[FFmpeg-devel] [PATCH] all: Replace assert by av_assert

[Andreas Rheinhardt]

Michael Niedermayer:
> On Sun, Mar 23, 2025 at 06:01:01PM +0100, Andreas Rheinhardt wrote:
>> Patch attached.
>>
>> - Andreas
> 
>>  libavcodec/aacpsy.c             |    7 ++++---
>>  libavcodec/encode.c             |    2 ++
>>  libavcodec/h264_direct.c        |   14 +++++++-------
>>  libavcodec/h264_refs.c          |   20 +++++++++-----------
>>  libavcodec/h264_slice.c         |    6 +++---
>>  libavcodec/mips/mpegvideo_mmi.c |    3 ++-
>>  libavcodec/mpegvideo_enc.c      |    7 ++++---
>>  libavcodec/mpegvideoencdsp.c    |    1 -
>>  libavcodec/sinewin_tablegen.h   |    9 +++++++++
>>  libavcodec/vdpau_mpeg12.c       |    4 +++-
>>  libavcodec/vdpau_mpeg4.c        |    6 ++++--
>>  libavcodec/vdpau_vc1.c          |    6 ++++--
>>  libavcodec/vorbisenc.c          |   14 ++++++--------
>>  libavfilter/vf_tonemap_opencl.c |    4 ++++
>>  libavformat/asfdec_f.c          |   15 ++++++---------
>>  libavutil/internal.h            |    5 -----
>>  16 files changed, 67 insertions(+), 56 deletions(-)
>> 4b05bc0165c919f3f4cfc4cee50ccd15a74cc4fe  0001-all-Replace-assert-by-av_assert.patch
>> From a6481493a5a9acf0968798dd5afd72239acca11c Mon Sep 17 00:00:00 2001
>> From: Andreas Rheinhardt <andreas.rheinhardt at outlook.com>
>> Date: Sun, 23 Mar 2025 17:50:54 +0100
>> Subject: [PATCH] all: Replace assert by av_assert
>>
>> Also remove the assert.h inclusion from lavu/internal.h.
> 
> breaks:
> 
> ./ffmpeg -i ~/tickets/2254/ttvHD_vlc_sample.ts -an -vframes 5 -bitexact -f framecrc -
> 
> Assertion h->cur_pic_ptr->reference != (1 << 2) failed at libavcodec/h264_slice.c:1451
> 
> I have more files hitting the same above
> 
> 
> i also see a Assertion b8_stride == 2 failed at libavcodec/h264_direct.c:441
> in freh5.264.SIGABRT.9f5.65

Interesting. It shows that nobody (not even the fuzzers) actually
defines DEBUG in order to stop disabling C asserts. I thought that these
asserts were tested which would make converting them to av_assert easy.
Honestly, I don't have time atm to dig into why exactly these asserts
exist and (if they are actually still necessary or not just superseded)
how to fix them. So this patch is put on hold.
Anyway, thanks for testing.

> i did not verify with bisect that this is from this patchset too

The underlying issue is definitely not from this issue.

- Andreas