[vlc-devel] [sout developers] VLC streaming violation of MPEG-TS discontinuity flag
ileoo at videolan.org
Wed Oct 26 11:00:19 CEST 2016
On Tue, Oct 25, 2016 at 04:04:34PM +0300, Rémi Denis-Courmont wrote:
> Le tiistaina 25. lokakuuta 2016, 14.51.35 EEST Francois Cartegnie a écrit :
> > Le 25/10/2016 à 14:28, Rémi Denis-Courmont a écrit :
> > > Discontinuity due to seeking is not the same as discontinuity due to
> > > silence in speech codec or due to packet loss. For instance, packetizers
> > > obviously handle the former and the later very differently.
> > - Discontinuity on seek.
> > I only see a flush requirement, which probably does not need to be done
> > indirectly through blocks.
> I suppose that I agree. Seek should not be handled through a block flag. The
> packetizer and decoder should be reset, by calling the flush callback, by
> destroying and recreating them, or whatever.
> That´s not what the patch series did.
Yep, the patch series mixed up the concepts between what discontinuity
actually should mean.
> I suppose so, although I suspect that some decoders could do away with the
> timeline entirely, and just copy the PTS to the output timestamp.
Most non video decoders most likely should be okay
More information about the vlc-devel