It doesn't work correctly. All componentes have to be instrumented for
msan to function properly. Changing linking mode to C++ exposes more
issues, valid as the memory comes from uninstrumented driver for
example.

Works around AV_NOWARN_DEPRECATED not being correctly defined for the
version of FFmpeg in this image.

Permanent solution would be to upgrade FFmpeg.

See-Also: videolan/libplacebo!510

No longer needed after 6097f324

Kacper Michajłow authored 1 year ago and Niklas Haas committed 1 year ago

Check for unguarded availability to avoid calling possiblly not
available weak symbols.

Fixes: 3fd65b26

Kacper Michajłow authored 1 year ago and Niklas Haas committed 1 year ago

For sanitizers to work/link properly some public symbols have to be available.

We do exception for MSAN to intentionally cripple it's interceptors, which
allows us to run it on CI with not all the external libraries instrumented.

This also needs static linking libplacebo into tests in MSAN case.

Kacper Michajłow authored 1 year ago and Niklas Haas committed 1 year ago

No need to disable them...

Kacper Michajłow authored 1 year ago and Niklas Haas committed 1 year ago

There is no reason to run those separately.

This is a fix for linking errors with MSAN/UBSAN when static runtime is
linked also into shared library, which causes weak symbols to me
incorrectly referenced by bfd. Just use lld, which is clever about this
case and we already force clang for those builds anyway.

Cosmetic change only.

We don't care about demos coverage.

This reverts commit 9d93981b.
This reverts commit 88c39881.

To speed up the abysmal amount of time it takes the CI runner to slog
its way through all these jobs.

Just testing whether our generated shaders also work when stripped, to
avoid mistakes like using %s instead of $ for identifiers.

A smarter man would probably disable testing older GLSL versions on
llvmpipe.

Hopefully the new semaphores will fix the llvmpipe regression.

Works around an llvmpipe crash in the latest version of mesa.

See-Also: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8203

Now we can finally actually compile this code. Yayifications?

Excluding the aarch64 images for now as a result of being blocked by
videolan/docker-images-aarch64!16

This should fix the sporadic CI failures.

This allows us to pull in new 3rdparty dependencies on the CI env.

Because the newer focal builds fail for some inscrutable reason.

This also fixes an issue with GPU tests where due to host machine
updates the /dev/dri/render* permissions were not adequate anymore.

Co-authored-by: Konstantin Pavlov <thresh@videolan.org>

The old runner stopped existing.

Avoid having to update a billion URLs

cobertura was generalized to coverage_report

This fowards display matrix metadata to the pl_frame.rotation field.

Update the aarch64 CI image since this requires pulling in libavformat.

Rather than fixing/determining the pl_glsl_version at compiler creation
time, this is now taken dynamically as a compilation parameter. In
addition to this, the exact API / target version to compile against will
be inferred from the GLSL description, rather than set explicitly. This
allows us to take into account some of these limits during shader
compilation.

Move stuff into the common glsl/ subdir, mostly to allow the creation of
glsl/utils.h which contains shared helpers between C and C++ that can't
be placed into spirv.h directly due to C99/C++ incompatibility.

Finally, rename some usage sites for consistency, enabling `stage` to
always refer to the shader stage, `shader` to refer to the shader
itself, and `glsl` to refer to the pl_glsl_version struct.

Requires bumping the shaderc dependency. Fortunately, none of the usual
candidates (debian stable etc.) even *package* ubuntu, so we're fine.

Similar to Dav1dPicAllocator, this allocates from DMA-visible GPU
memory. Especially useful on platforms that don't support arbitrary host
pointer imports.

Make `plplay` use this, for testing purposes.
Update the CI image to pull in avcodec headers.

Oops. I need some sleep.

This fails for reasons beyond my comprehension. Just disable it, since
it's not useful to build code we're not going to run.