Assertion `av1_is_subpelmv_in_range(&ms_params.m v_limits, start_mv)' failed.

[s1ddok]

What does it typically mean? I'm repacking huge amount of data and most of it proceeeds fine, but some files are crashing with this error:

avifenc: /home/avolodin/projects/libavif/build/_deps/libaom-src/av1/encoder/nonrd_pickmode.c:348: search_new_mv: Assertion `av1_is_subpelmv_in_range(&ms_params.m
v_limits, start_mv)' failed.       

died with <Signals.SIGABRT: 6>. 

I've built v1.1.1 myself from source, using it with libaom v.3.9.1

[y-guyon]

https://aomedia.googlesource.com/aom/+/refs/tags/v3.9.1/av1/encoder/nonrd_pickmode.c#348 does not provide more context.

Could you attach a file that reproduces the issue, along with the exact compile flags and avifenc command?

[s1ddok]

I think I can't as I'm repacking proprietary business data... I'm repacking a bunch of JPGs into an animated AVIF. What do you think I can do to further investigate the reason?

[y-guyon]

You can try filing a bug with libaom directly, see https://aomedia.googlesource.com/aom/#bug-reports.
Without a way to reproduce the issue, it is unlikely to be fixed though.

I am not familiar with this code. Assuming this assertion is safe to ignore, try building in Release to skip it.
Do you have issues only with Debug binaries?

[s1ddok]

I thought I'm running Release as I copied build command from README, do you think it is a debug?

[y-guyon]

You mean https://github.com/AOMediaCodec/libavif?tab=readme-ov-file#build-everything-from-scratch?

Could you give the exact command lines you used to build the binaries and the exact command lines you used to call the binaries please?

[s1ddok]

yes, I'm still struggling with it:

git clone -b v1.1.1 https://github.com/AOMediaCodec/libavif.git
cmake -S libavif -B libavif/build -DBUILD_SHARED_LIBS=OFF -DAVIF_CODEC_AOM=LOCAL -DAVIF_LIBYUV=LOCAL -DAVIF_LIBSHARPYUV=LOCAL -DAVIF_JPEG=LOCAL -DAVIF_ZLIBPNG=LOCAL -DAVIF_BUILD_APPS=ON -DCMAKE_C_FLAGS_RELEASE="-static" -DCMAKE_EXE_LINKER_FLAGS="-static"
cmake --build libavif/build --parallel

I followed build instrunctions exactly form the link you point to and then calling avifenc with the following command:

            command = [
                str(script_dir / "bin" / "linux" / "avifenc"),
                *file_paths,
                str(output_file),
                "-q", str(quality),
                "-a", "aq-mode=1",
                "-a", "deltaq-mode=3",
                "-a", "enable-qm=1",
                "-a", "sharpness=1",
                "-a", "enable-chroma-deltaq=1",
                "--speed", str(speed),
                "-k", "125",
                "--autotiling",
                "--fps", str(fps)
            ]

it works on majority of my jpg files but fails on some of them with the above error. I've tried passing -DCMAKE_BUILD_TYPE=Release but it didn't seem to change anything...

[y-guyon]

Thank you. I do not see any blatant problem with the command lines you shared.

I would suggest:

• Try to reproduce the issue with files that you can share, by first finding public data with permissive licenses. Once again a way to reproduce the issue is necessary to easily debug it and to properly fix it. If sharing a file privately is an option for you, you can also send it to me by email (yguyon at google dot com).
• Remove avifenc flags one by one in case there is a single culprit, which would help in identifying the root cause and maybe solve your issue in the short term if you can live without using that flag.

[s1ddok]

I think I can send data privately! Just to clarify: did the command I used for build (from README) builds a release build? if so, why does executable prints my local paths to files during the assertion? I've tried to build aom manually from ext/ folder but it didn't change anything.

[y-guyon]

Just to clarify: did the command I used for build (from README) builds a release build?

It should.

if so, why does executable prints my local paths to files during the assertion?

It should not. This is a separate issue filed as #2616.

[s1ddok]

@y-guyon I'm sorry for the delay -- I've sent the data sample and reproduction command on provided e-mail, I hope it will help to resolve the problem!