On Mon, Jun 26, 2023 at 6:08 PM Anton Khirnov <an...@khirnov.net> wrote:
> Quoting Fei Wang (2023-06-26 09:24:58) > > Filter may has a limited frame pool size. Do not always send frame to > > filters without reaping. > > > > Fix the regression of commit 5fa00b38e6. > > > > Example cmd: > > $ ffmpeg -threads 20 -init_hw_device vaapi=hw:/dev/dri/renderD128 \ > > -hwaccel_output_format vaapi -hwaccel vaapi -i avc_1080P.mp4 \ > > -vf 'scale_vaapi=w=720:h=480' -f null - > > > > Signed-off-by: Fei Wang <fei.w.w...@intel.com> > > --- > > Change: > > 1. rebase to master. > > > > fftools/ffmpeg_dec.c | 6 ++++++ > > 1 file changed, 6 insertions(+) > > > > diff --git a/fftools/ffmpeg_dec.c b/fftools/ffmpeg_dec.c > > index 85bf8dc536..a7ae51fed2 100644 > > --- a/fftools/ffmpeg_dec.c > > +++ b/fftools/ffmpeg_dec.c > > @@ -803,6 +803,12 @@ int dec_packet(InputStream *ist, const AVPacket > *pkt, int no_eof) > > av_frame_unref(d->frame); > > if (ret < 0) > > goto finish; > > + > > + // During flushing, break out to reap filter once send a frame > to filters to > > + // avoid drain out filter's output frame pool. Especially for > HW filters which > > + // always have a limited frame pool size. > > + if (!pkt) > > + return 0; > > This is wrong and will leave the decoder in an inconsistent state. > Two wrongs do not make things right. > > -- > Anton Khirnov > _______________________________________________ > 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". > _______________________________________________ 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".