[FFmpeg-devel] [PATCH] lavf/utils: has_decode_delay_been_guessed - return true if nb_decoded_frames >= 2, when there are no B-frames.

[Sasi Inguva]

yes. for the video you pointed the B-frame occurs as 6th frame, and until
then has_b_frames is zero.So we assume a wrong delay of zero with my patch .

Is there some other way we can make sure that , for videos with small
number of frames like 5 frames etc. DTS is filled out properly. For example
if av_find_stream_info reads all of the packets of the stream and reaches
EOF , then whatever the value of has_b_frames is , that is the delay and
has_decode_delay_been_guessed should return true .

On Mon, Mar 7, 2016 at 1:57 PM, Michael Niedermayer <michael at niedermayer.cc>
wrote:

> On Mon, Mar 07, 2016 at 11:10:16AM -0800, Sasi Inguva wrote:
> > Fixes DTS detection for any H264 file of less than 7 frames length.
> Without this DTS is N/A.
> >
> > Signed-off-by: Sasi Inguva <isasi at google.com>
> > ---
> >  libavformat/utils.c | 4 +++-
> >  1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/libavformat/utils.c b/libavformat/utils.c
> > index 5f48de1..5017937 100644
> > --- a/libavformat/utils.c
> > +++ b/libavformat/utils.c
> > @@ -877,7 +877,9 @@ static int has_decode_delay_been_guessed(AVStream
> *st)
> >         avpriv_h264_has_num_reorder_frames(st->codec) ==
> st->codec->has_b_frames)
> >          return 1;
> >  #endif
> > -    if (st->codec->has_b_frames<3)
> > +    if (st->codec->has_b_frames == 0)
> > +        return st->nb_decoded_frames >= 2;
> > +    else if (st->codec->has_b_frames<3)
> >          return st->nb_decoded_frames >= 7;
> >      else if (st->codec->has_b_frames<4)
> >          return st->nb_decoded_frames >= 18;
>
> this looks wrong
> also it breaks for example
> tickets/1242/sample.mkv
> try
> ./ffmpeg -i sample.mkv -c:v copy -an -t 1 -f framecrc -
> the DTS are wrong after this patch and there are error messages
>
> [framecrc @ 0x2a52ce0] Invalid DTS: 291 PTS: 209 in output stream 0:0,
> replacing by guess
> 0,        251,        251,       41,     4197, 0x3b9f74ad, F=0x0
> [framecrc @ 0x2a52ce0] Non-monotonous DTS in output stream 0:0; previous:
> 251, current: 209; changing to 251. This may result in incorrect timestamps
> in the output file.
>
> [...]
>
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> In a rich man's house there is no place to spit but his face.
> -- Diogenes of Sinope
>