[FFmpeg-devel] first_dts set to RELATIVE_TS_BASE in vorbis audio

[Dale Curtis]

On Fri, May 4, 2012 at 6:50 AM, Michael Niedermayer <michaelni at gmx.at>wrote:

> Hi Dale
>
> On Thu, May 03, 2012 at 05:10:59PM -0700, Dale Curtis wrote:
> > On Thu, Apr 12, 2012 at 4:05 PM, Dale Curtis <dalecurtis at chromium.org
> >wrote:
> >
> > > I resync'd and it works fine. Thanks Michael!
> > >
> > > - dale
> > >
> > >
> > On further testing this change breaks the ability to call av_seek_frame()
> > before an av_read_frame() on certain ogg files:
> >
> >
> http://commondatastorage.googleapis.com/dalecurtis-shared/ogg-ts-test.tar.bz2
> > http://commondatastorage.googleapis.com/dalecurtis-shared/bassdist_1.ogg
> >
> > Test case can be compiled with:
> > gcc -std=c99 -o ogg-ts ogg-ts.c -lstdc++ libavformat/libavformat.a
> > libavcodec/libavcodec.a libavutil/libavutil.a -lz -lbz2 -lpthread
> >
> > The test will fail at the first av_seek_frame() call when used with the
> > bassdist_1.ogg file.
> >
> > The test also highlights some issues with timestamps going backwards in
> > time (with test.ogv) that I've raised on libav-devel since the patch
>
> all fixed
>
> If you have further problems, please report them ASAP, ogg is a quite
> messy format and its easily possible that there are further issues.
> or that we misunderstood the spec at fine details, its not a very
> clear spec but then few specs are clearly written.
>
>
Thanks for the quick work Michael! We'll let you know as soon as we find
them :) Does your "all fixed" comment include the timestamps going backward
issue? I ask since, that problem is still present with an updated test case
(seek to 5 seconds instead of 0):

http://commondatastorage.googleapis.com/dalecurtis-shared/ogg-ts-test-2.tar.bz2

- dale