[FFmpeg-cvslog] r22010 - trunk/ffmpeg.c

Michael Niedermayer michaelni
Thu Jun 10 15:54:57 CEST 2010


On Wed, Jun 09, 2010 at 11:45:28PM -0700, Alexander Strange wrote:
> 
> On Feb 23, 2010, at 2:06 PM, michael wrote:
> 
> > Author: michael
> > Date: Tue Feb 23 23:06:36 2010
> > New Revision: 22010
> > 
> > Log:
> > Set ist->pts to something that isnt guranteed to entangle itself with stream copying b frames.
> 
> This breaks non-streamcopy, some files have their first frame given negative pts and thus dropped.

probably thats not it.
i would guess either AVFormatContext.start_time is wrong and larger than
the first pts. (demuxer bug)
or
has_b_frames is not set correctly (h264 spec bug, the element is optional)
its possible to improve has_b_frames detection of course but not 100%
but this case shouldnt cause a problem
this lead to the 3rd question, do pts when reordered through the decoder
match dts as in the input packets (if not theres a bug in either the file
or demuxer/parser/decoder

If the input stream is buggy or just compliant but wierd and the dts dont
match what they should then you can port the reordered_opaque code from
ffplay.c to ffmpeg.c
this will automatically select then the more correct timestamps of pts/dts


> 
> http://astrange.ithinksw.net/ffmpeg/5cm.h264
> > ./ffmpeg_g -i 5cm.h264 -an -f framecrc -y /dev/null
> ...
> frame= 1088 fps= 55 q=0.0 Lsize=       0kB time=45.33 bitrate=   0.0kbits/s dup=0 drop=1    
> 
> I would attach a patch but I can't quite tell what this was supposed to do.

see above, and sorry that theres no easy awnser
also if you give up working on this, please open an issue on the tracker so
its not forgotten

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

It is dangerous to be right in matters on which the established authorities
are wrong. -- Voltaire
-------------- 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-cvslog/attachments/20100610/6ca7d118/attachment.pgp>



More information about the ffmpeg-cvslog mailing list