they
+         * are always equal when the other bytes match, given that
+         * the hash keys are equal and that HASH_BITS >= 8.
+         */
+        scan += 2, match++;
+        Assert(*scan == *match, "match[2]?");
+
+        /* We check for insufficient lookahead only every 8th comparison;
+         * the 256th check will be made at strstart+258.
+         */
+        do {
+        } while (*++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 scan < strend);
+
+        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+        len = MAX_MATCH - (int)(strend - scan);
+        scan = strend - MAX_MATCH;
+
+#endif /* UNALIGNED_OK */
+
+        if (len > best_len) {
+            s->match_start = cur_match;
+            best_len = len;
+            if (len >= nice_match) break;
+#ifdef UNALIGNED_OK
+            scan_end = *(ushf*)(scan+best_len-1);
+#else
+            scan_end1  = scan[best_len-1];
+            scan_end   = scan[best_len];
+#endif
+        }
+    } while ((cur_match = prev[cur_match & wmask]) > limit
+             && --chain_length != 0);
+
+    if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
+    return s->lookahead;
+}
+
+#else /* FASTEST */
+/* ---------------------------------------------------------------------------
+ * Optimized version for level == 1 only
+ */
+local uInt longest_match(s, cur_match)
+    deflate_state *s;
+    IPos cur_match;                             /* current match */
+{
+    register Bytef *scan = s->window + s->strstart; /* current string */
+    register Bytef *match;                       /* matched string */
+    register int len;                           /* length of current match */
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+
+    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+     * It is easy to get rid of this optimization if necessary.
+     */
+    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+    Assert(cur_match < s->strstart, "no future");
+
+    match = s->window + cur_match;
+
+    /* Return failure if the match length is less than 2:
+     */
+    if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
+
+    /* The check at best_len-1 can be removed because it will be made
+     * again later. (This heuristic is not always a win.)
+     * It is not necessary to compare scan[2] and match[2] since they
+     * are always equal when the other bytes match, given that
+     * the hash keys are equal and that HASH_BITS >= 8.
+     */
+    scan += 2, match += 2;
+    Assert(*scan == *match, "match[2]?");
+
+    /* We check for insufficient lookahead only every 8th comparison;
+     * the 256th check will be made at strstart+258.
+     */
+    do {
+    } while (*++scan == *++match && *++scan == *++match &&
+	     *++scan == *++match && *++scan == *++match &&
+	     *++scan == *++match && *++scan == *++match &&
+	     *++scan == *++match && *++scan == *++match &&
+	     scan < strend);
+
+    Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+    len = MAX_MATCH - (int)(strend - scan);
+
+    if (len < MIN_MATCH) return MIN_MATCH - 1;
+
+    s->match_start = cur_match;
+    return len <= s->lookahead ? len : s->lookahead;
+}
+#endif /* FASTEST */
+#endif /* ASMV */
+
+#ifdef DEBUG
+/* ===========================================================================
+ * Check that the match at match_start is indeed a match.
+ */
+local void check_match(s, start, match, length)
+    deflate_state *s;
+    IPos start, match;
+    int length;
+{
+    /* check that the match is indeed a match */
+    if (zmemcmp(s->window + match,
+                s->window + start, length) != EQUAL) {
+        fprintf(stderr, " start %u, match %u, length %d\n",
+		start, match, length);
+        do {
+	    fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
+	} while (--length != 0);
+        z_error("invalid match");
+    }
+    if (z_verbose > 1) {
+        fprintf(stderr,"\\[%d,%d]", start-match, length);
+        do { putc(s->window[start++], stderr); } while (--length != 0);
+    }
+}
+#else
+#  define check_match(s, start, match, length)
+#endif
+
+/* ===========================================================================
+ * Fill the window when the lookahead becomes insufficient.
+ * Updates strstart and lookahead.
+ *
+ * IN assertion: lookahead < MIN_LOOKAHEAD
+ * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
+ *    At least one byte has been read, or avail_in == 0; reads are
+ *    performed for at least two bytes (required for the zip translate_eol
+ *    option -- not supported here).
+ */
+local void fill_window(s)
+    deflate_state *s;
+{
+    register unsigned n, m;
+    register Posf *p;
+    unsigned more;    /* Amount of free space at the end of the window. */
+    uInt wsize = s->w_size;
+
+    do {
+        more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
+
+        /* Deal with !@#$% 64K limit: */
+        if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
+            more = wsize;
+
+        } else if (more == (unsigned)(-1)) {
+            /* Very unlikely, but possible on 16 bit machine if strstart == 0
+             * and lookahead == 1 (input done one byte at time)
+             */
+            more--;
+
+        /* If the window is almost full and there is insufficient lookahead,
+         * move the upper half to the lower one to make room in the upper half.
+         */
+        } else if (s->strstart >= wsize+MAX_DIST(s)) {
+
+            zmemcpy(s->window, s->window+wsize, (unsigned)wsize);
+            s->match_start -= wsize;
+            s->strstart    -= wsize; /* we now have strstart >= MAX_DIST */
+            s->block_start -= (long) wsize;
+
+            /* Slide the hash table (could be avoided with 32 bit values
+               at the expense of memory usage). We slide even when level == 0
+               to keep the hash table consistent if we switch back to level > 0
+               later. (Using level 0 permanently is not an optimal usage of
+               zlib, so we don't care about this pathological case.)
+             */
+	    n = s->hash_size;
+	    p = &s->head[n];
+	    do {
+		m = *--p;
+		*p = (Pos)(m >= wsize ? m-wsize : NIL);
+	    } while (--n);
+
+	    n = wsize;
+#ifndef FASTEST
+	    p = &s->prev[n];
+	    do {
+		m = *--p;
+		*p = (Pos)(m >= wsize ? m-wsize : NIL);
+		/* If n is not on any hash chain, prev[n] is garbage but
+		 * its value will never be used.
+		 */
+	    } while (--n);
+#endif
+            more += wsize;
+        }
+        if (s->strm->avail_in == 0) return;
+
+        /* If there was no sliding:
+         *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
+         *    more == window_size - lookahead - strstart
+         * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
+         * => more >= window_size - 2*WSIZE + 2
+         * In the BIG_MEM or MMAP case (not yet supported),
+         *   window_size == input_size + MIN_LOOKAHEAD  &&
+         *   strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
+         * Otherwise, window_size == 2*WSIZE so more >= 2.
+         * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
+         */
+        Assert(more >= 2, "more < 2");
+
+        n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
+        s->lookahead += n;
+
+        /* Initialize the hash value now that we have some input: */
+        if (s->lookahead >= MIN_MATCH) {
+            s->ins_h = s->window[s->strstart];
+            UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
+#if MIN_MATCH != 3
+            Call UPDATE_HASH() MIN_MATCH-3 more times
+#endif
+        }
+        /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
+         * but this is not important since only literal bytes will be emitted.
+         */
+
+    } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
+}
+
+/* ===========================================================================
+ * Flush the current block, with given end-of-file flag.
+ * IN assertion: strstart is set to the end of the current match.
+ */
+#define FLUSH_BLOCK_ONLY(s, eof) { \
+   _tr_flush_block(s, (s->block_start >= 0L ? \
+                   (charf *)&s->window[(unsigned)s->block_start] : \
+                   (charf *)Z_NULL), \
+		(ulg)((long)s->strstart - s->block_start), \
+		(eof)); \
+   s->block_start = s->strstart; \
+   flush_pending(s->strm); \
+   Tracev((stderr,"[FLUSH]")); \
+}
+
+/* Same but force premature exit if necessary. */
+#define FLUSH_BLOCK(s, eof) { \
+   FLUSH_BLOCK_ONLY(s, eof); \
+   if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \
+}
+
+/* ===========================================================================
+ * Copy without compression as much as possible from the input stream, return
+ * the current block state.
+ * This function does not insert new strings in the dictionary since
+ * uncompressible data is probably not useful. This function is used
+ * only for the level=0 compression option.
+ * NOTE: this function should be optimized to avoid extra copying from
+ * window to pending_buf.
+ */
+local block_state deflate_stored(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    /* Stored blocks are limited to 0xffff bytes, pending_buf is limited
+     * to pending_buf_size, and each stored block has a 5 byte header:
+     */
+    ulg max_block_size = 0xffff;
+    ulg max_start;
+
+    if (max_block_size > s->pending_buf_size - 5) {
+        max_block_size = s->pending_buf_size - 5;
+    }
+
+    /* Copy as much as possible from input to output: */
+    for (;;) {
+        /* Fill the window as much as possible: */
+        if (s->lookahead <= 1) {
+
+            Assert(s->strstart < s->w_size+MAX_DIST(s) ||
+		   s->block_start >= (long)s->w_size, "slide too late");
+
+            fill_window(s);
+            if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more;
+
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+	Assert(s->block_start >= 0L, "block gone");
+
+	s->strstart += s->lookahead;
+	s->lookahead = 0;
+
+	/* Emit a stored block if pending_buf will be full: */
+ 	max_start = s->block_start + max_block_size;
+        if (s->strstart == 0 || (ulg)s->strstart >= max_start) {
+	    /* strstart == 0 is possible when wraparound on 16-bit machine */
+	    s->lookahead = (uInt)(s->strstart - max_start);
+	    s->strstart = (uInt)max_start;
+            FLUSH_BLOCK(s, 0);
+	}
+	/* Flush if we may have to slide, otherwise block_start may become
+         * negative and the data will be gone:
+         */
+        if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) {
+            FLUSH_BLOCK(s, 0);
+	}
+    }
+    FLUSH_BLOCK(s, flush == Z_FINISH);
+    return flush == Z_FINISH ? finish_done : block_done;
+}
+
+/* ===========================================================================
+ * Compress as much as possible from the input stream, return the current
+ * block state.
+ * This function does not perform lazy evaluation of matches and inserts
+ * new strings in the dictionary only for unmatched strings or for short
+ * matches. It is used only for the fast compression options.
+ */
+local block_state deflate_fast(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    IPos hash_head = NIL; /* head of the hash chain */
+    int bflush;           /* set if current block must be flushed */
+
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the next match, plus MIN_MATCH bytes to insert the
+         * string following the next match.
+         */
+        if (s->lookahead < MIN_LOOKAHEAD) {
+            fill_window(s);
+            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
+	        return need_more;
+	    }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* Insert the string window[strstart .. strstart+2] in the
+         * dictionary, and set hash_head to the head of the hash chain:
+         */
+        if (s->lookahead >= MIN_MATCH) {
+            INSERT_STRING(s, s->strstart, hash_head);
+        }
+
+        /* Find the longest match, discarding those <= prev_length.
+         * At this point we have always match_length < MIN_MATCH
+         */
+        if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
+            /* To simplify the code, we prevent matches with the string
+             * of window index 0 (in particular we have to avoid a match
+             * of the string with itself at the start of the input file).
+             */
+            if (s->strategy != Z_HUFFMAN_ONLY) {
+                s->match_length = longest_match (s, hash_head);
+            }
+            /* longest_match() sets match_start */
+        }
+        if (s->match_length >= MIN_MATCH) {
+            check_match(s, s->strstart, s->match_start, s->match_length);
+
+            _tr_tally_dist(s, s->strstart - s->match_start,
+                           s->match_length - MIN_MATCH, bflush);
+
+            s->lookahead -= s->match_length;
+
+            /* Insert new strings in the hash table only if the match length
+             * is not too large. This saves time but degrades compression.
+             */
+#ifndef FASTEST
+            if (s->match_length <= s->max_insert_length &&
+                s->lookahead >= MIN_MATCH) {
+                s->match_length--; /* string at strstart already in hash table */
+                do {
+                    s->strstart++;
+                    INSERT_STRING(s, s->strstart, hash_head);
+                    /* strstart never exceeds WSIZE-MAX_MATCH, so there are
+                     * always MIN_MATCH bytes ahead.
+                     */
+                } while (--s->match_length != 0);
+                s->strstart++; 
+            } else
+#endif
+	    {
+                s->strstart += s->match_length;
+                s->match_length = 0;
+                s->ins_h = s->window[s->strstart];
+                UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
+#if MIN_MATCH != 3
+                Call UPDATE_HASH() MIN_MATCH-3 more times
+#endif
+                /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
+                 * matter since it will be recomputed at next deflate call.
+                 */
+            }
+        } else {
+            /* No match, output a literal byte */
+            Tracevv((stderr,"%c", s->window[s->strstart]));
+            _tr_tally_lit (s, s->window[s->strstart], bflush);
+            s->lookahead--;
+            s->strstart++; 
+        }
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    FLUSH_BLOCK(s, flush == Z_FINISH);
+    return flush == Z_FINISH ? finish_done : block_done;
+}
+
+/* ===========================================================================
+ * Same as above, but achieves better compression. We use a lazy
+ * evaluation for matches: a match is finally adopted only if there is
+ * no better match at the next window position.
+ */
+local block_state deflate_slow(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    IPos hash_head = NIL;    /* head of hash chain */
+    int bflush;              /* set if current block must be flushed */
+
+    /* Process the input block. */
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the next match, plus MIN_MATCH bytes to insert the
+         * string following the next match.
+         */
+        if (s->lookahead < MIN_LOOKAHEAD) {
+            fill_window(s);
+            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
+	        return need_more;
+	    }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* Insert the string window[strstart .. strstart+2] in the
+         * dictionary, and set hash_head to the head of the hash chain:
+         */
+        if (s->lookahead >= MIN_MATCH) {
+            INSERT_STRING(s, s->strstart, hash_head);
+        }
+
+        /* Find the longest match, discarding those <= prev_length.
+         */
+        s->prev_length = s->match_length, s->prev_match = s->match_start;
+        s->match_length = MIN_MATCH-1;
+
+        if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
+            s->strstart - hash_head <= MAX_DIST(s)) {
+            /* To simplify the code, we prevent matches with the string
+             * of window index 0 (in particular we have to avoid a match
+             * of the string with itself at the start of the input file).
+             */
+            if (s->strategy != Z_HUFFMAN_ONLY) {
+                s->match_length = longest_match (s, hash_head);
+            }
+            /* longest_match() sets match_start */
+
+            if (s->match_length <= 5 && (s->strategy == Z_FILTERED ||
+                 (s->match_length == MIN_MATCH &&
+                  s->strstart - s->match_start > TOO_FAR))) {
+
+                /* If prev_match is also MIN_MATCH, match_start is garbage
+                 * but we will ignore the current match anyway.
+                 */
+                s->match_length = MIN_MATCH-1;
+            }
+        }
+        /* If there was a match at the previous step and the current
+         * match is not better, output the previous match:
+         */
+        if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
+            uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
+            /* Do not insert strings in hash table beyond this. */
+
+            check_match(s, s->strstart-1, s->prev_match, s->prev_length);
+
+            _tr_tally_dist(s, s->strstart -1 - s->prev_match,
+			   s->prev_length - MIN_MATCH, bflush);
+
+            /* Insert in hash table all strings up to the end of the match.
+             * strstart-1 and strstart are already inserted. If there is not
+             * enough lookahead, the last two strings are not inserted in
+             * the hash table.
+             */
+            s->lookahead -= s->prev_length-1;
+            s->prev_length -= 2;
+            do {
+                if (++s->strstart <= max_insert) {
+                    INSERT_STRING(s, s->strstart, hash_head);
+                }
+            } while (--s->prev_length != 0);
+            s->match_available = 0;
+            s->match_length = MIN_MATCH-1;
+            s->strstart++;
+
+            if (bflush) FLUSH_BLOCK(s, 0);
+
+        } else if (s->match_available) {
+            /* If there was no match at the previous position, output a
+             * single literal. If there was a match but the current match
+             * is longer, truncate the previous match to a single literal.
+             */
+            Tracevv((stderr,"%c", s->window[s->strstart-1]));
+	    _tr_tally_lit(s, s->window[s->strstart-1], bflush);
+	    if (bflush) {
+                FLUSH_BLOCK_ONLY(s, 0);
+            }
+            s->strstart++;
+            s->lookahead--;
+            if (s->strm->avail_out == 0) return need_more;
+        } else {
+            /* There is no previous match to compare with, wait for
+             * the next step to decide.
+             */
+            s->match_available = 1;
+            s->strstart++;
+            s->lookahead--;
+        }
+    }
+    Assert (flush != Z_NO_FLUSH, "no flush?");
+    if (s->match_available) {
+        Tracevv((stderr,"%c", s->window[s->strstart-1]));
+        _tr_tally_lit(s, s->window[s->strstart-1], bflush);
+        s->match_available = 0;
+    }
+    FLUSH_BLOCK(s, flush == Z_FINISH);
+    return flush == Z_FINISH ? finish_done : block_done;
+}
diff --git a/zlib/deflate.h b/zlib/deflate.h
new file mode 100644
index 000000000..b6fc79b1a
--- /dev/null
+++ b/zlib/deflate.h
@@ -0,0 +1,318 @@
+/* deflate.h -- internal compression state
+ * Copyright (C) 1995-2002 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* @(#) $Id: deflate.h,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#ifndef _DEFLATE_H
+#define _DEFLATE_H
+
+#include "zutil.h"
+
+/* ===========================================================================
+ * Internal compression state.
+ */
+
+#define LENGTH_CODES 29
+/* number of length codes, not counting the special END_BLOCK code */
+
+#define LITERALS  256
+/* number of literal bytes 0..255 */
+
+#define L_CODES (LITERALS+1+LENGTH_CODES)
+/* number of Literal or Length codes, including the END_BLOCK code */
+
+#define D_CODES   30
+/* number of distance codes */
+
+#define BL_CODES  19
+/* number of codes used to transfer the bit lengths */
+
+#define HEAP_SIZE (2*L_CODES+1)
+/* maximum heap size */
+
+#define MAX_BITS 15
+/* All codes must not exceed MAX_BITS bits */
+
+#define INIT_STATE    42
+#define BUSY_STATE   113
+#define FINISH_STATE 666
+/* Stream status */
+
+
+/* Data structure describing a single value and its code string. */
+typedef struct ct_data_s {
+    union {
+        ush  freq;       /* frequency count */
+        ush  code;       /* bit string */
+    } fc;
+    union {
+        ush  dad;        /* father node in Huffman tree */
+        ush  len;        /* length of bit string */
+    } dl;
+} FAR ct_data;
+
+#define Freq fc.freq
+#define Code fc.code
+#define Dad  dl.dad
+#define Len  dl.len
+
+typedef struct static_tree_desc_s  static_tree_desc;
+
+typedef struct tree_desc_s {
+    ct_data *dyn_tree;           /* the dynamic tree */
+    int     max_code;            /* largest code with non zero frequency */
+    static_tree_desc *stat_desc; /* the corresponding static tree */
+} FAR tree_desc;
+
+typedef ush Pos;
+typedef Pos FAR Posf;
+typedef unsigned IPos;
+
+/* A Pos is an index in the character window. We use short instead of int to
+ * save space in the various tables. IPos is used only for parameter passing.
+ */
+
+typedef struct internal_state {
+    z_streamp strm;      /* pointer back to this zlib stream */
+    int   status;        /* as the name implies */
+    Bytef *pending_buf;  /* output still pending */
+    ulg   pending_buf_size; /* size of pending_buf */
+    Bytef *pending_out;  /* next pending byte to output to the stream */
+    int   pending;       /* nb of bytes in the pending buffer */
+    int   noheader;      /* suppress zlib header and adler32 */
+    Byte  data_type;     /* UNKNOWN, BINARY or ASCII */
+    Byte  method;        /* STORED (for zip only) or DEFLATED */
+    int   last_flush;    /* value of flush param for previous deflate call */
+
+                /* used by deflate.c: */
+
+    uInt  w_size;        /* LZ77 window size (32K by default) */
+    uInt  w_bits;        /* log2(w_size)  (8..16) */
+    uInt  w_mask;        /* w_size - 1 */
+
+    Bytef *window;
+    /* Sliding window. Input bytes are read into the second half of the window,
+     * and move to the first half later to keep a dictionary of at least wSize
+     * bytes. With this organization, matches are limited to a distance of
+     * wSize-MAX_MATCH bytes, but this ensures that IO is always
+     * performed with a length multiple of the block size. Also, it limits
+     * the window size to 64K, which is quite useful on MSDOS.
+     * To do: use the user input buffer as sliding window.
+     */
+
+    ulg window_size;
+    /* Actual size of window: 2*wSize, except when the user input buffer
+     * is directly used as sliding window.
+     */
+
+    Posf *prev;
+    /* Link to older string with same hash index. To limit the size of this
+     * array to 64K, this link is maintained only for the last 32K strings.
+     * An index in this array is thus a window index modulo 32K.
+     */
+
+    Posf *head; /* Heads of the hash chains or NIL. */
+
+    uInt  ins_h;          /* hash index of string to be inserted */
+    uInt  hash_size;      /* number of elements in hash table */
+    uInt  hash_bits;      /* log2(hash_size) */
+    uInt  hash_mask;      /* hash_size-1 */
+
+    uInt  hash_shift;
+    /* Number of bits by which ins_h must be shifted at each input
+     * step. It must be such that after MIN_MATCH steps, the oldest
+     * byte no longer takes part in the hash key, that is:
+     *   hash_shift * MIN_MATCH >= hash_bits
+     */
+
+    long block_start;
+    /* Window position at the beginning of the current output block. Gets
+     * negative when the window is moved backwards.
+     */
+
+    uInt match_length;           /* length of best match */
+    IPos prev_match;             /* previous match */
+    int match_available;         /* set if previous match exists */
+    uInt strstart;               /* start of string to insert */
+    uInt match_start;            /* start of matching string */
+    uInt lookahead;              /* number of valid bytes ahead in window */
+
+    uInt prev_length;
+    /* Length of the best match at previous step. Matches not greater than this
+     * are discarded. This is used in the lazy match evaluation.
+     */
+
+    uInt max_chain_length;
+    /* To speed up deflation, hash chains are never searched beyond this
+     * length.  A higher limit improves compression ratio but degrades the
+     * speed.
+     */
+
+    uInt max_lazy_match;
+    /* Attempt to find a better match only when the current match is strictly
+     * smaller than this value. This mechanism is used only for compression
+     * levels >= 4.
+     */
+#   define max_insert_length  max_lazy_match
+    /* Insert new strings in the hash table only if the match length is not
+     * greater than this length. This saves time but degrades compression.
+     * max_insert_length is used only for compression levels <= 3.
+     */
+
+    int level;    /* compression level (1..9) */
+    int strategy; /* favor or force Huffman coding*/
+
+    uInt good_match;
+    /* Use a faster search when the previous match is longer than this */
+
+    int nice_match; /* Stop searching when current match exceeds this */
+
+                /* used by trees.c: */
+    /* Didn't use ct_data typedef below to supress compiler warning */
+    struct ct_data_s dyn_ltree[HEAP_SIZE];   /* literal and length tree */
+    struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
+    struct ct_data_s bl_tree[2*BL_CODES+1];  /* Huffman tree for bit lengths */
+
+    struct tree_desc_s l_desc;               /* desc. for literal tree */
+    struct tree_desc_s d_desc;               /* desc. for distance tree */
+    struct tree_desc_s bl_desc;              /* desc. for bit length tree */
+
+    ush bl_count[MAX_BITS+1];
+    /* number of codes at each bit length for an optimal tree */
+
+    int heap[2*L_CODES+1];      /* heap used to build the Huffman trees */
+    int heap_len;               /* number of elements in the heap */
+    int heap_max;               /* element of largest frequency */
+    /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
+     * The same heap array is used to build all trees.
+     */
+
+    uch depth[2*L_CODES+1];
+    /* Depth of each subtree used as tie breaker for trees of equal frequency
+     */
+
+    uchf *l_buf;          /* buffer for literals or lengths */
+
+    uInt  lit_bufsize;
+    /* Size of match buffer for literals/lengths.  There are 4 reasons for
+     * limiting lit_bufsize to 64K:
+     *   - frequencies can be kept in 16 bit counters
+     *   - if compression is not successful for the first block, all input
+     *     data is still in the window so we can still emit a stored block even
+     *     when input comes from standard input.  (This can also be done for
+     *     all blocks if lit_bufsize is not greater than 32K.)
+     *   - if compression is not successful for a file smaller than 64K, we can
+     *     even emit a stored file instead of a stored block (saving 5 bytes).
+     *     This is applicable only for zip (not gzip or zlib).
+     *   - creating new Huffman trees less frequently may not provide fast
+     *     adaptation to changes in the input data statistics. (Take for
+     *     example a binary file with poorly compressible code followed by
+     *     a highly compressible string table.) Smaller buffer sizes give
+     *     fast adaptation but have of course the overhead of transmitting
+     *     trees more frequently.
+     *   - I can't count above 4
+     */
+
+    uInt last_lit;      /* running index in l_buf */
+
+    ushf *d_buf;
+    /* Buffer for distances. To simplify the code, d_buf and l_buf have
+     * the same number of elements. To use different lengths, an extra flag
+     * array would be necessary.
+     */
+
+    ulg opt_len;        /* bit length of current block with optimal trees */
+    ulg static_len;     /* bit length of current block with static trees */
+    uInt matches;       /* number of string matches in current block */
+    int last_eob_len;   /* bit length of EOB code for last block */
+
+#ifdef DEBUG
+    ulg compressed_len; /* total bit length of compressed file mod 2^32 */
+    ulg bits_sent;      /* bit length of compressed data sent mod 2^32 */
+#endif
+
+    ush bi_buf;
+    /* Output buffer. bits are inserted starting at the bottom (least
+     * significant bits).
+     */
+    int bi_valid;
+    /* Number of valid bits in bi_buf.  All bits above the last valid bit
+     * are always zero.
+     */
+
+} FAR deflate_state;
+
+/* Output a byte on the stream.
+ * IN assertion: there is enough room in pending_buf.
+ */
+#define put_byte(s, c) {s->pending_buf[s->pending++] = (c);}
+
+
+#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
+/* Minimum amount of lookahead, except at the end of the input file.
+ * See deflate.c for comments about the MIN_MATCH+1.
+ */
+
+#define MAX_DIST(s)  ((s)->w_size-MIN_LOOKAHEAD)
+/* In order to simplify the code, particularly on 16 bit machines, match
+ * distances are limited to MAX_DIST instead of WSIZE.
+ */
+
+        /* in trees.c */
+void _tr_init         OF((deflate_state *s));
+int  _tr_tally        OF((deflate_state *s, unsigned dist, unsigned lc));
+void _tr_flush_block  OF((deflate_state *s, charf *buf, ulg stored_len,
+			  int eof));
+void _tr_align        OF((deflate_state *s));
+void _tr_stored_block OF((deflate_state *s, charf *buf, ulg stored_len,
+                          int eof));
+
+#define d_code(dist) \
+   ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
+/* Mapping from a distance to a distance code. dist is the distance - 1 and
+ * must not have side effects. _dist_code[256] and _dist_code[257] are never
+ * used.
+ */
+
+#ifndef DEBUG
+/* Inline versions of _tr_tally for speed: */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+  extern uch _length_code[];
+  extern uch _dist_code[];
+#else
+  extern const uch _length_code[];
+  extern const uch _dist_code[];
+#endif
+
+# define _tr_tally_lit(s, c, flush) \
+  { uch cc = (c); \
+    s->d_buf[s->last_lit] = 0; \
+    s->l_buf[s->last_lit++] = cc; \
+    s->dyn_ltree[cc].Freq++; \
+    flush = (s->last_lit == s->lit_bufsize-1); \
+   }
+# define _tr_tally_dist(s, distance, length, flush) \
+  { uch len = (length); \
+    ush dist = (distance); \
+    s->d_buf[s->last_lit] = dist; \
+    s->l_buf[s->last_lit++] = len; \
+    dist--; \
+    s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
+    s->dyn_dtree[d_code(dist)].Freq++; \
+    flush = (s->last_lit == s->lit_bufsize-1); \
+  }
+#else
+# define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
+# define _tr_tally_dist(s, distance, length, flush) \
+              flush = _tr_tally(s, distance, length) 
+#endif
+
+#endif
diff --git a/zlib/descrip.mms b/zlib/descrip.mms
new file mode 100644
index 000000000..9d364598a
--- /dev/null
+++ b/zlib/descrip.mms
@@ -0,0 +1,48 @@
+# descrip.mms: MMS description file for building zlib on VMS
+# written by Martin P.J. Zinser <m.zinser@gsi.de>
+
+cc_defs = 
+c_deb = 
+
+.ifdef __DECC__
+pref = /prefix=all
+.endif
+
+OBJS = adler32.obj, compress.obj, crc32.obj, gzio.obj, uncompr.obj,\
+       deflate.obj, trees.obj, zutil.obj, inflate.obj, infblock.obj,\
+       inftrees.obj, infcodes.obj, infutil.obj, inffast.obj
+
+CFLAGS= $(C_DEB) $(CC_DEFS) $(PREF)
+
+all : example.exe minigzip.exe
+        @ write sys$output " Example applications available"
+libz.olb : libz.olb($(OBJS))
+	@ write sys$output " libz available"
+
+example.exe : example.obj libz.olb
+              link example,libz.olb/lib
+
+minigzip.exe : minigzip.obj libz.olb
+              link minigzip,libz.olb/lib,x11vms:xvmsutils.olb/lib
+
+clean : 
+	delete *.obj;*,libz.olb;*
+
+
+# Other dependencies.
+adler32.obj : zutil.h zlib.h zconf.h
+compress.obj : zlib.h zconf.h
+crc32.obj : zutil.h zlib.h zconf.h
+deflate.obj : deflate.h zutil.h zlib.h zconf.h
+example.obj : zlib.h zconf.h
+gzio.obj : zutil.h zlib.h zconf.h
+infblock.obj : zutil.h zlib.h zconf.h infblock.h inftrees.h infcodes.h infutil.h
+infcodes.obj : zutil.h zlib.h zconf.h inftrees.h infutil.h infcodes.h inffast.h
+inffast.obj : zutil.h zlib.h zconf.h inftrees.h infutil.h inffast.h
+inflate.obj : zutil.h zlib.h zconf.h infblock.h
+inftrees.obj : zutil.h zlib.h zconf.h inftrees.h
+infutil.obj : zutil.h zlib.h zconf.h inftrees.h infutil.h
+minigzip.obj : zlib.h zconf.h
+trees.obj : deflate.h zutil.h zlib.h zconf.h
+uncompr.obj : zlib.h zconf.h
+zutil.obj : zutil.h zlib.h zconf.h
diff --git a/zlib/gzio.c b/zlib/gzio.c
new file mode 100644
index 000000000..1160bc127
--- /dev/null
+++ b/zlib/gzio.c
@@ -0,0 +1,875 @@
+/* gzio.c -- IO on .gz files
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ *
+ * Compile this file with -DNO_DEFLATE to avoid the compression code.
+ */
+
+/* @(#) $Id: gzio.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#include <stdio.h>
+
+#include "zutil.h"
+
+struct internal_state {int dummy;}; /* for buggy compilers */
+
+#ifndef Z_BUFSIZE
+#  ifdef MAXSEG_64K
+#    define Z_BUFSIZE 4096 /* minimize memory usage for 16-bit DOS */
+#  else
+#    define Z_BUFSIZE 16384
+#  endif
+#endif
+#ifndef Z_PRINTF_BUFSIZE
+#  define Z_PRINTF_BUFSIZE 4096
+#endif
+
+#define ALLOC(size) malloc(size)
+#define TRYFREE(p) {if (p) free(p);}
+
+static int gz_magic[2] = {0x1f, 0x8b}; /* gzip magic header */
+
+/* gzip flag byte */
+#define ASCII_FLAG   0x01 /* bit 0 set: file probably ascii text */
+#define HEAD_CRC     0x02 /* bit 1 set: header CRC present */
+#define EXTRA_FIELD  0x04 /* bit 2 set: extra field present */
+#define ORIG_NAME    0x08 /* bit 3 set: original file name present */
+#define COMMENT      0x10 /* bit 4 set: file comment present */
+#define RESERVED     0xE0 /* bits 5..7: reserved */
+
+typedef struct gz_stream {
+    z_stream stream;
+    int      z_err;   /* error code for last stream operation */
+    int      z_eof;   /* set if end of input file */
+    FILE     *file;   /* .gz file */
+    Byte     *inbuf;  /* input buffer */
+    Byte     *outbuf; /* output buffer */
+    uLong    crc;     /* crc32 of uncompressed data */
+    char     *msg;    /* error message */
+    char     *path;   /* path name for debugging only */
+    int      transparent; /* 1 if input file is not a .gz file */
+    char     mode;    /* 'w' or 'r' */
+    long     startpos; /* start of compressed data in file (header skipped) */
+} gz_stream;
+
+
+local gzFile gz_open      OF((const char *path, const char *mode, int  fd));
+local int do_flush        OF((gzFile file, int flush));
+local int    get_byte     OF((gz_stream *s));
+local void   check_header OF((gz_stream *s));
+local int    destroy      OF((gz_stream *s));
+local void   putLong      OF((FILE *file, uLong x));
+local uLong  getLong      OF((gz_stream *s));
+
+/* ===========================================================================
+     Opens a gzip (.gz) file for reading or writing. The mode parameter
+   is as in fopen ("rb" or "wb"). The file is given either by file descriptor
+   or path name (if fd == -1).
+     gz_open return NULL if the file could not be opened or if there was
+   insufficient memory to allocate the (de)compression state; errno
+   can be checked to distinguish the two cases (if errno is zero, the
+   zlib error is Z_MEM_ERROR).
+*/
+local gzFile gz_open (path, mode, fd)
+    const char *path;
+    const char *mode;
+    int  fd;
+{
+    int err;
+    int level = Z_DEFAULT_COMPRESSION; /* compression level */
+    int strategy = Z_DEFAULT_STRATEGY; /* compression strategy */
+    char *p = (char*)mode;
+    gz_stream *s;
+    char fmode[80]; /* copy of mode, without the compression level */
+    char *m = fmode;
+
+    if (!path || !mode) return Z_NULL;
+
+    s = (gz_stream *)ALLOC(sizeof(gz_stream));
+    if (!s) return Z_NULL;
+
+    s->stream.zalloc = (alloc_func)0;
+    s->stream.zfree = (free_func)0;
+    s->stream.opaque = (voidpf)0;
+    s->stream.next_in = s->inbuf = Z_NULL;
+    s->stream.next_out = s->outbuf = Z_NULL;
+    s->stream.avail_in = s->stream.avail_out = 0;
+    s->file = NULL;
+    s->z_err = Z_OK;
+    s->z_eof = 0;
+    s->crc = crc32(0L, Z_NULL, 0);
+    s->msg = NULL;
+    s->transparent = 0;
+
+    s->path = (char*)ALLOC(strlen(path)+1);
+    if (s->path == NULL) {
+        return destroy(s), (gzFile)Z_NULL;
+    }
+    strcpy(s->path, path); /* do this early for debugging */
+
+    s->mode = '\0';
+    do {
+        if (*p == 'r') s->mode = 'r';
+        if (*p == 'w' || *p == 'a') s->mode = 'w';
+        if (*p >= '0' && *p <= '9') {
+	    level = *p - '0';
+	} else if (*p == 'f') {
+	  strategy = Z_FILTERED;
+	} else if (*p == 'h') {
+	  strategy = Z_HUFFMAN_ONLY;
+	} else {
+	    *m++ = *p; /* copy the mode */
+	}
+    } while (*p++ && m != fmode + sizeof(fmode));
+    if (s->mode == '\0') return destroy(s), (gzFile)Z_NULL;
+    
+    if (s->mode == 'w') {
+#ifdef NO_DEFLATE
+        err = Z_STREAM_ERROR;
+#else
+        err = deflateInit2(&(s->stream), level,
+                           Z_DEFLATED, -MAX_WBITS, DEF_MEM_LEVEL, strategy);
+        /* windowBits is passed < 0 to suppress zlib header */
+
+        s->stream.next_out = s->outbuf = (Byte*)ALLOC(Z_BUFSIZE);
+#endif
+        if (err != Z_OK || s->outbuf == Z_NULL) {
+            return destroy(s), (gzFile)Z_NULL;
+        }
+    } else {
+        s->stream.next_in  = s->inbuf = (Byte*)ALLOC(Z_BUFSIZE);
+
+        err = inflateInit2(&(s->stream), -MAX_WBITS);
+        /* windowBits is passed < 0 to tell that there is no zlib header.
+         * Note that in this case inflate *requires* an extra "dummy" byte
+         * after the compressed stream in order to complete decompression and
+         * return Z_STREAM_END. Here the gzip CRC32 ensures that 4 bytes are
+         * present after the compressed stream.
+         */
+        if (err != Z_OK || s->inbuf == Z_NULL) {
+            return destroy(s), (gzFile)Z_NULL;
+        }
+    }
+    s->stream.avail_out = Z_BUFSIZE;
+
+    errno = 0;
+    s->file = fd < 0 ? F_OPEN(path, fmode) : (FILE*)fdopen(fd, fmode);
+
+    if (s->file == NULL) {
+        return destroy(s), (gzFile)Z_NULL;
+    }
+    if (s->mode == 'w') {
+        /* Write a very simple .gz header:
+         */
+        fprintf(s->file, "%c%c%c%c%c%c%c%c%c%c", gz_magic[0], gz_magic[1],
+             Z_DEFLATED, 0 /*flags*/, 0,0,0,0 /*time*/, 0 /*xflags*/, OS_CODE);
+	s->startpos = 10L;
+	/* We use 10L instead of ftell(s->file) to because ftell causes an
+         * fflush on some systems. This version of the library doesn't use
+         * startpos anyway in write mode, so this initialization is not
+         * necessary.
+         */
+    } else {
+	check_header(s); /* skip the .gz header */
+	s->startpos = (ftell(s->file) - s->stream.avail_in);
+    }
+    
+    return (gzFile)s;
+}
+
+/* ===========================================================================
+     Opens a gzip (.gz) file for reading or writing.
+*/
+gzFile ZEXPORT gzopen (path, mode)
+    const char *path;
+    const char *mode;
+{
+    return gz_open (path, mode, -1);
+}
+
+/* ===========================================================================
+     Associate a gzFile with the file descriptor fd. fd is not dup'ed here
+   to mimic the behavio(u)r of fdopen.
+*/
+gzFile ZEXPORT gzdopen (fd, mode)
+    int fd;
+    const char *mode;
+{
+    char name[20];
+
+    if (fd < 0) return (gzFile)Z_NULL;
+    sprintf(name, "<fd:%d>", fd); /* for debugging */
+
+    return gz_open (name, mode, fd);
+}
+
+/* ===========================================================================
+ * Update the compression level and strategy
+ */
+int ZEXPORT gzsetparams (file, level, strategy)
+    gzFile file;
+    int level;
+    int strategy;
+{
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
+
+    /* Make room to allow flushing */
+    if (s->stream.avail_out == 0) {
+
+	s->stream.next_out = s->outbuf;
+	if (fwrite(s->outbuf, 1, Z_BUFSIZE, s->file) != Z_BUFSIZE) {
+	    s->z_err = Z_ERRNO;
+	}
+	s->stream.avail_out = Z_BUFSIZE;
+    }
+
+    return deflateParams (&(s->stream), level, strategy);
+}
+
+/* ===========================================================================
+     Read a byte from a gz_stream; update next_in and avail_in. Return EOF
+   for end of file.
+   IN assertion: the stream s has been sucessfully opened for reading.
+*/
+local int get_byte(s)
+    gz_stream *s;
+{
+    if (s->z_eof) return EOF;
+    if (s->stream.avail_in == 0) {
+	errno = 0;
+	s->stream.avail_in = fread(s->inbuf, 1, Z_BUFSIZE, s->file);
+	if (s->stream.avail_in == 0) {
+	    s->z_eof = 1;
+	    if (ferror(s->file)) s->z_err = Z_ERRNO;
+	    return EOF;
+	}
+	s->stream.next_in = s->inbuf;
+    }
+    s->stream.avail_in--;
+    return *(s->stream.next_in)++;
+}
+
+/* ===========================================================================
+      Check the gzip header of a gz_stream opened for reading. Set the stream
+    mode to transparent if the gzip magic header is not present; set s->err
+    to Z_DATA_ERROR if the magic header is present but the rest of the header
+    is incorrect.
+    IN assertion: the stream s has already been created sucessfully;
+       s->stream.avail_in is zero for the first time, but may be non-zero
+       for concatenated .gz files.
+*/
+local void check_header(s)
+    gz_stream *s;
+{
+    int method; /* method byte */
+    int flags;  /* flags byte */
+    uInt len;
+    int c;
+
+    /* Check the gzip magic header */
+    for (len = 0; len < 2; len++) {
+	c = get_byte(s);
+	if (c != gz_magic[len]) {
+	    if (len != 0) s->stream.avail_in++, s->stream.next_in--;
+	    if (c != EOF) {
+		s->stream.avail_in++, s->stream.next_in--;
+		s->transparent = 1;
+	    }
+	    s->z_err = s->stream.avail_in != 0 ? Z_OK : Z_STREAM_END;
+	    return;
+	}
+    }
+    method = get_byte(s);
+    flags = get_byte(s);
+    if (method != Z_DEFLATED || (flags & RESERVED) != 0) {
+	s->z_err = Z_DATA_ERROR;
+	return;
+    }
+
+    /* Discard time, xflags and OS code: */
+    for (len = 0; len < 6; len++) (void)get_byte(s);
+
+    if ((flags & EXTRA_FIELD) != 0) { /* skip the extra field */
+	len  =  (uInt)get_byte(s);
+	len += ((uInt)get_byte(s))<<8;
+	/* len is garbage if EOF but the loop below will quit anyway */
+	while (len-- != 0 && get_byte(s) != EOF) ;
+    }
+    if ((flags & ORIG_NAME) != 0) { /* skip the original file name */
+	while ((c = get_byte(s)) != 0 && c != EOF) ;
+    }
+    if ((flags & COMMENT) != 0) {   /* skip the .gz file comment */
+	while ((c = get_byte(s)) != 0 && c != EOF) ;
+    }
+    if ((flags & HEAD_CRC) != 0) {  /* skip the header crc */
+	for (len = 0; len < 2; len++) (void)get_byte(s);
+    }
+    s->z_err = s->z_eof ? Z_DATA_ERROR : Z_OK;
+}
+
+ /* ===========================================================================
+ * Cleanup then free the given gz_stream. Return a zlib error code.
+   Try freeing in the reverse order of allocations.
+ */
+local int destroy (s)
+    gz_stream *s;
+{
+    int err = Z_OK;
+
+    if (!s) return Z_STREAM_ERROR;
+
+    TRYFREE(s->msg);
+
+    if (s->stream.state != NULL) {
+	if (s->mode == 'w') {
+#ifdef NO_DEFLATE
+	    err = Z_STREAM_ERROR;
+#else
+	    err = deflateEnd(&(s->stream));
+#endif
+	} else if (s->mode == 'r') {
+	    err = inflateEnd(&(s->stream));
+	}
+    }
+    if (s->file != NULL && fclose(s->file)) {
+#ifdef ESPIPE
+	if (errno != ESPIPE) /* fclose is broken for pipes in HP/UX */
+#endif
+	    err = Z_ERRNO;
+    }
+    if (s->z_err < 0) err = s->z_err;
+
+    TRYFREE(s->inbuf);
+    TRYFREE(s->outbuf);
+    TRYFREE(s->path);
+    TRYFREE(s);
+    return err;
+}
+
+/* ===========================================================================
+     Reads the given number of uncompressed bytes from the compressed file.
+   gzread returns the number of bytes actually read (0 for end of file).
+*/
+int ZEXPORT gzread (file, buf, len)
+    gzFile file;
+    voidp buf;
+    unsigned len;
+{
+    gz_stream *s = (gz_stream*)file;
+    Bytef *start = (Bytef*)buf; /* starting point for crc computation */
+    Byte  *next_out; /* == stream.next_out but not forced far (for MSDOS) */
+
+    if (s == NULL || s->mode != 'r') return Z_STREAM_ERROR;
+
+    if (s->z_err == Z_DATA_ERROR || s->z_err == Z_ERRNO) return -1;
+    if (s->z_err == Z_STREAM_END) return 0;  /* EOF */
+
+    next_out = (Byte*)buf;
+    s->stream.next_out = (Bytef*)buf;
+    s->stream.avail_out = len;
+
+    while (s->stream.avail_out != 0) {
+
+	if (s->transparent) {
+	    /* Copy first the lookahead bytes: */
+	    uInt n = s->stream.avail_in;
+	    if (n > s->stream.avail_out) n = s->stream.avail_out;
+	    if (n > 0) {
+		zmemcpy(s->stream.next_out, s->stream.next_in, n);
+		next_out += n;
+		s->stream.next_out = next_out;
+		s->stream.next_in   += n;
+		s->stream.avail_out -= n;
+		s->stream.avail_in  -= n;
+	    }
+	    if (s->stream.avail_out > 0) {
+		s->stream.avail_out -= fread(next_out, 1, s->stream.avail_out,
+					     s->file);
+	    }
+	    len -= s->stream.avail_out;
+	    s->stream.total_in  += (uLong)len;
+	    s->stream.total_out += (uLong)len;
+            if (len == 0) s->z_eof = 1;
+	    return (int)len;
+	}
+        if (s->stream.avail_in == 0 && !s->z_eof) {
+
+            errno = 0;
+            s->stream.avail_in = fread(s->inbuf, 1, Z_BUFSIZE, s->file);
+            if (s->stream.avail_in == 0) {
+                s->z_eof = 1;
+		if (ferror(s->file)) {
+		    s->z_err = Z_ERRNO;
+		    break;
+		}
+            }
+            s->stream.next_in = s->inbuf;
+        }
+        s->z_err = inflate(&(s->stream), Z_NO_FLUSH);
+
+	if (s->z_err == Z_STREAM_END) {
+	    /* Check CRC and original size */
+	    s->crc = crc32(s->crc, start, (uInt)(s->stream.next_out - start));
+	    start = s->stream.next_out;
+
+	    if (getLong(s) != s->crc) {
+		s->z_err = Z_DATA_ERROR;
+	    } else {
+	        (void)getLong(s);
+                /* The uncompressed length returned by above getlong() may
+                 * be different from s->stream.total_out) in case of
+		 * concatenated .gz files. Check for such files:
+		 */
+		check_header(s);
+		if (s->z_err == Z_OK) {
+		    uLong total_in = s->stream.total_in;
+		    uLong total_out = s->stream.total_out;
+
+		    inflateReset(&(s->stream));
+		    s->stream.total_in = total_in;
+		    s->stream.total_out = total_out;
+		    s->crc = crc32(0L, Z_NULL, 0);
+		}
+	    }
+	}
+	if (s->z_err != Z_OK || s->z_eof) break;
+    }
+    s->crc = crc32(s->crc, start, (uInt)(s->stream.next_out - start));
+
+    return (int)(len - s->stream.avail_out);
+}
+
+
+/* ===========================================================================
+      Reads one byte from the compressed file. gzgetc returns this byte
+   or -1 in case of end of file or error.
+*/
+int ZEXPORT gzgetc(file)
+    gzFile file;
+{
+    unsigned char c;
+
+    return gzread(file, &c, 1) == 1 ? c : -1;
+}
+
+
+/* ===========================================================================
+      Reads bytes from the compressed file until len-1 characters are
+   read, or a newline character is read and transferred to buf, or an
+   end-of-file condition is encountered.  The string is then terminated
+   with a null character.
+      gzgets returns buf, or Z_NULL in case of error.
+
+      The current implementation is not optimized at all.
+*/
+char * ZEXPORT gzgets(file, buf, len)
+    gzFile file;
+    char *buf;
+    int len;
+{
+    char *b = buf;
+    if (buf == Z_NULL || len <= 0) return Z_NULL;
+
+    while (--len > 0 && gzread(file, buf, 1) == 1 && *buf++ != '\n') ;
+    *buf = '\0';
+    return b == buf && len > 0 ? Z_NULL : b;
+}
+
+
+#ifndef NO_DEFLATE
+/* ===========================================================================
+     Writes the given number of uncompressed bytes into the compressed file.
+   gzwrite returns the number of bytes actually written (0 in case of error).
+*/
+int ZEXPORT gzwrite (file, buf, len)
+    gzFile file;
+    const voidp buf;
+    unsigned len;
+{
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
+
+    s->stream.next_in = (Bytef*)buf;
+    s->stream.avail_in = len;
+
+    while (s->stream.avail_in != 0) {
+
+        if (s->stream.avail_out == 0) {
+
+            s->stream.next_out = s->outbuf;
+            if (fwrite(s->outbuf, 1, Z_BUFSIZE, s->file) != Z_BUFSIZE) {
+                s->z_err = Z_ERRNO;
+                break;
+            }
+            s->stream.avail_out = Z_BUFSIZE;
+        }
+        s->z_err = deflate(&(s->stream), Z_NO_FLUSH);
+        if (s->z_err != Z_OK) break;
+    }
+    s->crc = crc32(s->crc, (const Bytef *)buf, len);
+
+    return (int)(len - s->stream.avail_in);
+}
+
+/* ===========================================================================
+     Converts, formats, and writes the args to the compressed file under
+   control of the format string, as in fprintf. gzprintf returns the number of
+   uncompressed bytes actually written (0 in case of error).
+*/
+#ifdef STDC
+#include <stdarg.h>
+
+int ZEXPORTVA gzprintf (gzFile file, const char *format, /* args */ ...)
+{
+    char buf[Z_PRINTF_BUFSIZE];
+    va_list va;
+    int len;
+
+    va_start(va, format);
+#ifdef HAS_vsnprintf
+    (void)vsnprintf(buf, sizeof(buf), format, va);
+#else
+    (void)vsprintf(buf, format, va);
+#endif
+    va_end(va);
+    len = strlen(buf); /* some *sprintf don't return the nb of bytes written */
+    if (len <= 0) return 0;
+
+    return gzwrite(file, buf, (unsigned)len);
+}
+#else /* not ANSI C */
+
+int ZEXPORTVA gzprintf (file, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10,
+	               a11, a12, a13, a14, a15, a16, a17, a18, a19, a20)
+    gzFile file;
+    const char *format;
+    int a1, a2, a3, a4, a5, a6, a7, a8, a9, a10,
+	a11, a12, a13, a14, a15, a16, a17, a18, a19, a20;
+{
+    char buf[Z_PRINTF_BUFSIZE];
+    int len;
+
+#ifdef HAS_snprintf
+    snprintf(buf, sizeof(buf), format, a1, a2, a3, a4, a5, a6, a7, a8,
+	     a9, a10, a11, a12, a13, a14, a15, a16, a17, a18, a19, a20);
+#else
+    sprintf(buf, format, a1, a2, a3, a4, a5, a6, a7, a8,
+	    a9, a10, a11, a12, a13, a14, a15, a16, a17, a18, a19, a20);
+#endif
+    len = strlen(buf); /* old sprintf doesn't return the nb of bytes written */
+    if (len <= 0) return 0;
+
+    return gzwrite(file, buf, len);
+}
+#endif
+
+/* ===========================================================================
+      Writes c, converted to an unsigned char, into the compressed file.
+   gzputc returns the value that was written, or -1 in case of error.
+*/
+int ZEXPORT gzputc(file, c)
+    gzFile file;
+    int c;
+{
+    unsigned char cc = (unsigned char) c; /* required for big endian systems */
+
+    return gzwrite(file, &cc, 1) == 1 ? (int)cc : -1;
+}
+
+
+/* ===========================================================================
+      Writes the given null-terminated string to the compressed file, excluding
+   the terminating null character.
+      gzputs returns the number of characters written, or -1 in case of error.
+*/
+int ZEXPORT gzputs(file, s)
+    gzFile file;
+    const char *s;
+{
+    return gzwrite(file, (char*)s, (unsigned)strlen(s));
+}
+
+
+/* ===========================================================================
+     Flushes all pending output into the compressed file. The parameter
+   flush is as in the deflate() function.
+*/
+local int do_flush (file, flush)
+    gzFile file;
+    int flush;
+{
+    uInt len;
+    int done = 0;
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
+
+    s->stream.avail_in = 0; /* should be zero already anyway */
+
+    for (;;) {
+        len = Z_BUFSIZE - s->stream.avail_out;
+
+        if (len != 0) {
+            if ((uInt)fwrite(s->outbuf, 1, len, s->file) != len) {
+                s->z_err = Z_ERRNO;
+                return Z_ERRNO;
+            }
+            s->stream.next_out = s->outbuf;
+            s->stream.avail_out = Z_BUFSIZE;
+        }
+        if (done) break;
+        s->z_err = deflate(&(s->stream), flush);
+
+	/* Ignore the second of two consecutive flushes: */
+	if (len == 0 && s->z_err == Z_BUF_ERROR) s->z_err = Z_OK;
+
+        /* deflate has finished flushing only when it hasn't used up
+         * all the available space in the output buffer: 
+         */
+        done = (s->stream.avail_out != 0 || s->z_err == Z_STREAM_END);
+ 
+        if (s->z_err != Z_OK && s->z_err != Z_STREAM_END) break;
+    }
+    return  s->z_err == Z_STREAM_END ? Z_OK : s->z_err;
+}
+
+int ZEXPORT gzflush (file, flush)
+     gzFile file;
+     int flush;
+{
+    gz_stream *s = (gz_stream*)file;
+    int err = do_flush (file, flush);
+
+    if (err) return err;
+    fflush(s->file);
+    return  s->z_err == Z_STREAM_END ? Z_OK : s->z_err;
+}
+#endif /* NO_DEFLATE */
+
+/* ===========================================================================
+      Sets the starting position for the next gzread or gzwrite on the given
+   compressed file. The offset represents a number of bytes in the
+      gzseek returns the resulting offset location as measured in bytes from
+   the beginning of the uncompressed stream, or -1 in case of error.
+      SEEK_END is not implemented, returns error.
+      In this version of the library, gzseek can be extremely slow.
+*/
+z_off_t ZEXPORT gzseek (file, offset, whence)
+    gzFile file;
+    z_off_t offset;
+    int whence;
+{
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL || whence == SEEK_END ||
+	s->z_err == Z_ERRNO || s->z_err == Z_DATA_ERROR) {
+	return -1L;
+    }
+    
+    if (s->mode == 'w') {
+#ifdef NO_DEFLATE
+	return -1L;
+#else
+	if (whence == SEEK_SET) {
+	    offset -= s->stream.total_in;
+	}
+	if (offset < 0) return -1L;
+
+	/* At this point, offset is the number of zero bytes to write. */
+	if (s->inbuf == Z_NULL) {
+	    s->inbuf = (Byte*)ALLOC(Z_BUFSIZE); /* for seeking */
+	    zmemzero(s->inbuf, Z_BUFSIZE);
+	}
+	while (offset > 0)  {
+	    uInt size = Z_BUFSIZE;
+	    if (offset < Z_BUFSIZE) size = (uInt)offset;
+
+	    size = gzwrite(file, s->inbuf, size);
+	    if (size == 0) return -1L;
+
+	    offset -= size;
+	}
+	return (z_off_t)s->stream.total_in;
+#endif
+    }
+    /* Rest of function is for reading only */
+
+    /* compute absolute position */
+    if (whence == SEEK_CUR) {
+	offset += s->stream.total_out;
+    }
+    if (offset < 0) return -1L;
+
+    if (s->transparent) {
+	/* map to fseek */
+	s->stream.avail_in = 0;
+	s->stream.next_in = s->inbuf;
+        if (fseek(s->file, offset, SEEK_SET) < 0) return -1L;
+
+	s->stream.total_in = s->stream.total_out = (uLong)offset;
+	return offset;
+    }
+
+    /* For a negative seek, rewind and use positive seek */
+    if ((uLong)offset >= s->stream.total_out) {
+	offset -= s->stream.total_out;
+    } else if (gzrewind(file) < 0) {
+	return -1L;
+    }
+    /* offset is now the number of bytes to skip. */
+
+    if (offset != 0 && s->outbuf == Z_NULL) {
+	s->outbuf = (Byte*)ALLOC(Z_BUFSIZE);
+    }
+    while (offset > 0)  {
+	int size = Z_BUFSIZE;
+	if (offset < Z_BUFSIZE) size = (int)offset;
+
+	size = gzread(file, s->outbuf, (uInt)size);
+	if (size <= 0) return -1L;
+	offset -= size;
+    }
+    return (z_off_t)s->stream.total_out;
+}
+
+/* ===========================================================================
+     Rewinds input file. 
+*/
+int ZEXPORT gzrewind (file)
+    gzFile file;
+{
+    gz_stream *s = (gz_stream*)file;
+    
+    if (s == NULL || s->mode != 'r') return -1;
+
+    s->z_err = Z_OK;
+    s->z_eof = 0;
+    s->stream.avail_in = 0;
+    s->stream.next_in = s->inbuf;
+    s->crc = crc32(0L, Z_NULL, 0);
+	
+    if (s->startpos == 0) { /* not a compressed file */
+	rewind(s->file);
+	return 0;
+    }
+
+    (void) inflateReset(&s->stream);
+    return fseek(s->file, s->startpos, SEEK_SET);
+}
+
+/* ===========================================================================
+     Returns the starting position for the next gzread or gzwrite on the
+   given compressed file. This position represents a number of bytes in the
+   uncompressed data stream.
+*/
+z_off_t ZEXPORT gztell (file)
+    gzFile file;
+{
+    return gzseek(file, 0L, SEEK_CUR);
+}
+
+/* ===========================================================================
+     Returns 1 when EOF has previously been detected reading the given
+   input stream, otherwise zero.
+*/
+int ZEXPORT gzeof (file)
+    gzFile file;
+{
+    gz_stream *s = (gz_stream*)file;
+    
+    return (s == NULL || s->mode != 'r') ? 0 : s->z_eof;
+}
+
+/* ===========================================================================
+   Outputs a long in LSB order to the given file
+*/
+local void putLong (file, x)
+    FILE *file;
+    uLong x;
+{
+    int n;
+    for (n = 0; n < 4; n++) {
+        fputc((int)(x & 0xff), file);
+        x >>= 8;
+    }
+}
+
+/* ===========================================================================
+   Reads a long in LSB order from the given gz_stream. Sets z_err in case
+   of error.
+*/
+local uLong getLong (s)
+    gz_stream *s;
+{
+    uLong x = (uLong)get_byte(s);
+    int c;
+
+    x += ((uLong)get_byte(s))<<8;
+    x += ((uLong)get_byte(s))<<16;
+    c = get_byte(s);
+    if (c == EOF) s->z_err = Z_DATA_ERROR;
+    x += ((uLong)c)<<24;
+    return x;
+}
+
+/* ===========================================================================
+     Flushes all pending output if necessary, closes the compressed file
+   and deallocates all the (de)compression state.
+*/
+int ZEXPORT gzclose (file)
+    gzFile file;
+{
+    int err;
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL) return Z_STREAM_ERROR;
+
+    if (s->mode == 'w') {
+#ifdef NO_DEFLATE
+	return Z_STREAM_ERROR;
+#else
+        err = do_flush (file, Z_FINISH);
+        if (err != Z_OK) return destroy((gz_stream*)file);
+
+        putLong (s->file, s->crc);
+        putLong (s->file, s->stream.total_in);
+#endif
+    }
+    return destroy((gz_stream*)file);
+}
+
+/* ===========================================================================
+     Returns the error message for the last error which occured on the
+   given compressed file. errnum is set to zlib error number. If an
+   error occured in the file system and not in the compression library,
+   errnum is set to Z_ERRNO and the application may consult errno
+   to get the exact error code.
+*/
+const char*  ZEXPORT gzerror (file, errnum)
+    gzFile file;
+    int *errnum;
+{
+    char *m;
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL) {
+        *errnum = Z_STREAM_ERROR;
+        return (const char*)ERR_MSG(Z_STREAM_ERROR);
+    }
+    *errnum = s->z_err;
+    if (*errnum == Z_OK) return (const char*)"";
+
+    m =  (char*)(*errnum == Z_ERRNO ? zstrerror(errno) : s->stream.msg);
+
+    if (m == NULL || *m == '\0') m = (char*)ERR_MSG(s->z_err);
+
+    TRYFREE(s->msg);
+    s->msg = (char*)ALLOC(strlen(s->path) + strlen(m) + 3);
+    strcpy(s->msg, s->path);
+    strcat(s->msg, ": ");
+    strcat(s->msg, m);
+    return (const char*)s->msg;
+}
diff --git a/zlib/infblock.c b/zlib/infblock.c
new file mode 100644
index 000000000..dd7a6d40a
--- /dev/null
+++ b/zlib/infblock.c
@@ -0,0 +1,403 @@
+/* infblock.c -- interpret and process block types to last block
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+#include "zutil.h"
+#include "infblock.h"
+#include "inftrees.h"
+#include "infcodes.h"
+#include "infutil.h"
+
+struct inflate_codes_state {int dummy;}; /* for buggy compilers */
+
+/* simplify the use of the inflate_huft type with some defines */
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+/* Table for deflate from PKZIP's appnote.txt. */
+local const uInt border[] = { /* Order of the bit length code lengths */
+        16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+/*
+   Notes beyond the 1.93a appnote.txt:
+
+   1. Distance pointers never point before the beginning of the output
+      stream.
+   2. Distance pointers can point back across blocks, up to 32k away.
+   3. There is an implied maximum of 7 bits for the bit length table and
+      15 bits for the actual data.
+   4. If only one code exists, then it is encoded using one bit.  (Zero
+      would be more efficient, but perhaps a little confusing.)  If two
+      codes exist, they are coded using one bit each (0 and 1).
+   5. There is no way of sending zero distance codes--a dummy must be
+      sent if there are none.  (History: a pre 2.0 version of PKZIP would
+      store blocks with no distance codes, but this was discovered to be
+      too harsh a criterion.)  Valid only for 1.93a.  2.04c does allow
+      zero distance codes, which is sent as one code of zero bits in
+      length.
+   6. There are up to 286 literal/length codes.  Code 256 represents the
+      end-of-block.  Note however that the static length tree defines
+      288 codes just to fill out the Huffman codes.  Codes 286 and 287
+      cannot be used though, since there is no length base or extra bits
+      defined for them.  Similarily, there are up to 30 distance codes.
+      However, static trees define 32 codes (all 5 bits) to fill out the
+      Huffman codes, but the last two had better not show up in the data.
+   7. Unzip can check dynamic Huffman blocks for complete code sets.
+      The exception is that a single code would not be complete (see #4).
+   8. The five bits following the block type is really the number of
+      literal codes sent minus 257.
+   9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits
+      (1+6+6).  Therefore, to output three times the length, you output
+      three codes (1+1+1), whereas to output four times the same length,
+      you only need two codes (1+3).  Hmm.
+  10. In the tree reconstruction algorithm, Code = Code + Increment
+      only if BitLength(i) is not zero.  (Pretty obvious.)
+  11. Correction: 4 Bits: # of Bit Length codes - 4     (4 - 19)
+  12. Note: length code 284 can represent 227-258, but length code 285
+      really is 258.  The last length deserves its own, short code
+      since it gets used a lot in very redundant files.  The length
+      258 is special since 258 - 3 (the min match length) is 255.
+  13. The literal/length and distance code bit lengths are read as a
+      single stream of lengths.  It is possible (and advantageous) for
+      a repeat code (16, 17, or 18) to go across the boundary between
+      the two sets of lengths.
+ */
+
+
+void inflate_blocks_reset(s, z, c)
+inflate_blocks_statef *s;
+z_streamp z;
+uLongf *c;
+{
+  if (c != Z_NULL)
+    *c = s->check;
+  if (s->mode == BTREE || s->mode == DTREE)
+    ZFREE(z, s->sub.trees.blens);
+  if (s->mode == CODES)
+    inflate_codes_free(s->sub.decode.codes, z);
+  s->mode = TYPE;
+  s->bitk = 0;
+  s->bitb = 0;
+  s->read = s->write = s->window;
+  if (s->checkfn != Z_NULL)
+    z->adler = s->check = (*s->checkfn)(0L, (const Bytef *)Z_NULL, 0);
+  Tracev((stderr, "inflate:   blocks reset\n"));
+}
+
+
+inflate_blocks_statef *inflate_blocks_new(z, c, w)
+z_streamp z;
+check_func c;
+uInt w;
+{
+  inflate_blocks_statef *s;
+
+  if ((s = (inflate_blocks_statef *)ZALLOC
+       (z,1,sizeof(struct inflate_blocks_state))) == Z_NULL)
+    return s;
+  if ((s->hufts =
+       (inflate_huft *)ZALLOC(z, sizeof(inflate_huft), MANY)) == Z_NULL)
+  {
+    ZFREE(z, s);
+    return Z_NULL;
+  }
+  if ((s->window = (Bytef *)ZALLOC(z, 1, w)) == Z_NULL)
+  {
+    ZFREE(z, s->hufts);
+    ZFREE(z, s);
+    return Z_NULL;
+  }
+  s->end = s->window + w;
+  s->checkfn = c;
+  s->mode = TYPE;
+  Tracev((stderr, "inflate:   blocks allocated\n"));
+  inflate_blocks_reset(s, z, Z_NULL);
+  return s;
+}
+
+
+int inflate_blocks(s, z, r)
+inflate_blocks_statef *s;
+z_streamp z;
+int r;
+{
+  uInt t;               /* temporary storage */
+  uLong b;              /* bit buffer */
+  uInt k;               /* bits in bit buffer */
+  Bytef *p;             /* input data pointer */
+  uInt n;               /* bytes available there */
+  Bytef *q;             /* output window write pointer */
+  uInt m;               /* bytes to end of window or read pointer */
+
+  /* copy input/output information to locals (UPDATE macro restores) */
+  LOAD
+
+  /* process input based on current state */
+  while (1) switch (s->mode)
+  {
+    case TYPE:
+      NEEDBITS(3)
+      t = (uInt)b & 7;
+      s->last = t & 1;
+      switch (t >> 1)
+      {
+        case 0:                         /* stored */
+          Tracev((stderr, "inflate:     stored block%s\n",
+                 s->last ? " (last)" : ""));
+          DUMPBITS(3)
+          t = k & 7;                    /* go to byte boundary */
+          DUMPBITS(t)
+          s->mode = LENS;               /* get length of stored block */
+          break;
+        case 1:                         /* fixed */
+          Tracev((stderr, "inflate:     fixed codes block%s\n",
+                 s->last ? " (last)" : ""));
+          {
+            uInt bl, bd;
+            inflate_huft *tl, *td;
+
+            inflate_trees_fixed(&bl, &bd, &tl, &td, z);
+            s->sub.decode.codes = inflate_codes_new(bl, bd, tl, td, z);
+            if (s->sub.decode.codes == Z_NULL)
+            {
+              r = Z_MEM_ERROR;
+              LEAVE
+            }
+          }
+          DUMPBITS(3)
+          s->mode = CODES;
+          break;
+        case 2:                         /* dynamic */
+          Tracev((stderr, "inflate:     dynamic codes block%s\n",
+                 s->last ? " (last)" : ""));
+          DUMPBITS(3)
+          s->mode = TABLE;
+          break;
+        case 3:                         /* illegal */
+          DUMPBITS(3)
+          s->mode = BAD;
+          z->msg = (char*)"invalid block type";
+          r = Z_DATA_ERROR;
+          LEAVE
+      }
+      break;
+    case LENS:
+      NEEDBITS(32)
+      if ((((~b) >> 16) & 0xffff) != (b & 0xffff))
+      {
+        s->mode = BAD;
+        z->msg = (char*)"invalid stored block lengths";
+        r = Z_DATA_ERROR;
+        LEAVE
+      }
+      s->sub.left = (uInt)b & 0xffff;
+      b = k = 0;                      /* dump bits */
+      Tracev((stderr, "inflate:       stored length %u\n", s->sub.left));
+      s->mode = s->sub.left ? STORED : (s->last ? DRY : TYPE);
+      break;
+    case STORED:
+      if (n == 0)
+        LEAVE
+      NEEDOUT
+      t = s->sub.left;
+      if (t > n) t = n;
+      if (t > m) t = m;
+      zmemcpy(q, p, t);
+      p += t;  n -= t;
+      q += t;  m -= t;
+      if ((s->sub.left -= t) != 0)
+        break;
+      Tracev((stderr, "inflate:       stored end, %lu total out\n",
+              z->total_out + (q >= s->read ? q - s->read :
+              (s->end - s->read) + (q - s->window))));
+      s->mode = s->last ? DRY : TYPE;
+      break;
+    case TABLE:
+      NEEDBITS(14)
+      s->sub.trees.table = t = (uInt)b & 0x3fff;
+#ifndef PKZIP_BUG_WORKAROUND
+      if ((t & 0x1f) > 29 || ((t >> 5) & 0x1f) > 29)
+      {
+        s->mode = BAD;
+        z->msg = (char*)"too many length or distance symbols";
+        r = Z_DATA_ERROR;
+        LEAVE
+      }
+#endif
+      t = 258 + (t & 0x1f) + ((t >> 5) & 0x1f);
+      if ((s->sub.trees.blens = (uIntf*)ZALLOC(z, t, sizeof(uInt))) == Z_NULL)
+      {
+        r = Z_MEM_ERROR;
+        LEAVE
+      }
+      DUMPBITS(14)
+      s->sub.trees.index = 0;
+      Tracev((stderr, "inflate:       table sizes ok\n"));
+      s->mode = BTREE;
+    case BTREE:
+      while (s->sub.trees.index < 4 + (s->sub.trees.table >> 10))
+      {
+        NEEDBITS(3)
+        s->sub.trees.blens[border[s->sub.trees.index++]] = (uInt)b & 7;
+        DUMPBITS(3)
+      }
+      while (s->sub.trees.index < 19)
+        s->sub.trees.blens[border[s->sub.trees.index++]] = 0;
+      s->sub.trees.bb = 7;
+      t = inflate_trees_bits(s->sub.trees.blens, &s->sub.trees.bb,
+                             &s->sub.trees.tb, s->hufts, z);
+      if (t != Z_OK)
+      {
+        r = t;
+        if (r == Z_DATA_ERROR)
+        {
+          ZFREE(z, s->sub.trees.blens);
+          s->mode = BAD;
+        }
+        LEAVE
+      }
+      s->sub.trees.index = 0;
+      Tracev((stderr, "inflate:       bits tree ok\n"));
+      s->mode = DTREE;
+    case DTREE:
+      while (t = s->sub.trees.table,
+             s->sub.trees.index < 258 + (t & 0x1f) + ((t >> 5) & 0x1f))
+      {
+        inflate_huft *h;
+        uInt i, j, c;
+
+        t = s->sub.trees.bb;
+        NEEDBITS(t)
+        h = s->sub.trees.tb + ((uInt)b & inflate_mask[t]);
+        t = h->bits;
+        c = h->base;
+        if (c < 16)
+        {
+          DUMPBITS(t)
+          s->sub.trees.blens[s->sub.trees.index++] = c;
+        }
+        else /* c == 16..18 */
+        {
+          i = c == 18 ? 7 : c - 14;
+          j = c == 18 ? 11 : 3;
+          NEEDBITS(t + i)
+          DUMPBITS(t)
+          j += (uInt)b & inflate_mask[i];
+          DUMPBITS(i)
+          i = s->sub.trees.index;
+          t = s->sub.trees.table;
+          if (i + j > 258 + (t & 0x1f) + ((t >> 5) & 0x1f) ||
+              (c == 16 && i < 1))
+          {
+            ZFREE(z, s->sub.trees.blens);
+            s->mode = BAD;
+            z->msg = (char*)"invalid bit length repeat";
+            r = Z_DATA_ERROR;
+            LEAVE
+          }
+          c = c == 16 ? s->sub.trees.blens[i - 1] : 0;
+          do {
+            s->sub.trees.blens[i++] = c;
+          } while (--j);
+          s->sub.trees.index = i;
+        }
+      }
+      s->sub.trees.tb = Z_NULL;
+      {
+        uInt bl, bd;
+        inflate_huft *tl, *td;
+        inflate_codes_statef *c;
+
+        bl = 9;         /* must be <= 9 for lookahead assumptions */
+        bd = 6;         /* must be <= 9 for lookahead assumptions */
+        t = s->sub.trees.table;
+        t = inflate_trees_dynamic(257 + (t & 0x1f), 1 + ((t >> 5) & 0x1f),
+                                  s->sub.trees.blens, &bl, &bd, &tl, &td,
+                                  s->hufts, z);
+        if (t != Z_OK)
+        {
+          if (t == (uInt)Z_DATA_ERROR)
+          {
+            ZFREE(z, s->sub.trees.blens);
+            s->mode = BAD;
+          }
+          r = t;
+          LEAVE
+        }
+        Tracev((stderr, "inflate:       trees ok\n"));
+        if ((c = inflate_codes_new(bl, bd, tl, td, z)) == Z_NULL)
+        {
+          r = Z_MEM_ERROR;
+          LEAVE
+        }
+        s->sub.decode.codes = c;
+      }
+      ZFREE(z, s->sub.trees.blens);
+      s->mode = CODES;
+    case CODES:
+      UPDATE
+      if ((r = inflate_codes(s, z, r)) != Z_STREAM_END)
+        return inflate_flush(s, z, r);
+      r = Z_OK;
+      inflate_codes_free(s->sub.decode.codes, z);
+      LOAD
+      Tracev((stderr, "inflate:       codes end, %lu total out\n",
+              z->total_out + (q >= s->read ? q - s->read :
+              (s->end - s->read) + (q - s->window))));
+      if (!s->last)
+      {
+        s->mode = TYPE;
+        break;
+      }
+      s->mode = DRY;
+    case DRY:
+      FLUSH
+      if (s->read != s->write)
+        LEAVE
+      s->mode = DONE;
+    case DONE:
+      r = Z_STREAM_END;
+      LEAVE
+    case BAD:
+      r = Z_DATA_ERROR;
+      LEAVE
+    default:
+      r = Z_STREAM_ERROR;
+      LEAVE
+  }
+}
+
+
+int inflate_blocks_free(s, z)
+inflate_blocks_statef *s;
+z_streamp z;
+{
+  inflate_blocks_reset(s, z, Z_NULL);
+  ZFREE(z, s->window);
+  ZFREE(z, s->hufts);
+  ZFREE(z, s);
+  Tracev((stderr, "inflate:   blocks freed\n"));
+  return Z_OK;
+}
+
+
+void inflate_set_dictionary(s, d, n)
+inflate_blocks_statef *s;
+const Bytef *d;
+uInt  n;
+{
+  zmemcpy(s->window, d, n);
+  s->read = s->write = s->window + n;
+}
+
+
+/* Returns true if inflate is currently at the end of a block generated
+ * by Z_SYNC_FLUSH or Z_FULL_FLUSH. 
+ * IN assertion: s != Z_NULL
+ */
+int inflate_blocks_sync_point(s)
+inflate_blocks_statef *s;
+{
+  return s->mode == LENS;
+}
diff --git a/zlib/infblock.h b/zlib/infblock.h
new file mode 100644
index 000000000..173b2267a
--- /dev/null
+++ b/zlib/infblock.h
@@ -0,0 +1,39 @@
+/* infblock.h -- header to use infblock.c
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+struct inflate_blocks_state;
+typedef struct inflate_blocks_state FAR inflate_blocks_statef;
+
+extern inflate_blocks_statef * inflate_blocks_new OF((
+    z_streamp z,
+    check_func c,               /* check function */
+    uInt w));                   /* window size */
+
+extern int inflate_blocks OF((
+    inflate_blocks_statef *,
+    z_streamp ,
+    int));                      /* initial return code */
+
+extern void inflate_blocks_reset OF((
+    inflate_blocks_statef *,
+    z_streamp ,
+    uLongf *));                  /* check value on output */
+
+extern int inflate_blocks_free OF((
+    inflate_blocks_statef *,
+    z_streamp));
+
+extern void inflate_set_dictionary OF((
+    inflate_blocks_statef *s,
+    const Bytef *d,  /* dictionary */
+    uInt  n));       /* dictionary length */
+
+extern int inflate_blocks_sync_point OF((
+    inflate_blocks_statef *s));
diff --git a/zlib/infcodes.c b/zlib/infcodes.c
new file mode 100644
index 000000000..9abe5412b
--- /dev/null
+++ b/zlib/infcodes.c
@@ -0,0 +1,251 @@
+/* infcodes.c -- process literals and length/distance pairs
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "infblock.h"
+#include "infcodes.h"
+#include "infutil.h"
+#include "inffast.h"
+
+/* simplify the use of the inflate_huft type with some defines */
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+typedef enum {        /* waiting for "i:"=input, "o:"=output, "x:"=nothing */
+      START,    /* x: set up for LEN */
+      LEN,      /* i: get length/literal/eob next */
+      LENEXT,   /* i: getting length extra (have base) */
+      DIST,     /* i: get distance next */
+      DISTEXT,  /* i: getting distance extra */
+      COPY,     /* o: copying bytes in window, waiting for space */
+      LIT,      /* o: got literal, waiting for output space */
+      WASH,     /* o: got eob, possibly still output waiting */
+      END,      /* x: got eob and all data flushed */
+      BADCODE}  /* x: got error */
+inflate_codes_mode;
+
+/* inflate codes private state */
+struct inflate_codes_state {
+
+  /* mode */
+  inflate_codes_mode mode;      /* current inflate_codes mode */
+
+  /* mode dependent information */
+  uInt len;
+  union {
+    struct {
+      inflate_huft *tree;       /* pointer into tree */
+      uInt need;                /* bits needed */
+    } code;             /* if LEN or DIST, where in tree */
+    uInt lit;           /* if LIT, literal */
+    struct {
+      uInt get;                 /* bits to get for extra */
+      uInt dist;                /* distance back to copy from */
+    } copy;             /* if EXT or COPY, where and how much */
+  } sub;                /* submode */
+
+  /* mode independent information */
+  Byte lbits;           /* ltree bits decoded per branch */
+  Byte dbits;           /* dtree bits decoder per branch */
+  inflate_huft *ltree;          /* literal/length/eob tree */
+  inflate_huft *dtree;          /* distance tree */
+
+};
+
+
+inflate_codes_statef *inflate_codes_new(bl, bd, tl, td, z)
+uInt bl, bd;
+inflate_huft *tl;
+inflate_huft *td; /* need separate declaration for Borland C++ */
+z_streamp z;
+{
+  inflate_codes_statef *c;
+
+  if ((c = (inflate_codes_statef *)
+       ZALLOC(z,1,sizeof(struct inflate_codes_state))) != Z_NULL)
+  {
+    c->mode = START;
+    c->lbits = (Byte)bl;
+    c->dbits = (Byte)bd;
+    c->ltree = tl;
+    c->dtree = td;
+    Tracev((stderr, "inflate:       codes new\n"));
+  }
+  return c;
+}
+
+
+int inflate_codes(s, z, r)
+inflate_blocks_statef *s;
+z_streamp z;
+int r;
+{
+  uInt j;               /* temporary storage */
+  inflate_huft *t;      /* temporary pointer */
+  uInt e;               /* extra bits or operation */
+  uLong b;              /* bit buffer */
+  uInt k;               /* bits in bit buffer */
+  Bytef *p;             /* input data pointer */
+  uInt n;               /* bytes available there */
+  Bytef *q;             /* output window write pointer */
+  uInt m;               /* bytes to end of window or read pointer */
+  Bytef *f;             /* pointer to copy strings from */
+  inflate_codes_statef *c = s->sub.decode.codes;  /* codes state */
+
+  /* copy input/output information to locals (UPDATE macro restores) */
+  LOAD
+
+  /* process input and output based on current state */
+  while (1) switch (c->mode)
+  {             /* waiting for "i:"=input, "o:"=output, "x:"=nothing */
+    case START:         /* x: set up for LEN */
+#ifndef SLOW
+      if (m >= 258 && n >= 10)
+      {
+        UPDATE
+        r = inflate_fast(c->lbits, c->dbits, c->ltree, c->dtree, s, z);
+        LOAD
+        if (r != Z_OK)
+        {
+          c->mode = r == Z_STREAM_END ? WASH : BADCODE;
+          break;
+        }
+      }
+#endif /* !SLOW */
+      c->sub.code.need = c->lbits;
+      c->sub.code.tree = c->ltree;
+      c->mode = LEN;
+    case LEN:           /* i: get length/literal/eob next */
+      j = c->sub.code.need;
+      NEEDBITS(j)
+      t = c->sub.code.tree + ((uInt)b & inflate_mask[j]);
+      DUMPBITS(t->bits)
+      e = (uInt)(t->exop);
+      if (e == 0)               /* literal */
+      {
+        c->sub.lit = t->base;
+        Tracevv((stderr, t->base >= 0x20 && t->base < 0x7f ?
+                 "inflate:         literal '%c'\n" :
+                 "inflate:         literal 0x%02x\n", t->base));
+        c->mode = LIT;
+        break;
+      }
+      if (e & 16)               /* length */
+      {
+        c->sub.copy.get = e & 15;
+        c->len = t->base;
+        c->mode = LENEXT;
+        break;
+      }
+      if ((e & 64) == 0)        /* next table */
+      {
+        c->sub.code.need = e;
+        c->sub.code.tree = t + t->base;
+        break;
+      }
+      if (e & 32)               /* end of block */
+      {
+        Tracevv((stderr, "inflate:         end of block\n"));
+        c->mode = WASH;
+        break;
+      }
+      c->mode = BADCODE;        /* invalid code */
+      z->msg = (char*)"invalid literal/length code";
+      r = Z_DATA_ERROR;
+      LEAVE
+    case LENEXT:        /* i: getting length extra (have base) */
+      j = c->sub.copy.get;
+      NEEDBITS(j)
+      c->len += (uInt)b & inflate_mask[j];
+      DUMPBITS(j)
+      c->sub.code.need = c->dbits;
+      c->sub.code.tree = c->dtree;
+      Tracevv((stderr, "inflate:         length %u\n", c->len));
+      c->mode = DIST;
+    case DIST:          /* i: get distance next */
+      j = c->sub.code.need;
+      NEEDBITS(j)
+      t = c->sub.code.tree + ((uInt)b & inflate_mask[j]);
+      DUMPBITS(t->bits)
+      e = (uInt)(t->exop);
+      if (e & 16)               /* distance */
+      {
+        c->sub.copy.get = e & 15;
+        c->sub.copy.dist = t->base;
+        c->mode = DISTEXT;
+        break;
+      }
+      if ((e & 64) == 0)        /* next table */
+      {
+        c->sub.code.need = e;
+        c->sub.code.tree = t + t->base;
+        break;
+      }
+      c->mode = BADCODE;        /* invalid code */
+      z->msg = (char*)"invalid distance code";
+      r = Z_DATA_ERROR;
+      LEAVE
+    case DISTEXT:       /* i: getting distance extra */
+      j = c->sub.copy.get;
+      NEEDBITS(j)
+      c->sub.copy.dist += (uInt)b & inflate_mask[j];
+      DUMPBITS(j)
+      Tracevv((stderr, "inflate:         distance %u\n", c->sub.copy.dist));
+      c->mode = COPY;
+    case COPY:          /* o: copying bytes in window, waiting for space */
+      f = q - c->sub.copy.dist;
+      while (f < s->window)             /* modulo window size-"while" instead */
+        f += s->end - s->window;        /* of "if" handles invalid distances */
+      while (c->len)
+      {
+        NEEDOUT
+        OUTBYTE(*f++)
+        if (f == s->end)
+          f = s->window;
+        c->len--;
+      }
+      c->mode = START;
+      break;
+    case LIT:           /* o: got literal, waiting for output space */
+      NEEDOUT
+      OUTBYTE(c->sub.lit)
+      c->mode = START;
+      break;
+    case WASH:          /* o: got eob, possibly more output */
+      if (k > 7)        /* return unused byte, if any */
+      {
+        Assert(k < 16, "inflate_codes grabbed too many bytes")
+        k -= 8;
+        n++;
+        p--;            /* can always return one */
+      }
+      FLUSH
+      if (s->read != s->write)
+        LEAVE
+      c->mode = END;
+    case END:
+      r = Z_STREAM_END;
+      LEAVE
+    case BADCODE:       /* x: got error */
+      r = Z_DATA_ERROR;
+      LEAVE
+    default:
+      r = Z_STREAM_ERROR;
+      LEAVE
+  }
+#ifdef NEED_DUMMY_RETURN
+  return Z_STREAM_ERROR;  /* Some dumb compilers complain without this */
+#endif
+}
+
+
+void inflate_codes_free(c, z)
+inflate_codes_statef *c;
+z_streamp z;
+{
+  ZFREE(z, c);
+  Tracev((stderr, "inflate:       codes free\n"));
+}
diff --git a/zlib/infcodes.h b/zlib/infcodes.h
new file mode 100644
index 000000000..46821a02b
--- /dev/null
+++ b/zlib/infcodes.h
@@ -0,0 +1,27 @@
+/* infcodes.h -- header to use infcodes.c
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+struct inflate_codes_state;
+typedef struct inflate_codes_state FAR inflate_codes_statef;
+
+extern inflate_codes_statef *inflate_codes_new OF((
+    uInt, uInt,
+    inflate_huft *, inflate_huft *,
+    z_streamp ));
+
+extern int inflate_codes OF((
+    inflate_blocks_statef *,
+    z_streamp ,
+    int));
+
+extern void inflate_codes_free OF((
+    inflate_codes_statef *,
+    z_streamp ));
+
diff --git a/zlib/inffast.c b/zlib/inffast.c
new file mode 100644
index 000000000..aa7f1d4d2
--- /dev/null
+++ b/zlib/inffast.c
@@ -0,0 +1,183 @@
+/* inffast.c -- process literals and length/distance pairs fast
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "infblock.h"
+#include "infcodes.h"
+#include "infutil.h"
+#include "inffast.h"
+
+struct inflate_codes_state {int dummy;}; /* for buggy compilers */
+
+/* simplify the use of the inflate_huft type with some defines */
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+/* macros for bit input with no checking and for returning unused bytes */
+#define GRABBITS(j) {while(k<(j)){b|=((uLong)NEXTBYTE)<<k;k+=8;}}
+#define UNGRAB {c=z->avail_in-n;c=(k>>3)<c?k>>3:c;n+=c;p-=c;k-=c<<3;}
+
+/* Called with number of bytes left to write in window at least 258
+   (the maximum string length) and number of input bytes available
+   at least ten.  The ten bytes are six bytes for the longest length/
+   distance pair plus four bytes for overloading the bit buffer. */
+
+int inflate_fast(bl, bd, tl, td, s, z)
+uInt bl, bd;
+inflate_huft *tl;
+inflate_huft *td; /* need separate declaration for Borland C++ */
+inflate_blocks_statef *s;
+z_streamp z;
+{
+  inflate_huft *t;      /* temporary pointer */
+  uInt e;               /* extra bits or operation */
+  uLong b;              /* bit buffer */
+  uInt k;               /* bits in bit buffer */
+  Bytef *p;             /* input data pointer */
+  uInt n;               /* bytes available there */
+  Bytef *q;             /* output window write pointer */
+  uInt m;               /* bytes to end of window or read pointer */
+  uInt ml;              /* mask for literal/length tree */
+  uInt md;              /* mask for distance tree */
+  uInt c;               /* bytes to copy */
+  uInt d;               /* distance back to copy from */
+  Bytef *r;             /* copy source pointer */
+
+  /* load input, output, bit values */
+  LOAD
+
+  /* initialize masks */
+  ml = inflate_mask[bl];
+  md = inflate_mask[bd];
+
+  /* do until not enough input or output space for fast loop */
+  do {                          /* assume called with m >= 258 && n >= 10 */
+    /* get literal/length code */
+    GRABBITS(20)                /* max bits for literal/length code */
+    if ((e = (t = tl + ((uInt)b & ml))->exop) == 0)
+    {
+      DUMPBITS(t->bits)
+      Tracevv((stderr, t->base >= 0x20 && t->base < 0x7f ?
+                "inflate:         * literal '%c'\n" :
+                "inflate:         * literal 0x%02x\n", t->base));
+      *q++ = (Byte)t->base;
+      m--;
+      continue;
+    }
+    do {
+      DUMPBITS(t->bits)
+      if (e & 16)
+      {
+        /* get extra bits for length */
+        e &= 15;
+        c = t->base + ((uInt)b & inflate_mask[e]);
+        DUMPBITS(e)
+        Tracevv((stderr, "inflate:         * length %u\n", c));
+
+        /* decode distance base of block to copy */
+        GRABBITS(15);           /* max bits for distance code */
+        e = (t = td + ((uInt)b & md))->exop;
+        do {
+          DUMPBITS(t->bits)
+          if (e & 16)
+          {
+            /* get extra bits to add to distance base */
+            e &= 15;
+            GRABBITS(e)         /* get extra bits (up to 13) */
+            d = t->base + ((uInt)b & inflate_mask[e]);
+            DUMPBITS(e)
+            Tracevv((stderr, "inflate:         * distance %u\n", d));
+
+            /* do the copy */
+            m -= c;
+            r = q - d;
+            if (r < s->window)                  /* wrap if needed */
+            {
+              do {
+                r += s->end - s->window;        /* force pointer in window */
+              } while (r < s->window);          /* covers invalid distances */
+              e = s->end - r;
+              if (c > e)
+              {
+                c -= e;                         /* wrapped copy */
+                do {
+                    *q++ = *r++;
+                } while (--e);
+                r = s->window;
+                do {
+                    *q++ = *r++;
+                } while (--c);
+              }
+              else                              /* normal copy */
+              {
+                *q++ = *r++;  c--;
+                *q++ = *r++;  c--;
+                do {
+                    *q++ = *r++;
+                } while (--c);
+              }
+            }
+            else                                /* normal copy */
+            {
+              *q++ = *r++;  c--;
+              *q++ = *r++;  c--;
+              do {
+                *q++ = *r++;
+              } while (--c);
+            }
+            break;
+          }
+          else if ((e & 64) == 0)
+          {
+            t += t->base;
+            e = (t += ((uInt)b & inflate_mask[e]))->exop;
+          }
+          else
+          {
+            z->msg = (char*)"invalid distance code";
+            UNGRAB
+            UPDATE
+            return Z_DATA_ERROR;
+          }
+        } while (1);
+        break;
+      }
+      if ((e & 64) == 0)
+      {
+        t += t->base;
+        if ((e = (t += ((uInt)b & inflate_mask[e]))->exop) == 0)
+        {
+          DUMPBITS(t->bits)
+          Tracevv((stderr, t->base >= 0x20 && t->base < 0x7f ?
+                    "inflate:         * literal '%c'\n" :
+                    "inflate:         * literal 0x%02x\n", t->base));
+          *q++ = (Byte)t->base;
+          m--;
+          break;
+        }
+      }
+      else if (e & 32)
+      {
+        Tracevv((stderr, "inflate:         * end of block\n"));
+        UNGRAB
+        UPDATE
+        return Z_STREAM_END;
+      }
+      else
+      {
+        z->msg = (char*)"invalid literal/length code";
+        UNGRAB
+        UPDATE
+        return Z_DATA_ERROR;
+      }
+    } while (1);
+  } while (m >= 258 && n >= 10);
+
+  /* not enough input or output--restore pointers and return */
+  UNGRAB
+  UPDATE
+  return Z_OK;
+}
diff --git a/zlib/inffast.h b/zlib/inffast.h
new file mode 100644
index 000000000..a31a4bbb0
--- /dev/null
+++ b/zlib/inffast.h
@@ -0,0 +1,17 @@
+/* inffast.h -- header to use inffast.c
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+extern int inflate_fast OF((
+    uInt,
+    uInt,
+    inflate_huft *,
+    inflate_huft *,
+    inflate_blocks_statef *,
+    z_streamp ));
diff --git a/zlib/inffixed.h b/zlib/inffixed.h
new file mode 100644
index 000000000..77f7e7631
--- /dev/null
+++ b/zlib/inffixed.h
@@ -0,0 +1,151 @@
+/* inffixed.h -- table for decoding fixed codes
+ * Generated automatically by the maketree.c program
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+local uInt fixed_bl = 9;
+local uInt fixed_bd = 5;
+local inflate_huft fixed_tl[] = {
+    {{{96,7}},256}, {{{0,8}},80}, {{{0,8}},16}, {{{84,8}},115},
+    {{{82,7}},31}, {{{0,8}},112}, {{{0,8}},48}, {{{0,9}},192},
+    {{{80,7}},10}, {{{0,8}},96}, {{{0,8}},32}, {{{0,9}},160},
+    {{{0,8}},0}, {{{0,8}},128}, {{{0,8}},64}, {{{0,9}},224},
+    {{{80,7}},6}, {{{0,8}},88}, {{{0,8}},24}, {{{0,9}},144},
+    {{{83,7}},59}, {{{0,8}},120}, {{{0,8}},56}, {{{0,9}},208},
+    {{{81,7}},17}, {{{0,8}},104}, {{{0,8}},40}, {{{0,9}},176},
+    {{{0,8}},8}, {{{0,8}},136}, {{{0,8}},72}, {{{0,9}},240},
+    {{{80,7}},4}, {{{0,8}},84}, {{{0,8}},20}, {{{85,8}},227},
+    {{{83,7}},43}, {{{0,8}},116}, {{{0,8}},52}, {{{0,9}},200},
+    {{{81,7}},13}, {{{0,8}},100}, {{{0,8}},36}, {{{0,9}},168},
+    {{{0,8}},4}, {{{0,8}},132}, {{{0,8}},68}, {{{0,9}},232},
+    {{{80,7}},8}, {{{0,8}},92}, {{{0,8}},28}, {{{0,9}},152},
+    {{{84,7}},83}, {{{0,8}},124}, {{{0,8}},60}, {{{0,9}},216},
+    {{{82,7}},23}, {{{0,8}},108}, {{{0,8}},44}, {{{0,9}},184},
+    {{{0,8}},12}, {{{0,8}},140}, {{{0,8}},76}, {{{0,9}},248},
+    {{{80,7}},3}, {{{0,8}},82}, {{{0,8}},18}, {{{85,8}},163},
+    {{{83,7}},35}, {{{0,8}},114}, {{{0,8}},50}, {{{0,9}},196},
+    {{{81,7}},11}, {{{0,8}},98}, {{{0,8}},34}, {{{0,9}},164},
+    {{{0,8}},2}, {{{0,8}},130}, {{{0,8}},66}, {{{0,9}},228},
+    {{{80,7}},7}, {{{0,8}},90}, {{{0,8}},26}, {{{0,9}},148},
+    {{{84,7}},67}, {{{0,8}},122}, {{{0,8}},58}, {{{0,9}},212},
+    {{{82,7}},19}, {{{0,8}},106}, {{{0,8}},42}, {{{0,9}},180},
+    {{{0,8}},10}, {{{0,8}},138}, {{{0,8}},74}, {{{0,9}},244},
+    {{{80,7}},5}, {{{0,8}},86}, {{{0,8}},22}, {{{192,8}},0},
+    {{{83,7}},51}, {{{0,8}},118}, {{{0,8}},54}, {{{0,9}},204},
+    {{{81,7}},15}, {{{0,8}},102}, {{{0,8}},38}, {{{0,9}},172},
+    {{{0,8}},6}, {{{0,8}},134}, {{{0,8}},70}, {{{0,9}},236},
+    {{{80,7}},9}, {{{0,8}},94}, {{{0,8}},30}, {{{0,9}},156},
+    {{{84,7}},99}, {{{0,8}},126}, {{{0,8}},62}, {{{0,9}},220},
+    {{{82,7}},27}, {{{0,8}},110}, {{{0,8}},46}, {{{0,9}},188},
+    {{{0,8}},14}, {{{0,8}},142}, {{{0,8}},78}, {{{0,9}},252},
+    {{{96,7}},256}, {{{0,8}},81}, {{{0,8}},17}, {{{85,8}},131},
+    {{{82,7}},31}, {{{0,8}},113}, {{{0,8}},49}, {{{0,9}},194},
+    {{{80,7}},10}, {{{0,8}},97}, {{{0,8}},33}, {{{0,9}},162},
+    {{{0,8}},1}, {{{0,8}},129}, {{{0,8}},65}, {{{0,9}},226},
+    {{{80,7}},6}, {{{0,8}},89}, {{{0,8}},25}, {{{0,9}},146},
+    {{{83,7}},59}, {{{0,8}},121}, {{{0,8}},57}, {{{0,9}},210},
+    {{{81,7}},17}, {{{0,8}},105}, {{{0,8}},41}, {{{0,9}},178},
+    {{{0,8}},9}, {{{0,8}},137}, {{{0,8}},73}, {{{0,9}},242},
+    {{{80,7}},4}, {{{0,8}},85}, {{{0,8}},21}, {{{80,8}},258},
+    {{{83,7}},43}, {{{0,8}},117}, {{{0,8}},53}, {{{0,9}},202},
+    {{{81,7}},13}, {{{0,8}},101}, {{{0,8}},37}, {{{0,9}},170},
+    {{{0,8}},5}, {{{0,8}},133}, {{{0,8}},69}, {{{0,9}},234},
+    {{{80,7}},8}, {{{0,8}},93}, {{{0,8}},29}, {{{0,9}},154},
+    {{{84,7}},83}, {{{0,8}},125}, {{{0,8}},61}, {{{0,9}},218},
+    {{{82,7}},23}, {{{0,8}},109}, {{{0,8}},45}, {{{0,9}},186},
+    {{{0,8}},13}, {{{0,8}},141}, {{{0,8}},77}, {{{0,9}},250},
+    {{{80,7}},3}, {{{0,8}},83}, {{{0,8}},19}, {{{85,8}},195},
+    {{{83,7}},35}, {{{0,8}},115}, {{{0,8}},51}, {{{0,9}},198},
+    {{{81,7}},11}, {{{0,8}},99}, {{{0,8}},35}, {{{0,9}},166},
+    {{{0,8}},3}, {{{0,8}},131}, {{{0,8}},67}, {{{0,9}},230},
+    {{{80,7}},7}, {{{0,8}},91}, {{{0,8}},27}, {{{0,9}},150},
+    {{{84,7}},67}, {{{0,8}},123}, {{{0,8}},59}, {{{0,9}},214},
+    {{{82,7}},19}, {{{0,8}},107}, {{{0,8}},43}, {{{0,9}},182},
+    {{{0,8}},11}, {{{0,8}},139}, {{{0,8}},75}, {{{0,9}},246},
+    {{{80,7}},5}, {{{0,8}},87}, {{{0,8}},23}, {{{192,8}},0},
+    {{{83,7}},51}, {{{0,8}},119}, {{{0,8}},55}, {{{0,9}},206},
+    {{{81,7}},15}, {{{0,8}},103}, {{{0,8}},39}, {{{0,9}},174},
+    {{{0,8}},7}, {{{0,8}},135}, {{{0,8}},71}, {{{0,9}},238},
+    {{{80,7}},9}, {{{0,8}},95}, {{{0,8}},31}, {{{0,9}},158},
+    {{{84,7}},99}, {{{0,8}},127}, {{{0,8}},63}, {{{0,9}},222},
+    {{{82,7}},27}, {{{0,8}},111}, {{{0,8}},47}, {{{0,9}},190},
+    {{{0,8}},15}, {{{0,8}},143}, {{{0,8}},79}, {{{0,9}},254},
+    {{{96,7}},256}, {{{0,8}},80}, {{{0,8}},16}, {{{84,8}},115},
+    {{{82,7}},31}, {{{0,8}},112}, {{{0,8}},48}, {{{0,9}},193},
+    {{{80,7}},10}, {{{0,8}},96}, {{{0,8}},32}, {{{0,9}},161},
+    {{{0,8}},0}, {{{0,8}},128}, {{{0,8}},64}, {{{0,9}},225},
+    {{{80,7}},6}, {{{0,8}},88}, {{{0,8}},24}, {{{0,9}},145},
+    {{{83,7}},59}, {{{0,8}},120}, {{{0,8}},56}, {{{0,9}},209},
+    {{{81,7}},17}, {{{0,8}},104}, {{{0,8}},40}, {{{0,9}},177},
+    {{{0,8}},8}, {{{0,8}},136}, {{{0,8}},72}, {{{0,9}},241},
+    {{{80,7}},4}, {{{0,8}},84}, {{{0,8}},20}, {{{85,8}},227},
+    {{{83,7}},43}, {{{0,8}},116}, {{{0,8}},52}, {{{0,9}},201},
+    {{{81,7}},13}, {{{0,8}},100}, {{{0,8}},36}, {{{0,9}},169},
+    {{{0,8}},4}, {{{0,8}},132}, {{{0,8}},68}, {{{0,9}},233},
+    {{{80,7}},8}, {{{0,8}},92}, {{{0,8}},28}, {{{0,9}},153},
+    {{{84,7}},83}, {{{0,8}},124}, {{{0,8}},60}, {{{0,9}},217},
+    {{{82,7}},23}, {{{0,8}},108}, {{{0,8}},44}, {{{0,9}},185},
+    {{{0,8}},12}, {{{0,8}},140}, {{{0,8}},76}, {{{0,9}},249},
+    {{{80,7}},3}, {{{0,8}},82}, {{{0,8}},18}, {{{85,8}},163},
+    {{{83,7}},35}, {{{0,8}},114}, {{{0,8}},50}, {{{0,9}},197},
+    {{{81,7}},11}, {{{0,8}},98}, {{{0,8}},34}, {{{0,9}},165},
+    {{{0,8}},2}, {{{0,8}},130}, {{{0,8}},66}, {{{0,9}},229},
+    {{{80,7}},7}, {{{0,8}},90}, {{{0,8}},26}, {{{0,9}},149},
+    {{{84,7}},67}, {{{0,8}},122}, {{{0,8}},58}, {{{0,9}},213},
+    {{{82,7}},19}, {{{0,8}},106}, {{{0,8}},42}, {{{0,9}},181},
+    {{{0,8}},10}, {{{0,8}},138}, {{{0,8}},74}, {{{0,9}},245},
+    {{{80,7}},5}, {{{0,8}},86}, {{{0,8}},22}, {{{192,8}},0},
+    {{{83,7}},51}, {{{0,8}},118}, {{{0,8}},54}, {{{0,9}},205},
+    {{{81,7}},15}, {{{0,8}},102}, {{{0,8}},38}, {{{0,9}},173},
+    {{{0,8}},6}, {{{0,8}},134}, {{{0,8}},70}, {{{0,9}},237},
+    {{{80,7}},9}, {{{0,8}},94}, {{{0,8}},30}, {{{0,9}},157},
+    {{{84,7}},99}, {{{0,8}},126}, {{{0,8}},62}, {{{0,9}},221},
+    {{{82,7}},27}, {{{0,8}},110}, {{{0,8}},46}, {{{0,9}},189},
+    {{{0,8}},14}, {{{0,8}},142}, {{{0,8}},78}, {{{0,9}},253},
+    {{{96,7}},256}, {{{0,8}},81}, {{{0,8}},17}, {{{85,8}},131},
+    {{{82,7}},31}, {{{0,8}},113}, {{{0,8}},49}, {{{0,9}},195},
+    {{{80,7}},10}, {{{0,8}},97}, {{{0,8}},33}, {{{0,9}},163},
+    {{{0,8}},1}, {{{0,8}},129}, {{{0,8}},65}, {{{0,9}},227},
+    {{{80,7}},6}, {{{0,8}},89}, {{{0,8}},25}, {{{0,9}},147},
+    {{{83,7}},59}, {{{0,8}},121}, {{{0,8}},57}, {{{0,9}},211},
+    {{{81,7}},17}, {{{0,8}},105}, {{{0,8}},41}, {{{0,9}},179},
+    {{{0,8}},9}, {{{0,8}},137}, {{{0,8}},73}, {{{0,9}},243},
+    {{{80,7}},4}, {{{0,8}},85}, {{{0,8}},21}, {{{80,8}},258},
+    {{{83,7}},43}, {{{0,8}},117}, {{{0,8}},53}, {{{0,9}},203},
+    {{{81,7}},13}, {{{0,8}},101}, {{{0,8}},37}, {{{0,9}},171},
+    {{{0,8}},5}, {{{0,8}},133}, {{{0,8}},69}, {{{0,9}},235},
+    {{{80,7}},8}, {{{0,8}},93}, {{{0,8}},29}, {{{0,9}},155},
+    {{{84,7}},83}, {{{0,8}},125}, {{{0,8}},61}, {{{0,9}},219},
+    {{{82,7}},23}, {{{0,8}},109}, {{{0,8}},45}, {{{0,9}},187},
+    {{{0,8}},13}, {{{0,8}},141}, {{{0,8}},77}, {{{0,9}},251},
+    {{{80,7}},3}, {{{0,8}},83}, {{{0,8}},19}, {{{85,8}},195},
+    {{{83,7}},35}, {{{0,8}},115}, {{{0,8}},51}, {{{0,9}},199},
+    {{{81,7}},11}, {{{0,8}},99}, {{{0,8}},35}, {{{0,9}},167},
+    {{{0,8}},3}, {{{0,8}},131}, {{{0,8}},67}, {{{0,9}},231},
+    {{{80,7}},7}, {{{0,8}},91}, {{{0,8}},27}, {{{0,9}},151},
+    {{{84,7}},67}, {{{0,8}},123}, {{{0,8}},59}, {{{0,9}},215},
+    {{{82,7}},19}, {{{0,8}},107}, {{{0,8}},43}, {{{0,9}},183},
+    {{{0,8}},11}, {{{0,8}},139}, {{{0,8}},75}, {{{0,9}},247},
+    {{{80,7}},5}, {{{0,8}},87}, {{{0,8}},23}, {{{192,8}},0},
+    {{{83,7}},51}, {{{0,8}},119}, {{{0,8}},55}, {{{0,9}},207},
+    {{{81,7}},15}, {{{0,8}},103}, {{{0,8}},39}, {{{0,9}},175},
+    {{{0,8}},7}, {{{0,8}},135}, {{{0,8}},71}, {{{0,9}},239},
+    {{{80,7}},9}, {{{0,8}},95}, {{{0,8}},31}, {{{0,9}},159},
+    {{{84,7}},99}, {{{0,8}},127}, {{{0,8}},63}, {{{0,9}},223},
+    {{{82,7}},27}, {{{0,8}},111}, {{{0,8}},47}, {{{0,9}},191},
+    {{{0,8}},15}, {{{0,8}},143}, {{{0,8}},79}, {{{0,9}},255}
+  };
+local inflate_huft fixed_td[] = {
+    {{{80,5}},1}, {{{87,5}},257}, {{{83,5}},17}, {{{91,5}},4097},
+    {{{81,5}},5}, {{{89,5}},1025}, {{{85,5}},65}, {{{93,5}},16385},
+    {{{80,5}},3}, {{{88,5}},513}, {{{84,5}},33}, {{{92,5}},8193},
+    {{{82,5}},9}, {{{90,5}},2049}, {{{86,5}},129}, {{{192,5}},24577},
+    {{{80,5}},2}, {{{87,5}},385}, {{{83,5}},25}, {{{91,5}},6145},
+    {{{81,5}},7}, {{{89,5}},1537}, {{{85,5}},97}, {{{93,5}},24577},
+    {{{80,5}},4}, {{{88,5}},769}, {{{84,5}},49}, {{{92,5}},12289},
+    {{{82,5}},13}, {{{90,5}},3073}, {{{86,5}},193}, {{{192,5}},24577}
+  };
diff --git a/zlib/inflate.c b/zlib/inflate.c
new file mode 100644
index 000000000..dfb2e867d
--- /dev/null
+++ b/zlib/inflate.c
@@ -0,0 +1,366 @@
+/* inflate.c -- zlib interface to inflate modules
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+#include "zutil.h"
+#include "infblock.h"
+
+struct inflate_blocks_state {int dummy;}; /* for buggy compilers */
+
+typedef enum {
+      METHOD,   /* waiting for method byte */
+      FLAG,     /* waiting for flag byte */
+      DICT4,    /* four dictionary check bytes to go */
+      DICT3,    /* three dictionary check bytes to go */
+      DICT2,    /* two dictionary check bytes to go */
+      DICT1,    /* one dictionary check byte to go */
+      DICT0,    /* waiting for inflateSetDictionary */
+      BLOCKS,   /* decompressing blocks */
+      CHECK4,   /* four check bytes to go */
+      CHECK3,   /* three check bytes to go */
+      CHECK2,   /* two check bytes to go */
+      CHECK1,   /* one check byte to go */
+      DONE,     /* finished check, done */
+      BAD}      /* got an error--stay here */
+inflate_mode;
+
+/* inflate private state */
+struct internal_state {
+
+  /* mode */
+  inflate_mode  mode;   /* current inflate mode */
+
+  /* mode dependent information */
+  union {
+    uInt method;        /* if FLAGS, method byte */
+    struct {
+      uLong was;                /* computed check value */
+      uLong need;               /* stream check value */
+    } check;            /* if CHECK, check values to compare */
+    uInt marker;        /* if BAD, inflateSync's marker bytes count */
+  } sub;        /* submode */
+
+  /* mode independent information */
+  int  nowrap;          /* flag for no wrapper */
+  uInt wbits;           /* log2(window size)  (8..15, defaults to 15) */
+  inflate_blocks_statef 
+    *blocks;            /* current inflate_blocks state */
+
+};
+
+
+int ZEXPORT inflateReset(z)
+z_streamp z;
+{
+  if (z == Z_NULL || z->state == Z_NULL)
+    return Z_STREAM_ERROR;
+  z->total_in = z->total_out = 0;
+  z->msg = Z_NULL;
+  z->state->mode = z->state->nowrap ? BLOCKS : METHOD;
+  inflate_blocks_reset(z->state->blocks, z, Z_NULL);
+  Tracev((stderr, "inflate: reset\n"));
+  return Z_OK;
+}
+
+
+int ZEXPORT inflateEnd(z)
+z_streamp z;
+{
+  if (z == Z_NULL || z->state == Z_NULL || z->zfree == Z_NULL)
+    return Z_STREAM_ERROR;
+  if (z->state->blocks != Z_NULL)
+    inflate_blocks_free(z->state->blocks, z);
+  ZFREE(z, z->state);
+  z->state = Z_NULL;
+  Tracev((stderr, "inflate: end\n"));
+  return Z_OK;
+}
+
+
+int ZEXPORT inflateInit2_(z, w, version, stream_size)
+z_streamp z;
+int w;
+const char *version;
+int stream_size;
+{
+  if (version == Z_NULL || version[0] != ZLIB_VERSION[0] ||
+      stream_size != sizeof(z_stream))
+      return Z_VERSION_ERROR;
+
+  /* initialize state */
+  if (z == Z_NULL)
+    return Z_STREAM_ERROR;
+  z->msg = Z_NULL;
+  if (z->zalloc == Z_NULL)
+  {
+    z->zalloc = zcalloc;
+    z->opaque = (voidpf)0;
+  }
+  if (z->zfree == Z_NULL) z->zfree = zcfree;
+  if ((z->state = (struct internal_state FAR *)
+       ZALLOC(z,1,sizeof(struct internal_state))) == Z_NULL)
+    return Z_MEM_ERROR;
+  z->state->blocks = Z_NULL;
+
+  /* handle undocumented nowrap option (no zlib header or check) */
+  z->state->nowrap = 0;
+  if (w < 0)
+  {
+    w = - w;
+    z->state->nowrap = 1;
+  }
+
+  /* set window size */
+  if (w < 8 || w > 15)
+  {
+    inflateEnd(z);
+    return Z_STREAM_ERROR;
+  }
+  z->state->wbits = (uInt)w;
+
+  /* create inflate_blocks state */
+  if ((z->state->blocks =
+      inflate_blocks_new(z, z->state->nowrap ? Z_NULL : adler32, (uInt)1 << w))
+      == Z_NULL)
+  {
+    inflateEnd(z);
+    return Z_MEM_ERROR;
+  }
+  Tracev((stderr, "inflate: allocated\n"));
+
+  /* reset state */
+  inflateReset(z);
+  return Z_OK;
+}
+
+
+int ZEXPORT inflateInit_(z, version, stream_size)
+z_streamp z;
+const char *version;
+int stream_size;
+{
+  return inflateInit2_(z, DEF_WBITS, version, stream_size);
+}
+
+
+#define NEEDBYTE {if(z->avail_in==0)return r;r=f;}
+#define NEXTBYTE (z->avail_in--,z->total_in++,*z->next_in++)
+
+int ZEXPORT inflate(z, f)
+z_streamp z;
+int f;
+{
+  int r;
+  uInt b;
+
+  if (z == Z_NULL || z->state == Z_NULL || z->next_in == Z_NULL)
+    return Z_STREAM_ERROR;
+  f = f == Z_FINISH ? Z_BUF_ERROR : Z_OK;
+  r = Z_BUF_ERROR;
+  while (1) switch (z->state->mode)
+  {
+    case METHOD:
+      NEEDBYTE
+      if (((z->state->sub.method = NEXTBYTE) & 0xf) != Z_DEFLATED)
+      {
+        z->state->mode = BAD;
+        z->msg = (char*)"unknown compression method";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      if ((z->state->sub.method >> 4) + 8 > z->state->wbits)
+      {
+        z->state->mode = BAD;
+        z->msg = (char*)"invalid window size";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      z->state->mode = FLAG;
+    case FLAG:
+      NEEDBYTE
+      b = NEXTBYTE;
+      if (((z->state->sub.method << 8) + b) % 31)
+      {
+        z->state->mode = BAD;
+        z->msg = (char*)"incorrect header check";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      Tracev((stderr, "inflate: zlib header ok\n"));
+      if (!(b & PRESET_DICT))
+      {
+        z->state->mode = BLOCKS;
+        break;
+      }
+      z->state->mode = DICT4;
+    case DICT4:
+      NEEDBYTE
+      z->state->sub.check.need = (uLong)NEXTBYTE << 24;
+      z->state->mode = DICT3;
+    case DICT3:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE << 16;
+      z->state->mode = DICT2;
+    case DICT2:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE << 8;
+      z->state->mode = DICT1;
+    case DICT1:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE;
+      z->adler = z->state->sub.check.need;
+      z->state->mode = DICT0;
+      return Z_NEED_DICT;
+    case DICT0:
+      z->state->mode = BAD;
+      z->msg = (char*)"need dictionary";
+      z->state->sub.marker = 0;       /* can try inflateSync */
+      return Z_STREAM_ERROR;
+    case BLOCKS:
+      r = inflate_blocks(z->state->blocks, z, r);
+      if (r == Z_DATA_ERROR)
+      {
+        z->state->mode = BAD;
+        z->state->sub.marker = 0;       /* can try inflateSync */
+        break;
+      }
+      if (r == Z_OK)
+        r = f;
+      if (r != Z_STREAM_END)
+        return r;
+      r = f;
+      inflate_blocks_reset(z->state->blocks, z, &z->state->sub.check.was);
+      if (z->state->nowrap)
+      {
+        z->state->mode = DONE;
+        break;
+      }
+      z->state->mode = CHECK4;
+    case CHECK4:
+      NEEDBYTE
+      z->state->sub.check.need = (uLong)NEXTBYTE << 24;
+      z->state->mode = CHECK3;
+    case CHECK3:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE << 16;
+      z->state->mode = CHECK2;
+    case CHECK2:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE << 8;
+      z->state->mode = CHECK1;
+    case CHECK1:
+      NEEDBYTE
+      z->state->sub.check.need += (uLong)NEXTBYTE;
+
+      if (z->state->sub.check.was != z->state->sub.check.need)
+      {
+        z->state->mode = BAD;
+        z->msg = (char*)"incorrect data check";
+        z->state->sub.marker = 5;       /* can't try inflateSync */
+        break;
+      }
+      Tracev((stderr, "inflate: zlib check ok\n"));
+      z->state->mode = DONE;
+    case DONE:
+      return Z_STREAM_END;
+    case BAD:
+      return Z_DATA_ERROR;
+    default:
+      return Z_STREAM_ERROR;
+  }
+#ifdef NEED_DUMMY_RETURN
+  return Z_STREAM_ERROR;  /* Some dumb compilers complain without this */
+#endif
+}
+
+
+int ZEXPORT inflateSetDictionary(z, dictionary, dictLength)
+z_streamp z;
+const Bytef *dictionary;
+uInt  dictLength;
+{
+  uInt length = dictLength;
+
+  if (z == Z_NULL || z->state == Z_NULL || z->state->mode != DICT0)
+    return Z_STREAM_ERROR;
+
+  if (adler32(1L, dictionary, dictLength) != z->adler) return Z_DATA_ERROR;
+  z->adler = 1L;
+
+  if (length >= ((uInt)1<<z->state->wbits))
+  {
+    length = (1<<z->state->wbits)-1;
+    dictionary += dictLength - length;
+  }
+  inflate_set_dictionary(z->state->blocks, dictionary, length);
+  z->state->mode = BLOCKS;
+  return Z_OK;
+}
+
+
+int ZEXPORT inflateSync(z)
+z_streamp z;
+{
+  uInt n;       /* number of bytes to look at */
+  Bytef *p;     /* pointer to bytes */
+  uInt m;       /* number of marker bytes found in a row */
+  uLong r, w;   /* temporaries to save total_in and total_out */
+
+  /* set up */
+  if (z == Z_NULL || z->state == Z_NULL)
+    return Z_STREAM_ERROR;
+  if (z->state->mode != BAD)
+  {
+    z->state->mode = BAD;
+    z->state->sub.marker = 0;
+  }
+  if ((n = z->avail_in) == 0)
+    return Z_BUF_ERROR;
+  p = z->next_in;
+  m = z->state->sub.marker;
+
+  /* search */
+  while (n && m < 4)
+  {
+    static const Byte mark[4] = {0, 0, 0xff, 0xff};
+    if (*p == mark[m])
+      m++;
+    else if (*p)
+      m = 0;
+    else
+      m = 4 - m;
+    p++, n--;
+  }
+
+  /* restore */
+  z->total_in += p - z->next_in;
+  z->next_in = p;
+  z->avail_in = n;
+  z->state->sub.marker = m;
+
+  /* return no joy or set up to restart on a new block */
+  if (m != 4)
+    return Z_DATA_ERROR;
+  r = z->total_in;  w = z->total_out;
+  inflateReset(z);
+  z->total_in = r;  z->total_out = w;
+  z->state->mode = BLOCKS;
+  return Z_OK;
+}
+
+
+/* Returns true if inflate is currently at the end of a block generated
+ * by Z_SYNC_FLUSH or Z_FULL_FLUSH. This function is used by one PPP
+ * implementation to provide an additional safety check. PPP uses Z_SYNC_FLUSH
+ * but removes the length bytes of the resulting empty stored block. When
+ * decompressing, PPP checks that at the end of input packet, inflate is
+ * waiting for these length bytes.
+ */
+int ZEXPORT inflateSyncPoint(z)
+z_streamp z;
+{
+  if (z == Z_NULL || z->state == Z_NULL || z->state->blocks == Z_NULL)
+    return Z_STREAM_ERROR;
+  return inflate_blocks_sync_point(z->state->blocks);
+}
diff --git a/zlib/inftrees.c b/zlib/inftrees.c
new file mode 100644
index 000000000..4c32ca30d
--- /dev/null
+++ b/zlib/inftrees.c
@@ -0,0 +1,454 @@
+/* inftrees.c -- generate Huffman trees for efficient decoding
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+
+#if !defined(BUILDFIXED) && !defined(STDC)
+#  define BUILDFIXED   /* non ANSI compilers may not accept inffixed.h */
+#endif
+
+const char inflate_copyright[] =
+   " inflate 1.1.4 Copyright 1995-2002 Mark Adler ";
+/*
+  If you use the zlib library in a product, an acknowledgment is welcome
+  in the documentation of your product. If for some reason you cannot
+  include such an acknowledgment, I would appreciate that you keep this
+  copyright string in the executable of your product.
+ */
+struct internal_state  {int dummy;}; /* for buggy compilers */
+
+/* simplify the use of the inflate_huft type with some defines */
+#define exop word.what.Exop
+#define bits word.what.Bits
+
+
+local int huft_build OF((
+    uIntf *,            /* code lengths in bits */
+    uInt,               /* number of codes */
+    uInt,               /* number of "simple" codes */
+    const uIntf *,      /* list of base values for non-simple codes */
+    const uIntf *,      /* list of extra bits for non-simple codes */
+    inflate_huft * FAR*,/* result: starting table */
+    uIntf *,            /* maximum lookup bits (returns actual) */
+    inflate_huft *,     /* space for trees */
+    uInt *,             /* hufts used in space */
+    uIntf * ));         /* space for values */
+
+/* Tables for deflate from PKZIP's appnote.txt. */
+local const uInt cplens[31] = { /* Copy lengths for literal codes 257..285 */
+        3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
+        35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
+        /* see note #13 above about 258 */
+local const uInt cplext[31] = { /* Extra bits for literal codes 257..285 */
+        0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
+        3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 112, 112}; /* 112==invalid */
+local const uInt cpdist[30] = { /* Copy offsets for distance codes 0..29 */
+        1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
+        257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
+        8193, 12289, 16385, 24577};
+local const uInt cpdext[30] = { /* Extra bits for distance codes */
+        0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
+        7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
+        12, 12, 13, 13};
+
+/*
+   Huffman code decoding is performed using a multi-level table lookup.
+   The fastest way to decode is to simply build a lookup table whose
+   size is determined by the longest code.  However, the time it takes
+   to build this table can also be a factor if the data being decoded
+   is not very long.  The most common codes are necessarily the
+   shortest codes, so those codes dominate the decoding time, and hence
+   the speed.  The idea is you can have a shorter table that decodes the
+   shorter, more probable codes, and then point to subsidiary tables for
+   the longer codes.  The time it costs to decode the longer codes is
+   then traded against the time it takes to make longer tables.
+
+   This results of this trade are in the variables lbits and dbits
+   below.  lbits is the number of bits the first level table for literal/
+   length codes can decode in one step, and dbits is the same thing for
+   the distance codes.  Subsequent tables are also less than or equal to
+   those sizes.  These values may be adjusted either when all of the
+   codes are shorter than that, in which case the longest code length in
+   bits is used, or when the shortest code is *longer* than the requested
+   table size, in which case the length of the shortest code in bits is
+   used.
+
+   There are two different values for the two tables, since they code a
+   different number of possibilities each.  The literal/length table
+   codes 286 possible values, or in a flat code, a little over eight
+   bits.  The distance table codes 30 possible values, or a little less
+   than five bits, flat.  The optimum values for speed end up being
+   about one bit more than those, so lbits is 8+1 and dbits is 5+1.
+   The optimum values may differ though from machine to machine, and
+   possibly even between compilers.  Your mileage may vary.
+ */
+
+
+/* If BMAX needs to be larger than 16, then h and x[] should be uLong. */
+#define BMAX 15         /* maximum bit length of any code */
+
+local int huft_build(b, n, s, d, e, t, m, hp, hn, v)
+uIntf *b;               /* code lengths in bits (all assumed <= BMAX) */
+uInt n;                 /* number of codes (assumed <= 288) */
+uInt s;                 /* number of simple-valued codes (0..s-1) */
+const uIntf *d;         /* list of base values for non-simple codes */
+const uIntf *e;         /* list of extra bits for non-simple codes */
+inflate_huft * FAR *t;  /* result: starting table */
+uIntf *m;               /* maximum lookup bits, returns actual */
+inflate_huft *hp;       /* space for trees */
+uInt *hn;               /* hufts used in space */
+uIntf *v;               /* working area: values in order of bit length */
+/* Given a list of code lengths and a maximum table size, make a set of
+   tables to decode that set of codes.  Return Z_OK on success, Z_BUF_ERROR
+   if the given code set is incomplete (the tables are still built in this
+   case), or Z_DATA_ERROR if the input is invalid. */
+{
+
+  uInt a;                       /* counter for codes of length k */
+  uInt c[BMAX+1];               /* bit length count table */
+  uInt f;                       /* i repeats in table every f entries */
+  int g;                        /* maximum code length */
+  int h;                        /* table level */
+  register uInt i;              /* counter, current code */
+  register uInt j;              /* counter */
+  register int k;               /* number of bits in current code */
+  int l;                        /* bits per table (returned in m) */
+  uInt mask;                    /* (1 << w) - 1, to avoid cc -O bug on HP */
+  register uIntf *p;            /* pointer into c[], b[], or v[] */
+  inflate_huft *q;              /* points to current table */
+  struct inflate_huft_s r;      /* table entry for structure assignment */
+  inflate_huft *u[BMAX];        /* table stack */
+  register int w;               /* bits before this table == (l * h) */
+  uInt x[BMAX+1];               /* bit offsets, then code stack */
+  uIntf *xp;                    /* pointer into x */
+  int y;                        /* number of dummy codes added */
+  uInt z;                       /* number of entries in current table */
+
+
+  /* Generate counts for each bit length */
+  p = c;
+#define C0 *p++ = 0;
+#define C2 C0 C0 C0 C0
+#define C4 C2 C2 C2 C2
+  C4                            /* clear c[]--assume BMAX+1 is 16 */
+  p = b;  i = n;
+  do {
+    c[*p++]++;                  /* assume all entries <= BMAX */
+  } while (--i);
+  if (c[0] == n)                /* null input--all zero length codes */
+  {
+    *t = (inflate_huft *)Z_NULL;
+    *m = 0;
+    return Z_OK;
+  }
+
+
+  /* Find minimum and maximum length, bound *m by those */
+  l = *m;
+  for (j = 1; j <= BMAX; j++)
+    if (c[j])
+      break;
+  k = j;                        /* minimum code length */
+  if ((uInt)l < j)
+    l = j;
+  for (i = BMAX; i; i--)
+    if (c[i])
+      break;
+  g = i;                        /* maximum code length */
+  if ((uInt)l > i)
+    l = i;
+  *m = l;
+
+
+  /* Adjust last length count to fill out codes, if needed */
+  for (y = 1 << j; j < i; j++, y <<= 1)
+    if ((y -= c[j]) < 0)
+      return Z_DATA_ERROR;
+  if ((y -= c[i]) < 0)
+    return Z_DATA_ERROR;
+  c[i] += y;
+
+
+  /* Generate starting offsets into the value table for each length */
+  x[1] = j = 0;
+  p = c + 1;  xp = x + 2;
+  while (--i) {                 /* note that i == g from above */
+    *xp++ = (j += *p++);
+  }
+
+
+  /* Make a table of values in order of bit lengths */
+  p = b;  i = 0;
+  do {
+    if ((j = *p++) != 0)
+      v[x[j]++] = i;
+  } while (++i < n);
+  n = x[g];                     /* set n to length of v */
+
+
+  /* Generate the Huffman codes and for each, make the table entries */
+  x[0] = i = 0;                 /* first Huffman code is zero */
+  p = v;                        /* grab values in bit order */
+  h = -1;                       /* no tables yet--level -1 */
+  w = -l;                       /* bits decoded == (l * h) */
+  u[0] = (inflate_huft *)Z_NULL;        /* just to keep compilers happy */
+  q = (inflate_huft *)Z_NULL;   /* ditto */
+  z = 0;                        /* ditto */
+
+  /* go through the bit lengths (k already is bits in shortest code) */
+  for (; k <= g; k++)
+  {
+    a = c[k];
+    while (a--)
+    {
+      /* here i is the Huffman code of length k bits for value *p */
+      /* make tables up to required level */
+      while (k > w + l)
+      {
+        h++;
+        w += l;                 /* previous table always l bits */
+
+        /* compute minimum size table less than or equal to l bits */
+        z = g - w;
+        z = z > (uInt)l ? l : z;        /* table size upper limit */
+        if ((f = 1 << (j = k - w)) > a + 1)     /* try a k-w bit table */
+        {                       /* too few codes for k-w bit table */
+          f -= a + 1;           /* deduct codes from patterns left */
+          xp = c + k;
+          if (j < z)
+            while (++j < z)     /* try smaller tables up to z bits */
+            {
+              if ((f <<= 1) <= *++xp)
+                break;          /* enough codes to use up j bits */
+              f -= *xp;         /* else deduct codes from patterns */
+            }
+        }
+        z = 1 << j;             /* table entries for j-bit table */
+
+        /* allocate new table */
+        if (*hn + z > MANY)     /* (note: doesn't matter for fixed) */
+          return Z_DATA_ERROR;  /* overflow of MANY */
+        u[h] = q = hp + *hn;
+        *hn += z;
+
+        /* connect to last table, if there is one */
+        if (h)
+        {
+          x[h] = i;             /* save pattern for backing up */
+          r.bits = (Byte)l;     /* bits to dump before this table */
+          r.exop = (Byte)j;     /* bits in this table */
+          j = i >> (w - l);
+          r.base = (uInt)(q - u[h-1] - j);   /* offset to this table */
+          u[h-1][j] = r;        /* connect to last table */
+        }
+        else
+          *t = q;               /* first table is returned result */
+      }
+
+      /* set up table entry in r */
+      r.bits = (Byte)(k - w);
+      if (p >= v + n)
+        r.exop = 128 + 64;      /* out of values--invalid code */
+      else if (*p < s)
+      {
+        r.exop = (Byte)(*p < 256 ? 0 : 32 + 64);     /* 256 is end-of-block */
+        r.base = *p++;          /* simple code is just the value */
+      }
+      else
+      {
+        r.exop = (Byte)(e[*p - s] + 16 + 64);/* non-simple--look up in lists */
+        r.base = d[*p++ - s];
+      }
+
+      /* fill code-like entries with r */
+      f = 1 << (k - w);
+      for (j = i >> w; j < z; j += f)
+        q[j] = r;
+
+      /* backwards increment the k-bit code i */
+      for (j = 1 << (k - 1); i & j; j >>= 1)
+        i ^= j;
+      i ^= j;
+
+      /* backup over finished tables */
+      mask = (1 << w) - 1;      /* needed on HP, cc -O bug */
+      while ((i & mask) != x[h])
+      {
+        h--;                    /* don't need to update q */
+        w -= l;
+        mask = (1 << w) - 1;
+      }
+    }
+  }
+
+
+  /* Return Z_BUF_ERROR if we were given an incomplete table */
+  return y != 0 && g != 1 ? Z_BUF_ERROR : Z_OK;
+}
+
+
+int inflate_trees_bits(c, bb, tb, hp, z)
+uIntf *c;               /* 19 code lengths */
+uIntf *bb;              /* bits tree desired/actual depth */
+inflate_huft * FAR *tb; /* bits tree result */
+inflate_huft *hp;       /* space for trees */
+z_streamp z;            /* for messages */
+{
+  int r;
+  uInt hn = 0;          /* hufts used in space */
+  uIntf *v;             /* work area for huft_build */
+
+  if ((v = (uIntf*)ZALLOC(z, 19, sizeof(uInt))) == Z_NULL)
+    return Z_MEM_ERROR;
+  r = huft_build(c, 19, 19, (uIntf*)Z_NULL, (uIntf*)Z_NULL,
+                 tb, bb, hp, &hn, v);
+  if (r == Z_DATA_ERROR)
+    z->msg = (char*)"oversubscribed dynamic bit lengths tree";
+  else if (r == Z_BUF_ERROR || *bb == 0)
+  {
+    z->msg = (char*)"incomplete dynamic bit lengths tree";
+    r = Z_DATA_ERROR;
+  }
+  ZFREE(z, v);
+  return r;
+}
+
+
+int inflate_trees_dynamic(nl, nd, c, bl, bd, tl, td, hp, z)
+uInt nl;                /* number of literal/length codes */
+uInt nd;                /* number of distance codes */
+uIntf *c;               /* that many (total) code lengths */
+uIntf *bl;              /* literal desired/actual bit depth */
+uIntf *bd;              /* distance desired/actual bit depth */
+inflate_huft * FAR *tl; /* literal/length tree result */
+inflate_huft * FAR *td; /* distance tree result */
+inflate_huft *hp;       /* space for trees */
+z_streamp z;            /* for messages */
+{
+  int r;
+  uInt hn = 0;          /* hufts used in space */
+  uIntf *v;             /* work area for huft_build */
+
+  /* allocate work area */
+  if ((v = (uIntf*)ZALLOC(z, 288, sizeof(uInt))) == Z_NULL)
+    return Z_MEM_ERROR;
+
+  /* build literal/length tree */
+  r = huft_build(c, nl, 257, cplens, cplext, tl, bl, hp, &hn, v);
+  if (r != Z_OK || *bl == 0)
+  {
+    if (r == Z_DATA_ERROR)
+      z->msg = (char*)"oversubscribed literal/length tree";
+    else if (r != Z_MEM_ERROR)
+    {
+      z->msg = (char*)"incomplete literal/length tree";
+      r = Z_DATA_ERROR;
+    }
+    ZFREE(z, v);
+    return r;
+  }
+
+  /* build distance tree */
+  r = huft_build(c + nl, nd, 0, cpdist, cpdext, td, bd, hp, &hn, v);
+  if (r != Z_OK || (*bd == 0 && nl > 257))
+  {
+    if (r == Z_DATA_ERROR)
+      z->msg = (char*)"oversubscribed distance tree";
+    else if (r == Z_BUF_ERROR) {
+#ifdef PKZIP_BUG_WORKAROUND
+      r = Z_OK;
+    }
+#else
+      z->msg = (char*)"incomplete distance tree";
+      r = Z_DATA_ERROR;
+    }
+    else if (r != Z_MEM_ERROR)
+    {
+      z->msg = (char*)"empty distance tree with lengths";
+      r = Z_DATA_ERROR;
+    }
+    ZFREE(z, v);
+    return r;
+#endif
+  }
+
+  /* done */
+  ZFREE(z, v);
+  return Z_OK;
+}
+
+
+/* build fixed tables only once--keep them here */
+#ifdef BUILDFIXED
+local int fixed_built = 0;
+#define FIXEDH 544      /* number of hufts used by fixed tables */
+local inflate_huft fixed_mem[FIXEDH];
+local uInt fixed_bl;
+local uInt fixed_bd;
+local inflate_huft *fixed_tl;
+local inflate_huft *fixed_td;
+#else
+#include "inffixed.h"
+#endif
+
+
+int inflate_trees_fixed(bl, bd, tl, td, z)
+uIntf *bl;               /* literal desired/actual bit depth */
+uIntf *bd;               /* distance desired/actual bit depth */
+inflate_huft * FAR *tl;  /* literal/length tree result */
+inflate_huft * FAR *td;  /* distance tree result */
+z_streamp z;             /* for memory allocation */
+{
+#ifdef BUILDFIXED
+  /* build fixed tables if not already */
+  if (!fixed_built)
+  {
+    int k;              /* temporary variable */
+    uInt f = 0;         /* number of hufts used in fixed_mem */
+    uIntf *c;           /* length list for huft_build */
+    uIntf *v;           /* work area for huft_build */
+
+    /* allocate memory */
+    if ((c = (uIntf*)ZALLOC(z, 288, sizeof(uInt))) == Z_NULL)
+      return Z_MEM_ERROR;
+    if ((v = (uIntf*)ZALLOC(z, 288, sizeof(uInt))) == Z_NULL)
+    {
+      ZFREE(z, c);
+      return Z_MEM_ERROR;
+    }
+
+    /* literal table */
+    for (k = 0; k < 144; k++)
+      c[k] = 8;
+    for (; k < 256; k++)
+      c[k] = 9;
+    for (; k < 280; k++)
+      c[k] = 7;
+    for (; k < 288; k++)
+      c[k] = 8;
+    fixed_bl = 9;
+    huft_build(c, 288, 257, cplens, cplext, &fixed_tl, &fixed_bl,
+               fixed_mem, &f, v);
+
+    /* distance table */
+    for (k = 0; k < 30; k++)
+      c[k] = 5;
+    fixed_bd = 5;
+    huft_build(c, 30, 0, cpdist, cpdext, &fixed_td, &fixed_bd,
+               fixed_mem, &f, v);
+
+    /* done */
+    ZFREE(z, v);
+    ZFREE(z, c);
+    fixed_built = 1;
+  }
+#endif
+  *bl = fixed_bl;
+  *bd = fixed_bd;
+  *tl = fixed_tl;
+  *td = fixed_td;
+  return Z_OK;
+}
diff --git a/zlib/inftrees.h b/zlib/inftrees.h
new file mode 100644
index 000000000..04b73b729
--- /dev/null
+++ b/zlib/inftrees.h
@@ -0,0 +1,58 @@
+/* inftrees.h -- header to use inftrees.c
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* Huffman code lookup table entry--this entry is four bytes for machines
+   that have 16-bit pointers (e.g. PC's in the small or medium model). */
+
+typedef struct inflate_huft_s FAR inflate_huft;
+
+struct inflate_huft_s {
+  union {
+    struct {
+      Byte Exop;        /* number of extra bits or operation */
+      Byte Bits;        /* number of bits in this code or subcode */
+    } what;
+    uInt pad;           /* pad structure to a power of 2 (4 bytes for */
+  } word;               /*  16-bit, 8 bytes for 32-bit int's) */
+  uInt base;            /* literal, length base, distance base,
+                           or table offset */
+};
+
+/* Maximum size of dynamic tree.  The maximum found in a long but non-
+   exhaustive search was 1004 huft structures (850 for length/literals
+   and 154 for distances, the latter actually the result of an
+   exhaustive search).  The actual maximum is not known, but the
+   value below is more than safe. */
+#define MANY 1440
+
+extern int inflate_trees_bits OF((
+    uIntf *,                    /* 19 code lengths */
+    uIntf *,                    /* bits tree desired/actual depth */
+    inflate_huft * FAR *,       /* bits tree result */
+    inflate_huft *,             /* space for trees */
+    z_streamp));                /* for messages */
+
+extern int inflate_trees_dynamic OF((
+    uInt,                       /* number of literal/length codes */
+    uInt,                       /* number of distance codes */
+    uIntf *,                    /* that many (total) code lengths */
+    uIntf *,                    /* literal desired/actual bit depth */
+    uIntf *,                    /* distance desired/actual bit depth */
+    inflate_huft * FAR *,       /* literal/length tree result */
+    inflate_huft * FAR *,       /* distance tree result */
+    inflate_huft *,             /* space for trees */
+    z_streamp));                /* for messages */
+
+extern int inflate_trees_fixed OF((
+    uIntf *,                    /* literal desired/actual bit depth */
+    uIntf *,                    /* distance desired/actual bit depth */
+    inflate_huft * FAR *,       /* literal/length tree result */
+    inflate_huft * FAR *,       /* distance tree result */
+    z_streamp));                /* for memory allocation */
diff --git a/zlib/infutil.c b/zlib/infutil.c
new file mode 100644
index 000000000..9a076221f
--- /dev/null
+++ b/zlib/infutil.c
@@ -0,0 +1,87 @@
+/* inflate_util.c -- data and routines common to blocks and codes
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+#include "zutil.h"
+#include "infblock.h"
+#include "inftrees.h"
+#include "infcodes.h"
+#include "infutil.h"
+
+struct inflate_codes_state {int dummy;}; /* for buggy compilers */
+
+/* And'ing with mask[n] masks the lower n bits */
+uInt inflate_mask[17] = {
+    0x0000,
+    0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
+    0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
+};
+
+
+/* copy as much as possible from the sliding window to the output area */
+int inflate_flush(s, z, r)
+inflate_blocks_statef *s;
+z_streamp z;
+int r;
+{
+  uInt n;
+  Bytef *p;
+  Bytef *q;
+
+  /* local copies of source and destination pointers */
+  p = z->next_out;
+  q = s->read;
+
+  /* compute number of bytes to copy as far as end of window */
+  n = (uInt)((q <= s->write ? s->write : s->end) - q);
+  if (n > z->avail_out) n = z->avail_out;
+  if (n && r == Z_BUF_ERROR) r = Z_OK;
+
+  /* update counters */
+  z->avail_out -= n;
+  z->total_out += n;
+
+  /* update check information */
+  if (s->checkfn != Z_NULL)
+    z->adler = s->check = (*s->checkfn)(s->check, q, n);
+
+  /* copy as far as end of window */
+  zmemcpy(p, q, n);
+  p += n;
+  q += n;
+
+  /* see if more to copy at beginning of window */
+  if (q == s->end)
+  {
+    /* wrap pointers */
+    q = s->window;
+    if (s->write == s->end)
+      s->write = s->window;
+
+    /* compute bytes to copy */
+    n = (uInt)(s->write - q);
+    if (n > z->avail_out) n = z->avail_out;
+    if (n && r == Z_BUF_ERROR) r = Z_OK;
+
+    /* update counters */
+    z->avail_out -= n;
+    z->total_out += n;
+
+    /* update check information */
+    if (s->checkfn != Z_NULL)
+      z->adler = s->check = (*s->checkfn)(s->check, q, n);
+
+    /* copy */
+    zmemcpy(p, q, n);
+    p += n;
+    q += n;
+  }
+
+  /* update pointers */
+  z->next_out = p;
+  s->read = q;
+
+  /* done */
+  return r;
+}
diff --git a/zlib/infutil.h b/zlib/infutil.h
new file mode 100644
index 000000000..4401df82f
--- /dev/null
+++ b/zlib/infutil.h
@@ -0,0 +1,98 @@
+/* infutil.h -- types and macros common to blocks and codes
+ * Copyright (C) 1995-2002 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+#ifndef _INFUTIL_H
+#define _INFUTIL_H
+
+typedef enum {
+      TYPE,     /* get type bits (3, including end bit) */
+      LENS,     /* get lengths for stored */
+      STORED,   /* processing stored block */
+      TABLE,    /* get table lengths */
+      BTREE,    /* get bit lengths tree for a dynamic block */
+      DTREE,    /* get length, distance trees for a dynamic block */
+      CODES,    /* processing fixed or dynamic block */
+      DRY,      /* output remaining window bytes */
+      DONE,     /* finished last block, done */
+      BAD}      /* got a data error--stuck here */
+inflate_block_mode;
+
+/* inflate blocks semi-private state */
+struct inflate_blocks_state {
+
+  /* mode */
+  inflate_block_mode  mode;     /* current inflate_block mode */
+
+  /* mode dependent information */
+  union {
+    uInt left;          /* if STORED, bytes left to copy */
+    struct {
+      uInt table;               /* table lengths (14 bits) */
+      uInt index;               /* index into blens (or border) */
+      uIntf *blens;             /* bit lengths of codes */
+      uInt bb;                  /* bit length tree depth */
+      inflate_huft *tb;         /* bit length decoding tree */
+    } trees;            /* if DTREE, decoding info for trees */
+    struct {
+      inflate_codes_statef 
+         *codes;
+    } decode;           /* if CODES, current state */
+  } sub;                /* submode */
+  uInt last;            /* true if this block is the last block */
+
+  /* mode independent information */
+  uInt bitk;            /* bits in bit buffer */
+  uLong bitb;           /* bit buffer */
+  inflate_huft *hufts;  /* single malloc for tree space */
+  Bytef *window;        /* sliding window */
+  Bytef *end;           /* one byte after sliding window */
+  Bytef *read;          /* window read pointer */
+  Bytef *write;         /* window write pointer */
+  check_func checkfn;   /* check function */
+  uLong check;          /* check on output */
+
+};
+
+
+/* defines for inflate input/output */
+/*   update pointers and return */
+#define UPDBITS {s->bitb=b;s->bitk=k;}
+#define UPDIN {z->avail_in=n;z->total_in+=p-z->next_in;z->next_in=p;}
+#define UPDOUT {s->write=q;}
+#define UPDATE {UPDBITS UPDIN UPDOUT}
+#define LEAVE {UPDATE return inflate_flush(s,z,r);}
+/*   get bytes and bits */
+#define LOADIN {p=z->next_in;n=z->avail_in;b=s->bitb;k=s->bitk;}
+#define NEEDBYTE {if(n)r=Z_OK;else LEAVE}
+#define NEXTBYTE (n--,*p++)
+#define NEEDBITS(j) {while(k<(j)){NEEDBYTE;b|=((uLong)NEXTBYTE)<<k;k+=8;}}
+#define DUMPBITS(j) {b>>=(j);k-=(j);}
+/*   output bytes */
+#define WAVAIL (uInt)(q<s->read?s->read-q-1:s->end-q)
+#define LOADOUT {q=s->write;m=(uInt)WAVAIL;}
+#define WRAP {if(q==s->end&&s->read!=s->window){q=s->window;m=(uInt)WAVAIL;}}
+#define FLUSH {UPDOUT r=inflate_flush(s,z,r); LOADOUT}
+#define NEEDOUT {if(m==0){WRAP if(m==0){FLUSH WRAP if(m==0) LEAVE}}r=Z_OK;}
+#define OUTBYTE(a) {*q++=(Byte)(a);m--;}
+/*   load local pointers */
+#define LOAD {LOADIN LOADOUT}
+
+/* masks for lower bits (size given to avoid silly warnings with Visual C++) */
+extern uInt inflate_mask[17];
+
+/* copy as much as possible from the sliding window to the output area */
+extern int inflate_flush OF((
+    inflate_blocks_statef *,
+    z_streamp ,
+    int));
+
+struct internal_state      {int dummy;}; /* for buggy compilers */
+
+#endif
diff --git a/zlib/makedepend b/zlib/makedepend
new file mode 100644
index 000000000..b5c4a9f96
--- /dev/null
+++ b/zlib/makedepend
@@ -0,0 +1,18 @@
+# DO NOT DELETE
+
+adler32.o: zlib.h zconf.h
+compress.o: zlib.h zconf.h
+crc32.o: zlib.h zconf.h
+gzio.o: zutil.h zlib.h zconf.h
+uncompr.o: zlib.h zconf.h
+deflate.o: deflate.h zutil.h zlib.h zconf.h
+trees.o: deflate.h zutil.h zlib.h zconf.h trees.h
+zutil.o: zutil.h zlib.h zconf.h
+inflate.o: zutil.h zlib.h zconf.h infblock.h
+infblock.o: zutil.h zlib.h zconf.h infblock.h inftrees.h infcodes.h infutil.h
+inftrees.o: zutil.h zlib.h zconf.h inftrees.h inffixed.h
+infcodes.o: zutil.h zlib.h zconf.h inftrees.h infblock.h infcodes.h infutil.h
+infcodes.o: inffast.h
+infutil.o: zutil.h zlib.h zconf.h infblock.h inftrees.h infcodes.h infutil.h
+inffast.o: zutil.h zlib.h zconf.h inftrees.h infblock.h infcodes.h infutil.h
+inffast.o: inffast.h
diff --git a/zlib/trees.c b/zlib/trees.c
new file mode 100644
index 000000000..88af872b1
--- /dev/null
+++ b/zlib/trees.c
@@ -0,0 +1,1214 @@
+/* trees.c -- output deflated data using Huffman coding
+ * Copyright (C) 1995-2002 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/*
+ *  ALGORITHM
+ *
+ *      The "deflation" process uses several Huffman trees. The more
+ *      common source values are represented by shorter bit sequences.
+ *
+ *      Each code tree is stored in a compressed form which is itself
+ * a Huffman encoding of the lengths of all the code strings (in
+ * ascending order by source values).  The actual code strings are
+ * reconstructed from the lengths in the inflate process, as described
+ * in the deflate specification.
+ *
+ *  REFERENCES
+ *
+ *      Deutsch, L.P.,"'Deflate' Compressed Data Format Specification".
+ *      Available in ftp.uu.net:/pub/archiving/zip/doc/deflate-1.1.doc
+ *
+ *      Storer, James A.
+ *          Data Compression:  Methods and Theory, pp. 49-50.
+ *          Computer Science Press, 1988.  ISBN 0-7167-8156-5.
+ *
+ *      Sedgewick, R.
+ *          Algorithms, p290.
+ *          Addison-Wesley, 1983. ISBN 0-201-06672-6.
+ */
+
+/* @(#) $Id: trees.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+/* #define GEN_TREES_H */
+
+#include "deflate.h"
+
+#ifdef DEBUG
+#  include <ctype.h>
+#endif
+
+/* ===========================================================================
+ * Constants
+ */
+
+#define MAX_BL_BITS 7
+/* Bit length codes must not exceed MAX_BL_BITS bits */
+
+#define END_BLOCK 256
+/* end of block literal code */
+
+#define REP_3_6      16
+/* repeat previous bit length 3-6 times (2 bits of repeat count) */
+
+#define REPZ_3_10    17
+/* repeat a zero length 3-10 times  (3 bits of repeat count) */
+
+#define REPZ_11_138  18
+/* repeat a zero length 11-138 times  (7 bits of repeat count) */
+
+local const int extra_lbits[LENGTH_CODES] /* extra bits for each length code */
+   = {0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0};
+
+local const int extra_dbits[D_CODES] /* extra bits for each distance code */
+   = {0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
+
+local const int extra_blbits[BL_CODES]/* extra bits for each bit length code */
+   = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,3,7};
+
+local const uch bl_order[BL_CODES]
+   = {16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15};
+/* The lengths of the bit length codes are sent in order of decreasing
+ * probability, to avoid transmitting the lengths for unused bit length codes.
+ */
+
+#define Buf_size (8 * 2*sizeof(char))
+/* Number of bits used within bi_buf. (bi_buf might be implemented on
+ * more than 16 bits on some systems.)
+ */
+
+/* ===========================================================================
+ * Local data. These are initialized only once.
+ */
+
+#define DIST_CODE_LEN  512 /* see definition of array dist_code below */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+/* non ANSI compilers may not accept trees.h */
+
+local ct_data static_ltree[L_CODES+2];
+/* The static literal tree. Since the bit lengths are imposed, there is no
+ * need for the L_CODES extra codes used during heap construction. However
+ * The codes 286 and 287 are needed to build a canonical tree (see _tr_init
+ * below).
+ */
+
+local ct_data static_dtree[D_CODES];
+/* The static distance tree. (Actually a trivial tree since all codes use
+ * 5 bits.)
+ */
+
+uch _dist_code[DIST_CODE_LEN];
+/* Distance codes. The first 256 values correspond to the distances
+ * 3 .. 258, the last 256 values correspond to the top 8 bits of
+ * the 15 bit distances.
+ */
+
+uch _length_code[MAX_MATCH-MIN_MATCH+1];
+/* length code for each normalized match length (0 == MIN_MATCH) */
+
+local int base_length[LENGTH_CODES];
+/* First normalized length for each code (0 = MIN_MATCH) */
+
+local int base_dist[D_CODES];
+/* First normalized distance for each code (0 = distance of 1) */
+
+#else
+#  include "trees.h"
+#endif /* GEN_TREES_H */
+
+struct static_tree_desc_s {
+    const ct_data *static_tree;  /* static tree or NULL */
+    const intf *extra_bits;      /* extra bits for each code or NULL */
+    int     extra_base;          /* base index for extra_bits */
+    int     elems;               /* max number of elements in the tree */
+    int     max_length;          /* max bit length for the codes */
+};
+
+local static_tree_desc  static_l_desc =
+{static_ltree, extra_lbits, LITERALS+1, L_CODES, MAX_BITS};
+
+local static_tree_desc  static_d_desc =
+{static_dtree, extra_dbits, 0,          D_CODES, MAX_BITS};
+
+local static_tree_desc  static_bl_desc =
+{(const ct_data *)0, extra_blbits, 0,   BL_CODES, MAX_BL_BITS};
+
+/* ===========================================================================
+ * Local (static) routines in this file.
+ */
+
+local void tr_static_init OF((void));
+local void init_block     OF((deflate_state *s));
+local void pqdownheap     OF((deflate_state *s, ct_data *tree, int k));
+local void gen_bitlen     OF((deflate_state *s, tree_desc *desc));
+local void gen_codes      OF((ct_data *tree, int max_code, ushf *bl_count));
+local void build_tree     OF((deflate_state *s, tree_desc *desc));
+local void scan_tree      OF((deflate_state *s, ct_data *tree, int max_code));
+local void send_tree      OF((deflate_state *s, ct_data *tree, int max_code));
+local int  build_bl_tree  OF((deflate_state *s));
+local void send_all_trees OF((deflate_state *s, int lcodes, int dcodes,
+                              int blcodes));
+local void compress_block OF((deflate_state *s, ct_data *ltree,
+                              ct_data *dtree));
+local void set_data_type  OF((deflate_state *s));
+local unsigned bi_reverse OF((unsigned value, int length));
+local void bi_windup      OF((deflate_state *s));
+local void bi_flush       OF((deflate_state *s));
+local void copy_block     OF((deflate_state *s, charf *buf, unsigned len,
+                              int header));
+
+#ifdef GEN_TREES_H
+local void gen_trees_header OF((void));
+#endif
+
+#ifndef DEBUG
+#  define send_code(s, c, tree) send_bits(s, tree[c].Code, tree[c].Len)
+   /* Send a code of the given tree. c and tree must not have side effects */
+
+#else /* DEBUG */
+#  define send_code(s, c, tree) \
+     { if (z_verbose>2) fprintf(stderr,"\ncd %3d ",(c)); \
+       send_bits(s, tree[c].Code, tree[c].Len); }
+#endif
+
+/* ===========================================================================
+ * Output a short LSB first on the stream.
+ * IN assertion: there is enough room in pendingBuf.
+ */
+#define put_short(s, w) { \
+    put_byte(s, (uch)((w) & 0xff)); \
+    put_byte(s, (uch)((ush)(w) >> 8)); \
+}
+
+/* ===========================================================================
+ * Send a value on a given number of bits.
+ * IN assertion: length <= 16 and value fits in length bits.
+ */
+#ifdef DEBUG
+local void send_bits      OF((deflate_state *s, int value, int length));
+
+local void send_bits(s, value, length)
+    deflate_state *s;
+    int value;  /* value to send */
+    int length; /* number of bits */
+{
+    Tracevv((stderr," l %2d v %4x ", length, value));
+    Assert(length > 0 && length <= 15, "invalid length");
+    s->bits_sent += (ulg)length;
+
+    /* If not enough room in bi_buf, use (valid) bits from bi_buf and
+     * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
+     * unused bits in value.
+     */
+    if (s->bi_valid > (int)Buf_size - length) {
+        s->bi_buf |= (value << s->bi_valid);
+        put_short(s, s->bi_buf);
+        s->bi_buf = (ush)value >> (Buf_size - s->bi_valid);
+        s->bi_valid += length - Buf_size;
+    } else {
+        s->bi_buf |= value << s->bi_valid;
+        s->bi_valid += length;
+    }
+}
+#else /* !DEBUG */
+
+#define send_bits(s, value, length) \
+{ int len = length;\
+  if (s->bi_valid > (int)Buf_size - len) {\
+    int val = value;\
+    s->bi_buf |= (val << s->bi_valid);\
+    put_short(s, s->bi_buf);\
+    s->bi_buf = (ush)val >> (Buf_size - s->bi_valid);\
+    s->bi_valid += len - Buf_size;\
+  } else {\
+    s->bi_buf |= (value) << s->bi_valid;\
+    s->bi_valid += len;\
+  }\
+}
+#endif /* DEBUG */
+
+
+#define MAX(a,b) (a >= b ? a : b)
+/* the arguments must not have side effects */
+
+/* ===========================================================================
+ * Initialize the various 'constant' tables.
+ */
+local void tr_static_init()
+{
+#if defined(GEN_TREES_H) || !defined(STDC)
+    static int static_init_done = 0;
+    int n;        /* iterates over tree elements */
+    int bits;     /* bit counter */
+    int length;   /* length value */
+    int code;     /* code value */
+    int dist;     /* distance index */
+    ush bl_count[MAX_BITS+1];
+    /* number of codes at each bit length for an optimal tree */
+
+    if (static_init_done) return;
+
+    /* For some embedded targets, global variables are not initialized: */
+    static_l_desc.static_tree = static_ltree;
+    static_l_desc.extra_bits = extra_lbits;
+    static_d_desc.static_tree = static_dtree;
+    static_d_desc.extra_bits = extra_dbits;
+    static_bl_desc.extra_bits = extra_blbits;
+
+    /* Initialize the mapping length (0..255) -> length code (0..28) */
+    length = 0;
+    for (code = 0; code < LENGTH_CODES-1; code++) {
+        base_length[code] = length;
+        for (n = 0; n < (1<<extra_lbits[code]); n++) {
+            _length_code[length++] = (uch)code;
+        }
+    }
+    Assert (length == 256, "tr_static_init: length != 256");
+    /* Note that the length 255 (match length 258) can be represented
+     * in two different ways: code 284 + 5 bits or code 285, so we
+     * overwrite length_code[255] to use the best encoding:
+     */
+    _length_code[length-1] = (uch)code;
+
+    /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
+    dist = 0;
+    for (code = 0 ; code < 16; code++) {
+        base_dist[code] = dist;
+        for (n = 0; n < (1<<extra_dbits[code]); n++) {
+            _dist_code[dist++] = (uch)code;
+        }
+    }
+    Assert (dist == 256, "tr_static_init: dist != 256");
+    dist >>= 7; /* from now on, all distances are divided by 128 */
+    for ( ; code < D_CODES; code++) {
+        base_dist[code] = dist << 7;
+        for (n = 0; n < (1<<(extra_dbits[code]-7)); n++) {
+            _dist_code[256 + dist++] = (uch)code;
+        }
+    }
+    Assert (dist == 256, "tr_static_init: 256+dist != 512");
+
+    /* Construct the codes of the static literal tree */
+    for (bits = 0; bits <= MAX_BITS; bits++) bl_count[bits] = 0;
+    n = 0;
+    while (n <= 143) static_ltree[n++].Len = 8, bl_count[8]++;
+    while (n <= 255) static_ltree[n++].Len = 9, bl_count[9]++;
+    while (n <= 279) static_ltree[n++].Len = 7, bl_count[7]++;
+    while (n <= 287) static_ltree[n++].Len = 8, bl_count[8]++;
+    /* Codes 286 and 287 do not exist, but we must include them in the
+     * tree construction to get a canonical Huffman tree (longest code
+     * all ones)
+     */
+    gen_codes((ct_data *)static_ltree, L_CODES+1, bl_count);
+
+    /* The static distance tree is trivial: */
+    for (n = 0; n < D_CODES; n++) {
+        static_dtree[n].Len = 5;
+        static_dtree[n].Code = bi_reverse((unsigned)n, 5);
+    }
+    static_init_done = 1;
+
+#  ifdef GEN_TREES_H
+    gen_trees_header();
+#  endif
+#endif /* defined(GEN_TREES_H) || !defined(STDC) */
+}
+
+/* ===========================================================================
+ * Genererate the file trees.h describing the static trees.
+ */
+#ifdef GEN_TREES_H
+#  ifndef DEBUG
+#    include <stdio.h>
+#  endif
+
+#  define SEPARATOR(i, last, width) \
+      ((i) == (last)? "\n};\n\n" :    \
+       ((i) % (width) == (width)-1 ? ",\n" : ", "))
+
+void gen_trees_header()
+{
+    FILE *header = fopen("trees.h", "w");
+    int i;
+
+    Assert (header != NULL, "Can't open trees.h");
+    fprintf(header,
+	    "/* header created automatically with -DGEN_TREES_H */\n\n");
+
+    fprintf(header, "local const ct_data static_ltree[L_CODES+2] = {\n");
+    for (i = 0; i < L_CODES+2; i++) {
+	fprintf(header, "{{%3u},{%3u}}%s", static_ltree[i].Code,
+		static_ltree[i].Len, SEPARATOR(i, L_CODES+1, 5));
+    }
+
+    fprintf(header, "local const ct_data static_dtree[D_CODES] = {\n");
+    for (i = 0; i < D_CODES; i++) {
+	fprintf(header, "{{%2u},{%2u}}%s", static_dtree[i].Code,
+		static_dtree[i].Len, SEPARATOR(i, D_CODES-1, 5));
+    }
+
+    fprintf(header, "const uch _dist_code[DIST_CODE_LEN] = {\n");
+    for (i = 0; i < DIST_CODE_LEN; i++) {
+	fprintf(header, "%2u%s", _dist_code[i],
+		SEPARATOR(i, DIST_CODE_LEN-1, 20));
+    }
+
+    fprintf(header, "const uch _length_code[MAX_MATCH-MIN_MATCH+1]= {\n");
+    for (i = 0; i < MAX_MATCH-MIN_MATCH+1; i++) {
+	fprintf(header, "%2u%s", _length_code[i],
+		SEPARATOR(i, MAX_MATCH-MIN_MATCH, 20));
+    }
+
+    fprintf(header, "local const int base_length[LENGTH_CODES] = {\n");
+    for (i = 0; i < LENGTH_CODES; i++) {
+	fprintf(header, "%1u%s", base_length[i],
+		SEPARATOR(i, LENGTH_CODES-1, 20));
+    }
+
+    fprintf(header, "local const int base_dist[D_CODES] = {\n");
+    for (i = 0; i < D_CODES; i++) {
+	fprintf(header, "%5u%s", base_dist[i],
+		SEPARATOR(i, D_CODES-1, 10));
+    }
+
+    fclose(header);
+}
+#endif /* GEN_TREES_H */
+
+/* ===========================================================================
+ * Initialize the tree data structures for a new zlib stream.
+ */
+void _tr_init(s)
+    deflate_state *s;
+{
+    tr_static_init();
+
+    s->l_desc.dyn_tree = s->dyn_ltree;
+    s->l_desc.stat_desc = &static_l_desc;
+
+    s->d_desc.dyn_tree = s->dyn_dtree;
+    s->d_desc.stat_desc = &static_d_desc;
+
+    s->bl_desc.dyn_tree = s->bl_tree;
+    s->bl_desc.stat_desc = &static_bl_desc;
+
+    s->bi_buf = 0;
+    s->bi_valid = 0;
+    s->last_eob_len = 8; /* enough lookahead for inflate */
+#ifdef DEBUG
+    s->compressed_len = 0L;
+    s->bits_sent = 0L;
+#endif
+
+    /* Initialize the first block of the first file: */
+    init_block(s);
+}
+
+/* ===========================================================================
+ * Initialize a new block.
+ */
+local void init_block(s)
+    deflate_state *s;
+{
+    int n; /* iterates over tree elements */
+
+    /* Initialize the trees. */
+    for (n = 0; n < L_CODES;  n++) s->dyn_ltree[n].Freq = 0;
+    for (n = 0; n < D_CODES;  n++) s->dyn_dtree[n].Freq = 0;
+    for (n = 0; n < BL_CODES; n++) s->bl_tree[n].Freq = 0;
+
+    s->dyn_ltree[END_BLOCK].Freq = 1;
+    s->opt_len = s->static_len = 0L;
+    s->last_lit = s->matches = 0;
+}
+
+#define SMALLEST 1
+/* Index within the heap array of least frequent node in the Huffman tree */
+
+
+/* ===========================================================================
+ * Remove the smallest element from the heap and recreate the heap with
+ * one less element. Updates heap and heap_len.
+ */
+#define pqremove(s, tree, top) \
+{\
+    top = s->heap[SMALLEST]; \
+    s->heap[SMALLEST] = s->heap[s->heap_len--]; \
+    pqdownheap(s, tree, SMALLEST); \
+}
+
+/* ===========================================================================
+ * Compares to subtrees, using the tree depth as tie breaker when
+ * the subtrees have equal frequency. This minimizes the worst case length.
+ */
+#define smaller(tree, n, m, depth) \
+   (tree[n].Freq < tree[m].Freq || \
+   (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
+
+/* ===========================================================================
+ * Restore the heap property by moving down the tree starting at node k,
+ * exchanging a node with the smallest of its two sons if necessary, stopping
+ * when the heap property is re-established (each father smaller than its
+ * two sons).
+ */
+local void pqdownheap(s, tree, k)
+    deflate_state *s;
+    ct_data *tree;  /* the tree to restore */
+    int k;               /* node to move down */
+{
+    int v = s->heap[k];
+    int j = k << 1;  /* left son of k */
+    while (j <= s->heap_len) {
+        /* Set j to the smallest of the two sons: */
+        if (j < s->heap_len &&
+            smaller(tree, s->heap[j+1], s->heap[j], s->depth)) {
+            j++;
+        }
+        /* Exit if v is smaller than both sons */
+        if (smaller(tree, v, s->heap[j], s->depth)) break;
+
+        /* Exchange v with the smallest son */
+        s->heap[k] = s->heap[j];  k = j;
+
+        /* And continue down the tree, setting j to the left son of k */
+        j <<= 1;
+    }
+    s->heap[k] = v;
+}
+
+/* ===========================================================================
+ * Compute the optimal bit lengths for a tree and update the total bit length
+ * for the current block.
+ * IN assertion: the fields freq and dad are set, heap[heap_max] and
+ *    above are the tree nodes sorted by increasing frequency.
+ * OUT assertions: the field len is set to the optimal bit length, the
+ *     array bl_count contains the frequencies for each bit length.
+ *     The length opt_len is updated; static_len is also updated if stree is
+ *     not null.
+ */
+local void gen_bitlen(s, desc)
+    deflate_state *s;
+    tree_desc *desc;    /* the tree descriptor */
+{
+    ct_data *tree        = desc->dyn_tree;
+    int max_code         = desc->max_code;
+    const ct_data *stree = desc->stat_desc->static_tree;
+    const intf *extra    = desc->stat_desc->extra_bits;
+    int base             = desc->stat_desc->extra_base;
+    int max_length       = desc->stat_desc->max_length;
+    int h;              /* heap index */
+    int n, m;           /* iterate over the tree elements */
+    int bits;           /* bit length */
+    int xbits;          /* extra bits */
+    ush f;              /* frequency */
+    int overflow = 0;   /* number of elements with bit length too large */
+
+    for (bits = 0; bits <= MAX_BITS; bits++) s->bl_count[bits] = 0;
+
+    /* In a first pass, compute the optimal bit lengths (which may
+     * overflow in the case of the bit length tree).
+     */
+    tree[s->heap[s->heap_max]].Len = 0; /* root of the heap */
+
+    for (h = s->heap_max+1; h < HEAP_SIZE; h++) {
+        n = s->heap[h];
+        bits = tree[tree[n].Dad].Len + 1;
+        if (bits > max_length) bits = max_length, overflow++;
+        tree[n].Len = (ush)bits;
+        /* We overwrite tree[n].Dad which is no longer needed */
+
+        if (n > max_code) continue; /* not a leaf node */
+
+        s->bl_count[bits]++;
+        xbits = 0;
+        if (n >= base) xbits = extra[n-base];
+        f = tree[n].Freq;
+        s->opt_len += (ulg)f * (bits + xbits);
+        if (stree) s->static_len += (ulg)f * (stree[n].Len + xbits);
+    }
+    if (overflow == 0) return;
+
+    Trace((stderr,"\nbit length overflow\n"));
+    /* This happens for example on obj2 and pic of the Calgary corpus */
+
+    /* Find the first bit length which could increase: */
+    do {
+        bits = max_length-1;
+        while (s->bl_count[bits] == 0) bits--;
+        s->bl_count[bits]--;      /* move one leaf down the tree */
+        s->bl_count[bits+1] += 2; /* move one overflow item as its brother */
+        s->bl_count[max_length]--;
+        /* The brother of the overflow item also moves one step up,
+         * but this does not affect bl_count[max_length]
+         */
+        overflow -= 2;
+    } while (overflow > 0);
+
+    /* Now recompute all bit lengths, scanning in increasing frequency.
+     * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
+     * lengths instead of fixing only the wrong ones. This idea is taken
+     * from 'ar' written by Haruhiko Okumura.)
+     */
+    for (bits = max_length; bits != 0; bits--) {
+        n = s->bl_count[bits];
+        while (n != 0) {
+            m = s->heap[--h];
+            if (m > max_code) continue;
+            if (tree[m].Len != (unsigned) bits) {
+                Trace((stderr,"code %d bits %d->%d\n", m, tree[m].Len, bits));
+                s->opt_len += ((long)bits - (long)tree[m].Len)
+                              *(long)tree[m].Freq;
+                tree[m].Len = (ush)bits;
+            }
+            n--;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Generate the codes for a given tree and bit counts (which need not be
+ * optimal).
+ * IN assertion: the array bl_count contains the bit length statistics for
+ * the given tree and the field len is set for all tree elements.
+ * OUT assertion: the field code is set for all tree elements of non
+ *     zero code length.
+ */
+local void gen_codes (tree, max_code, bl_count)
+    ct_data *tree;             /* the tree to decorate */
+    int max_code;              /* largest code with non zero frequency */
+    ushf *bl_count;            /* number of codes at each bit length */
+{
+    ush next_code[MAX_BITS+1]; /* next code value for each bit length */
+    ush code = 0;              /* running code value */
+    int bits;                  /* bit index */
+    int n;                     /* code index */
+
+    /* The distribution counts are first used to generate the code values
+     * without bit reversal.
+     */
+    for (bits = 1; bits <= MAX_BITS; bits++) {
+        next_code[bits] = code = (code + bl_count[bits-1]) << 1;
+    }
+    /* Check that the bit counts in bl_count are consistent. The last code
+     * must be all ones.
+     */
+    Assert (code + bl_count[MAX_BITS]-1 == (1<<MAX_BITS)-1,
+            "inconsistent bit counts");
+    Tracev((stderr,"\ngen_codes: max_code %d ", max_code));
+
+    for (n = 0;  n <= max_code; n++) {
+        int len = tree[n].Len;
+        if (len == 0) continue;
+        /* Now reverse the bits */
+        tree[n].Code = bi_reverse(next_code[len]++, len);
+
+        Tracecv(tree != static_ltree, (stderr,"\nn %3d %c l %2d c %4x (%x) ",
+             n, (isgraph(n) ? n : ' '), len, tree[n].Code, next_code[len]-1));
+    }
+}
+
+/* ===========================================================================
+ * Construct one Huffman tree and assigns the code bit strings and lengths.
+ * Update the total bit length for the current block.
+ * IN assertion: the field freq is set for all tree elements.
+ * OUT assertions: the fields len and code are set to the optimal bit length
+ *     and corresponding code. The length opt_len is updated; static_len is
+ *     also updated if stree is not null. The field max_code is set.
+ */
+local void build_tree(s, desc)
+    deflate_state *s;
+    tree_desc *desc; /* the tree descriptor */
+{
+    ct_data *tree         = desc->dyn_tree;
+    const ct_data *stree  = desc->stat_desc->static_tree;
+    int elems             = desc->stat_desc->elems;
+    int n, m;          /* iterate over heap elements */
+    int max_code = -1; /* largest code with non zero frequency */
+    int node;          /* new node being created */
+
+    /* Construct the initial heap, with least frequent element in
+     * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
+     * heap[0] is not used.
+     */
+    s->heap_len = 0, s->heap_max = HEAP_SIZE;
+
+    for (n = 0; n < elems; n++) {
+        if (tree[n].Freq != 0) {
+            s->heap[++(s->heap_len)] = max_code = n;
+            s->depth[n] = 0;
+        } else {
+            tree[n].Len = 0;
+        }
+    }
+
+    /* The pkzip format requires that at least one distance code exists,
+     * and that at least one bit should be sent even if there is only one
+     * possible code. So to avoid special checks later on we force at least
+     * two codes of non zero frequency.
+     */
+    while (s->heap_len < 2) {
+        node = s->heap[++(s->heap_len)] = (max_code < 2 ? ++max_code : 0);
+        tree[node].Freq = 1;
+        s->depth[node] = 0;
+        s->opt_len--; if (stree) s->static_len -= stree[node].Len;
+        /* node is 0 or 1 so it does not have extra bits */
+    }
+    desc->max_code = max_code;
+
+    /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
+     * establish sub-heaps of increasing lengths:
+     */
+    for (n = s->heap_len/2; n >= 1; n--) pqdownheap(s, tree, n);
+
+    /* Construct the Huffman tree by repeatedly combining the least two
+     * frequent nodes.
+     */
+    node = elems;              /* next internal node of the tree */
+    do {
+        pqremove(s, tree, n);  /* n = node of least frequency */
+        m = s->heap[SMALLEST]; /* m = node of next least frequency */
+
+        s->heap[--(s->heap_max)] = n; /* keep the nodes sorted by frequency */
+        s->heap[--(s->heap_max)] = m;
+
+        /* Create a new node father of n and m */
+        tree[node].Freq = tree[n].Freq + tree[m].Freq;
+        s->depth[node] = (uch) (MAX(s->depth[n], s->depth[m]) + 1);
+        tree[n].Dad = tree[m].Dad = (ush)node;
+#ifdef DUMP_BL_TREE
+        if (tree == s->bl_tree) {
+            fprintf(stderr,"\nnode %d(%d), sons %d(%d) %d(%d)",
+                    node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
+        }
+#endif
+        /* and insert the new node in the heap */
+        s->heap[SMALLEST] = node++;
+        pqdownheap(s, tree, SMALLEST);
+
+    } while (s->heap_len >= 2);
+
+    s->heap[--(s->heap_max)] = s->heap[SMALLEST];
+
+    /* At this point, the fields freq and dad are set. We can now
+     * generate the bit lengths.
+     */
+    gen_bitlen(s, (tree_desc *)desc);
+
+    /* The field len is now set, we can generate the bit codes */
+    gen_codes ((ct_data *)tree, max_code, s->bl_count);
+}
+
+/* ===========================================================================
+ * Scan a literal or distance tree to determine the frequencies of the codes
+ * in the bit length tree.
+ */
+local void scan_tree (s, tree, max_code)
+    deflate_state *s;
+    ct_data *tree;   /* the tree to be scanned */
+    int max_code;    /* and its largest code of non zero frequency */
+{
+    int n;                     /* iterates over all tree elements */
+    int prevlen = -1;          /* last emitted length */
+    int curlen;                /* length of current code */
+    int nextlen = tree[0].Len; /* length of next code */
+    int count = 0;             /* repeat count of the current code */
+    int max_count = 7;         /* max repeat count */
+    int min_count = 4;         /* min repeat count */
+
+    if (nextlen == 0) max_count = 138, min_count = 3;
+    tree[max_code+1].Len = (ush)0xffff; /* guard */
+
+    for (n = 0; n <= max_code; n++) {
+        curlen = nextlen; nextlen = tree[n+1].Len;
+        if (++count < max_count && curlen == nextlen) {
+            continue;
+        } else if (count < min_count) {
+            s->bl_tree[curlen].Freq += count;
+        } else if (curlen != 0) {
+            if (curlen != prevlen) s->bl_tree[curlen].Freq++;
+            s->bl_tree[REP_3_6].Freq++;
+        } else if (count <= 10) {
+            s->bl_tree[REPZ_3_10].Freq++;
+        } else {
+            s->bl_tree[REPZ_11_138].Freq++;
+        }
+        count = 0; prevlen = curlen;
+        if (nextlen == 0) {
+            max_count = 138, min_count = 3;
+        } else if (curlen == nextlen) {
+            max_count = 6, min_count = 3;
+        } else {
+            max_count = 7, min_count = 4;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Send a literal or distance tree in compressed form, using the codes in
+ * bl_tree.
+ */
+local void send_tree (s, tree, max_code)
+    deflate_state *s;
+    ct_data *tree; /* the tree to be scanned */
+    int max_code;       /* and its largest code of non zero frequency */
+{
+    int n;                     /* iterates over all tree elements */
+    int prevlen = -1;          /* last emitted length */
+    int curlen;                /* length of current code */
+    int nextlen = tree[0].Len; /* length of next code */
+    int count = 0;             /* repeat count of the current code */
+    int max_count = 7;         /* max repeat count */
+    int min_count = 4;         /* min repeat count */
+
+    /* tree[max_code+1].Len = -1; */  /* guard already set */
+    if (nextlen == 0) max_count = 138, min_count = 3;
+
+    for (n = 0; n <= max_code; n++) {
+        curlen = nextlen; nextlen = tree[n+1].Len;
+        if (++count < max_count && curlen == nextlen) {
+            continue;
+        } else if (count < min_count) {
+            do { send_code(s, curlen, s->bl_tree); } while (--count != 0);
+
+        } else if (curlen != 0) {
+            if (curlen != prevlen) {
+                send_code(s, curlen, s->bl_tree); count--;
+            }
+            Assert(count >= 3 && count <= 6, " 3_6?");
+            send_code(s, REP_3_6, s->bl_tree); send_bits(s, count-3, 2);
+
+        } else if (count <= 10) {
+            send_code(s, REPZ_3_10, s->bl_tree); send_bits(s, count-3, 3);
+
+        } else {
+            send_code(s, REPZ_11_138, s->bl_tree); send_bits(s, count-11, 7);
+        }
+        count = 0; prevlen = curlen;
+        if (nextlen == 0) {
+            max_count = 138, min_count = 3;
+        } else if (curlen == nextlen) {
+            max_count = 6, min_count = 3;
+        } else {
+            max_count = 7, min_count = 4;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Construct the Huffman tree for the bit lengths and return the index in
+ * bl_order of the last bit length code to send.
+ */
+local int build_bl_tree(s)
+    deflate_state *s;
+{
+    int max_blindex;  /* index of last bit length code of non zero freq */
+
+    /* Determine the bit length frequencies for literal and distance trees */
+    scan_tree(s, (ct_data *)s->dyn_ltree, s->l_desc.max_code);
+    scan_tree(s, (ct_data *)s->dyn_dtree, s->d_desc.max_code);
+
+    /* Build the bit length tree: */
+    build_tree(s, (tree_desc *)(&(s->bl_desc)));
+    /* opt_len now includes the length of the tree representations, except
+     * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
+     */
+
+    /* Determine the number of bit length codes to send. The pkzip format
+     * requires that at least 4 bit length codes be sent. (appnote.txt says
+     * 3 but the actual value used is 4.)
+     */
+    for (max_blindex = BL_CODES-1; max_blindex >= 3; max_blindex--) {
+        if (s->bl_tree[bl_order[max_blindex]].Len != 0) break;
+    }
+    /* Update opt_len to include the bit length tree and counts */
+    s->opt_len += 3*(max_blindex+1) + 5+5+4;
+    Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld",
+            s->opt_len, s->static_len));
+
+    return max_blindex;
+}
+
+/* ===========================================================================
+ * Send the header for a block using dynamic Huffman trees: the counts, the
+ * lengths of the bit length codes, the literal tree and the distance tree.
+ * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
+ */
+local void send_all_trees(s, lcodes, dcodes, blcodes)
+    deflate_state *s;
+    int lcodes, dcodes, blcodes; /* number of codes for each tree */
+{
+    int rank;                    /* index in bl_order */
+
+    Assert (lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
+    Assert (lcodes <= L_CODES && dcodes <= D_CODES && blcodes <= BL_CODES,
+            "too many codes");
+    Tracev((stderr, "\nbl counts: "));
+    send_bits(s, lcodes-257, 5); /* not +255 as stated in appnote.txt */
+    send_bits(s, dcodes-1,   5);
+    send_bits(s, blcodes-4,  4); /* not -3 as stated in appnote.txt */
+    for (rank = 0; rank < blcodes; rank++) {
+        Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
+        send_bits(s, s->bl_tree[bl_order[rank]].Len, 3);
+    }
+    Tracev((stderr, "\nbl tree: sent %ld", s->bits_sent));
+
+    send_tree(s, (ct_data *)s->dyn_ltree, lcodes-1); /* literal tree */
+    Tracev((stderr, "\nlit tree: sent %ld", s->bits_sent));
+
+    send_tree(s, (ct_data *)s->dyn_dtree, dcodes-1); /* distance tree */
+    Tracev((stderr, "\ndist tree: sent %ld", s->bits_sent));
+}
+
+/* ===========================================================================
+ * Send a stored block
+ */
+void _tr_stored_block(s, buf, stored_len, eof)
+    deflate_state *s;
+    charf *buf;       /* input block */
+    ulg stored_len;   /* length of input block */
+    int eof;          /* true if this is the last block for a file */
+{
+    send_bits(s, (STORED_BLOCK<<1)+eof, 3);  /* send block type */
+#ifdef DEBUG
+    s->compressed_len = (s->compressed_len + 3 + 7) & (ulg)~7L;
+    s->compressed_len += (stored_len + 4) << 3;
+#endif
+    copy_block(s, buf, (unsigned)stored_len, 1); /* with header */
+}
+
+/* ===========================================================================
+ * Send one empty static block to give enough lookahead for inflate.
+ * This takes 10 bits, of which 7 may remain in the bit buffer.
+ * The current inflate code requires 9 bits of lookahead. If the
+ * last two codes for the previous block (real code plus EOB) were coded
+ * on 5 bits or less, inflate may have only 5+3 bits of lookahead to decode
+ * the last real code. In this case we send two empty static blocks instead
+ * of one. (There are no problems if the previous block is stored or fixed.)
+ * To simplify the code, we assume the worst case of last real code encoded
+ * on one bit only.
+ */
+void _tr_align(s)
+    deflate_state *s;
+{
+    send_bits(s, STATIC_TREES<<1, 3);
+    send_code(s, END_BLOCK, static_ltree);
+#ifdef DEBUG
+    s->compressed_len += 10L; /* 3 for block type, 7 for EOB */
+#endif
+    bi_flush(s);
+    /* Of the 10 bits for the empty block, we have already sent
+     * (10 - bi_valid) bits. The lookahead for the last real code (before
+     * the EOB of the previous block) was thus at least one plus the length
+     * of the EOB plus what we have just sent of the empty static block.
+     */
+    if (1 + s->last_eob_len + 10 - s->bi_valid < 9) {
+        send_bits(s, STATIC_TREES<<1, 3);
+        send_code(s, END_BLOCK, static_ltree);
+#ifdef DEBUG
+        s->compressed_len += 10L;
+#endif
+        bi_flush(s);
+    }
+    s->last_eob_len = 7;
+}
+
+/* ===========================================================================
+ * Determine the best encoding for the current block: dynamic trees, static
+ * trees or store, and output the encoded block to the zip file.
+ */
+void _tr_flush_block(s, buf, stored_len, eof)
+    deflate_state *s;
+    charf *buf;       /* input block, or NULL if too old */
+    ulg stored_len;   /* length of input block */
+    int eof;          /* true if this is the last block for a file */
+{
+    ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
+    int max_blindex = 0;  /* index of last bit length code of non zero freq */
+
+    /* Build the Huffman trees unless a stored block is forced */
+    if (s->level > 0) {
+
+	 /* Check if the file is ascii or binary */
+	if (s->data_type == Z_UNKNOWN) set_data_type(s);
+
+	/* Construct the literal and distance trees */
+	build_tree(s, (tree_desc *)(&(s->l_desc)));
+	Tracev((stderr, "\nlit data: dyn %ld, stat %ld", s->opt_len,
+		s->static_len));
+
+	build_tree(s, (tree_desc *)(&(s->d_desc)));
+	Tracev((stderr, "\ndist data: dyn %ld, stat %ld", s->opt_len,
+		s->static_len));
+	/* At this point, opt_len and static_len are the total bit lengths of
+	 * the compressed block data, excluding the tree representations.
+	 */
+
+	/* Build the bit length tree for the above two trees, and get the index
+	 * in bl_order of the last bit length code to send.
+	 */
+	max_blindex = build_bl_tree(s);
+
+	/* Determine the best encoding. Compute first the block length in bytes*/
+	opt_lenb = (s->opt_len+3+7)>>3;
+	static_lenb = (s->static_len+3+7)>>3;
+
+	Tracev((stderr, "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u ",
+		opt_lenb, s->opt_len, static_lenb, s->static_len, stored_len,
+		s->last_lit));
+
+	if (static_lenb <= opt_lenb) opt_lenb = static_lenb;
+
+    } else {
+        Assert(buf != (char*)0, "lost buf");
+	opt_lenb = static_lenb = stored_len + 5; /* force a stored block */
+    }
+
+#ifdef FORCE_STORED
+    if (buf != (char*)0) { /* force stored block */
+#else
+    if (stored_len+4 <= opt_lenb && buf != (char*)0) {
+                       /* 4: two words for the lengths */
+#endif
+        /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
+         * Otherwise we can't have processed more than WSIZE input bytes since
+         * the last block flush, because compression would have been
+         * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
+         * transform a block into a stored block.
+         */
+        _tr_stored_block(s, buf, stored_len, eof);
+
+#ifdef FORCE_STATIC
+    } else if (static_lenb >= 0) { /* force static trees */
+#else
+    } else if (static_lenb == opt_lenb) {
+#endif
+        send_bits(s, (STATIC_TREES<<1)+eof, 3);
+        compress_block(s, (ct_data *)static_ltree, (ct_data *)static_dtree);
+#ifdef DEBUG
+        s->compressed_len += 3 + s->static_len;
+#endif
+    } else {
+        send_bits(s, (DYN_TREES<<1)+eof, 3);
+        send_all_trees(s, s->l_desc.max_code+1, s->d_desc.max_code+1,
+                       max_blindex+1);
+        compress_block(s, (ct_data *)s->dyn_ltree, (ct_data *)s->dyn_dtree);
+#ifdef DEBUG
+        s->compressed_len += 3 + s->opt_len;
+#endif
+    }
+    Assert (s->compressed_len == s->bits_sent, "bad compressed size");
+    /* The above check is made mod 2^32, for files larger than 512 MB
+     * and uLong implemented on 32 bits.
+     */
+    init_block(s);
+
+    if (eof) {
+        bi_windup(s);
+#ifdef DEBUG
+        s->compressed_len += 7;  /* align on byte boundary */
+#endif
+    }
+    Tracev((stderr,"\ncomprlen %lu(%lu) ", s->compressed_len>>3,
+           s->compressed_len-7*eof));
+}
+
+/* ===========================================================================
+ * Save the match info and tally the frequency counts. Return true if
+ * the current block must be flushed.
+ */
+int _tr_tally (s, dist, lc)
+    deflate_state *s;
+    unsigned dist;  /* distance of matched string */
+    unsigned lc;    /* match length-MIN_MATCH or unmatched char (if dist==0) */
+{
+    s->d_buf[s->last_lit] = (ush)dist;
+    s->l_buf[s->last_lit++] = (uch)lc;
+    if (dist == 0) {
+        /* lc is the unmatched char */
+        s->dyn_ltree[lc].Freq++;
+    } else {
+        s->matches++;
+        /* Here, lc is the match length - MIN_MATCH */
+        dist--;             /* dist = match distance - 1 */
+        Assert((ush)dist < (ush)MAX_DIST(s) &&
+               (ush)lc <= (ush)(MAX_MATCH-MIN_MATCH) &&
+               (ush)d_code(dist) < (ush)D_CODES,  "_tr_tally: bad match");
+
+        s->dyn_ltree[_length_code[lc]+LITERALS+1].Freq++;
+        s->dyn_dtree[d_code(dist)].Freq++;
+    }
+
+#ifdef TRUNCATE_BLOCK
+    /* Try to guess if it is profitable to stop the current block here */
+    if ((s->last_lit & 0x1fff) == 0 && s->level > 2) {
+        /* Compute an upper bound for the compressed length */
+        ulg out_length = (ulg)s->last_lit*8L;
+        ulg in_length = (ulg)((long)s->strstart - s->block_start);
+        int dcode;
+        for (dcode = 0; dcode < D_CODES; dcode++) {
+            out_length += (ulg)s->dyn_dtree[dcode].Freq *
+                (5L+extra_dbits[dcode]);
+        }
+        out_length >>= 3;
+        Tracev((stderr,"\nlast_lit %u, in %ld, out ~%ld(%ld%%) ",
+               s->last_lit, in_length, out_length,
+               100L - out_length*100L/in_length));
+        if (s->matches < s->last_lit/2 && out_length < in_length/2) return 1;
+    }
+#endif
+    return (s->last_lit == s->lit_bufsize-1);
+    /* We avoid equality with lit_bufsize because of wraparound at 64K
+     * on 16 bit machines and because stored blocks are restricted to
+     * 64K-1 bytes.
+     */
+}
+
+/* ===========================================================================
+ * Send the block data compressed using the given Huffman trees
+ */
+local void compress_block(s, ltree, dtree)
+    deflate_state *s;
+    ct_data *ltree; /* literal tree */
+    ct_data *dtree; /* distance tree */
+{
+    unsigned dist;      /* distance of matched string */
+    int lc;             /* match length or unmatched char (if dist == 0) */
+    unsigned lx = 0;    /* running index in l_buf */
+    unsigned code;      /* the code to send */
+    int extra;          /* number of extra bits to send */
+
+    if (s->last_lit != 0) do {
+        dist = s->d_buf[lx];
+        lc = s->l_buf[lx++];
+        if (dist == 0) {
+            send_code(s, lc, ltree); /* send a literal byte */
+            Tracecv(isgraph(lc), (stderr," '%c' ", lc));
+        } else {
+            /* Here, lc is the match length - MIN_MATCH */
+            code = _length_code[lc];
+            send_code(s, code+LITERALS+1, ltree); /* send the length code */
+            extra = extra_lbits[code];
+            if (extra != 0) {
+                lc -= base_length[code];
+                send_bits(s, lc, extra);       /* send the extra length bits */
+            }
+            dist--; /* dist is now the match distance - 1 */
+            code = d_code(dist);
+            Assert (code < D_CODES, "bad d_code");
+
+            send_code(s, code, dtree);       /* send the distance code */
+            extra = extra_dbits[code];
+            if (extra != 0) {
+                dist -= base_dist[code];
+                send_bits(s, dist, extra);   /* send the extra distance bits */
+            }
+        } /* literal or match pair ? */
+
+        /* Check that the overlay between pending_buf and d_buf+l_buf is ok: */
+        Assert(s->pending < s->lit_bufsize + 2*lx, "pendingBuf overflow");
+
+    } while (lx < s->last_lit);
+
+    send_code(s, END_BLOCK, ltree);
+    s->last_eob_len = ltree[END_BLOCK].Len;
+}
+
+/* ===========================================================================
+ * Set the data type to ASCII or BINARY, using a crude approximation:
+ * binary if more than 20% of the bytes are <= 6 or >= 128, ascii otherwise.
+ * IN assertion: the fields freq of dyn_ltree are set and the total of all
+ * frequencies does not exceed 64K (to fit in an int on 16 bit machines).
+ */
+local void set_data_type(s)
+    deflate_state *s;
+{
+    int n = 0;
+    unsigned ascii_freq = 0;
+    unsigned bin_freq = 0;
+    while (n < 7)        bin_freq += s->dyn_ltree[n++].Freq;
+    while (n < 128)    ascii_freq += s->dyn_ltree[n++].Freq;
+    while (n < LITERALS) bin_freq += s->dyn_ltree[n++].Freq;
+    s->data_type = (Byte)(bin_freq > (ascii_freq >> 2) ? Z_BINARY : Z_ASCII);
+}
+
+/* ===========================================================================
+ * Reverse the first len bits of a code, using straightforward code (a faster
+ * method would use a table)
+ * IN assertion: 1 <= len <= 15
+ */
+local unsigned bi_reverse(code, len)
+    unsigned code; /* the value to invert */
+    int len;       /* its bit length */
+{
+    register unsigned res = 0;
+    do {
+        res |= code & 1;
+        code >>= 1, res <<= 1;
+    } while (--len > 0);
+    return res >> 1;
+}
+
+/* ===========================================================================
+ * Flush the bit buffer, keeping at most 7 bits in it.
+ */
+local void bi_flush(s)
+    deflate_state *s;
+{
+    if (s->bi_valid == 16) {
+        put_short(s, s->bi_buf);
+        s->bi_buf = 0;
+        s->bi_valid = 0;
+    } else if (s->bi_valid >= 8) {
+        put_byte(s, (Byte)s->bi_buf);
+        s->bi_buf >>= 8;
+        s->bi_valid -= 8;
+    }
+}
+
+/* ===========================================================================
+ * Flush the bit buffer and align the output on a byte boundary
+ */
+local void bi_windup(s)
+    deflate_state *s;
+{
+    if (s->bi_valid > 8) {
+        put_short(s, s->bi_buf);
+    } else if (s->bi_valid > 0) {
+        put_byte(s, (Byte)s->bi_buf);
+    }
+    s->bi_buf = 0;
+    s->bi_valid = 0;
+#ifdef DEBUG
+    s->bits_sent = (s->bits_sent+7) & ~7;
+#endif
+}
+
+/* ===========================================================================
+ * Copy a stored block, storing first the length and its
+ * one's complement if requested.
+ */
+local void copy_block(s, buf, len, header)
+    deflate_state *s;
+    charf    *buf;    /* the input data */
+    unsigned len;     /* its length */
+    int      header;  /* true if block header must be written */
+{
+    bi_windup(s);        /* align on byte boundary */
+    s->last_eob_len = 8; /* enough lookahead for inflate */
+
+    if (header) {
+        put_short(s, (ush)len);   
+        put_short(s, (ush)~len);
+#ifdef DEBUG
+        s->bits_sent += 2*16;
+#endif
+    }
+#ifdef DEBUG
+    s->bits_sent += (ulg)len<<3;
+#endif
+    while (len--) {
+        put_byte(s, *buf++);
+    }
+}
diff --git a/zlib/trees.h b/zlib/trees.h
new file mode 100644
index 000000000..72facf900
--- /dev/null
+++ b/zlib/trees.h
@@ -0,0 +1,128 @@
+/* header created automatically with -DGEN_TREES_H */
+
+local const ct_data static_ltree[L_CODES+2] = {
+{{ 12},{  8}}, {{140},{  8}}, {{ 76},{  8}}, {{204},{  8}}, {{ 44},{  8}},
+{{172},{  8}}, {{108},{  8}}, {{236},{  8}}, {{ 28},{  8}}, {{156},{  8}},
+{{ 92},{  8}}, {{220},{  8}}, {{ 60},{  8}}, {{188},{  8}}, {{124},{  8}},
+{{252},{  8}}, {{  2},{  8}}, {{130},{  8}}, {{ 66},{  8}}, {{194},{  8}},
+{{ 34},{  8}}, {{162},{  8}}, {{ 98},{  8}}, {{226},{  8}}, {{ 18},{  8}},
+{{146},{  8}}, {{ 82},{  8}}, {{210},{  8}}, {{ 50},{  8}}, {{178},{  8}},
+{{114},{  8}}, {{242},{  8}}, {{ 10},{  8}}, {{138},{  8}}, {{ 74},{  8}},
+{{202},{  8}}, {{ 42},{  8}}, {{170},{  8}}, {{106},{  8}}, {{234},{  8}},
+{{ 26},{  8}}, {{154},{  8}}, {{ 90},{  8}}, {{218},{  8}}, {{ 58},{  8}},
+{{186},{  8}}, {{122},{  8}}, {{250},{  8}}, {{  6},{  8}}, {{134},{  8}},
+{{ 70},{  8}}, {{198},{  8}}, {{ 38},{  8}}, {{166},{  8}}, {{102},{  8}},
+{{230},{  8}}, {{ 22},{  8}}, {{150},{  8}}, {{ 86},{  8}}, {{214},{  8}},
+{{ 54},{  8}}, {{182},{  8}}, {{118},{  8}}, {{246},{  8}}, {{ 14},{  8}},
+{{142},{  8}}, {{ 78},{  8}}, {{206},{  8}}, {{ 46},{  8}}, {{174},{  8}},
+{{110},{  8}}, {{238},{  8}}, {{ 30},{  8}}, {{158},{  8}}, {{ 94},{  8}},
+{{222},{  8}}, {{ 62},{  8}}, {{190},{  8}}, {{126},{  8}}, {{254},{  8}},
+{{  1},{  8}}, {{129},{  8}}, {{ 65},{  8}}, {{193},{  8}}, {{ 33},{  8}},
+{{161},{  8}}, {{ 97},{  8}}, {{225},{  8}}, {{ 17},{  8}}, {{145},{  8}},
+{{ 81},{  8}}, {{209},{  8}}, {{ 49},{  8}}, {{177},{  8}}, {{113},{  8}},
+{{241},{  8}}, {{  9},{  8}}, {{137},{  8}}, {{ 73},{  8}}, {{201},{  8}},
+{{ 41},{  8}}, {{169},{  8}}, {{105},{  8}}, {{233},{  8}}, {{ 25},{  8}},
+{{153},{  8}}, {{ 89},{  8}}, {{217},{  8}}, {{ 57},{  8}}, {{185},{  8}},
+{{121},{  8}}, {{249},{  8}}, {{  5},{  8}}, {{133},{  8}}, {{ 69},{  8}},
+{{197},{  8}}, {{ 37},{  8}}, {{165},{  8}}, {{101},{  8}}, {{229},{  8}},
+{{ 21},{  8}}, {{149},{  8}}, {{ 85},{  8}}, {{213},{  8}}, {{ 53},{  8}},
+{{181},{  8}}, {{117},{  8}}, {{245},{  8}}, {{ 13},{  8}}, {{141},{  8}},
+{{ 77},{  8}}, {{205},{  8}}, {{ 45},{  8}}, {{173},{  8}}, {{109},{  8}},
+{{237},{  8}}, {{ 29},{  8}}, {{157},{  8}}, {{ 93},{  8}}, {{221},{  8}},
+{{ 61},{  8}}, {{189},{  8}}, {{125},{  8}}, {{253},{  8}}, {{ 19},{  9}},
+{{275},{  9}}, {{147},{  9}}, {{403},{  9}}, {{ 83},{  9}}, {{339},{  9}},
+{{211},{  9}}, {{467},{  9}}, {{ 51},{  9}}, {{307},{  9}}, {{179},{  9}},
+{{435},{  9}}, {{115},{  9}}, {{371},{  9}}, {{243},{  9}}, {{499},{  9}},
+{{ 11},{  9}}, {{267},{  9}}, {{139},{  9}}, {{395},{  9}}, {{ 75},{  9}},
+{{331},{  9}}, {{203},{  9}}, {{459},{  9}}, {{ 43},{  9}}, {{299},{  9}},
+{{171},{  9}}, {{427},{  9}}, {{107},{  9}}, {{363},{  9}}, {{235},{  9}},
+{{491},{  9}}, {{ 27},{  9}}, {{283},{  9}}, {{155},{  9}}, {{411},{  9}},
+{{ 91},{  9}}, {{347},{  9}}, {{219},{  9}}, {{475},{  9}}, {{ 59},{  9}},
+{{315},{  9}}, {{187},{  9}}, {{443},{  9}}, {{123},{  9}}, {{379},{  9}},
+{{251},{  9}}, {{507},{  9}}, {{  7},{  9}}, {{263},{  9}}, {{135},{  9}},
+{{391},{  9}}, {{ 71},{  9}}, {{327},{  9}}, {{199},{  9}}, {{455},{  9}},
+{{ 39},{  9}}, {{295},{  9}}, {{167},{  9}}, {{423},{  9}}, {{103},{  9}},
+{{359},{  9}}, {{231},{  9}}, {{487},{  9}}, {{ 23},{  9}}, {{279},{  9}},
+{{151},{  9}}, {{407},{  9}}, {{ 87},{  9}}, {{343},{  9}}, {{215},{  9}},
+{{471},{  9}}, {{ 55},{  9}}, {{311},{  9}}, {{183},{  9}}, {{439},{  9}},
+{{119},{  9}}, {{375},{  9}}, {{247},{  9}}, {{503},{  9}}, {{ 15},{  9}},
+{{271},{  9}}, {{143},{  9}}, {{399},{  9}}, {{ 79},{  9}}, {{335},{  9}},
+{{207},{  9}}, {{463},{  9}}, {{ 47},{  9}}, {{303},{  9}}, {{175},{  9}},
+{{431},{  9}}, {{111},{  9}}, {{367},{  9}}, {{239},{  9}}, {{495},{  9}},
+{{ 31},{  9}}, {{287},{  9}}, {{159},{  9}}, {{415},{  9}}, {{ 95},{  9}},
+{{351},{  9}}, {{223},{  9}}, {{479},{  9}}, {{ 63},{  9}}, {{319},{  9}},
+{{191},{  9}}, {{447},{  9}}, {{127},{  9}}, {{383},{  9}}, {{255},{  9}},
+{{511},{  9}}, {{  0},{  7}}, {{ 64},{  7}}, {{ 32},{  7}}, {{ 96},{  7}},
+{{ 16},{  7}}, {{ 80},{  7}}, {{ 48},{  7}}, {{112},{  7}}, {{  8},{  7}},
+{{ 72},{  7}}, {{ 40},{  7}}, {{104},{  7}}, {{ 24},{  7}}, {{ 88},{  7}},
+{{ 56},{  7}}, {{120},{  7}}, {{  4},{  7}}, {{ 68},{  7}}, {{ 36},{  7}},
+{{100},{  7}}, {{ 20},{  7}}, {{ 84},{  7}}, {{ 52},{  7}}, {{116},{  7}},
+{{  3},{  8}}, {{131},{  8}}, {{ 67},{  8}}, {{195},{  8}}, {{ 35},{  8}},
+{{163},{  8}}, {{ 99},{  8}}, {{227},{  8}}
+};
+
+local const ct_data static_dtree[D_CODES] = {
+{{ 0},{ 5}}, {{16},{ 5}}, {{ 8},{ 5}}, {{24},{ 5}}, {{ 4},{ 5}},
+{{20},{ 5}}, {{12},{ 5}}, {{28},{ 5}}, {{ 2},{ 5}}, {{18},{ 5}},
+{{10},{ 5}}, {{26},{ 5}}, {{ 6},{ 5}}, {{22},{ 5}}, {{14},{ 5}},
+{{30},{ 5}}, {{ 1},{ 5}}, {{17},{ 5}}, {{ 9},{ 5}}, {{25},{ 5}},
+{{ 5},{ 5}}, {{21},{ 5}}, {{13},{ 5}}, {{29},{ 5}}, {{ 3},{ 5}},
+{{19},{ 5}}, {{11},{ 5}}, {{27},{ 5}}, {{ 7},{ 5}}, {{23},{ 5}}
+};
+
+const uch _dist_code[DIST_CODE_LEN] = {
+ 0,  1,  2,  3,  4,  4,  5,  5,  6,  6,  6,  6,  7,  7,  7,  7,  8,  8,  8,  8,
+ 8,  8,  8,  8,  9,  9,  9,  9,  9,  9,  9,  9, 10, 10, 10, 10, 10, 10, 10, 10,
+10, 10, 10, 10, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13, 13, 13, 13,
+13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+13, 13, 13, 13, 13, 13, 13, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  0,  0, 16, 17,
+18, 18, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22,
+23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+24, 24, 24, 24, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29
+};
+
+const uch _length_code[MAX_MATCH-MIN_MATCH+1]= {
+ 0,  1,  2,  3,  4,  5,  6,  7,  8,  8,  9,  9, 10, 10, 11, 11, 12, 12, 12, 12,
+13, 13, 13, 13, 14, 14, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, 16, 16, 16, 16,
+17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 19, 19, 19, 19,
+19, 19, 19, 19, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20,
+21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22,
+22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23,
+23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,
+25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 28
+};
+
+local const int base_length[LENGTH_CODES] = {
+0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56,
+64, 80, 96, 112, 128, 160, 192, 224, 0
+};
+
+local const int base_dist[D_CODES] = {
+    0,     1,     2,     3,     4,     6,     8,    12,    16,    24,
+   32,    48,    64,    96,   128,   192,   256,   384,   512,   768,
+ 1024,  1536,  2048,  3072,  4096,  6144,  8192, 12288, 16384, 24576
+};
+
diff --git a/zlib/uncompr.c b/zlib/uncompr.c
new file mode 100644
index 000000000..35578f93c
--- /dev/null
+++ b/zlib/uncompr.c
@@ -0,0 +1,58 @@
+/* uncompr.c -- decompress a memory buffer
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* @(#) $Id: uncompr.c,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#include "zlib.h"
+
+/* ===========================================================================
+     Decompresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer. Upon entry, destLen is the total
+   size of the destination buffer, which must be large enough to hold the
+   entire uncompressed data. (The size of the uncompressed data must have
+   been saved previously by the compressor and transmitted to the decompressor
+   by some mechanism outside the scope of this compression library.)
+   Upon exit, destLen is the actual size of the compressed buffer.
+     This function can be used to decompress a whole file at once if the
+   input file is mmap'ed.
+
+     uncompress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer, or Z_DATA_ERROR if the input data was corrupted.
+*/
+int ZEXPORT uncompress (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+{
+    z_stream stream;
+    int err;
+
+    stream.next_in = (Bytef*)source;
+    stream.avail_in = (uInt)sourceLen;
+    /* Check for source > 64K on 16-bit machine: */
+    if ((uLong)stream.avail_in != sourceLen) return Z_BUF_ERROR;
+
+    stream.next_out = dest;
+    stream.avail_out = (uInt)*destLen;
+    if ((uLong)stream.avail_out != *destLen) return Z_BUF_ERROR;
+
+    stream.zalloc = (alloc_func)0;
+    stream.zfree = (free_func)0;
+
+    err = inflateInit(&stream);
+    if (err != Z_OK) return err;
+
+    err = inflate(&stream, Z_FINISH);
+    if (err != Z_STREAM_END) {
+        inflateEnd(&stream);
+        return err == Z_OK ? Z_BUF_ERROR : err;
+    }
+    *destLen = stream.total_out;
+
+    err = inflateEnd(&stream);
+    return err;
+}
diff --git a/zlib/zconf.h b/zlib/zconf.h
new file mode 100644
index 000000000..9615f74b8
--- /dev/null
+++ b/zlib/zconf.h
@@ -0,0 +1,279 @@
+/* zconf.h -- configuration of the zlib compression library
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* @(#) $Id: zconf.h,v 1.1.2.1 2004/07/05 14:12:38 easysw Exp $ */
+
+#ifndef _ZCONF_H
+#define _ZCONF_H
+
+/*
+ * If you *really* need a unique prefix for all types and library functions,
+ * compile with -DZ_PREFIX. The "standard" zlib should be compiled without it.
+ */
+#ifdef Z_PREFIX
+#  define deflateInit_	z_deflateInit_
+#  define deflate	z_deflate
+#  define deflateEnd	z_deflateEnd
+#  define inflateInit_ 	z_inflateInit_
+#  define inflate	z_inflate
+#  define inflateEnd	z_inflateEnd
+#  define deflateInit2_	z_deflateInit2_
+#  define deflateSetDictionary z_deflateSetDictionary
+#  define deflateCopy	z_deflateCopy
+#  define deflateReset	z_deflateReset
+#  define deflateParams	z_deflateParams
+#  define inflateInit2_	z_inflateInit2_
+#  define inflateSetDictionary z_inflateSetDictionary
+#  define inflateSync	z_inflateSync
+#  define inflateSyncPoint z_inflateSyncPoint
+#  define inflateReset	z_inflateReset
+#  define compress	z_compress
+#  define compress2	z_compress2
+#  define uncompress	z_uncompress
+#  define adler32	z_adler32
+#  define crc32		z_crc32
+#  define get_crc_table z_get_crc_table
+
+#  define Byte		z_Byte
+#  define uInt		z_uInt
+#  define uLong		z_uLong
+#  define Bytef	        z_Bytef
+#  define charf		z_charf
+#  define intf		z_intf
+#  define uIntf		z_uIntf
+#  define uLongf	z_uLongf
+#  define voidpf	z_voidpf
+#  define voidp		z_voidp
+#endif
+
+#if (defined(_WIN32) || defined(__WIN32__)) && !defined(WIN32)
+#  define WIN32
+#endif
+#if defined(__GNUC__) || defined(WIN32) || defined(__386__) || defined(i386)
+#  ifndef __32BIT__
+#    define __32BIT__
+#  endif
+#endif
+#if defined(__MSDOS__) && !defined(MSDOS)
+#  define MSDOS
+#endif
+
+/*
+ * Compile with -DMAXSEG_64K if the alloc function cannot allocate more
+ * than 64k bytes at a time (needed on systems with 16-bit int).
+ */
+#if defined(MSDOS) && !defined(__32BIT__)
+#  define MAXSEG_64K
+#endif
+#ifdef MSDOS
+#  define UNALIGNED_OK
+#endif
+
+#if (defined(MSDOS) || defined(_WINDOWS) || defined(WIN32))  && !defined(STDC)
+#  define STDC
+#endif
+#if defined(__STDC__) || defined(__cplusplus) || defined(__OS2__)
+#  ifndef STDC
+#    define STDC
+#  endif
+#endif
+
+#ifndef STDC
+#  ifndef const /* cannot use !defined(STDC) && !defined(const) on Mac */
+#    define const
+#  endif
+#endif
+
+/* Some Mac compilers merge all .h files incorrectly: */
+#if defined(__MWERKS__) || defined(applec) ||defined(THINK_C) ||defined(__SC__)
+#  define NO_DUMMY_DECL
+#endif
+
+/* Old Borland C incorrectly complains about missing returns: */
+#if defined(__BORLANDC__) && (__BORLANDC__ < 0x500)
+#  define NEED_DUMMY_RETURN
+#endif
+
+
+/* Maximum value for memLevel in deflateInit2 */
+#ifndef MAX_MEM_LEVEL
+#  ifdef MAXSEG_64K
+#    define MAX_MEM_LEVEL 8
+#  else
+#    define MAX_MEM_LEVEL 9
+#  endif
+#endif
+
+/* Maximum value for windowBits in deflateInit2 and inflateInit2.
+ * WARNING: reducing MAX_WBITS makes minigzip unable to extract .gz files
+ * created by gzip. (Files created by minigzip can still be extracted by
+ * gzip.)
+ */
+#ifndef MAX_WBITS
+#  define MAX_WBITS   15 /* 32K LZ77 window */
+#endif
+
+/* The memory requirements for deflate are (in bytes):
+            (1 << (windowBits+2)) +  (1 << (memLevel+9))
+ that is: 128K for windowBits=15  +  128K for memLevel = 8  (default values)
+ plus a few kilobytes for small objects. For example, if you want to reduce
+ the default memory requirements from 256K to 128K, compile with
+     make CFLAGS="-O -DMAX_WBITS=14 -DMAX_MEM_LEVEL=7"
+ Of course this will generally degrade compression (there's no free lunch).
+
+   The memory requirements for inflate are (in bytes) 1 << windowBits
+ that is, 32K for windowBits=15 (default value) plus a few kilobytes
+ for small objects.
+*/
+
+                        /* Type declarations */
+
+#ifndef OF /* function prototypes */
+#  ifdef STDC
+#    define OF(args)  args
+#  else
+#    define OF(args)  ()
+#  endif
+#endif
+
+/* The following definitions for FAR are needed only for MSDOS mixed
+ * model programming (small or medium model with some far allocations).
+ * This was tested only with MSC; for other MSDOS compilers you may have
+ * to define NO_MEMCPY in zutil.h.  If you don't need the mixed model,
+ * just define FAR to be empty.
+ */
+#if (defined(M_I86SM) || defined(M_I86MM)) && !defined(__32BIT__)
+   /* MSC small or medium model */
+#  define SMALL_MEDIUM
+#  ifdef _MSC_VER
+#    define FAR _far
+#  else
+#    define FAR far
+#  endif
+#endif
+#if defined(__BORLANDC__) && (defined(__SMALL__) || defined(__MEDIUM__))
+#  ifndef __32BIT__
+#    define SMALL_MEDIUM
+#    define FAR _far
+#  endif
+#endif
+
+/* Compile with -DZLIB_DLL for Windows DLL support */
+#if defined(ZLIB_DLL)
+#  if defined(_WINDOWS) || defined(WINDOWS)
+#    ifdef FAR
+#      undef FAR
+#    endif
+#    include <windows.h>
+#    define ZEXPORT  WINAPI
+#    ifdef WIN32
+#      define ZEXPORTVA  WINAPIV
+#    else
+#      define ZEXPORTVA  FAR _cdecl _export
+#    endif
+#  endif
+#  if defined (__BORLANDC__)
+#    if (__BORLANDC__ >= 0x0500) && defined (WIN32)
+#      include <windows.h>
+#      define ZEXPORT __declspec(dllexport) WINAPI
+#      define ZEXPORTRVA __declspec(dllexport) WINAPIV
+#    else
+#      if defined (_Windows) && defined (__DLL__)
+#        define ZEXPORT _export
+#        define ZEXPORTVA _export
+#      endif
+#    endif
+#  endif
+#endif
+
+#if defined (__BEOS__)
+#  if defined (ZLIB_DLL)
+#    define ZEXTERN extern __declspec(dllexport)
+#  else
+#    define ZEXTERN extern __declspec(dllimport)
+#  endif
+#endif
+
+#ifndef ZEXPORT
+#  define ZEXPORT
+#endif
+#ifndef ZEXPORTVA
+#  define ZEXPORTVA
+#endif
+#ifndef ZEXTERN
+#  define ZEXTERN extern
+#endif
+
+#ifndef FAR
+#   define FAR
+#endif
+
+#if !defined(MACOS) && !defined(TARGET_OS_MAC)
+typedef unsigned char  Byte;  /* 8 bits */
+#endif
+typedef unsigned int   uInt;  /* 16 bits or more */
+typedef unsigned long  uLong; /* 32 bits or more */
+
+#ifdef SMALL_MEDIUM
+   /* Borland C/C++ and some old MSC versions ignore FAR inside typedef */
+#  define Bytef Byte FAR
+#else
+   typedef Byte  FAR Bytef;
+#endif
+typedef char  FAR charf;
+typedef int   FAR intf;
+typedef uInt  FAR uIntf;
+typedef uLong FAR uLongf;
+
+#ifdef STDC
+   typedef void FAR *voidpf;
+   typedef void     *voidp;
+#else
+   typedef Byte FAR *voidpf;
+   typedef Byte     *voidp;
+#endif
+
+#ifdef HAVE_UNISTD_H
+#  include <sys/types.h> /* for off_t */
+#  include <unistd.h>    /* for SEEK_* and off_t */
+#  define z_off_t  off_t
+#endif
+#ifndef SEEK_SET
+#  define SEEK_SET        0       /* Seek from beginning of file.  */
+#  define SEEK_CUR        1       /* Seek from current position.  */
+#  define SEEK_END        2       /* Set file pointer to EOF plus "offset" */
+#endif
+#ifndef z_off_t
+#  define  z_off_t long
+#endif
+
+/* MVS linker does not support external names larger than 8 bytes */
+#if defined(__MVS__)
+#   pragma map(deflateInit_,"DEIN")
+#   pragma map(deflateInit2_,"DEIN2")
+#   pragma map(deflateEnd,"DEEND")
+#   pragma map(inflateInit_,"ININ")
+#   pragma map(inflateInit2_,"ININ2")
+#   pragma map(inflateEnd,"INEND")
+#   pragma map(inflateSync,"INSY")
+#   pragma map(inflateSetDictionary,"INSEDI")
+#   pragma map(inflate_blocks,"INBL")
+#   pragma map(inflate_blocks_new,"INBLNE")
+#   pragma map(inflate_blocks_free,"INBLFR")
+#   pragma map(inflate_blocks_reset,"INBLRE")
+#   pragma map(inflate_codes_free,"INCOFR")
+#   pragma map(inflate_codes,"INCO")
+#   pragma map(inflate_fast,"INFA")
+#   pragma map(inflate_flush,"INFLU")
+#   pragma map(inflate_mask,"INMA")
+#   pragma map(inflate_set_dictionary,"INSEDI2")
+#   pragma map(inflate_copyright,"INCOPY")
+#   pragma map(inflate_trees_bits,"INTRBI")
+#   pragma map(inflate_trees_dynamic,"INTRDY")
+#   pragma map(inflate_trees_fixed,"INTRFI")
+#   pragma map(inflate_trees_free,"INTRFR")
+#endif
+
+#endif /* _ZCONF_H */
diff --git a/zlib/zlib.3 b/zlib/zlib.3
new file mode 100644
index 000000000..3a6e45047
--- /dev/null
+++ b/zlib/zlib.3
@@ -0,0 +1,107 @@
+.TH ZLIB 3 "11 March 2002"
+.SH NAME
+zlib \- compression/decompression library
+.SH SYNOPSIS
+[see
+.I zlib.h
+for full description]
+.SH DESCRIPTION
+The
+.I zlib
+library is a general purpose data compression library.
+The code is thread safe.
+It provides in-memory compression and decompression functions,
+including integrity checks of the uncompressed data.
+This version of the library supports only one compression method (deflation)
+but other algorithms will be added later and will have the same stream interface.
+.LP
+Compression can be done in a single step if the buffers are large enough
+(for example if an input file is mmap'ed),
+or can be done by repeated calls of the compression function.
+In the latter case,
+the application must provide more input and/or consume the output
+(providing more output space) before each call.
+.LP
+The library also supports reading and writing files in
+.I gzip
+(.gz) format
+with an interface similar to that of stdio.
+.LP
+The library does not install any signal handler. The decoder checks
+the consistency of the compressed data, so the library should never
+crash even in case of corrupted input.
+.LP
+All functions of the compression library are documented in the file
+.IR zlib.h.
+The distribution source includes examples of use of the library
+the files
+.I example.c
+and
+.IR minigzip.c .
+.LP
+A Java implementation of
+.IR zlib
+is available in the Java Development Kit 1.1
+.IP
+http://www.javasoft.com/products/JDK/1.1/docs/api/Package-java.util.zip.html
+.LP
+A Perl interface to
+.IR zlib ,
+written by Paul Marquess (pmarquess@bfsec.bt.co.uk)
+is available at CPAN (Comprehensive Perl Archive Network) sites,
+such as:
+.IP
+ftp://ftp.cis.ufl.edu/pub/perl/CPAN/modules/by-module/Compress/Compress-Zlib*
+.LP
+A Python interface to
+.IR zlib
+written by A.M. Kuchling <amk@magnet.com>
+is available from the Python Software Association sites, such as:
+.IP
+ftp://ftp.python.org/pub/python/contrib/Encoding/zlib*.tar.gz
+.SH "SEE ALSO"
+Questions about zlib should be sent to:
+.IP
+zlib@quest.jpl.nasa.gov
+or, if this fails, to the author addresses given below.
+The zlib home page is:
+.IP
+http://www.cdrom.com/pub/infozip/zlib/
+.LP
+The data format used by the zlib library is described by RFC
+(Request for Comments) 1950 to 1952 in the files: 
+.IP
+ftp://ds.internic.net/rfc/rfc1950.txt (zlib format)
+.br
+rfc1951.txt (deflate format)
+.br
+rfc1952.txt (gzip format)
+.LP
+These documents are also available in other formats from:
+.IP
+ftp://ftp.uu.net/graphics/png/documents/zlib/zdoc-index.html
+.SH AUTHORS
+Version 1.1.4
+Copyright (C) 1995-2002 Jean-loup Gailly (jloup@gzip.org)
+and Mark Adler (madler@alumni.caltech.edu).
+.LP
+This software is provided "as-is,"
+without any express or implied warranty.
+In no event will the authors be held liable for any damages
+arising from the use of this software.
+See the distribution directory with respect to requirements
+governing redistribution.
+The deflate format used by
+.I zlib
+was defined by Phil Katz.
+The deflate and
+.I zlib
+specifications were written by L. Peter Deutsch.
+Thanks to all the people who reported problems and suggested various
+improvements in
+.IR zlib ;
+who are too numerous to cite here.
+.LP
+UNIX manual page by R. P. C. Rodgers,
+U.S. National Library of Medicine (rodgers@nlm.nih.gov).
+.\" end of man page
diff --git a/zlib/zlib.h b/zlib/zlib.h
new file mode 100644
index 000000000..52cb529f6
--- /dev/null
+++ b/zlib/zlib.h
@@ -0,0 +1,893 @@
+/* zlib.h -- interface of the 'zlib' general purpose compression library
+  version 1.1.4, March 11th, 2002
+
+  Copyright (C) 1995-2002 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu
+
+
+  The data format used by the zlib library is described by RFCs (Request for
+  Comments) 1950 to 1952 in the files ftp://ds.internic.net/rfc/rfc1950.txt
+  (zlib format), rfc1951.txt (deflate format) and rfc1952.txt (gzip format).
+*/
+
+#ifndef _ZLIB_H
+#define _ZLIB_H
+
+#include "zconf.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZLIB_VERSION "1.1.4"
+
+/* 
+     The 'zlib' compression library provides in-memory compression and
+  decompression functions, including integrity checks of the uncompressed
+  data.  This version of the library supports only one compression method
+  (deflation) but other algorithms will be added later and will have the same
+  stream interface.
+
+     Compression can be done in a single step if the buffers are large
+  enough (for example if an input file is mmap'ed), or can be done by
+  repeated calls of the compression function.  In the latter case, the
+  application must provide more input and/or consume the output
+  (providing more output space) before each call.
+
+     The library also supports reading and writing files in gzip (.gz) format
+  with an interface similar to that of stdio.
+
+     The library does not install any signal handler. The decoder checks
+  the consistency of the compressed data, so the library should never
+  crash even in case of corrupted input.
+*/
+
+typedef voidpf (*alloc_func) OF((voidpf opaque, uInt items, uInt size));
+typedef void   (*free_func)  OF((voidpf opaque, voidpf address));
+
+struct internal_state;
+
+typedef struct z_stream_s {
+    Bytef    *next_in;  /* next input byte */
+    uInt     avail_in;  /* number of bytes available at next_in */
+    uLong    total_in;  /* total nb of input bytes read so far */
+
+    Bytef    *next_out; /* next output byte should be put there */
+    uInt     avail_out; /* remaining free space at next_out */
+    uLong    total_out; /* total nb of bytes output so far */
+
+    char     *msg;      /* last error message, NULL if no error */
+    struct internal_state FAR *state; /* not visible by applications */
+
+    alloc_func zalloc;  /* used to allocate the internal state */
+    free_func  zfree;   /* used to free the internal state */
+    voidpf     opaque;  /* private data object passed to zalloc and zfree */
+
+    int     data_type;  /* best guess about the data type: ascii or binary */
+    uLong   adler;      /* adler32 value of the uncompressed data */
+    uLong   reserved;   /* reserved for future use */
+} z_stream;
+
+typedef z_stream FAR *z_streamp;
+
+/*
+   The application must update next_in and avail_in when avail_in has
+   dropped to zero. It must update next_out and avail_out when avail_out
+   has dropped to zero. The application must initialize zalloc, zfree and
+   opaque before calling the init function. All other fields are set by the
+   compression library and must not be updated by the application.
+
+   The opaque value provided by the application will be passed as the first
+   parameter for calls of zalloc and zfree. This can be useful for custom
+   memory management. The compression library attaches no meaning to the
+   opaque value.
+
+   zalloc must return Z_NULL if there is not enough memory for the object.
+   If zlib is used in a multi-threaded application, zalloc and zfree must be
+   thread safe.
+
+   On 16-bit systems, the functions zalloc and zfree must be able to allocate
+   exactly 65536 bytes, but will not be required to allocate more than this
+   if the symbol MAXSEG_64K is defined (see zconf.h). WARNING: On MSDOS,
+   pointers returned by zalloc for objects of exactly 65536 bytes *must*
+   have their offset normalized to zero. The default allocation function
+   provided by this library ensures this (see zutil.c). To reduce memory
+   requirements and avoid any allocation of 64K objects, at the expense of
+   compression ratio, compile the library with -DMAX_WBITS=14 (see zconf.h).
+
+   The fields total_in and total_out can be used for statistics or
+   progress reports. After compression, total_in holds the total size of
+   the uncompressed data and may be saved for use in the decompressor
+   (particularly if the decompressor wants to decompress everything in
+   a single step).
+*/
+
+                        /* constants */
+
+#define Z_NO_FLUSH      0
+#define Z_PARTIAL_FLUSH 1 /* will be removed, use Z_SYNC_FLUSH instead */
+#define Z_SYNC_FLUSH    2
+#define Z_FULL_FLUSH    3
+#define Z_FINISH        4
+/* Allowed flush values; see deflate() below for details */
+
+#define Z_OK            0
+#define Z_STREAM_END    1
+#define Z_NEED_DICT     2
+#define Z_ERRNO        (-1)
+#define Z_STREAM_ERROR (-2)
+#define Z_DATA_ERROR   (-3)
+#define Z_MEM_ERROR    (-4)
+#define Z_BUF_ERROR    (-5)
+#define Z_VERSION_ERROR (-6)
+/* Return codes for the compression/decompression functions. Negative
+ * values are errors, positive values are used for special but normal events.
+ */
+
+#define Z_NO_COMPRESSION         0
+#define Z_BEST_SPEED             1
+#define Z_BEST_COMPRESSION       9
+#define Z_DEFAULT_COMPRESSION  (-1)
+/* compression levels */
+
+#define Z_FILTERED            1
+#define Z_HUFFMAN_ONLY        2
+#define Z_DEFAULT_STRATEGY    0
+/* compression strategy; see deflateInit2() below for details */
+
+#define Z_BINARY   0
+#define Z_ASCII    1
+#define Z_UNKNOWN  2
+/* Possible values of the data_type field */
+
+#define Z_DEFLATED   8
+/* The deflate compression method (the only one supported in this version) */
+
+#define Z_NULL  0  /* for initializing zalloc, zfree, opaque */
+
+#define zlib_version zlibVersion()
+/* for compatibility with versions < 1.0.2 */
+
+                        /* basic functions */
+
+ZEXTERN const char * ZEXPORT zlibVersion OF((void));
+/* The application can compare zlibVersion and ZLIB_VERSION for consistency.
+   If the first character differs, the library code actually used is
+   not compatible with the zlib.h header file used by the application.
+   This check is automatically made by deflateInit and inflateInit.
+ */
+
+/* 
+ZEXTERN int ZEXPORT deflateInit OF((z_streamp strm, int level));
+
+     Initializes the internal stream state for compression. The fields
+   zalloc, zfree and opaque must be initialized before by the caller.
+   If zalloc and zfree are set to Z_NULL, deflateInit updates them to
+   use default allocation functions.
+
+     The compression level must be Z_DEFAULT_COMPRESSION, or between 0 and 9:
+   1 gives best speed, 9 gives best compression, 0 gives no compression at
+   all (the input data is simply copied a block at a time).
+   Z_DEFAULT_COMPRESSION requests a default compromise between speed and
+   compression (currently equivalent to level 6).
+
+     deflateInit returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_STREAM_ERROR if level is not a valid compression level,
+   Z_VERSION_ERROR if the zlib library version (zlib_version) is incompatible
+   with the version assumed by the caller (ZLIB_VERSION).
+   msg is set to null if there is no error message.  deflateInit does not
+   perform any compression: this will be done by deflate().
+*/
+
+
+ZEXTERN int ZEXPORT deflate OF((z_streamp strm, int flush));
+/*
+    deflate compresses as much data as possible, and stops when the input
+  buffer becomes empty or the output buffer becomes full. It may introduce some
+  output latency (reading input without producing any output) except when
+  forced to flush.
+
+    The detailed semantics are as follows. deflate performs one or both of the
+  following actions:
+
+  - Compress more input starting at next_in and update next_in and avail_in
+    accordingly. If not all input can be processed (because there is not
+    enough room in the output buffer), next_in and avail_in are updated and
+    processing will resume at this point for the next call of deflate().
+
+  - Provide more output starting at next_out and update next_out and avail_out
+    accordingly. This action is forced if the parameter flush is non zero.
+    Forcing flush frequently degrades the compression ratio, so this parameter
+    should be set only when necessary (in interactive applications).
+    Some output may be provided even if flush is not set.
+
+  Before the call of deflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming
+  more output, and updating avail_in or avail_out accordingly; avail_out
+  should never be zero before the call. The application can consume the
+  compressed output when it wants, for example when the output buffer is full
+  (avail_out == 0), or after each call of deflate(). If deflate returns Z_OK
+  and with zero avail_out, it must be called again after making room in the
+  output buffer because there might be more output pending.
+
+    If the parameter flush is set to Z_SYNC_FLUSH, all pending output is
+  flushed to the output buffer and the output is aligned on a byte boundary, so
+  that the decompressor can get all input data available so far. (In particular
+  avail_in is zero after the call if enough output space has been provided
+  before the call.)  Flushing may degrade compression for some compression
+  algorithms and so it should be used only when necessary.
+
+    If flush is set to Z_FULL_FLUSH, all output is flushed as with
+  Z_SYNC_FLUSH, and the compression state is reset so that decompression can
+  restart from this point if previous compressed data has been damaged or if
+  random access is desired. Using Z_FULL_FLUSH too often can seriously degrade
+  the compression.
+
+    If deflate returns with avail_out == 0, this function must be called again
+  with the same value of the flush parameter and more output space (updated
+  avail_out), until the flush is complete (deflate returns with non-zero
+  avail_out).
+
+    If the parameter flush is set to Z_FINISH, pending input is processed,
+  pending output is flushed and deflate returns with Z_STREAM_END if there
+  was enough output space; if deflate returns with Z_OK, this function must be
+  called again with Z_FINISH and more output space (updated avail_out) but no
+  more input data, until it returns with Z_STREAM_END or an error. After
+  deflate has returned Z_STREAM_END, the only possible operations on the
+  stream are deflateReset or deflateEnd.
+  
+    Z_FINISH can be used immediately after deflateInit if all the compression
+  is to be done in a single step. In this case, avail_out must be at least
+  0.1% larger than avail_in plus 12 bytes.  If deflate does not return
+  Z_STREAM_END, then it must be called again as described above.
+
+    deflate() sets strm->adler to the adler32 checksum of all input read
+  so far (that is, total_in bytes).
+
+    deflate() may update data_type if it can make a good guess about
+  the input data type (Z_ASCII or Z_BINARY). In doubt, the data is considered
+  binary. This field is only for information purposes and does not affect
+  the compression algorithm in any manner.
+
+    deflate() returns Z_OK if some progress has been made (more input
+  processed or more output produced), Z_STREAM_END if all input has been
+  consumed and all output has been produced (only when flush is set to
+  Z_FINISH), Z_STREAM_ERROR if the stream state was inconsistent (for example
+  if next_in or next_out was NULL), Z_BUF_ERROR if no progress is possible
+  (for example avail_in or avail_out was zero).
+*/
+
+
+ZEXTERN int ZEXPORT deflateEnd OF((z_streamp strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any
+   pending output.
+
+     deflateEnd returns Z_OK if success, Z_STREAM_ERROR if the
+   stream state was inconsistent, Z_DATA_ERROR if the stream was freed
+   prematurely (some input or output was discarded). In the error case,
+   msg may be set but then points to a static string (which must not be
+   deallocated).
+*/
+
+
+/* 
+ZEXTERN int ZEXPORT inflateInit OF((z_streamp strm));
+
+     Initializes the internal stream state for decompression. The fields
+   next_in, avail_in, zalloc, zfree and opaque must be initialized before by
+   the caller. If next_in is not Z_NULL and avail_in is large enough (the exact
+   value depends on the compression method), inflateInit determines the
+   compression method from the zlib header and allocates all data structures
+   accordingly; otherwise the allocation will be deferred to the first call of
+   inflate.  If zalloc and zfree are set to Z_NULL, inflateInit updates them to
+   use default allocation functions.
+
+     inflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_VERSION_ERROR if the zlib library version is incompatible with the
+   version assumed by the caller.  msg is set to null if there is no error
+   message. inflateInit does not perform any decompression apart from reading
+   the zlib header if present: this will be done by inflate().  (So next_in and
+   avail_in may be modified, but next_out and avail_out are unchanged.)
+*/
+
+
+ZEXTERN int ZEXPORT inflate OF((z_streamp strm, int flush));
+/*
+    inflate decompresses as much data as possible, and stops when the input
+  buffer becomes empty or the output buffer becomes full. It may some
+  introduce some output latency (reading input without producing any output)
+  except when forced to flush.
+
+  The detailed semantics are as follows. inflate performs one or both of the
+  following actions:
+
+  - Decompress more input starting at next_in and update next_in and avail_in
+    accordingly. If not all input can be processed (because there is not
+    enough room in the output buffer), next_in is updated and processing
+    will resume at this point for the next call of inflate().
+
+  - Provide more output starting at next_out and update next_out and avail_out
+    accordingly.  inflate() provides as much output as possible, until there
+    is no more input data or no more space in the output buffer (see below
+    about the flush parameter).
+
+  Before the call of inflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming
+  more output, and updating the next_* and avail_* values accordingly.
+  The application can consume the uncompressed output when it wants, for
+  example when the output buffer is full (avail_out == 0), or after each
+  call of inflate(). If inflate returns Z_OK and with zero avail_out, it
+  must be called again after making room in the output buffer because there
+  might be more output pending.
+
+    If the parameter flush is set to Z_SYNC_FLUSH, inflate flushes as much
+  output as possible to the output buffer. The flushing behavior of inflate is
+  not specified for values of the flush parameter other than Z_SYNC_FLUSH
+  and Z_FINISH, but the current implementation actually flushes as much output
+  as possible anyway.
+
+    inflate() should normally be called until it returns Z_STREAM_END or an
+  error. However if all decompression is to be performed in a single step
+  (a single call of inflate), the parameter flush should be set to
+  Z_FINISH. In this case all pending input is processed and all pending
+  output is flushed; avail_out must be large enough to hold all the
+  uncompressed data. (The size of the uncompressed data may have been saved
+  by the compressor for this purpose.) The next operation on this stream must
+  be inflateEnd to deallocate the decompression state. The use of Z_FINISH
+  is never required, but can be used to inform inflate that a faster routine
+  may be used for the single inflate() call.
+
+     If a preset dictionary is needed at this point (see inflateSetDictionary
+  below), inflate sets strm-adler to the adler32 checksum of the
+  dictionary chosen by the compressor and returns Z_NEED_DICT; otherwise 
+  it sets strm->adler to the adler32 checksum of all output produced
+  so far (that is, total_out bytes) and returns Z_OK, Z_STREAM_END or
+  an error code as described below. At the end of the stream, inflate()
+  checks that its computed adler32 checksum is equal to that saved by the
+  compressor and returns Z_STREAM_END only if the checksum is correct.
+
+    inflate() returns Z_OK if some progress has been made (more input processed
+  or more output produced), Z_STREAM_END if the end of the compressed data has
+  been reached and all uncompressed output has been produced, Z_NEED_DICT if a
+  preset dictionary is needed at this point, Z_DATA_ERROR if the input data was
+  corrupted (input stream not conforming to the zlib format or incorrect
+  adler32 checksum), Z_STREAM_ERROR if the stream structure was inconsistent
+  (for example if next_in or next_out was NULL), Z_MEM_ERROR if there was not
+  enough memory, Z_BUF_ERROR if no progress is possible or if there was not
+  enough room in the output buffer when Z_FINISH is used. In the Z_DATA_ERROR
+  case, the application may then call inflateSync to look for a good
+  compression block.
+*/
+
+
+ZEXTERN int ZEXPORT inflateEnd OF((z_streamp strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any
+   pending output.
+
+     inflateEnd returns Z_OK if success, Z_STREAM_ERROR if the stream state
+   was inconsistent. In the error case, msg may be set but then points to a
+   static string (which must not be deallocated).
+*/
+
+                        /* Advanced functions */
+
+/*
+    The following functions are needed only in some special applications.
+*/
+
+/*   
+ZEXTERN int ZEXPORT deflateInit2 OF((z_streamp strm,
+                                     int  level,
+                                     int  method,
+                                     int  windowBits,
+                                     int  memLevel,
+                                     int  strategy));
+
+     This is another version of deflateInit with more compression options. The
+   fields next_in, zalloc, zfree and opaque must be initialized before by
+   the caller.
+
+     The method parameter is the compression method. It must be Z_DEFLATED in
+   this version of the library.
+
+     The windowBits parameter is the base two logarithm of the window size
+   (the size of the history buffer).  It should be in the range 8..15 for this
+   version of the library. Larger values of this parameter result in better
+   compression at the expense of memory usage. The default value is 15 if
+   deflateInit is used instead.
+
+     The memLevel parameter specifies how much memory should be allocated
+   for the internal compression state. memLevel=1 uses minimum memory but
+   is slow and reduces compression ratio; memLevel=9 uses maximum memory
+   for optimal speed. The default value is 8. See zconf.h for total memory
+   usage as a function of windowBits and memLevel.
+
+     The strategy parameter is used to tune the compression algorithm. Use the
+   value Z_DEFAULT_STRATEGY for normal data, Z_FILTERED for data produced by a
+   filter (or predictor), or Z_HUFFMAN_ONLY to force Huffman encoding only (no
+   string match).  Filtered data consists mostly of small values with a
+   somewhat random distribution. In this case, the compression algorithm is
+   tuned to compress them better. The effect of Z_FILTERED is to force more
+   Huffman coding and less string matching; it is somewhat intermediate
+   between Z_DEFAULT and Z_HUFFMAN_ONLY. The strategy parameter only affects
+   the compression ratio but not the correctness of the compressed output even
+   if it is not set appropriately.
+
+      deflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_STREAM_ERROR if a parameter is invalid (such as an invalid
+   method). msg is set to null if there is no error message.  deflateInit2 does
+   not perform any compression: this will be done by deflate().
+*/
+                            
+ZEXTERN int ZEXPORT deflateSetDictionary OF((z_streamp strm,
+                                             const Bytef *dictionary,
+                                             uInt  dictLength));
+/*
+     Initializes the compression dictionary from the given byte sequence
+   without producing any compressed output. This function must be called
+   immediately after deflateInit, deflateInit2 or deflateReset, before any
+   call of deflate. The compressor and decompressor must use exactly the same
+   dictionary (see inflateSetDictionary).
+
+     The dictionary should consist of strings (byte sequences) that are likely
+   to be encountered later in the data to be compressed, with the most commonly
+   used strings preferably put towards the end of the dictionary. Using a
+   dictionary is most useful when the data to be compressed is short and can be
+   predicted with good accuracy; the data can then be compressed better than
+   with the default empty dictionary.
+
+     Depending on the size of the compression data structures selected by
+   deflateInit or deflateInit2, a part of the dictionary may in effect be
+   discarded, for example if the dictionary is larger than the window size in
+   deflate or deflate2. Thus the strings most likely to be useful should be
+   put at the end of the dictionary, not at the front.
+
+     Upon return of this function, strm->adler is set to the Adler32 value
+   of the dictionary; the decompressor may later use this value to determine
+   which dictionary has been used by the compressor. (The Adler32 value
+   applies to the whole dictionary even if only a subset of the dictionary is
+   actually used by the compressor.)
+
+     deflateSetDictionary returns Z_OK if success, or Z_STREAM_ERROR if a
+   parameter is invalid (such as NULL dictionary) or the stream state is
+   inconsistent (for example if deflate has already been called for this stream
+   or if the compression method is bsort). deflateSetDictionary does not
+   perform any compression: this will be done by deflate().
+*/
+
+ZEXTERN int ZEXPORT deflateCopy OF((z_streamp dest,
+                                    z_streamp source));
+/*
+     Sets the destination stream as a complete copy of the source stream.
+
+     This function can be useful when several compression strategies will be
+   tried, for example when there are several ways of pre-processing the input
+   data with a filter. The streams that will be discarded should then be freed
+   by calling deflateEnd.  Note that deflateCopy duplicates the internal
+   compression state which can be quite large, so this strategy is slow and
+   can consume lots of memory.
+
+     deflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_STREAM_ERROR if the source stream state was inconsistent
+   (such as zalloc being NULL). msg is left unchanged in both source and
+   destination.
+*/
+
+ZEXTERN int ZEXPORT deflateReset OF((z_streamp strm));
+/*
+     This function is equivalent to deflateEnd followed by deflateInit,
+   but does not free and reallocate all the internal compression state.
+   The stream will keep the same compression level and any other attributes
+   that may have been set by deflateInit2.
+
+      deflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being NULL).
+*/
+
+ZEXTERN int ZEXPORT deflateParams OF((z_streamp strm,
+				      int level,
+				      int strategy));
+/*
+     Dynamically update the compression level and compression strategy.  The
+   interpretation of level and strategy is as in deflateInit2.  This can be
+   used to switch between compression and straight copy of the input data, or
+   to switch to a different kind of input data requiring a different
+   strategy. If the compression level is changed, the input available so far
+   is compressed with the old level (and may be flushed); the new level will
+   take effect only at the next call of deflate().
+
+     Before the call of deflateParams, the stream state must be set as for
+   a call of deflate(), since the currently available input may have to
+   be compressed and flushed. In particular, strm->avail_out must be non-zero.
+
+     deflateParams returns Z_OK if success, Z_STREAM_ERROR if the source
+   stream state was inconsistent or if a parameter was invalid, Z_BUF_ERROR
+   if strm->avail_out was zero.
+*/
+
+/*   
+ZEXTERN int ZEXPORT inflateInit2 OF((z_streamp strm,
+                                     int  windowBits));
+
+     This is another version of inflateInit with an extra parameter. The
+   fields next_in, avail_in, zalloc, zfree and opaque must be initialized
+   before by the caller.
+
+     The windowBits parameter is the base two logarithm of the maximum window
+   size (the size of the history buffer).  It should be in the range 8..15 for
+   this version of the library. The default value is 15 if inflateInit is used
+   instead. If a compressed stream with a larger window size is given as
+   input, inflate() will return with the error code Z_DATA_ERROR instead of
+   trying to allocate a larger window.
+
+      inflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_STREAM_ERROR if a parameter is invalid (such as a negative
+   memLevel). msg is set to null if there is no error message.  inflateInit2
+   does not perform any decompression apart from reading the zlib header if
+   present: this will be done by inflate(). (So next_in and avail_in may be
+   modified, but next_out and avail_out are unchanged.)
+*/
+
+ZEXTERN int ZEXPORT inflateSetDictionary OF((z_streamp strm,
+                                             const Bytef *dictionary,
+                                             uInt  dictLength));
+/*
+     Initializes the decompression dictionary from the given uncompressed byte
+   sequence. This function must be called immediately after a call of inflate
+   if this call returned Z_NEED_DICT. The dictionary chosen by the compressor
+   can be determined from the Adler32 value returned by this call of
+   inflate. The compressor and decompressor must use exactly the same
+   dictionary (see deflateSetDictionary).
+
+     inflateSetDictionary returns Z_OK if success, Z_STREAM_ERROR if a
+   parameter is invalid (such as NULL dictionary) or the stream state is
+   inconsistent, Z_DATA_ERROR if the given dictionary doesn't match the
+   expected one (incorrect Adler32 value). inflateSetDictionary does not
+   perform any decompression: this will be done by subsequent calls of
+   inflate().
+*/
+
+ZEXTERN int ZEXPORT inflateSync OF((z_streamp strm));
+/* 
+    Skips invalid compressed data until a full flush point (see above the
+  description of deflate with Z_FULL_FLUSH) can be found, or until all
+  available input is skipped. No output is provided.
+
+    inflateSync returns Z_OK if a full flush point has been found, Z_BUF_ERROR
+  if no more input was provided, Z_DATA_ERROR if no flush point has been found,
+  or Z_STREAM_ERROR if the stream structure was inconsistent. In the success
+  case, the application may save the current current value of total_in which
+  indicates where valid compressed data was found. In the error case, the
+  application may repeatedly call inflateSync, providing more input each time,
+  until success or end of the input data.
+*/
+
+ZEXTERN int ZEXPORT inflateReset OF((z_streamp strm));
+/*
+     This function is equivalent to inflateEnd followed by inflateInit,
+   but does not free and reallocate all the internal decompression state.
+   The stream will keep attributes that may have been set by inflateInit2.
+
+      inflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being NULL).
+*/
+
+
+                        /* utility functions */
+
+/*
+     The following utility functions are implemented on top of the
+   basic stream-oriented functions. To simplify the interface, some
+   default options are assumed (compression level and memory usage,
+   standard memory allocation functions). The source code of these
+   utility functions can easily be modified if you need special options.
+*/
+
+ZEXTERN int ZEXPORT compress OF((Bytef *dest,   uLongf *destLen,
+                                 const Bytef *source, uLong sourceLen));
+/*
+     Compresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer. Upon entry, destLen is the total
+   size of the destination buffer, which must be at least 0.1% larger than
+   sourceLen plus 12 bytes. Upon exit, destLen is the actual size of the
+   compressed buffer.
+     This function can be used to compress a whole file at once if the
+   input file is mmap'ed.
+     compress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer.
+*/
+
+ZEXTERN int ZEXPORT compress2 OF((Bytef *dest,   uLongf *destLen,
+                                  const Bytef *source, uLong sourceLen,
+                                  int level));
+/*
+     Compresses the source buffer into the destination buffer. The level
+   parameter has the same meaning as in deflateInit.  sourceLen is the byte
+   length of the source buffer. Upon entry, destLen is the total size of the
+   destination buffer, which must be at least 0.1% larger than sourceLen plus
+   12 bytes. Upon exit, destLen is the actual size of the compressed buffer.
+
+     compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer,
+   Z_STREAM_ERROR if the level parameter is invalid.
+*/
+
+ZEXTERN int ZEXPORT uncompress OF((Bytef *dest,   uLongf *destLen,
+                                   const Bytef *source, uLong sourceLen));
+/*
+     Decompresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer. Upon entry, destLen is the total
+   size of the destination buffer, which must be large enough to hold the
+   entire uncompressed data. (The size of the uncompressed data must have
+   been saved previously by the compressor and transmitted to the decompressor
+   by some mechanism outside the scope of this compression library.)
+   Upon exit, destLen is the actual size of the compressed buffer.
+     This function can be used to decompress a whole file at once if the
+   input file is mmap'ed.
+
+     uncompress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer, or Z_DATA_ERROR if the input data was corrupted.
+*/
+
+
+typedef voidp gzFile;
+
+ZEXTERN gzFile ZEXPORT gzopen  OF((const char *path, const char *mode));
+/*
+     Opens a gzip (.gz) file for reading or writing. The mode parameter
+   is as in fopen ("rb" or "wb") but can also include a compression level
+   ("wb9") or a strategy: 'f' for filtered data as in "wb6f", 'h' for
+   Huffman only compression as in "wb1h". (See the description
+   of deflateInit2 for more information about the strategy parameter.)
+
+     gzopen can be used to read a file which is not in gzip format; in this
+   case gzread will directly read from the file without decompression.
+
+     gzopen returns NULL if the file could not be opened or if there was
+   insufficient memory to allocate the (de)compression state; errno
+   can be checked to distinguish the two cases (if errno is zero, the
+   zlib error is Z_MEM_ERROR).  */
+
+ZEXTERN gzFile ZEXPORT gzdopen  OF((int fd, const char *mode));
+/*
+     gzdopen() associates a gzFile with the file descriptor fd.  File
+   descriptors are obtained from calls like open, dup, creat, pipe or
+   fileno (in the file has been previously opened with fopen).
+   The mode parameter is as in gzopen.
+     The next call of gzclose on the returned gzFile will also close the
+   file descriptor fd, just like fclose(fdopen(fd), mode) closes the file
+   descriptor fd. If you want to keep fd open, use gzdopen(dup(fd), mode).
+     gzdopen returns NULL if there was insufficient memory to allocate
+   the (de)compression state.
+*/
+
+ZEXTERN int ZEXPORT gzsetparams OF((gzFile file, int level, int strategy));
+/*
+     Dynamically update the compression level or strategy. See the description
+   of deflateInit2 for the meaning of these parameters.
+     gzsetparams returns Z_OK if success, or Z_STREAM_ERROR if the file was not
+   opened for writing.
+*/
+
+ZEXTERN int ZEXPORT    gzread  OF((gzFile file, voidp buf, unsigned len));
+/*
+     Reads the given number of uncompressed bytes from the compressed file.
+   If the input file was not in gzip format, gzread copies the given number
+   of bytes into the buffer.
+     gzread returns the number of uncompressed bytes actually read (0 for
+   end of file, -1 for error). */
+
+ZEXTERN int ZEXPORT    gzwrite OF((gzFile file, 
+				   const voidp buf, unsigned len));
+/*
+     Writes the given number of uncompressed bytes into the compressed file.
+   gzwrite returns the number of uncompressed bytes actually written
+   (0 in case of error).
+*/
+
+ZEXTERN int ZEXPORTVA   gzprintf OF((gzFile file, const char *format, ...));
+/*
+     Converts, formats, and writes the args to the compressed file under
+   control of the format string, as in fprintf. gzprintf returns the number of
+   uncompressed bytes actually written (0 in case of error).
+*/
+
+ZEXTERN int ZEXPORT gzputs OF((gzFile file, const char *s));
+/*
+      Writes the given null-terminated string to the compressed file, excluding
+   the terminating null character.
+      gzputs returns the number of characters written, or -1 in case of error.
+*/
+
+ZEXTERN char * ZEXPORT gzgets OF((gzFile file, char *buf, int len));
+/*
+      Reads bytes from the compressed file until len-1 characters are read, or
+   a newline character is read and transferred to buf, or an end-of-file
+   condition is encountered.  The string is then terminated with a null
+   character.
+      gzgets returns buf, or Z_NULL in case of error.
+*/
+
+ZEXTERN int ZEXPORT    gzputc OF((gzFile file, int c));
+/*
+      Writes c, converted to an unsigned char, into the compressed file.
+   gzputc returns the value that was written, or -1 in case of error.
+*/
+
+ZEXTERN int ZEXPORT    gzgetc OF((gzFile file));
+/*
+      Reads one byte from the compressed file. gzgetc returns this byte
+   or -1 in case of end of file or error.
+*/
+
+ZEXTERN int ZEXPORT    gzflush OF((gzFile file, int flush));
+/*
+     Flushes all pending output into the compressed file. The parameter
+   flush is as in the deflate() function. The return value is the zlib
+   error number (see function gzerror below). gzflush returns Z_OK if
+   the flush parameter is Z_FINISH and all output could be flushed.
+     gzflush should be called only when strictly necessary because it can
+   degrade compression.
+*/
+
+ZEXTERN z_off_t ZEXPORT    gzseek OF((gzFile file,
+				      z_off_t offset, int whence));
+/* 
+      Sets the starting position for the next gzread or gzwrite on the
+   given compressed file. The offset represents a number of bytes in the
+   uncompressed data stream. The whence parameter is defined as in lseek(2);
+   the value SEEK_END is not supported.
+     If the file is opened for reading, this function is emulated but can be
+   extremely slow. If the file is opened for writing, only forward seeks are
+   supported; gzseek then compresses a sequence of zeroes up to the new
+   starting position.
+
+      gzseek returns the resulting offset location as measured in bytes from
+   the beginning of the uncompressed stream, or -1 in case of error, in
+   particular if the file is opened for writing and the new starting position
+   would be before the current position.
+*/
+
+ZEXTERN int ZEXPORT    gzrewind OF((gzFile file));
+/*
+     Rewinds the given file. This function is supported only for reading.
+
+   gzrewind(file) is equivalent to (int)gzseek(file, 0L, SEEK_SET)
+*/
+
+ZEXTERN z_off_t ZEXPORT    gztell OF((gzFile file));
+/*
+     Returns the starting position for the next gzread or gzwrite on the
+   given compressed file. This position represents a number of bytes in the
+   uncompressed data stream.
+
+   gztell(file) is equivalent to gzseek(file, 0L, SEEK_CUR)
+*/
+
+ZEXTERN int ZEXPORT gzeof OF((gzFile file));
+/*
+     Returns 1 when EOF has previously been detected reading the given
+   input stream, otherwise zero.
+*/
+
+ZEXTERN int ZEXPORT    gzclose OF((gzFile file));
+/*
+     Flushes all pending output if necessary, closes the compressed file
+   and deallocates all the (de)compression state. The return value is the zlib
+   error number (see function gzerror below).
+*/
+
+ZEXTERN const char * ZEXPORT gzerror OF((gzFile file, int *errnum));
+/*
+     Returns the error message for the last error which occurred on the
+   given compressed file. errnum is set to zlib error number. If an
+   error occurred in the file system and not in the compression library,
+   errnum is set to Z_ERRNO and the application may consult errno
+   to get the exact error code.
+*/
+
+                        /* checksum functions */
+
+/*
+     These functions are not related to compression but are exported
+   anyway because they might be useful in applications using the
+   compression library.
+*/
+
+ZEXTERN uLong ZEXPORT adler32 OF((uLong adler, const Bytef *buf, uInt len));
+
+/*
+     Update a running Adler-32 checksum with the bytes buf[0..len-1] and
+   return the updated checksum. If buf is NULL, this function returns
+   the required initial value for the checksum.
+   An Adler-32 checksum is almost as reliable as a CRC32 but can be computed
+   much faster. Usage example:
+
+     uLong adler = adler32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       adler = adler32(adler, buffer, length);
+     }
+     if (adler != original_adler) error();
+*/
+
+ZEXTERN uLong ZEXPORT crc32   OF((uLong crc, const Bytef *buf, uInt len));
+/*
+     Update a running crc with the bytes buf[0..len-1] and return the updated
+   crc. If buf is NULL, this function returns the required initial value
+   for the crc. Pre- and post-conditioning (one's complement) is performed
+   within this function so it shouldn't be done by the application.
+   Usage example:
+
+     uLong crc = crc32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       crc = crc32(crc, buffer, length);
+     }
+     if (crc != original_crc) error();
+*/
+
+
+                        /* various hacks, don't look :) */
+
+/* deflateInit and inflateInit are macros to allow checking the zlib version
+ * and the compiler's view of z_stream:
+ */
+ZEXTERN int ZEXPORT deflateInit_ OF((z_streamp strm, int level,
+                                     const char *version, int stream_size));
+ZEXTERN int ZEXPORT inflateInit_ OF((z_streamp strm,
+                                     const char *version, int stream_size));
+ZEXTERN int ZEXPORT deflateInit2_ OF((z_streamp strm, int  level, int  method,
+                                      int windowBits, int memLevel,
+                                      int strategy, const char *version,
+                                      int stream_size));
+ZEXTERN int ZEXPORT inflateInit2_ OF((z_streamp strm, int  windowBits,
+                                      const char *version, int stream_size));
+#define deflateInit(strm, level) \
+        deflateInit_((strm), (level),       ZLIB_VERSION, sizeof(z_stream))
+#define inflateInit(strm) \
+        inflateInit_((strm),                ZLIB_VERSION, sizeof(z_stream))
+#define deflateInit2(strm, level, method, windowBits, memLevel, strategy) \
+        deflateInit2_((strm),(level),(method),(windowBits),(memLevel),\
+                      (strategy),           ZLIB_VERSION, sizeof(z_stream))
+#define inflateInit2(strm, windowBits) \
+        inflateInit2_((strm), (windowBits), ZLIB_VERSION, sizeof(z_stream))
+
+
+#if !defined(_Z_UTIL_H) && !defined(NO_DUMMY_DECL)
+    struct internal_state {int dummy;}; /* hack for buggy compilers */
+#endif
+
+ZEXTERN const char   * ZEXPORT zError           OF((int err));
+ZEXTERN int            ZEXPORT inflateSyncPoint OF((z_streamp z));
+ZEXTERN const uLongf * ZEXPORT get_crc_table    OF((void));
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _ZLIB_H */
diff --git a/zlib/zutil.c b/zlib/zutil.c
new file mode 100644
index 000000000..e33bae954
--- /dev/null
+++ b/zlib/zutil.c
@@ -0,0 +1,225 @@
+/* zutil.c -- target dependent utility functions for the compression library
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h 
+ */
+
+/* @(#) $Id: zutil.c,v 1.1.2.1 2004/07/05 14:12:39 easysw Exp $ */
+
+#include "zutil.h"
+
+struct internal_state      {int dummy;}; /* for buggy compilers */
+
+#ifndef STDC
+extern void exit OF((int));
+#endif
+
+const char *z_errmsg[10] = {
+"need dictionary",     /* Z_NEED_DICT       2  */
+"stream end",          /* Z_STREAM_END      1  */
+"",                    /* Z_OK              0  */
+"file error",          /* Z_ERRNO         (-1) */
+"stream error",        /* Z_STREAM_ERROR  (-2) */
+"data error",          /* Z_DATA_ERROR    (-3) */
+"insufficient memory", /* Z_MEM_ERROR     (-4) */
+"buffer error",        /* Z_BUF_ERROR     (-5) */
+"incompatible version",/* Z_VERSION_ERROR (-6) */
+""};
+
+
+const char * ZEXPORT zlibVersion()
+{
+    return ZLIB_VERSION;
+}
+
+#ifdef DEBUG
+
+#  ifndef verbose
+#    define verbose 0
+#  endif
+int z_verbose = verbose;
+
+void z_error (m)
+    char *m;
+{
+    fprintf(stderr, "%s\n", m);
+    exit(1);
+}
+#endif
+
+/* exported to allow conversion of error code to string for compress() and
+ * uncompress()
+ */
+const char * ZEXPORT zError(err)
+    int err;
+{
+    return ERR_MSG(err);
+}
+
+
+#ifndef HAVE_MEMCPY
+
+void zmemcpy(dest, source, len)
+    Bytef* dest;
+    const Bytef* source;
+    uInt  len;
+{
+    if (len == 0) return;
+    do {
+        *dest++ = *source++; /* ??? to be unrolled */
+    } while (--len != 0);
+}
+
+int zmemcmp(s1, s2, len)
+    const Bytef* s1;
+    const Bytef* s2;
+    uInt  len;
+{
+    uInt j;
+
+    for (j = 0; j < len; j++) {
+        if (s1[j] != s2[j]) return 2*(s1[j] > s2[j])-1;
+    }
+    return 0;
+}
+
+void zmemzero(dest, len)
+    Bytef* dest;
+    uInt  len;
+{
+    if (len == 0) return;
+    do {
+        *dest++ = 0;  /* ??? to be unrolled */
+    } while (--len != 0);
+}
+#endif
+
+#ifdef __TURBOC__
+#if (defined( __BORLANDC__) || !defined(SMALL_MEDIUM)) && !defined(__32BIT__)
+/* Small and medium model in Turbo C are for now limited to near allocation
+ * with reduced MAX_WBITS and MAX_MEM_LEVEL
+ */
+#  define MY_ZCALLOC
+
+/* Turbo C malloc() does not allow dynamic allocation of 64K bytes
+ * and farmalloc(64K) returns a pointer with an offset of 8, so we
+ * must fix the pointer. Warning: the pointer must be put back to its
+ * original form in order to free it, use zcfree().
+ */
+
+#define MAX_PTR 10
+/* 10*64K = 640K */
+
+local int next_ptr = 0;
+
+typedef struct ptr_table_s {
+    voidpf org_ptr;
+    voidpf new_ptr;
+} ptr_table;
+
+local ptr_table table[MAX_PTR];
+/* This table is used to remember the original form of pointers
+ * to large buffers (64K). Such pointers are normalized with a zero offset.
+ * Since MSDOS is not a preemptive multitasking OS, this table is not
+ * protected from concurrent access. This hack doesn't work anyway on
+ * a protected system like OS/2. Use Microsoft C instead.
+ */
+
+voidpf zcalloc (voidpf opaque, unsigned items, unsigned size)
+{
+    voidpf buf = opaque; /* just to make some compilers happy */
+    ulg bsize = (ulg)items*size;
+
+    /* If we allocate less than 65520 bytes, we assume that farmalloc
+     * will return a usable pointer which doesn't have to be normalized.
+     */
+    if (bsize < 65520L) {
+        buf = farmalloc(bsize);
+        if (*(ush*)&buf != 0) return buf;
+    } else {
+        buf = farmalloc(bsize + 16L);
+    }
+    if (buf == NULL || next_ptr >= MAX_PTR) return NULL;
+    table[next_ptr].org_ptr = buf;
+
+    /* Normalize the pointer to seg:0 */
+    *((ush*)&buf+1) += ((ush)((uch*)buf-0) + 15) >> 4;
+    *(ush*)&buf = 0;
+    table[next_ptr++].new_ptr = buf;
+    return buf;
+}
+
+void  zcfree (voidpf opaque, voidpf ptr)
+{
+    int n;
+    if (*(ush*)&ptr != 0) { /* object < 64K */
+        farfree(ptr);
+        return;
+    }
+    /* Find the original pointer */
+    for (n = 0; n < next_ptr; n++) {
+        if (ptr != table[n].new_ptr) continue;
+
+        farfree(table[n].org_ptr);
+        while (++n < next_ptr) {
+            table[n-1] = table[n];
+        }
+        next_ptr--;
+        return;
+    }
+    ptr = opaque; /* just to make some compilers happy */
+    Assert(0, "zcfree: ptr not found");
+}
+#endif
+#endif /* __TURBOC__ */
+
+
+#if defined(M_I86) && !defined(__32BIT__)
+/* Microsoft C in 16-bit mode */
+
+#  define MY_ZCALLOC
+
+#if (!defined(_MSC_VER) || (_MSC_VER <= 600))
+#  define _halloc  halloc
+#  define _hfree   hfree
+#endif
+
+voidpf zcalloc (voidpf opaque, unsigned items, unsigned size)
+{
+    if (opaque) opaque = 0; /* to make compiler happy */
+    return _halloc((long)items, size);
+}
+
+void  zcfree (voidpf opaque, voidpf ptr)
+{
+    if (opaque) opaque = 0; /* to make compiler happy */
+    _hfree(ptr);
+}
+
+#endif /* MSC */
+
+
+#ifndef MY_ZCALLOC /* Any system without a special alloc function */
+
+#ifndef STDC
+extern voidp  calloc OF((uInt items, uInt size));
+extern void   free   OF((voidpf ptr));
+#endif
+
+voidpf zcalloc (opaque, items, size)
+    voidpf opaque;
+    unsigned items;
+    unsigned size;
+{
+    if (opaque) items += size - size; /* make compiler happy */
+    return (voidpf)calloc(items, size);
+}
+
+void  zcfree (opaque, ptr)
+    voidpf opaque;
+    voidpf ptr;
+{
+    free(ptr);
+    if (opaque) return; /* make compiler happy */
+}
+
+#endif /* MY_ZCALLOC */
diff --git a/zlib/zutil.h b/zlib/zutil.h
new file mode 100644
index 000000000..a318e9b83
--- /dev/null
+++ b/zlib/zutil.h
@@ -0,0 +1,220 @@
+/* zutil.h -- internal interface and configuration of the compression library
+ * Copyright (C) 1995-2002 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* @(#) $Id: zutil.h,v 1.1.2.1 2004/07/05 14:12:39 easysw Exp $ */
+
+#ifndef _Z_UTIL_H
+#define _Z_UTIL_H
+
+#include "zlib.h"
+
+#ifdef STDC
+#  include <stddef.h>
+#  include <string.h>
+#  include <stdlib.h>
+#endif
+#ifdef NO_ERRNO_H
+    extern int errno;
+#else
+#   include <errno.h>
+#endif
+
+#ifndef local
+#  define local static
+#endif
+/* compile with -Dlocal if your debugger can't find static symbols */
+
+typedef unsigned char  uch;
+typedef uch FAR uchf;
+typedef unsigned short ush;
+typedef ush FAR ushf;
+typedef unsigned long  ulg;
+
+extern const char *z_errmsg[10]; /* indexed by 2-zlib_error */
+/* (size given to avoid silly warnings with Visual C++) */
+
+#define ERR_MSG(err) z_errmsg[Z_NEED_DICT-(err)]
+
+#define ERR_RETURN(strm,err) \
+  return (strm->msg = (char*)ERR_MSG(err), (err))
+/* To be used only when the state is known to be valid */
+
+        /* common constants */
+
+#ifndef DEF_WBITS
+#  define DEF_WBITS MAX_WBITS
+#endif
+/* default windowBits for decompression. MAX_WBITS is for compression only */
+
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+/* default memLevel */
+
+#define STORED_BLOCK 0
+#define STATIC_TREES 1
+#define DYN_TREES    2
+/* The three kinds of block type */
+
+#define MIN_MATCH  3
+#define MAX_MATCH  258
+/* The minimum and maximum match lengths */
+
+#define PRESET_DICT 0x20 /* preset dictionary flag in zlib header */
+
+        /* target dependencies */
+
+#ifdef MSDOS
+#  define OS_CODE  0x00
+#  if defined(__TURBOC__) || defined(__BORLANDC__)
+#    if(__STDC__ == 1) && (defined(__LARGE__) || defined(__COMPACT__))
+       /* Allow compilation with ANSI keywords only enabled */
+       void _Cdecl farfree( void *block );
+       void *_Cdecl farmalloc( unsigned long nbytes );
+#    else
+#     include <alloc.h>
+#    endif
+#  else /* MSC or DJGPP */
+#    include <malloc.h>
+#  endif
+#endif
+
+#ifdef OS2
+#  define OS_CODE  0x06
+#endif
+
+#ifdef WIN32 /* Window 95 & Windows NT */
+#  define OS_CODE  0x0b
+#endif
+
+#if defined(VAXC) || defined(VMS)
+#  define OS_CODE  0x02
+#  define F_OPEN(name, mode) \
+     fopen((name), (mode), "mbc=60", "ctx=stm", "rfm=fix", "mrs=512")
+#endif
+
+#ifdef AMIGA
+#  define OS_CODE  0x01
+#endif
+
+#if defined(ATARI) || defined(atarist)
+#  define OS_CODE  0x05
+#endif
+
+#if defined(MACOS) || defined(TARGET_OS_MAC)
+#  define OS_CODE  0x07
+#  if defined(__MWERKS__) && __dest_os != __be_os && __dest_os != __win32_os
+#    include <unix.h> /* for fdopen */
+#  else
+#    ifndef fdopen
+#      define fdopen(fd,mode) NULL /* No fdopen() */
+#    endif
+#  endif
+#endif
+
+#ifdef __50SERIES /* Prime/PRIMOS */
+#  define OS_CODE  0x0F
+#endif
+
+#ifdef TOPS20
+#  define OS_CODE  0x0a
+#endif
+
+#if defined(_BEOS_) || defined(RISCOS)
+#  define fdopen(fd,mode) NULL /* No fdopen() */
+#endif
+
+#if (defined(_MSC_VER) && (_MSC_VER > 600))
+#  define fdopen(fd,type)  _fdopen(fd,type)
+#endif
+
+
+        /* Common defaults */
+
+#ifndef OS_CODE
+#  define OS_CODE  0x03  /* assume Unix */
+#endif
+
+#ifndef F_OPEN
+#  define F_OPEN(name, mode) fopen((name), (mode))
+#endif
+
+         /* functions */
+
+#ifdef HAVE_STRERROR
+   extern char *strerror OF((int));
+#  define zstrerror(errnum) strerror(errnum)
+#else
+#  define zstrerror(errnum) ""
+#endif
+
+#if defined(pyr)
+#  define NO_MEMCPY
+#endif
+#if defined(SMALL_MEDIUM) && !defined(_MSC_VER) && !defined(__SC__)
+ /* Use our own functions for small and medium model with MSC <= 5.0.
+  * You may have to use the same strategy for Borland C (untested).
+  * The __SC__ check is for Symantec.
+  */
+#  define NO_MEMCPY
+#endif
+#if defined(STDC) && !defined(HAVE_MEMCPY) && !defined(NO_MEMCPY)
+#  define HAVE_MEMCPY
+#endif
+#ifdef HAVE_MEMCPY
+#  ifdef SMALL_MEDIUM /* MSDOS small or medium model */
+#    define zmemcpy _fmemcpy
+#    define zmemcmp _fmemcmp
+#    define zmemzero(dest, len) _fmemset(dest, 0, len)
+#  else
+#    define zmemcpy memcpy
+#    define zmemcmp memcmp
+#    define zmemzero(dest, len) memset(dest, 0, len)
+#  endif
+#else
+   extern void zmemcpy  OF((Bytef* dest, const Bytef* source, uInt len));
+   extern int  zmemcmp  OF((const Bytef* s1, const Bytef* s2, uInt len));
+   extern void zmemzero OF((Bytef* dest, uInt len));
+#endif
+
+/* Diagnostic functions */
+#ifdef DEBUG
+#  include <stdio.h>
+   extern int z_verbose;
+   extern void z_error    OF((char *m));
+#  define Assert(cond,msg) {if(!(cond)) z_error(msg);}
+#  define Trace(x) {if (z_verbose>=0) fprintf x ;}
+#  define Tracev(x) {if (z_verbose>0) fprintf x ;}
+#  define Tracevv(x) {if (z_verbose>1) fprintf x ;}
+#  define Tracec(c,x) {if (z_verbose>0 && (c)) fprintf x ;}
+#  define Tracecv(c,x) {if (z_verbose>1 && (c)) fprintf x ;}
+#else
+#  define Assert(cond,msg)
+#  define Trace(x)
+#  define Tracev(x)
+#  define Tracevv(x)
+#  define Tracec(c,x)
+#  define Tracecv(c,x)
+#endif
+
+
+typedef uLong (ZEXPORT *check_func) OF((uLong check, const Bytef *buf,
+				       uInt len));
+voidpf zcalloc OF((voidpf opaque, unsigned items, unsigned size));
+void   zcfree  OF((voidpf opaque, voidpf ptr));
+
+#define ZALLOC(strm, items, size) \
+           (*((strm)->zalloc))((strm)->opaque, (items), (size))
+#define ZFREE(strm, addr)  (*((strm)->zfree))((strm)->opaque, (voidpf)(addr))
+#define TRY_FREE(s, p) {if (p) ZFREE(s, p);}
+
+#endif /* _Z_UTIL_H */