[MPlayer-dev-eng] [PATCH] Read normal duration when demuxing subtitles through lavf (v2)

[Alexander Strasser]

Alexander Strasser wrote:
> Philip Langdale wrote:
> > On Wed, 1 Aug 2012 22:03:37 +0200
> > Nicolas George <nicolas.george at normalesup.org> wrote:
> > 
> > > Le quintidi 15 thermidor, an CCXX, Alexander Strasser a écrit :
> > > >   Maybe I am confused, but when testing I could not find a sample
> > > > that triggered the (convergence_duration > 0) condition. But I
> > > > rather saw many samples where (convergence_duration == 0) and
> > > > (duration != 0). Seems recent lavf does not set convergence_duration
> > > > that often.
> > > 
> > > Try putting that in a text file:
> > > 
> > > # 1
> > > # 00:00:01,000 --> 00:00:03,000
> > > # Hello world.
> > > # 
> > > 
> > > And then using mkvmerge to turn it into Matroska.
> > 
> > Yeah. The only place convergence_duration is used for subtitles is
> > in matroska UTF8/TEXT subtitles.
> 
>   Oh; seems I wasn't able to get/produce those.
> 
> > This is a very common format, so
> > people see it all the time. Matroska ASS subtitles go through a
> > different path in mplayer, so are not effected.
> > 
> > In practice, when you look at what's actually in use, there are
> > very few formats that have text subtitles that pass through this
> > path in mplayer - mp4 files with timed text is one of them, and I
> > happen to have been working on those in ffmpeg, so I noticed they
> > were broken in mplayer.
> > 
> > It is, perhaps, a bold claim, but I firmly believe this change won't
> > make any subtitles worse.
> 
>   Thanks to both of you for shedding a light on this!
> 
>   I think too that it is time to get this in. As soon as I have
> confirmed, I will commit. So if nothing unexpected appears it
> will be tomorrow evening.

  After examining the history I committed something different that
seemed more correct to me.

  Philip, please report if it works for you and/or introduces any
regressions.

Thanks,
  Alexander