3 Feb
2009
3 Feb
'09
9:30 a.m.
Paul Davis wrote:
On Tue, 2009-01-27 at 13:49 +0100, Olivier Guilyardi
wrote:
I just wanted to let you know that an interesting discussion is going on the
ffmpeg-devel mailing list, about choosing a time source, NTP effects, and A/V
synchronization. It is named "libavdevice: PTS stability and risks of drift".
--
Olivier
For example, in FFmpeg, each audio and video
packets are tagged using a PTS
which is computed using av_gettime() which calls gettimeofday().
this should be considered a bug and reported as such to the FFmpeg group
(preferably with a patch to make it use clock_gettime() if available).
gettimeofday is a non-linear, non-mononotic clock if ntp is used. it is
not suitable for timestamping in this way. without ntp, it will work
better, but it is still subject to potential non-monotonicity.