[FFmpeg-devel] [RFC] Are those workarounds really needed?

[Alexander Strange]

On Sat, Feb 12, 2011 at 9:39 PM, Luca Barbato <lu_zero at gentoo.org> wrote:
> On 02/13/2011 01:37 AM, M?ns Rullg?rd wrote:
>> Did you figure out why this causes problems with your sample?
>
> in the sample pts are always the same as the dts, sometimes we got
>
> st->codec->has_b_frames == 1 (while pc->pict_type != FF_B_TYPE) and that
> triggers that code...

Are there B-frames in the stream according to 'ffplay -debug 1'?
Either way, how does it behave with -flags +nofillin?

That code was added for the file in here:
http://samples.mplayerhq.hu/ffmpeg-bugs/roundup/issue171_mpeg2_decoding_issue/
which is 403, so I can't comment on it.

Also triggers on the first frame here:
http://samples.mplayerhq.hu/archive/all/mpeg%2bmpeg2video%2bac3%2b%2bnon_monotone_timestamps.mpg
which has dts=pts on the first frame (certainly wrong), but the effect
of the code is just to delete the first pts, which seems unhelpful.