[MPlayer-dev-eng] ogm seeking

[D Richard Felker III]

On Mon, Dec 09, 2002 at 10:38:39AM +0000, gabor wrote:
> On Sun, 2002-12-08 at 07:38, D Richard Felker III wrote:
> > Anyone know what the issue is with seeking in ogm files? It seems
> > mplayer sometimes seeks to non-keyframes, and in some cases it refuses
> > to seek at all. Also audio PTS jumps to 0 momentarily after seeks,
> > causing A-V correction to try to compensate and thus desyncing A-V
> > until PTS gets correctly identified again. I assume this is similar to
> > the way audio PTS always goes to 0 when seeking ogg files. Anyone know
> > what's wrong and if there's a simple way to fix it? ...or at least get
> > things working a little better? Nowadays I'm using ogm for all my
> > encodes so I can fit decent video bitrate on a single CD, and the only
> > thing I miss is clean seeking.
> 
> i was asking for correcting this bag at least a million times, but the
> developers simply don't have time...:-(
> 
> btw. xine does it correctly, if you want to see some working  code, but
> i'm not sure how close the two projects sources are..

I've been looking at the code some, and it's definitely a mess. IMHO
the correct way to fix seeking in ogm (and in all the other formats
its currently broken or semi-broken in) is to put a dynamic index
builder in the demuxer layer and always use an index for seeking. This
way you always get exact seeking to the time you want, rather than
random broken seeking with non-indexed vbr formats. Comments, anyone?

Rich