[MPlayer-dev-eng] A-V sync handling

[Uoti Urpala]

On Sun, 2009-10-04 at 18:20 +0200, Dan Oscarsson wrote:
> sön 2009-10-04 klockan 19:07 +0300 skrev Uoti Urpala:
> > On Sun, 2009-10-04 at 17:33 +0200, Dan Oscarsson wrote:
> > > I have tested changing the code to use pts values instead and so far it
> > > works fine, and A-V syncing is handled as part of the sleep time
> > > calculation instead of separately. Are there any problems doing it this
> > > way instead of using "mpctx->delay"?
> > 
> > Details of course depend on exactly what your "instead" is, but I
> > suspect you've at least made it less robust against inconsistent
> > timestamps, especially pts resets.
> 
> For every frame, mplayer will adjust "delay" when it compares pts values
> (in adjust_sync_and_print_status) to correct the A-V sync. If you get
> inconsistent timestamps or resets, you should get the same problem
> there. Or am I missing something?

You are missing something, perhaps that the adjustment does not try to
force immediate match in audio and video pts like straightforward timing
code using the pts values directly would do.