The internal read code change to stop sharing the palette was incompletely
implemented. The result is that unless palette index checking is turned off and
there are no read transformations the png_info palette gets deleted when the
png_struct is deleted. This is normally harmless (png_info gets deleted first)
but in the case of pngcp it results in use-after-free of the palette and,
therefore, palette corruption and maybe on some operating systems and access
violation.
This also updated pngcp 'search' mode to check a restricted range of memLevels;
there is an unrelated bug which means that lower zlib memLevels result in memory
corruption under some circumstances, probably less often than 1:1000.
Signed-off-by: John Bowler <jbowler@acm.org>
Also change the order of the 'level' and 'windowBits' searches to seach
windowBits first; this favours windowBits optimizations over compression level
ones on the basis that the latter should only affect the write code.
Signed-off-by: John Bowler <jbowler@acm.org>
The sequential read code failed to read to the end of the IDAT stream in about
1/820 cases, resulting in a spurious warning. The
png_set_compression_buffer_size API also would not work (or do bad things) if
the size of a zlib uInt was less than 32 bits.
This includes a quiet API change to alter png_set_compression_buffer_size to use
a png_alloc_size_t, not png_size_t and implement the correct checks.
Signed-off-by: John Bowler <jbowler@acm.org>
png_set_compression_buffer_size would result in a spurious debug assert if the
compression buffer size was set to something other than a multiple of
PNG_ROW_BUFFER_SIZE; the debug test failed to add the buffer 'start'
Signed-off-by: John Bowler <jbowler@acm.org>
This is still a work-in-progress but it seems fairly stable (if not exactly 100%
optimal). pngcp now allows 'all' for some options which iterates through all
possible settings (this reliably produces the smallest IDAT that libpng can
produce with those settings.) It also contains a --search command line option
which attempts to optimize this by skipping pointless tests; it is close, most
of the time, but not perfect.
Signed-off-by: John Bowler <jbowler@acm.org>
There are two separate problems. The first is that the CMINFO optimization code
gets run twice on any PNG IDAT stream longer than 2048 bytes and the second time
can overwrite bytes 2048,2049 destroying the output.
The second is that one of the (debug) checks was slightly wrong (< when <=
should have been used) and this causes write to abort maybe 1/2048 times.
Signed-off-by: John Bowler <jbowler@acm.org>
The handling of tIME and text chunks on read now records the location of the
chunks relative to PLTE and IDAT. Behavior on write is unchanged except that if
the position was recorded on read it will be re-used.
This involves an ABI change to the png_text_struct; a one byte location field is
added (with the same meaning as the one used to record unknown chunk location.)
Because this field is only used on read there is no API change unless a png_info
from a libpng read is passed to a subsequent libpng write (this did not work
very well before 1.7; the tIME chunk could get duplicated.)
png_set_text ignores the new field, resetting it to the current position in the
read or write stream. On write the position is set to the next location to be
written unless the write has not started (the position is before the signature)
in which case the location is set to PNG_HAVE_PLTE|PNG_AFTER_IDAT. When the
chunk is written the position is set to the actual write location (effectively
the position is frozen.)
Signed-off-by: John Bowler <jbowler@acm.org>
This adds code to iterate through the command line options when non-list options
have the 'all' parameter. Used to rapidly test the effect of different zlib
options.
Signed-off-by: John Bowler <jbowler@acm.org>
This seems safer; in fact a NULL dereference never happens because the test ends
up just doing arithmetic on NULL but it seems possible that the undefined
arithmetic can still happen if there is a bug elsewhere.
Signed-off-by: John Bowler <jbowler@acm.org>
Undefined behaviour sanitizer currently fires for this test
with an error about member access within null pointer to
struct; put the check that relies on pz->list being non-NULL
second.
Also add an example program, pngcp.c, which illustrates the problem when used to
copy pngtest.png (the result is an invalid PNG because the tIME chunk is
duplicated.)
Signed-off-by: John Bowler <jbowler@acm.org>
This rewrites the code used previously in the heuristics to make it easier to
debug and introduces the 'methodical' method, which is intended to be an
expensive but reliable way of reducing image size.
The code in this commit does not work; the 'methodical' test for success does
not take account of data buffered inside zlib and, anyway, it changes the
results of pngtest so that the test fails. This commit is just a checkpoint of
the current state; another commit will temporarily disable the 'methodical'
code.
Signed-off-by: John Bowler <jbowler@acm.org>
Undefined because (png_byte) << shift is undefined if it changes the signed bit
(because png_byte is promoted to int). The libpng exported functions
png_get_uint_32 and png_get_uint_16 handle this.
Bug reported by David Drysdale as a result of reports from UBSAN in clang 3.8.
This changes pngvalid to use BE random numbers; this used to produce errors but
these should not be fixed as a result of the previous changes.
Signed-off-by: John Bowler <jbowler@acm.org>
This widens the 'limit' check on the internally calculated error limits in the
'DIGITIZE' case (the code used prior to 1.7 for rgb_to_gray error checks) and it
changes the check to only operate in non-release builds (base build type not RC
or RELEASE.)
Signed-off-by: John Bowler <jbowler@acm.org>
Fixes to pngvalid for various reduced build configurations (eliminate unused
statics) and a fix for the case in rgb_to_gray when the digitize option reduces
graylo to 0, producing a large error.
Signed-off-by: John Bowler <jbowler@acm.org>
This 'fixes' the current problem in the BE tests by not testing it, making the
BE code the same as the LE version.
Signed-off-by: John Bowler <jbowler@acm.org>
The filter selection options are made more intuitive (IMO, but I am not an
Intuitive User). Dummy code added for methodical filter selection.
Signed-off-by: John Bowler <jbowler@acm.org>
This implements the heuristic part of filter selction, the methodic testing
approach is still not implemented. png_set_option methods are incomplete.
Signed-off-by: John Bowler <jbowler@acm.org>
This removes the side-effect on the png_struct palette of calling png_set_PLTE
or png_set_tRNS. NOTE: this is a quiet API change, it was possible before to
alter the palette on a PNG image by using png_set_PLTE, but this was unintended
and inconsistent with the other png_set APIs.
Fix a bug in palette index checking; png_struct::num_palette could, in
principle, get changed by the transformations (e.g. png_set_quantize) and this
would invalidate the check. The palette checking init function now makes a copy
of png_struct::num_palette.
Fix a bug in pngvalid error handling. A png_error in png_write_info is not
continuable (a valid image cannot necessarily be written afterward) because the
png_error aborts the write of subsequent pre-IDAT chunks. In particular an
abort as a result of a bogus colorspace information (gAMA, cHRM, sBIT etc)
prevents the write of the PLTE chunk.
Signed-off-by: John Bowler <jbowler@acm.org>
Unify the compression code so that inflate calls are localized to a common
routine. Ground work for filter selection support. Minor API changes to use
void* not byte* for data parameters. Unification of some of the compression
code with the decompression code; IDAT_size replaces IDAT_read_size and
zbuffer_size, IDAT reading and writing is no longer controlled by the size of
the compression buffer.
Signed-off-by: John Bowler <jbowler@acm.org>
Fixes the case where IDAT is treated as unknown and saved; the first IDAT chunk
ends up 'before' IDAT, the rest 'after', pngunknown has to take this into
account since it checks before and after.
This commit moves code round and changes the filter write interfaces that took
png_uint_32 buffer pixel counts to unsigned int. Also moves compression code
and definitions into pngwutil.c so that the compression code is isolated from
other definitions.
Signed-off-by: John Bowler <jbowler@acm.org>
Improve pngvalid coverage of filter combinations, remove the (new in 1.7) code
which disabled previous-row filters on the first row of an image in some cases.
Signed-off-by: John Bowler <jbowler@acm.org>
This tests the 'size' images in the sequential reader, the existing test case
only uses the progressive reader which means that important parts of the filter
code never got tested.
Signed-off-by: John Bowler <jbowler@acm.org>
The code failed to zero out the row buffer at the start of a pass where the AVG
filter is used because 'row_number' is not always zero (on interlaced images)
and, apparently, the assertion that the buffer had been zeroed when allocated
was not always correct for non-interlaced images. Revealed by pngvalid --size
with a hack to make it always use AVG.
Signed-off-by: John Bowler <jbowler@acm.org>
This error prevents the read of non-interlaced PNG files with height 1. A
simple bracket problem caused by hasty multiple edits.
Signed-off-by: John Bowler <jbowler@acm.org>
This fix is to the PNG_MAX_GAMMA_8 handling and png_set_rgb_to_gray, which had
bugs which were likely to expose end cases of rgb-to-gray conversion errors.
This might explain some of the machine math dependencies we are seeing
(*might*).
Signed-off-by: John Bowler <jbowler@acm.org>
This only changes the detection of invalid application behavior (too many calls
to png_read_row) from an affirm to a png_app_error.
Signed-off-by: John Bowler <jbowler@acm.org>
The reliance of png_read_png on interlace handling and some minor issues in the
test programs where they failed to correctly check for interlace handling were
exposed by the ability to write interlaced images even if WRITE_INTERLACING is
turned off. This is fixed here.
Signed-off-by: John Bowler <jbowler@acm.org>
The removal of png_struct::row_buffer and png_struct::row_format from write
builds configured without filter or transform support (respectively) wasn't
complete; some of the png.c and pngwrite.c cleanup code needed to handle the
potentially removed members.
Signed-off-by: John Bowler <jbowler@acm.org>
Now that the code works consistently so that these just switch off the support
for having libpng do the interlace/deinterlace the old names make more sense,
restoring them avoids cruft in the configuration file and avoids an unnecessary
version specific change.
Signed-off-by: John Bowler <jbowler@acm.org>
commit fc4b42b1d56f95efeb1b9fe42dc35b7d98d246bb
Merge: 74516c7 9eb1413
Author: John Bowler <jbowler@acm.org>
Date: Sun Nov 22 19:37:54 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 74516c7257f1a28a69985684c5673caa390c700a
Author: John Bowler <jbowler@acm.org>
Date: Sun Nov 22 19:32:43 2015 -0800
Make check full pass on gcc/g++ x86_64
Signed-off-by: John Bowler <jbowler@acm.org>
commit e891e34737fc0bc9ee873a5d56b83c1e777b990c
Author: John Bowler <jbowler@acm.org>
Date: Sun Nov 22 12:01:37 2015 -0800
Checkpoint: write buffering changes
This version fails in pngvalid --size because of an error handling very narrow
images, otherwise a standard build passes make check.
Signed-off-by: John Bowler <jbowler@acm.org>
commit 457a046ebdab737eefb477126cf855e49df6de50
Author: John Bowler <jbowler@acm.org>
Date: Sun Nov 22 06:39:36 2015 -0800
Fix previous bad merge
commit b4f426c97267317637d43f41fe0b05d1659bc63d
Merge: 07b9b90 a3458a6
Author: John Bowler <jbowler@acm.org>
Date: Sun Nov 22 06:32:34 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 07b9b90dfd653b744dbc3710b096facf6b4605f6
Merge: ed43306 5592e0b
Author: John Bowler <jbowler@acm.org>
Date: Sat Nov 21 17:07:23 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit ed43306599f7039a90187862db82273fca3a4c3d
Merge: 772aed7 d3c0359
Author: John Bowler <jbowler@acm.org>
Date: Tue Nov 17 17:47:26 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 772aed72378df9c8fccc5a4594b095d02d410a9c
Merge: 73ae431 801608f
Author: John Bowler <jbowler@acm.org>
Date: Mon Nov 16 14:26:38 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 73ae4316cb6db7d7f6756583a1c213c35ca4e3f4
Merge: 687e6e3 c09b3ab
Author: John Bowler <jbowler@acm.org>
Date: Sun Nov 15 09:31:30 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 687e6e393e9d0220c2a12ec474aa01b83c5e9f25
Merge: fedd6da e916d9b
Author: John Bowler <jbowler@acm.org>
Date: Thu Nov 5 08:45:14 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit fedd6da8798a14b2e002b0bc1379f5a09a03598a
Merge: 2e2fc5f ea41fd2
Author: John Bowler <jbowler@acm.org>
Date: Tue Nov 3 21:05:01 2015 -0800
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 2e2fc5f6d7678b710c52b7ea081ac4add677d8d5
Merge: 990d5f8 5b05197
Author: John Bowler <jbowler@acm.org>
Date: Mon Oct 12 08:28:30 2015 -0700
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 990d5f88688635dc0888657b689e30ffe7e7a7b3
Author: John Bowler <jbowler@acm.org>
Date: Sun Oct 4 17:04:53 2015 -0700
Read row buffer changes
The read code now allocates one row buffer of the size of the input PNG row and,
only if required, one buffer of the size of the output.
The output buffer is required for the progressive reader (always) and for the
sequential reader if libpng is de-interlacing an image (because the output row
is used multiple times if png_read_row is called with a display row parameter.)
This should reduce memory utilization by libpng significantly, but it has no
detectable effect on overall performance figures of the test programs, these are
probably dominated by memory allocations for the whole image within the test
programs.
Signed-off-by: John Bowler <jbowler@acm.org>
commit 527bf989bf0e30440f9e07a5544a6ebb1d6fd039
Merge: 50ebbc2 9099254
Author: John Bowler <jbowler@acm.org>
Date: Sat Oct 3 13:39:17 2015 -0700
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 50ebbc2c9a24cf1a6b428db53d55fbd5af4d6be6
Merge: 21a7f40 2cd6d56
Author: John Bowler <jbowler@acm.org>
Date: Sat Oct 3 11:16:32 2015 -0700
Merge branch 'libpng17' into libpng17-filter-enhancements
commit 21a7f401ab40c79ead9e35882a8066e2cf1d6902
Merge: b512e1c 15a143e
Author: John Bowler <jbowler@acm.org>
Date: Wed Sep 30 19:01:23 2015 -0700
Merge branch 'libpng17' into libpng17-filter-enhancements
commit b512e1c2c5bfe6df8b6dca32f862d325ec22115e
Author: John Bowler <jbowler@acm.org>
Date: Wed Sep 30 17:33:34 2015 -0700
Transform rewrite: perform transforms in small chunks
The intent of this change is to reduce the memory footprint during transform
sequences by performing transforms in fixed (small) sized blocks of pixels.
The change is incomplete; the filter code still works row-by-row, so the whole
tranform also works row-by-row, the intent is to fix this so that everything
works in small(ish) chunks.
At present the change has no discernable effect on pngvalid --speed or pngstest
with (e.g.) rgb-8-1.8.png; user time and (minor) page faults are the same in old
and new versions. The same applies to real-world 15MP PNG images; even on these
the presence of the filter code causes a cyclical progress through memory which
will interfere with any caching otherwise possible (useful word, 'otherwise'.)
Signed-off-by: John Bowler <jbowler@acm.org>
commit 781cb3699b92beb0e6bc5e03cef8fba820267082
Author: John Bowler <jbowler@acm.org>
Date: Wed Sep 30 17:12:53 2015 -0700
Fix NO_WRITE_INTERLACE in pngvalid.c
The support for writing interlaced images directly from libpng 1.7 was
unintentionally disabled (INTERLACE_LAST was defined incorrectly, excluding the
interlaced images). This obscured the fact that the transform and error test
case generators lacked the support for writing interlaced images from libpng.
Signed-off-by: John Bowler <jbowler@acm.org>
commit 406ee2fd7946a384f1d7713712dc646080c5c52c
Author: John Bowler <jbowler@acm.org>
Date: Wed Sep 30 17:11:40 2015 -0700
Add pngvalid --transform --interlace test
This increases code coverage by generating test cases with smaller length rows
as a result of the interlacing. Without this packswap handling was incompletely
tested.
Signed-off-by: John Bowler <jbowler@acm.org>
Signed-off-by: John Bowler <jbowler@acm.org>
It is legal for 'pow' to be a macro, therefore the argument list cannot contain
preprocessing directives. Make sure pow is a function where this happens (this
is a minimal safe fix, the issue only arises in non-performance-critical code.)
Signed-off-by: John Bowler <jbowler@acm.org>
The previous attempts to ignore the errors in the code aren't quite enough to
deal with the 'channel selection' encoding added to libpng 1.7; abort.
pngvalid.c is changed to drop this encoding in prior versions.
Signed-off-by: John Bowler <jbowler@acm.org>
Fixes for sBIT handling in the low-bit-depth gray and pngstest cases, extends
sBIT handling to the colormap code in the simplified API which is separately
implemented in pngread.c
Signed-off-by: John Bowler <jbowler@acm.org>
tests/pngstest-* are changed so that the new test files are divided into 8
groups by gamma and alpha channel. This makes each test take just under 5
seconds on a moderately up-to-date x86_64 system using libpng16 (test times are
somewhat variable under libpng17 at present because of the lack of optimization
in some gamma processing paths.)
These tests have considerably better code and pixel-value coverage than
contrib/pngsuite however coverage is still incomplete.
Signed-off-by: John Bowler <jbowler@acm.org>
This fixes the code that optimizes RGB to Gray transformations that only
selected on channel and clarifies the handling of max_depth in the transform
code. It eliminates some UNTESTED cases and removes the write 'invert alpha'
UNTESTED macro because, while there are no test cases for it, code review
some months after writing it suggests it is ok.
Signed-off-by: John Bowler <jbowler@acm.org>
commit bf51c644f7
Author: John Bowler <jbowler@acm.org>
Date: Tue Sep 22 09:05:13 2015 -0700
Updated contrib/testpngs/*.png
Corrected copyright/licensing information, corrected filtering.
Signed-off-by: John Bowler <jbowler@acm.org>
commit 05399c8c82
Author: John Bowler <jbowler@acm.org>
Date: Tue Sep 22 09:04:03 2015 -0700
makepng.c: correct the first row filter setting
It needs to include the "UP" filter so that libpng knows to retain the first
line for the UP filter on the next one. Without this libpng ended up using NONE
for the second line.
Signed-off-by: John Bowler <jbowler@acm.org>
commit 8fae6609c0
Author: John Bowler <jbowler@acm.org>
Date: Mon Sep 21 17:15:45 2015 -0700
makepng: spelling corrections, single test output
Corrections to some obvious mistakes, plus one output file.
Signed-off-by: John Bowler <jbowler@acm.org>
commit a2245a2c54
Author: John Bowler <jbowler@acm.org>
Date: Mon Sep 21 13:39:48 2015 -0700
Updated testpngs files with new Licensing tag
Signed-off-by: John Bowler <jbowler@acm.org>
commit 46d3f863ca
Author: John Bowler <jbowler@acm.org>
Date: Mon Sep 21 13:38:34 2015 -0700
Updated makepng.c
This supports more test cases and a reduced licence keyword (now licensing).
The extra text cases require more slack in pngstest.c
Signed-off-by: John Bowler <jbowler@acm.org>
commit 7c9a08b8bc
Author: John Bowler <jbowler@acm.org>
Date: Sun Sep 20 20:55:19 2015 -0700
Test PNG files
These files were generated by contrib/testpngs/makepngs.sh --small with MAKEPNG
set to the makepng compiled from the current contrib/libtests/makepngs.sh
Signed-off-by: John Bowler <jbowler@acm.org>
commit c2a8a01b2a
Author: John Bowler <jbowler@acm.org>
Date: Sun Sep 20 20:51:06 2015 -0700
makepng --small
This patch adds code to 'makepng' to allow the generation of smaller images that
also give better test case coverage. It also adds copyright and licence to the
generated images, reflecting the fact that they are original works of the author
of the code. The licence used is the Creative Commons Public Domain one,
appropriate to a non-source-code work.
The resultant image were used to generate the updated pngstest-errors.h file in
a prior commit.
Signed-off-by: John Bowler <jbowler@acm.org>
commit 8c025bcf3f
Author: John Bowler <jbowler@acm.org>
Date: Sun Sep 20 20:40:52 2015 -0700
Simplified API sRGB gamma handling correction
If a gamma encoded file that has a gamma not matching that of sRGB is passed to
the simplified API the previous code simply interpreted it as a power law
encoding. However old Mac files had a power law correction of 1.45 built in to
the encoding, even though the display devices were consistent with sRGB.
Assuming a power law encoding results in substantial differences in the
interpretation of low 8-bit values; below 10. For example an Apple '5' which is
equivalent to an sRGB '17' ends up as the value '8'.
This patch provides some measure of correction for this by making the gamma
correction done within the simplified API assume that any encoded data is
encoded relative to an sRGB-like transfer function; the data is corrected back
to the PNG-nominal 2.2 value then decoded to linear (if required) using the sRGB
transfer function.
This reduces the errors reported by pngstest for such files (colormapped ones)
but still leaves the issue with files where the standard libpng code does the
gamma decoding. To cope with the latter cases the patch also includes a new
pngstest-errors which allows the result, however this is still a
work-in-progress; a better solution is possible.
Signed-off-by: John Bowler <jbowler@acm.org>
It needs to include the "UP" filter so that libpng knows to retain the first
line for the UP filter on the next one. Without this libpng ended up using NONE
for the second line.
Signed-off-by: John Bowler <jbowler@acm.org>
This supports more test cases and a reduced licence keyword (now licensing).
The extra text cases require more slack in pngstest.c
Signed-off-by: John Bowler <jbowler@acm.org>
These files were generated by contrib/testpngs/makepngs.sh --small with MAKEPNG
set to the makepng compiled from the current contrib/libtests/makepngs.sh
Signed-off-by: John Bowler <jbowler@acm.org>
This patch adds code to 'makepng' to allow the generation of smaller images that
also give better test case coverage. It also adds copyright and licence to the
generated images, reflecting the fact that they are original works of the author
of the code. The licence used is the Creative Commons Public Domain one,
appropriate to a non-source-code work.
The resultant image were used to generate the updated pngstest-errors.h file in
a prior commit.
Signed-off-by: John Bowler <jbowler@acm.org>
If a gamma encoded file that has a gamma not matching that of sRGB is passed to
the simplified API the previous code simply interpreted it as a power law
encoding. However old Mac files had a power law correction of 1.45 built in to
the encoding, even though the display devices were consistent with sRGB.
Assuming a power law encoding results in substantial differences in the
interpretation of low 8-bit values; below 10. For example an Apple '5' which is
equivalent to an sRGB '17' ends up as the value '8'.
This patch provides some measure of correction for this by making the gamma
correction done within the simplified API assume that any encoded data is
encoded relative to an sRGB-like transfer function; the data is corrected back
to the PNG-nominal 2.2 value then decoded to linear (if required) using the sRGB
transfer function.
This reduces the errors reported by pngstest for such files (colormapped ones)
but still leaves the issue with files where the standard libpng code does the
gamma decoding. To cope with the latter cases the patch also includes a new
pngstest-errors which allows the result, however this is still a
work-in-progress; a better solution is possible.
Signed-off-by: John Bowler <jbowler@acm.org>
One serious bug; 8 was forced as a bit depth when caching a palette, even though
the palette have been expanded. One less serious bug, the optimization for
avoiding gamma correction introduced significant errors if the app subsequently
did further corrections, only shown by using gAMA 1/1.52 files in pngstest.
Signed-off-by: John Bowler <jbowler@acm.org>
Splitting the machine generated error structs out to a file allows the values to
be updated without changing pngstest.c itself, sine libpng 1.6 and 1.7 have
slightly different error limits this simplifies maintenance.
makepngs.sh has also been updated to more accurately reflect current problems in
libpng 1.7
Signed-off-by: John Bowler <jbowler@acm.org>
Also fix the previous fix attempt; the intent was to limit the calls to crc32 to
ZLIB_IO_MAX, although it is not clear that this is optimal.
Signed-off-by: John Bowler <jbowler@acm.org>
These should fix most of the reported Coverity issues. The remaining issues
should be the back_b etc assignments, which look like a Coverity bug, and
passing a pointer to a byte to a function that expects a pointer to one or more
bytes, which should (I believe) be fixed in one case and not the other
(next_filter) case; the latter case will probably go away as I am going to
rewrite that piece of code to avoid a spurious buffer allocation.
Signed-off-by: John Bowler <jbowler@acm.org>
Replaced by a structure copy which is safer since it doesn't depend on knowing
the first member to be copied, also the copies are improved to copy the
transform args too; not required at present but it may prevent a bug being
introduced in the future.
Signed-off-by: John Bowler <jbowler@acm.org>
The low-bit-depth gray tests were disabled in prior versions of libpng because
of problems which should have been fixed by the recent changes to libpng17,
enabling the tests reveals bugs in those changes which are fixed by this commit.
Signed-off-by: John Bowler <jbowler@acm.org>
This implements a new chunk parse implementation that can be shared, it
is currently shared by the progressive reader and the sequential one
(not, yet, the writer).
The patch also implements shared transform handling that is used
throughout.
Signed-off-by: John Bowler <jbowler@acm.org>
Most of these are back-portable to earlier versions (contrib/libtests
should just work with earlier versions), however the 1.7 specific
changes in pngvalid mean that it probably won't work against 1.7 without
the commits following this one.
Signed-off-by: John Bowler <jbowler@acm.org>
pngimage: add --list-combos
pngunknown: add --strict to catch warnings
pngvalid-standard: remove the spurious --progressive read (compare with
pngvalid-progressive-standard!)
This patch probably applies to 1.5 and 1.6 (where the tests exist) too.
Signed-off-by: John Bowler <jbowler@acm.org>
This patch contains changes to the build (configure) system and the code
required to support the move. The patch is provided to isolate the
changes from the others that follow it; the code won't necessarily
compile after this patch (the next changes are required.)
Signed-off-by: John Bowler <jbowler@acm.org>
libpng will not compile if ARM NEON support is enabled with this commit;
the commit is provided to isolate the rename of the directory from the
rest of the required changes.
Signed-off-by: John Bowler <jbowler@acm.org>
because usleep() is deprecated (port from libpng16).
Fixed potential leak of png_pixels in contrib/pngminus/pnm2png.c
Fixed uninitialized variable in contrib/gregbook/rpng2-x.c
handling in contrib/libtests/pngstest.c; possible overflow of
unsigned char in contrib/tools/png-fix-itxt.c). To use the "secure"
file handling, define PNG_USE_MKSTEMP, otherwise "tmpfile()" will
be used.
function has apparently never been used. It was implemented
to support back-door modification of png_struct in libpng-1.4.x
but (because it does nothing and cannot do anything) was apparently
never tested (John Bowler).
depend on READ_TRANSFORMS and moves all the transform gamma (READ_GAMMA)
code from elsewhere to png_tran.c. There are no code changes.
There is one remaining use of the gamma (16-bit) code in the simplified
API in pngread.c but that is because of a long-standing libpng bug,
namely that the gamma corrected palette is no produced by
png_read_update_info (John Bowler).
pngstest.c, and pngimage.c. Most seem harmless, but png-fix-itxt
would only work with iTXt chunks with length 255 or less.
Fixed cexcept.h in which GCC 5 reported that one of the auto
variables in the Try macro needs to be volatile to prevent value
being lost over the setjmp, and fixed g++ build breaks (John Bowler).
compiled library size. It never worked properly and as far as we can
tell, no one uses it. The png_set_filter_heuristics() and
png_set_filter_heuristics_fixed() APIs are retained but deprecated.
internal struct (png_transform_control) to replace row_info and uses
that to implement affirms correctly. The change also adds checks on
the rowbytes calculation and additional checks on most transform
implementations.
Added png_uint_16 range checking, pngvalid tRNS, fixed png_uint_16:
review of previous checks, removal of some where SAFE. pngvalid: add
testing of tRNS for better code coverage pngvalid: correct rgb-to-gray
error calculations. Code coverage is still incomplete: see /*UNTESTED*/
in pngrtran.c
added RELEASE/!RELEASE convenience macros. png_muldiv_warn was used in
only one place, and the overflow condition is a genuine warning not
an internal error. Four macros allow code or function parameters to be
condition on RELEASE (or not) builds and tidy up the #ifdef handling of
functions.
where the code depended on the build base type and can be defined on
the command line, allowing testing in beta builds. Overflows previously
handled by png_warning in png.c have been changed to use new
macros/affirm functions so that beta builds will abort on overflow and
release builds will quietly ignore it. This avoids release builds
producing warnings that are of no use to end users.
Added 'assert' based overflow checking for debug builds for char
and short assignments. Simply ignore the error in release builds
(a truncated value will be used without warning). Controlled by
PNG_RANGE_CHECK_SUPPORTED.
single try_row buffer and in cases where two or more of those are
being tested, a second tst_row buffer. This improves CPU speed
over that achieved by libpng-1.7.0beta49.
pnglibconf.dfa, that can be reset by the user at build time or run time.
This provides a more robust defense against DOS and as-yet undiscovered
overflows.
similarly-named macros were already defined in makefiles. This behavior
is different from POSIX make and other make programs. Surround the
macro definitions with ifndef guards (Cosmin).
seem to generate warnings when an unsigned value is implicitly
converted to double. This is probably a GCC bug but this change
avoids the issue by explicitly converting to (int) where safe.
Free all allocated memory in pngimage. The file buffer cache was left
allocated at the end of the program, harmless but it causes memory
leak reports from clang.
Fixed array size calculations to avoid warnings. At various points
in the code the number of elements in an array is calculated using
sizeof. This generates a compile time constant of type (size_t) which
is then typically assigned to an (unsigned int) or (int). Some versions
of GCC on 64-bit systems warn about the apparent narrowing, even though
the same compiler does apparently generate the correct, in-range,
numeric constant. This adds appropriate, safe, casts to make the
warnings go away.
Removed "option READ_COMPRESSED_TEXT enables READ_TEXT" from pnglibconf.dfa,
to make it possible to configure a libpng that supports iCCP but not TEXT.
Removed "option WRITE_COMPRESSED_TEXT enables WRITE_TEXT" from pnglibconf.dfa
Only mark text chunks as written after successfully writing them.
Added "option READ_iCCP enables READ_COMPRESSED_TEXT" to pnglibconf.dfa
Removed unused "text_len" parameter from private function png_write_zTXt().
Conditionally compile some code in png_deflate_claim(), when
PNG_WARNINGS_SUPPORTED and PNG_ERROR_TEXT_SUPPORTED are disabled.
Replaced repeated code in pngpread.c with PNG_PUSH_SAVE_BUFFER_IF_FULL.
Added "chunk iTXt enables TEXT" and "chunk zTXt enables TEXT"
to pnglibconf.dfa.
Edit and fix typos in comments.
Silence 'unused parameter' build warnings (Cosmin).
$(CP) is now used alongside $(RM_F). Also, use 'copy' instead
of 'cp' where applicable, and applied other minor makefile changes.
Don't warn about invalid dimensions exceeding user limits.
png_push_read_chunk after recognizing the IDAT chunk, which avoids an
infinite loop while reading a datastream whose first IDAT chunk is of
zero-length.
to run all the png_handle_*() and depend on them to return if info_ptr == NULL.
This improves the performance of png_read_end(png_ptr, NULL) and makes
it more robust against future programming errors.
implementation of the ARM NEON discovery function to be set at
build-time and provide sample implementations from the current code in the
contrib/arm-neon subdirectory. The __linux__ code has also been changed to
compile and link on Android by using /proc/cpuinfo, and the old linux code
is in contrib/arm-neon/linux-auxv.c. The new code avoids POSIX and Linux
dependencies apart from opening /proc/cpuinfo and is C90 compliant.
PNG_TRANSFORM_* values are always defined in png.h and, because they
are used for both read and write in some cases, it is not reliable
to #if out ones that are totally unsupported. This change adds error
detection in png_read_image() and png_write_image() to do a
png_app_error() if the app requests something that cannot be done
and it adds corresponding code to pngimage.c to handle such options
by not attempting to test them.
values on write, so a malicious PNG writer could therefore cause
the read code to return an invalid sBIT chunk, which might lead to
application errors or crashes. Such chunks are now skipped (with
chunk_benign_error).
scripts. Fixed combination of ~alpha with shift. On read invert alpha,
processing occurred after shift processing, which causes the final values to be
outside the range that should be produced by the shift. Reversing the
order on read makes the two transforms work together correctly and mirrors
the order used on write.
This is a work-in-progress; no tests are run automatically at present and
the program by virtue of exhaustively testing all the transforms is
very slow.
configure. In configure builds 'make install' changes/adds links like png.h
and libpng.a to point to the newly installed, versioned, files (e.g.
libpng17/png.h and libpng17.a). Three new configure options and some
rearrangement of Makefile.am allow creation of these links to be disabled.
instead of png_chunk_report(), which by default issues a warning
rather than an error, leading to later reading from a NULL pointer
(png_ptr->palette) in png_do_expand_palette().
internal functions called by png_do_{read|write}_transformations static.
On an x86-64 DLL build (Gentoo Linux) this reduces the size of the text
segment of the DLL by 1208 bytes, about 0.6%. It also simplifies
maintenance by removing the declarations from pngpriv.h and allowing
easier changes to the internal interfaces.
function definitions before the place where they are called so that
they can be masde static. Move the intrapixel functions and the
grayscale palette builder out of the png?tran.c files. The latter
isn't a transform function and is no longer used internally, and the
former MNG specific functions are better placed in pngread/pngwrite.c
Fixed 'minimal' builds. Various obviously useful minimal configurations
don't build because of missing contrib/libtests test programs and overly
complex dependencies in scripts/pnglibconf.dfa. This change adds
contrib/conftest/*.dfa files that can be used in automatic build
scripts to ensure that these configurations continue to build.
Enabled WRITE_INVERT and WRITE_PACK in contrib/pngminim/encoder.
the useful side effect of avoiding a bogus warning generated by the latest
version of the Intel C compiler (it objects to
condition ? string-literal : string-literal).
1) Added the 'compile' program to the list of programs that must be cleaned
in autogen.sh
2) Added 'subdir-objects' which causes .c files in sub-directories to be
compiled such that the corresponding .o files are also in the
sub-directory. This is because automake 1.14 warns that the
current behavior of compiling to the top level directory may be removed
in the future.
3) Updated dependencies on pnglibconf.h to match the new .o locations and
added all the files in contrib/libtests and contrib/tools that depend
on pnglibconf.h
4) Added 'BUILD_SOURCES = pnglibconf.h'; this is the automake recommended
way of handling the dependencies of sources that are machine generated;
unfortunately it only works if the user does 'make all' or 'make check',
so the dependencies (3) are still required.
png_set of the transform function, but that doesn't matter unless the
transform function changes the rowbuf size, and that is only valid if
transform_info is called.
library (Claudio Bley):
Introduced a PNG_LIB_TARGETS variable that lists all activated library
targets. It is an error if this variable ends up empty, ie. you have
to build at least one library variant.
Made the *_COPY targets only depend on library targets actually being build.
Use PNG_LIB_TARGETS to unify a code path.
Changed the CREATE_SYMLINK macro to expect the full path to a file as the
first argument. When symlinking the filename component of that path is
determined and used as the link target.
Use copy_if_different in the CREATE_SYMLINK macro.
a flag to change default. In 1.6.0 when the simplified API was used
to produce color-mapped output from an input image with no gamma
information the gamma assumed for the input could be different from
that assumed for non-color-mapped output. In particular 16-bit depth
input files were assumed to be sRGB encoded, whereas in the 'direct'
case they were assumed to have linear data. This was an error. The
fix makes the simplified API treat all input files the same way and
adds a new flag to the png_image::flags member to allow the
application/user to specify that 16-bit files contain sRGB data
rather than the default linear.
Fixed bugs in the pngpixel and makepng test programs.
prefixing. This adds an API to set optimization options externally,
providing an alternative and general solution for the non-portable
run-time tests used by the ARM Neon code. It also makes those tests
compile and link on Android. The order of settings vs options in
pnglibconf.h is reversed to allow settings to depend on options and
options can now set (or override) the defaults for settings.
defaults. Previously a return of 0 from a user chunk callback would result in
the chunk being saved (if this was safe, even if the chunk was unsafe-to-copy);
this change respects the defaults set by the application, so unknown chunks
can be discarded by default and known-safe ones preserved.
Allow run-time ARM NEON checking to be disabled. A new configure option:
--enable-arm-neon=always will stop the run-time checks. New checks
within arm/arm_init.c will cause the code not to be compiled unless
__ARM_NEON__ is set. This should make it fail safe (if someone asks
for it on then the build will fail if it can't be done.)
internal 'prefix.h' file can only be generated correctly after pnglibconf.h,
however the dependency was not in Makefile.am. The symptoms are
unpredictable depending on the order make chooses to build pngprefix.h and
pnglibconf.h, often the error goes unnoticed because there is a system
pnglibconf.h to use instead.
being included because the whole of the 'tests' directory is included,
fixed to just include the actual test files and tests/pngstest. Also
reverted the previous erroneous change to remove the ${srcdir} from the
front of contrib, scripts, projects and added an explanation of what is
going on to Makefile.am
the massive speed improvements use a make capable of parallel builds
on a multi-CPU machine and pass the right arguments to make (-j10000
for GNU make) to get the build to run in parallel.
on SUN OS cc 5.9, which demonstrates the tokenization problem previously
avoided by using /lib/cpp. Since all .dfn output is now protected
in double quotes unless it is to be macro substituted the fix should work
everywhere.
by earlier versions of GCC fixed for Cygwin and Min/GW (which both use old GCCs.)
ARM support is enabled by default in zlib.props (unsupported by Microsoft) and
ARM compilation is made possible by deleting the check for x86. The test programs
cannot be run because they are not signed.
sCAL APIs; some of these use floating point arithmetic so they need to be
disabled if floating point arithmetic is switched off. This is a quiet API
change - previously it appeared that the APIs were supported if fixed point
arithmetic was used internally, however they required certain APIs (floor,
modf, frexp, atof) that are part of C floating point support. Changed
png_fixed and the gamma code specific version of the same to avoid floor(),
which may be a library function (not an intrinsic). Removed unused #if 0
code.
links and tests against zlib with a prefix; tests have been clarified; and
irrelevant or obsolete things (as defined by the autotools man page) have
been removed.
png_malloc_default png_free_default.
Updated some left over "1.6.0beta32" in code sources.
Fixed a "png_structp" prototype (should be png_structrp) in arm_init.c
Updated the version-number hack in pngvalid.c
handling png_struct members rearranged - partly to reorder to avoid packing,
partly to put frequently accessed members at the start and partly to make
the grouping more clear. png_set_filter code has been rewritten and the
code shared with png_write_start_row moved to a common function. Comments
in png.h have been made more clear. Minor fixes to
contrib/libtests/timepng.c and some of the png_*_tRNS logic, including
more error detection in png_set_tRNS.
These changes cause 16-bit arithmetic to be used for 8-bit data in the gamma
corrected compose and grayscale operations. The arithmetic errors have
three sources all of which are fixed in this commit:
1) 8-bit linear calculations produce massive errors for lower intensity
values.
2) The old 16-bit "16 to 8" gamma table code erroneously wrote the lowest
output value into a table entry which corresponded to multiple output
values (so where the value written should have been the closest to the
transformed input value.)
3) In a number of cases the code to access the 16-bit table did not round;
it did a simple shift, which was wrong and made the side effects of (2)
even worse.
The new gamma code does not have the 16-to-8 problem at the cost of slighly
more calculations and the algorithm used to minimize the number of
calculations has been extended to all the 16-bit tables; it has advantages
for any significant gamma correction.
READ_GAMMA. Prior to 1.6.0 switching off READ_GAMMA did unpredictable things
to the interfaces that use it (specifically, png_do_background in 1.4 would
simply display composite for grayscale images but do composition
with the incorrect arithmetic for color ones). In 1.6 the semantic
of -DPNG_NO_READ_GAMMA is changed to simply disable any interface that
depends on it; this obliges people who set it to consider whether they
really want it off if they happen to use any of the interfaces in
question (typically most users who disable it won't).
management system (CMS). In practice it is not possible to obtain cHRM
information reliably using just the end-point tags because they do not exist
in the vast majority of profiles. Instead it is necessary to run the endpoints
through the ICC colorimetric intent transform (as described in the v4
spec). Since this is likely to be too much code inside libpng for too
little gain (it implies a fairly complete CMS implementation) the code
has been changed to allow an external CMS to be used. This code is
temporarily disabled until a suitable set of test cases using one or more
external CMS implementations have been implemented.
type rules. By the ANSI-C standard the new types should be identical to the
previous versions, and all known versions of gcc tested with the previous
versions except for GCC-4.2.1 work with this version. The change makes the
ANSI-C rule that const applied to an array of elements applies instead to the
elements in the array moot by explicitly applying const to the base
elements of the png_icc_matrix and png_icc_vector types. The accidental
(harmless) 'const' previously applied to the parameters of two of the
functions have also been removed.
Glenn thought the "adopted" were typos and changed them all to "adapted".
This reverts the spelling and adds parenthetical comments when each
word first appears in png.h.
fixed Cygwin+MSVC build errors. The ICC profile handling now includes more
checking. Several errors that caused rejection of the profile are now
handled with a warning in such a way that the invalid profiles will be
read by default in release (but not pre-RC) builds but will not be
written by default. The easy part of handling a cHRM chunk is written,
where the ICC profile contains the required data. The more difficult
part plus guessing a gAMA value requires code to pass selected RGB values
through the profile.
option to pngtest. The "--relaxed" option turns off the benign errors that are
enabled by default in pre-RC builds. makepng can now write ICC profiles
where the length has not been extended to a multiple of 4, and pngtest
now intercepts all libpng errors, allowing the previously-introduced
"--strict test" on no warnings to actually work.
contrib/libtests/makepng: corrected 'rgb' and 'gray' cases. profile_error
messages could be truncated; made a correct buffer size calculation and
adjusted pngerror.c appropriately. png_icc_check_* checking improved;
changed the functions to receive the correct color type of the PNG on read
or write and check that it matches the color space of the profile (despite
what the comments said before, there is danger in assuming the app will
cope correctly with an RGB profile on a grayscale image and, since it
violates the PNG spec, allowing it is certain to produce inconsistent
app behavior and might even cause app crashes.) Check that profiles
contain the tags needed to process the PNG (tags all required by the ICC
spec). Removed unused PNG_STATIC from pngpriv.h.
png_handle_unknown() when SAVE support is absent. Moved the shell test
scripts for contrib/libtests from the libpng top directory to contrib/libtests.
png_handle_unknown() must always read or skip the chunk, if
SAVE_UNKNOWN_CHUNKS is turned off *and* the application does not set
a user callback an unknown chunk will not be read, leading to a read
error, which was revealed by the "tunknown" test.
libpng handling of unknown chunks other than vpAg and sTER (including
unsafe-to-copy chunks which were dropped before) and eliminates the
repositioning of vpAg and sTER in pngtest.png by changing pngtest.png
(so the chunks are where libpng would put them).
png_set_keep_unknown_chunks() to be turned off if not required and causes
both read and write to behave appropriately (on read this is only possible
if the user callback is used to handle unknown chunks). The change
also removes the support for storing unknown chunks in the info_struct
if the only unknown handling enabled is via the callback, allowing libpng
to be configured with callback reading and none of the unnecessary code.
control of the unknown handling, corrects the pre-existing bug where
the per-chunk 'keep' setting is ignored and makes it possible to skip
IDAT chunks in the sequential reader (broken in earlier 1.6 versions).
There is a new test program, test-unknown.c, which is a work in progress
(not currently part of the test suite). Comments in the header files now
explain how the unknown handling works.
support. Now "make check" will succeed even if libpng is compiled with
-DPNG_NO_READ or -DPNG_NO_WRITE. The tests performed are reduced, but the
basic reading and writing of a PNG file is always tested by one or more of
the tests.
errors) and cause it not to fail at the comparison step if libpng lacks support
for writing chunks that it reads from the input (currently only implemented
for compressed text chunks).
on read and write. In libpng 1.6 the two cases can be distinguished and
considerable code cleanup, and extra error checking, is possible. This
makes calls on the write side that have no effect be ignored with a
png_app_error(), which can be disabled in the app using
png_set_benign_errors(), and removes the spurious use of usr_channels
on the read side.
Because of the changes to support symbol prefixing PNG_INTERNAL_FUNCTION
declares floating point APIs during libpng builds even if they are completely
disabled. This requires the png floating point types (png_double*) to be
declared even though the functions are never actually defined. This
change provides a dummy definition so that the declarations work, yet any
implementation will fail to compile because of an incomplete type.
ignore all unknown chunks and all known chunks except for IHDR, PLTE, tRNS,
IDAT, and IEND. Previously it only meant ignore all unknown chunks, the
same as num_chunks == 0. Revised png_image_skip_unused_chunks() to
provide a list of chunks to be processed instead of a list of chunks to
ignore. Revised contrib/gregbook/readpng2.c accordingly.
The compiler performs an optimization which is safe; however it then warns
about it. Changing the type of 'palette_number' in pngvalid.c removes the
warning.
on alignment increasing pointer casts when -Wcast-align is passed. This
fixes the cases clang warns about either (pngread.c) by eliminating the
casts from png_bytep to png_uint_16p or, for pngrutil.c where the cast
is previously verified or pngstest.c where it is OK by introducing new
png_aligncast macros to do the cast in a way that clang accepts.
around architecture list; add missing ppc architecture; add architecture
options to shared library link; don't try to create a shared lib based on
missing RELEASE variable (Toby Thain).
checks. The code now does a png_error if an attempt is made to do the
row initialization twice; this is an application error and it has
serious consequences because the transform data in png_struct is
changed by each call.
This turns on PNG_FLAG_DETECT_UNINITIALIZED to detect app code that
fails to call one of the 'start' routines (not enabled in libpng-1.5
because it is technically an API change, since it did normally work
before.) It also makes duplicate calls to png_read_start_row (an
internal function called at the start of the image read) benign, as
they were before changes to use png_inflate_claim. Somehow webkit is
causing this to happen; this is probably a mis-feature in the zlib
changes so this commit is only a work-round.
default. Also changed some warnings in the iCCP and sRGB handling
from to benign errors. Configuration now makes read benign
errors warnings and write benign errors to errors by default (thus
changing the behavior on read). The simplified API always forces
read benign errors to errors (regardless of the system default, unless
this is disabled in which case the simplified API can't be built.)
This is necessary because the VS2010 output window otherwise simply loses
the error messages on error (they weren't flushed to the window before
the process exited, apparently!)
inflate. This works round a bug introduced in 1.5.4 where compressed ancillary
chunks could end up with a too-small windowBits value in the deflate
header.
functions [rw]util.c. A new shared keyword check routine was also added
and the 'zbuf' is no longer allocated on progressive read. It is now
possible to call png_inflate() incrementally.
read code now claims and releases png_ptr->zstream, like the write code.
The bug whereby the progressive reader failed to release the zstream
is now fixed, all initialization is delayed, and the code checks for
changed parameters on deflate rather than always calling
deflatedEnd/deflateInit.
problems. This is an intermediate check-in that solves the immediate problems
and introduces one performance improvement (avoiding a copy via png_ptr->zbuf.)
Further changes will be made to make ICC profile handling more secure.
options declares 'index' as a global, causing a warning if it is used as a
local variable. GCC 64-bit warns about assigning a (size_t) (unsigned 64-bit)
to an (int) (signed 32-bit). MSVC, however, warns about using the
unary '-' operator on an unsigned value (even though it is well defined
by ANSI-C to be ~x+1). The padding calculation was changed to use a
different method. Removed the tests on png_ptr->pass.
all potentially valid ICC profiles and reject obviously invalid ones.
It now uses png_error() to do so rather than casually writing a PNG
without the necessary color data.
generator. The generator generates images that test the full range of sample
values, allow the error numbers in pngstest to be tuned and checked. makepng
also allows generation of images with extra chunks, although this is
still work-in-progress.
store intermediate files, or intermediate in-memory data, while processing
image data with the simplified API. The option makes the files larger
but faster to write and read. pngstest now uses this by default; this
can be disabled with the --slow option.
to reset the user limits to safe ones if PNG_SAFE_LIMITS is defined.
To enable, use CPPFLAGS=-DPNG_SAFE_LIMITS on the configure command
or put #define PNG_SAFE_LIMITS_SUPPORTED in pnglibconf.h.prebuilt.
(Reverted previous implementation of PNG_SECURE.)
pngpread.c and use the sequential png_handle_tEXt, etc., in pngrutil.c;
now that png_ptr->buffer is inaccessible to applications, the special
handling is no longer useful.
FLOATING_POINT options were switched off, png.h ended up with lone ';'
characters. This is not valid ANSI-C outside a function. The ';'
characters have been moved inside the definition of PNG_FP_EXPORT and
PNG_FIXED_EXPORT. 2) If either option was switched off, the declaration
of the corresponding functions were completely omitted, even though some
of them are still used internally. The result is still valid, but
produces warnings from gcc with some warning options (including -Wall). The
fix is to cause png.h to declare the functions with PNG_INTERNAL_FUNCTION
when png.h is included from pngpriv.h.
Fixed a memory overwrite bug in simplified read of RGB PNG with
non-linear gamma Also bugs in the error checking in pngread.c and changed
quite a lot of the checks in pngstest.c to be correct; either correctly
written or not over-optimistic. The pngstest changes are insufficient to
allow all possible RGB transforms to be passed; pngstest cmppixel needs
to be rewritten to make it clearer which errors it allows and then changed
to permit known inaccuracies.
The API is extremely inaccurate for sRGB conversions because it uses an 8-bit
intermediate linear value and it does not use the sRGB transform, so it
suffers from the known instability in gamma transforms for values close
to 0 (see Poynton). The net result is that the calculation has a maximum
error of 14.99/255; 0.5/255^(1/2.2). pngstest now uses 15 for the
permitted 8-bit error. This may still not be enough because of arithmetic
error.
to the background parameter of png_image_finish_read. The --background
option is now done automagically only when required, so that commandline
option no longer exists.
Also some layout regularization and removal of introduced tab characters
(replaced with 3-character indentation). Obsolete macros identified by
autoupdate have been removed; the replacements are all in 2.59 so
the pre-req hasn't been changed. --enable-werror checks for support
for -Werror (or the given argument) in the compiler. This mimics the
gcc configure option by allowing -Werror to be turned on safely; without
the option the tests written in configure itself fail compilation because
they cause compiler warnings.
compiler issues slightly different warnings from those issued by the
current vesions of GCC. This eliminates those warnings by
adding/removing casts and small code rewrites.
those uses of png_memcpy that were doing a simple assignment to assignments
(all those cases where the thing being copied is a non-array C L-value.)
Added some error checking to png_set_*() routines and removed the
reference to the non-exported function png_memcpy() from example.c. Fixed
the Visual C 64-bit build - it requires jmp_buf to be aligned, but it had
become misaligned.
Added new "png_structrp" typedef. Because of the
way libpng works both png_info and png_struct are always accessed via a
single pointer. This means adding C99 'restrict' to the pointer gives
the compiler some opportunity to optimize the code. This change allows that.
Previously it was disabled whenever internal fixed point arithmetic was
selected, which meant it didn't exist even on systems where FP was available
but not preferred.
changes alter how the tricky allocation of the initial png_struct and png_info
structures are handled. png_info is now handled in pretty much the same
way as everything else, except that the allocations handle NULL return
silently. png_struct is changed in a similar way on allocation and on
deallocation a 'safety' error handler is put in place (which should never
be required). The error handler itself is changed to permit mismatches
in the application and libpng error buffer size; however, this means a
silent change to the API to return the jmp_buf if the size doesn't match
the size from the libpng compilation; libpng now allocates the memory and
this may fail. Overall these changes result in slight code size
reductions; however, this is a reduction in code that is always executed
so is particularly valuable. Overall on a 64-bit system the libpng DLL
decreases in code size by 1733 bytes. pngerror.o increases in size by
about 465 bytes because of the new functionality.
These changes alter how the tricky allocation of the initial png_struct and
png_info structures are handled. png_info is now handled in pretty much the
same way as everything else, except that the allocations handle NULL return
silently. png_struct is changed in a similar way on allocation and on
deallocation a 'safety' error handler is put in place (which should never
be required). The error handler itself is changed to permit mismatches
in the application and libpng error buffer size; however, this means a
silent change to the API to return the jmp_buf if the size doesn't match
the size from the libpng compilation; libpng now allocates the memory and
this may fail. Overall these changes result in slight code size
reductions; however, this is a reduction in code that is always executed
so is particularly valuable. Overall on a 64-bit system the libpng DLL
decreases in code size by 1733 bytes. pngerror.o increases in size by
about 465 bytes because of the new functionality.
of parallelism during 'make' the use of the same temporary file names such
as 'dfn*' can result in a race where a temporary file from one arm of the
build is deleted or overwritten in another arm. This changes the
temporary files for suffix rules to always use $* and ensures that the
non-suffix rules use unique file names.
Some compilers fault 'extern const' data declarations (because the data is
not initialized); this turns on const-ness only for compilers where
this is known to work.
using g++. The compiler imposes C++ rules on the C source; thus it
is desireable to make the source work with either C or C++ rules
without throwing away useful error information. This change adds
png_voidcast to allow C semantic (void*) cases or the corresponding
C++ static_cast operation, as appropriate.
The slightly modified tables reduce the number of 16-bit values that
convert to an off-by-one 8-bit value. The "makesRGB.c" code that was used
to generate the tables is now in a contrib/sRGBtables sub-directory.
the sBIT fields in the test pixel as 0, which resulted in a floating
point division by zero which was irrelevant but causes systems where
FP exceptions cause a crash. Added code to pngvalid to turn on FP
exceptions if the appropriate glibc support is there to ensure this is
tested in the future.
casts. The compression_type parameter is always assigned to, so must
be non-NULL. The cast of the profile length potentially truncated the
value unnecessarily on a 16-bit int system, so the cast of the (byte)
compression type to (int) is specified by ANSI-C anyway.
changes the 'Paeth' reconstruction function to improve the GCC code
generation on x86. The changes are only part of the suggested ones;
just the changes that definitely improve speed and remain simple.
The changes also slightly increase the clarity of the code. In a
in pngmem.c; pngvalid would attempt to call png_error() if the allocation
of a png_struct or png_info failed. This would probably have led to a
crash. The pngmem.c implementation of png_malloc() included a cast
to png_size_t which would fail on large allocations on 16-bit systems.
The logical shift fix for Microsoft Visual C is required by other compilers,
so this enables that fix for all compilers when using compile-time constants.
Under MSYS 'byte' is a name declared in a system header file, so we
changed the name of a local variable to avoid the warnings that result.
is not byte aligned, while reading. Prior to libpng-1.5.6 libpng would
overwrite the end of the image if the row width is not an exact multiple
of 8 bits and the image is not interlaced.
These fixes attend to most of the errors revealed in pngvalid, however doing
the gamma work twice results in inaccuracies that can't be easily fixed.
There is now a warning in the code if this is going to happen.
It is too risky. Exactly how png_ptr->zbuf is used is under control of the
individual chunk implementation and there could easily be significant changes
within a major release.
small percentage for 16-bit and 32-bit pixels in the typical case where the
output row buffers are appropriately aligned. The optimization was not
previously possible because the png_struct buffer was always misaligned.
caused by a bug in the code that attempted to align it; the code needs to
subtract one from the pointer to take account of the filter byte prepended to
each row.
generality of the code, allowing it to be optimized for Adam7 interlace. The
masks passed to png_combine_row() are now generated internally, avoiding
some code duplication and localizing the interlace handling somewhat.
to conditions where types that are 32 bits in the GCC 32-bit
world (uLong and png_size_t) become 64 bits in the 64-bit
world. This produces potential truncation errors which the
compiler correctly flags.
recognize the file extension, so it always compiles with C++ rules. Made minor
changes to pngrutil.c to cast results where C++ expects it but C does not.
1.5.3beta06
Also removed a spurious (totally unused and confusing) member from png_info.
The palette expand optimization prevented expansion to an intermediate RGBA
form if tRNS was present but alpha was marked to be stripped; this exposed
a check for tRNS in png_do_expand_palette() which is inconsistent with the
code elsewhere in libpng.
scripts/makefile.watcom. This project works with OpenWatcom 1.9. The
IDE autogenerates appropriate makefiles (libpng.mk) for batch processing.
The project is configurable, unlike the Visual Studio project, so long
as the developer has an awk.
wrong (high by one) 25% of the time. Dividing by 257 with rounding is
wrong in 128 out of 65536 cases. Getting the right answer all the time
without division is easy.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.