[FFmpeg-devel] [PATCH 1/2] avcodec/decode: Workaround for media100 decoding failure issue

[Zhao Zhili]

> On Jun 11, 2025, at 13:55, Andreas Rheinhardt <andreas.rheinhardt at outlook.com> wrote:
> 
> Zhao Zhili:
>> From: Zhao Zhili <zhilizhao at tencent.com>
>> 
>> This is a regression since 1c170613.
>> ---
>> libavcodec/decode.c | 11 ++++++++---
>> 1 file changed, 8 insertions(+), 3 deletions(-)
>> 
>> diff --git a/libavcodec/decode.c b/libavcodec/decode.c
>> index ef09568381..7e47a2971c 100644
>> --- a/libavcodec/decode.c
>> +++ b/libavcodec/decode.c
>> @@ -204,9 +204,14 @@ static int decode_bsfs_init(AVCodecContext *avctx)
>>     ret = av_bsf_init(avci->bsf);
>>     if (ret < 0)
>>         goto fail;
>> -    ret = avcodec_parameters_to_context(avctx, avci->bsf->par_out);
>> -    if (ret < 0)
>> -        goto fail;
>> +    /* Don't copy par_out to avcodec context if bsf changed codec ID,
>> +     * e.g., media100_to_mjpegb
>> +     */
>> +    if (avctx->codec_id == avci->bsf->par_out->codec_id) {
>> +        ret = avcodec_parameters_to_context(avctx, avci->bsf->par_out);
>> +        if (ret < 0)
>> +            goto fail;
>> +    }
>> 
>>     return 0;
>> fail:
> 
> We should rather use a clean revert; after all, the internal BSF is
> supposed to be internal and not change the user-visible extradata etc.

Then how to let decoder see new extradata? Access bsf directly?
I’m OK to revert this after extradata can be synced to decoder properly.

AVCodecContext::extradata is owned by codec. Updating extradata can
be a little surprise, but it’s less surprise than change codec id in bsf.

What are the boundaries/limits of bsf’s functionality? Is it allowed for a bsf
to update media type? I remember there is a trac with samples which
can be handled with a video to audio bsf.

> 
> - Andreas
> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".