[FFmpeg-devel] [PATCH] vp9: ignore frame_context_index field in key-/intraonly-frames.

Ronald S. Bultje rsbultje at gmail.com
Sun Apr 17 15:47:28 CEST 2016


Hi,

On Sun, Apr 17, 2016 at 6:17 AM, wm4 <nfxjfg at googlemail.com> wrote:

> On Sat, 16 Apr 2016 13:56:18 -0400
> "Ronald S. Bultje" <rsbultje at gmail.com> wrote:
>
> > Reproduces a bug to remain consistent with libvpx' behaviour.
> > ---
> >  libavcodec/vp9.c | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/libavcodec/vp9.c b/libavcodec/vp9.c
> > index 5c6f176..70d9c4b 100644
> > --- a/libavcodec/vp9.c
> > +++ b/libavcodec/vp9.c
> > @@ -642,6 +642,8 @@ static int decode_frame_header(AVCodecContext *ctx,
> >      s->s.h.refreshctx   = s->s.h.errorres ? 0 : get_bits1(&s->gb);
> >      s->s.h.parallelmode = s->s.h.errorres ? 1 : get_bits1(&s->gb);
> >      s->s.h.framectxid   = c = get_bits(&s->gb, 2);
> > +    if (s->s.h.keyframe || s->s.h.intraonly)
> > +        s->s.h.framectxid = 0; // BUG: ignore field in keyframes
> >
> >      /* loopfilter header data */
> >      if (s->s.h.keyframe || s->s.h.errorres || s->s.h.intraonly) {
>
> Might be a good idea to indicate in the comment that it's not a lavc
> bug, but reproduces a vpx one.


Hm, you're right. I pushed it with a slightly less ambiguous comment.

Ronald


More information about the ffmpeg-devel mailing list