[Ffmpeg-devel] Hitting Assertion on mov.c line 1889

[Tom Hayden]

On 3/22/06, Michael Niedermayer <michaelni at gmx.at> wrote:
>
> Hi
>
> On Wed, Mar 22, 2006 at 06:21:09AM -0500, Tom Hayden wrote:
> > Good Morning All,
> >
> > I'm somewhat new to ffmpeg and I think this is a solid peice of
> software.
> > However, I'm having some problems, namely hitting the assertion on
> > mov.cline 1889. I'm trying to convert 3gp files to flv and have ffmpeg
> > compiled
> > with these options: --enable-amr_nb --enable-amr_wb --enable-mp3lame
> > --enable-gpl --enable-amr_wb --enable-pp --enable-a52
> >
> > When I comment out the assert() on line 1889 of mov.c then it converts
> my
> > video, however the time seems to be off since my video playback is at
> > roughly 75% of the original speed (both audio and video remain in sync).
> You
> > can find the videos below:
> >
> > http://35.11.78.70/tom.flv
> > http://35.11.78.70/tom.3gp
>
> not reproduceable


I have moved to a new box and updated recompiled with the CVS from about 10
minutes ago and now I do not seem to encounter this problem.  However, the
video when encoded using "ffmpeg -i tom.3gp -acodec mp3 tom.flv" still
yields a video at about 75% of the original speed ( you can hear how
everything moves/sounds slower) in the flv.

>
> > I'd like to to just fix this myself (and contribute to the code), but I
> > guess I need a little background from one of the experts on what is
> going on
> > here.  I suspect this time_base.num is probably linked to the fact that
> my
> > videos are being encoded at a slower rate.
> >
> > assert(pts % st->time_base.num == 0);
> > assert(dts % st->time_base.num == 0);
>
> well, pts and dts are timestamps, they MUST be in exact units of the
> time_base otherwise the mov demuxer would output wrong timestamps
> the bug here is/was either in the calculation of the timestamps or
> time_base
>
> [...]
>
> --
> Michael


Perhaps this belongs on the -user list.  Let me know if you have any ideas.

--
Tom