[vlc-devel] [PATCH] pulse: avoid overflows

[Thomas Guillem]

By the way, I think everything is working more by chance than anything.

There is no clear rules that demuxes/decodeurs should not sent more than 10sec (maxlength pa buffer for me) of data with one block ? We also let pulseaudio decide its max buffer length. I guess this can be configured by the user in order to use less memory.
With current master, I could have a delay of 300ms or 2sec, depending of the input.

On Mon, Feb 11, 2019, at 18:07, Thomas Guillem wrote:
> 
> On Mon, Feb 11, 2019, at 17:51, Rémi Denis-Courmont wrote:
> > Le maanantaina 11. helmikuuta 2019, 18.44.24 EET Thomas Guillem a écrit :
> > > > Currently the PulseAudio output is configured with buffers larger than the
> > > > maximum advance guaranteed by VLC core.
> > > 
> > > Do you know where it is done ? I'm currently trying to fix pulse underflow
> > > with ogg files using the new output clock. I tweaked the input buffering
> > > and ended up having overflow...
> > 
> > Look for AOUT_MAX_PREPARE_TIME, I guess.
> 
> There is a define, but like AOUT_MAX_ADVANCE_TIME, it is only used by 
> aout modules. Nothing in the core.
> 
> The audio output buffer value is directly linked to the jitter value of 
> the input clock. It is coming from the demux pts_delay that is 300ms 
> for file, and 3000ms for network by default.
> 
> What is really puzzling me is that 300ms is not enough anymore using 
> the new output clock (only with some files).
> After doing some test, a value of 1000ms is way better.
> 
> 
> > 
> > -- 
> > Реми Дёни-Курмон
> > http://www.remlab.net/
> > 
> > 
> > 
> > _______________________________________________
> > vlc-devel mailing list
> > To unsubscribe or modify your subscription options:
> > https://mailman.videolan.org/listinfo/vlc-devel
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel