In "test: Add consistency checks for the PNG_LIBPNG_VER* number" [0] the
`STDERR` macro was moved from outside an `ifdef` to inside an `ifdef`.
This broke the code in the `else` of this `ifdef` which also uses the
`STDERR` macro. Move `STDERR` back to where it was to avoid compile
errors in the `else` case.
[0] cc8006c48dFixes: #560
Reviewed-by: Cosmin Truta <ctruta@gmail.com>
Signed-off-by: Cosmin Truta <ctruta@gmail.com>
ACES AP1 has a red endpoint with a negative Z, this triggers the checks
in libpng that ensure that x, y and z (chromaticities) are all >=0.
This removes the checks on the sign of the chromaticities since it is
valid to use negative values for any of them and converts the "internal"
error code return to external (because the internal cases correspond to
negative x, y or z.)
Reviewed-by: Cosmin Truta <ctruta@gmail.com>
Signed-off-by: John Bowler <jbowler@acm.org>
Signed-off-by: Cosmin Truta <ctruta@gmail.com>
Considering that a non-trivial amount of libpng code is arch-specific,
we should perform cross-platform builds (with cross-platform toolchains)
and test runs (on emulated architectures) in our routine verification.
The content of ci/targets/ shall consist of target description files,
written in the standard shell language. These files may be source'd as
needed, before running the verification scripts ci/ci_verify_*.sh.
Here is the initial list of target systems:
Android, Cygwin, FreeBSD, Linux, MSDOS, Windows.
And here is the initial list of target architectures:
ARM, MIPS, PowerPC, RISC-V, x86.
Because of a missing "amd64" string (in lowercase) in a regex match,
the CMake build was unable to pick up the PNG_HARDWARE_OPTIMIZATIONS
flag on FreeBSD/amd64 (and possibly other amd64 systems as well).
Rename the target arch variable from TARGET_ARCH to a more idiomatic
PNG_TARGET_ARCHITECTURE, and set it to an always-lowercase string.
The follow-on checks are now simpler and easier to get right.
Initialize the arch-specific MSYSTEM environment variable, to ensure
that msys2 bash picks up and executes /etc/profile correctly.
Install and use the host-specific cmake and ninja, to ensure that
msys2 cmake picks up the host-specific zlib build correctly.
In the previous commit 9e538750d99c8f1accf7e93878e4fde47c069908
we removed the obsolete assembler implementation `filter_neon.S`.
In this commit we add a stand-in for the original file, restoring
the original source tree structure, for the benefit of continuing
hassle-free libpng source upgrades in the 1.6.x line.
This file contains hand-coded assembler implementations of the filter
functions for 32-bit Arm platforms. These are only used when the
compiler doesn't support neon intrinsics (added to GCC 4.3 in 2008) or
is exactly GCC 4.5.4 (released 2012), both of which are sufficiently
unlikely to be true that it's fair to say the assembler is no longer
used.
This commit deletes filter_neon.S and removes the now obsolete
preprocessor logic in pngpriv.h.
Signed-off-by: Bill Roberts <bill.roberts@arm.com>
Signed-off-by: Cosmin Truta <ctruta@gmail.com>
This fixes the bug https://github.com/pnggroup/libpng/issues/505
"libpng does not support PAC/BTI on aarch64 targets" which arises
because the build mechanisms (both cmake and configure) assemble
arm/filter_neon.S even though it ends up completely empty. The empty
file effectively poisons the so that the PAC/BTI support gets disabled.
The fix is minimal; it simply removes arm/filter_neon.S from the list of
sources included in the 64-bit ARM builds build. Note that this was
already done in cmake for MSVC - it's not clear whether this change was
a partial fix for the same issue.
This version of the fix ONLY affects aarch64 (arm64) builds; 32-bit ARM
systems can still invoke the assembler if required and, indeed, there
should be no change whatsover to those builds.
The assembler code could not be used on 64-bit systems in any case so
in practice there is no material change to 64-bit builds either.
TESTING: pull the changes then type "autoreconf" if using configure (not
required for cmake).
TESTS: cmake has not been tested because cross-builds with cmake
currently fail to find the zlib installation from the cmake system root
path. The following has been tested with configure cross builds:
armv7-linux-gnueabi [no neon support]
armv7a-linux-gnueabi [no neon support]
armv7a-hardfloat-linux-gnueabi [neon support not enabled]
armv7a-hardfloat-linux-gnueabi -mfpu=neon [uses intrinics]
armv7a-hardfloat-linux-gnueabi -mfpu=neon
-DPNG_ARM_NEON_IMPLEMENTATION=2 [uses assembler]
aarch64-linux-gnu [uses intrinsics]
aarch64-linux-gnu -DPNG_ARM_NEON_OPT=0 [neon support disabled]
Signed-off-by: John Bowler <jbowler@acm.org>
We used this experimental project in the development of the PNG-EXIF
("eXIf") specification, back in 2017. The project evolved together
with the draft specification, which was finalized on 2017-Jun-15 and
approved by the PNG Group on 2017-Jul-13.
The EXIF specification, outside of the scope of PNG and libpng, is
quite complex. The libpng implementation cannot grow too much beyond
performing basic integrity checks on top of serialization. In order
to create and manipulate PNG-EXIF image files, the use of external
libraries and tools such as ExifTool is necessary.
Now, with the addition of contrib/pngexif to the libpng repository,
offline tasks like metadata inspection and linting can be performed
without importing external dependencies.
Modern compilers can disable the warnings that originate from system
headers. This change allows them to do so with the libpng headers.
Signed-off-by: Cosmin Truta <ctruta@gmail.com>
In the configure script, checking whether the LoongArch LSX intrinsics
are supported by the compiler was done unconditionally, regardless of
the targetted host platform. Compared to how we support the other SIMD
platforms and compilers, this is rather unconventional.
We are placing this check under the guard of its own platform, for the
time being. A full solution, in line with the rest of the configure.ac
patterns concering SIMD optimizations, is TODO.
We also do an overall cleanup in the SIMD section of configure.ac, and,
finally, we regenerate the configure script.
Declare AWK explicitly via the AC_ARG_VAR directive, in order to make
it "precious", and to include it in the list of influential variables
at the end of the configure help text.
Rephrase a few comments and config traces.
Finally, regenerate the configure script.
Update the PNG_TOOLS option: set it to OFF by default when the
target is an embedded system, yet still allow it to be overridden.
Update the PNG_FRAMEWORK option: force it back to OFF and print a
warning if the option was ON but the target is not an Apple system.
This fixes commit 4edbb4da81626a7342a22824d7a8f60a3ea71bd0.
During the move of CMake scripts to the scripts/cmake/ subdirectory,
some of the workflows have been broken.
Signed-off-by: Cosmin Truta <ctruta@gmail.com>
This fixes commit dddaf0c625a8daea4d027cb57380b7fac6f58285.
The way to reliably `find` executable files is different on BSD, Mac
and Linux, unfortunately.
The variable `CI_LINT_COUNTER` was incremented inside subshells, but
remained unchanged in the main shell process. The errors detected by
the internal linters remained unreported by the main script. (Oopsie!)
Besides fixing this defect, considering that only a pass/fail status
is needed, we are replacing `CI_LINT_COUNTER` with `CI_LINT_STATUS`.
Introduce the environment option CI_FORCE:
* ci_verify_configure.sh is known to fail if an existing build
configuration is found in the top-level directory.
Setting CI_FORCE=1 will run `make distclean` before verification.
* ci_verify_makefiles.sh cannot be reliably executed if random
object files are found in the top-level directory.
Setting CI_FORCE=1 will run `rm *.o *.obj` before verification.
* ci_verify_cmake.sh is not known at this time to fail for similar
reasons; but if it does, we will use CI_FORCE to trigger any
necessary pre-build cleanup.
By definition, `sizeof(png_byte)` is 1.
Remove all the occurences of `sizeof(png_byte)` from the code, and fix
a related typo in the libpng manual.
Also update the main .editorconfig file to reflect the fixing expected
by a FIXME note.
Work around a limitation in the `shellcheck source` directive, which
does not recognize quotes in shellcheck versions older than 0.9.
Also extend the checks for YAML files over the entire source tree, in
preparation for the introduction of the GitHub Actions config file.
`PNG_LIBPNG_VER_BUILD` should be zero for public releases and non-zero
for development versions. The ci_verify_version.sh script should check
this requirement as such.
Split AR_RC into AR and ARFLAGS. The variables AR and ARFLAGS are
de-facto standards (like CC and CFLAGS, LD and LDFLAGS, etc.) that
may be overridden when running make. Moreover, configuring CC, LD,
AR, etc., to point to a cross-platform compiler, linker, librarian,
etc., is a de-facto standard practice as well.
Also remove the MKDIR_P variable definitions from all makefiles.
They've been leftovers from the removal of the "install*" targets.
This program verifies the libpng source tree, expecting consistent
definitions of version numbers in the C source code, in the Autoconf
scripts, and in the CMake scripts.
The version verification is performed as follows. (Please note that
the version definitions in png.h are checked twice.)
* The files png.h, configure.ac and CMakeLists.txt are checked by
the ci_verify_version.sh program.
* The files png.h, png.c and pngtest.c are checked by the pngtest
program.
The version macros `PNG_LIBPNG_VER_SONUM` and `PNG_LIBPNG_VER_DLLNUM`
weren't always in sync, but they should be, going forward.
Or, better yet, we should keep them in, deprecated, and introduce
`PNG_LIBPNG_VER_SHAREDLIB` for all shared library builds of all kinds
on all platforms.
It is unknown how many user applications have been using these macros.
We have been using `PNG_LIBPNG_VER_DLLNUM` for pngwin.rc, for example.
Which, by the way, was last updated in 2009.
We have been running address-sanitized CI verifications for a while.
We can finally afford to simplify pngtest.c by removing a compile-time
option and the associated code branches that used to serve (only to a
limited extent) the purpose of bounds checking.
Also change the brief description of pngtest.c at the top of the file.
This is no longer just a simple test program.
Improve:
The pngtest program used to be rather relaxed upon seeing invalid
parameters in callbacks: it either ignored them, or it bailed out
of the callbacks, essentially sweeping the bug under the rug.
But no more. Now it terminates with a severe `png_error`, in which
it says what's broken and where.
Improve:
`PNG_DEBUG`, defined externally at build time, and defaulting to zero,
was assumed to be non-negative. Now it's checked.
Clean up:
In a very distant past, the pngtest program used to "travel" across
libpng versions, on its own, not necessarily accompanied by the actual
library version that it was meant to test. However, this stopped being
the case, and now is as good a time as any to remove the compatibility
workarounds that had made the aforementioned "travel" possible.
Other chores include:
* The refactoring of the user-defined chunk handling routines;
* The cleanup of an unnecessary use of volatile;
* The various cosmetic improvements of code and comments.
This adds a new configuration file, 'contrib/conftest/basic.dfa' which
disables unused APIs on a test Linux-based system. So support the
configuration several fixes were necessary in the test programs so that
the tests are skipped correctly when APIs are not available.
The configuration has been tested on a range of common Linux apps
including web browser code (qtwebengine), image processing code (e.g.
ImageMagick) and general display code (X11, Qt5 and Qt6, KDE). Overall
this first step reduces libpng linked code and data size to about 2/3 of
the full configuration.
To use the new test simply copy basic.dfa to 'pngusr.dfa' in the root of
the source directory and build.
Signed-off-by: John Bowler <jbowler@acm.org>
This changes the mask used in the IHDR width check from ~7U to
~(png_alloc_size_t)7 which is a quantity at least as big as both
png_uint_32 and size_t whereas "7U" will be 16 bits on a 16 bit system.
The change both corrects a bug in the code (on 16 bit systems) and
removes compiler warnings about the test always being false on 64-bit
architectures.
Signed-off-by: John Bowler <jbowler@acm.org>
