James Almer: > On 6/17/2019 12:42 AM, Andreas Rheinhardt wrote: >> ff_cbs_delete_unit never fails if the index of the unit to delete is >> valid; document this behaviour explicitly and remove the checks for >> whether ff_cbs_delete_unit failed, because all the callers of >> ff_cbs_delete_unit already make sure that the index is valid. > > Add a comment about why you're ignoring the return values in all three > filters. If for whatever reason the filters are changed in the future > and it can no longer be ensured the call will never fail, the developer > should be aware of it. > Ok. I just noticed that the call for deleting AV1 temporal delimiters doesn't check whether the fragment has a unit at all (in fact, the check before that call already presupposes it). The only way for ff_cbs_read_packet to return an empty fragment and no error is if the packet's buf has a size of zero (this is legal according to the API if the packet has side-data). What should be done about packets that don't result in a single unit and more specifically about zero-sized packets? The obvious answer would be to simply add a check to av1_metadata like the already existing in h264_metadata; but discarding side-data only packets is no good. (If the zero-sized packet had new extradata as side-data, then that's something that may be parsed and rewritten; the other side-data stuff should simply be passed through.)
- Andreas _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
