GNOME Bugzilla – Bug 325733
[oggdemux] seeking to wrong keyframe with ogm
Last modified: 2013-07-17 10:59:37 UTC
With my usual test program : http://www.iecn.u-nancy.fr/~torri/files/gstreamer_seek/seek.c with an ogm file, and when I seek to times 4'80, then the frame timestamp is equal to 9'08. You can test that with the program above and the file (use frame #120): http://www.iecn.u-nancy.fr/~torri/files/gstreamer_pb/blood.ogm i've tried to find the problem without success.
oggdemux does not even attempt to locate keyframes yet. Requires adding minimal codec support in oggdemux to parse and use the granulepos to locate keyframes. This is on the TODO list.
*** Bug 338804 has been marked as a duplicate of this bug. ***
isn't this a duplicate of #344013 ?
I'll check that next week-end (I'm away for some days)
Vincent?
Seeking to a keyframe in ogm is problematic as you can't use the granulepos to locate keyframes, this means you need to scan all packets in search for keyframes.
so it means to do something like what avisynth does for mpeg: parsing first all the file, then using the gathered informations ?
Vincent, is this behaviour still a problem in 1.x ? The file in comment #0 is not accessible anymore.
i don't know if it's still a problem in 1.0. You can find the files here : http://www.maths.univ-evry.fr/pages_perso/vtorri/files/gstreamer_test_seek/blood.ogm http://www.maths.univ-evry.fr/pages_perso/vtorri/files/gstreamer_test_seek/seek_ogm.c
I can't remember any changes related to ogm, it's probably still broken. However I also didn't see any ogm files out there in the last years so it doesn't feel very important right now
That test program needs a special sink
Seems to work good enough in totem btw, seeks are not instant but playback starts about at the correct position. Please reopen if there's still something to fix and there's a good test case
here are the files http://www.maths.univ-evry.fr/pages_perso/vtorri/files/gstreamer_test_seek.tar.bz2