[FFmpeg-devel] Problem with seeking in avi file.

[Michael Niedermayer]

Hi

On Tue, Aug 14, 2007 at 03:26:38AM +0200, Michael Niedermayer wrote:
> Hi
> 
> On Sat, Aug 11, 2007 at 10:45:45AM +1000, Neil Brown wrote:
> > 
> > I have another interesting movie file that ffmpeg doesn't quite work
> > properly with.
> > 
> > This one is an avi file containing dvvideo extracted off a digital
> > camcorder (I can find out what program was used to exact it if that
> > might be of interest).
> > 
> > Anyway, the problem is that the first time that I seek into the file,
> > if I aim for beyond 4:57, I alway arrive at 4:57.  Subsequent seeks
> > work as expected.
> > e.g.
> > 
> >   ffmpeg -ss 400 filename.avi
> > 
> > starts at 4:57, as does any number of seconds greater than 300.\
> > 
> > The file starts with RIFF with a size of 1069618820.
> > This covers about 1/4 of the file (5 minutes of about 20).
> > There are 3 more RIFF headers of 1069611496, then
> > one of 63795560 which completes the file.
> > 
> > The file contains a 'dmlh' tag, so is_odml is set.
> > The file contains a 'LIST' tag with a 'movi' subtag and a size of
> > 1069467476.  This just covers the first of 5 RIFF sections.
> > 
> > So ->movi_end gets set to 1/4 of the way through the file - the end of
> > the first section.
> > 
> > In avi_read_packet, after a seek further into the file,
> > 'i' is the current position in the file, so this code:
> > 
> >         if (i >= avi->movi_end) {
> >             if (avi->is_odml) {
> >                 url_fskip(pb, avi->riff_end - i);
> >                 avi->riff_end = avi->movi_end = url_fsize(pb);
> >             } else
> >                 break;
> >         }
> > 
> > seeks back to the end of the first RIFF, then sets movi_end to the
> > correct end of the file.
> > It seems to me that if 'odml' allows as to set movi_end to fsize at
> > this point, it should allow movi_end to be set to fsize right at the
> > start.
> 
> yes, setting it to fsize is wrong no matter where its done, it breaks
> avi_load_index()
> the code quoted above should be removed its broken
> you seem pretty good at spotting totally broken code :)
> ill try to fix this one ...

should be fixed

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

No human being will ever know the Truth, for even if they happen to say it
by chance, they would not even known they had done so. -- Xenophanes
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20070814/943f85b3/attachment.pgp>