[FFmpeg-devel] About guess_correct_pts / AVFrame.best_effort_timestamp

[Alexander Strange]

On Feb 16, 2011, at 9:17 AM, M?ns Rullg?rd wrote:

> Nicolas George <nicolas.george at normalesup.org> writes:
> 
>> L'octidi 28 pluvi?se, an CCXIX, M?ns Rullg?rd a ?crit :
>>> What did you "check"?
>> 
>> That guess_correct_pts gives coherent timestamps.
>> 
>>> 			Please try this command:
>>> 
>>> ffmpeg -i foo.mkv -vcodec copy -acodec copy foo.mp4
>> 
>> The code we are discussing is not used in copy mode, as it is only called
>> during avcodec_decode_video2.
> 
> With the same bad inputs as trigger the error you see in copy mode.

Copied packets need pts and dts (matroskadec only sets one). Decoded frames only need one timestamp.
So the question of giving lavc users an easier time with their timestamps only partially overlaps this, and it would still be present if streamcopy was fixed.

This patch is exactly what I proposed as a next step when I last improved sync in ffmpeg.c, but nobody complained then.
I'll explain the issue completely in a bit with the samples I used to check it then, but it will take a while to finish reading my email backlog.

> 
>> There is indeed an issue there, and it certainly needs fixing, but this is a
>> completely unrelated issue to the topic of this thread, and it is not a
>> valid reason to reject the patch.
> 
> The topic of the thread is timestamps.  Your patch does not fix any
> existing problem in the timestamp handling.  All it does is make some
> broken guesswork part of the API.  On those grounds, the patch is
> rejected.
> 
> -- 
> M?ns Rullg?rd
> mans at mansr.com
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel