search

ffvvc / FFmpeg

VVC Decoder for ffmpeg
Other
50 stars 12 forks source link

random md5 mismatch issue #107

Closed nuomi2021 closed 10 months ago

nuomi2021 commented 1 year ago

TEMPSCAL_A_Panasonic_4.bit: https://github.com/ffvvc/FFmpeg/actions/runs/5445305861/jobs/9904363017 c code only https://github.com/ffvvc/FFmpeg/actions/runs/5865552084/job/15902678600 c code only

TREE_C_HHI_3.bit: https://github.com/ffvvc/FFmpeg/actions/runs/5538781956/jobs/10109072165

frankplow commented 1 year ago

The TREE_C_HHI failures are all timeouts:

There is a pending PR to add a timeout option to the test runner (https://github.com/ffvvc/tests/pull/14).

The TEMPSCAL_A_Panasonic_4 failure I'm not sure about. It's only occurred once, and I am unable to reproduce the error on my system. I have run it several hundred times using the environment provided by nektos/act, which is intended to imitate a GitHub Actions runner, and it hasn't failed once.

nuomi2021 commented 1 year ago

Highly possible we have some random issues in TREE_C_HHI. so it can't finish decoding at some point. Increasing timeout value will not fix the issue. It's usually hard to fix this kind of issue. :(

nuomi2021 commented 1 year ago

If you can help use https://learn.microsoft.com/en-us/windows-hardware/drivers/debugger/time-travel-debugging-overview or https://rr-project.org/ to capture the "crime scene". it will help us replay the issue.

nuomi2021 commented 10 months ago

Since we added LUMA_EXTRA_AFTER at https://github.com/ffvvc/FFmpeg/blob/main/libavcodec/vvc/vvc_thread.c#L287, we never see the random error now. Let us close this temporarily. We will reopen if we still have issue