[vlc-devel] [PATCH] Revert "pulse: use pa_stream_begin_write()" (fixes #25187)

Thomas Guillem thomas at gllm.fr
Mon Feb 8 09:05:17 UTC 2021



On Sat, Feb 6, 2021, at 17:57, Rémi Denis-Courmont wrote:
> Le lauantaina 6. helmikuuta 2021, 18.40.18 EET Thomas Guillem a écrit :
> > Hello,
> > 
> > My comment in the trac issue is still now answered.
> 
> ???

*not*

> 
> > Pulseaudio is the only aout module that flush in case of overflow.
> 
> Originally, the PulseAudio parameters were picked such that overflow would not 
> normally occur, and so it could be treated as a major failure to recover from. 
> There's no big secret there.
> 
> > I really think it's should be fixed (by waiting from Play()).
> 
> Waiting in Play() will make volume, mute and migration unresponsive, so it is 
> undesirable.

I think it's OK since it's not the nominal case. It's better to wait 2 seconds for volume change than thrashing the audio every 2 seconds.

I think we all agree to properly rework audio output asynchronicity for VLC 5.0. Meanwhile, preventing overflow for non-nominal case should be fixed in VLC 4.0 (and have every aout modules behaving the same way).

> 
> I don't see what this had to do with this patch either way.

This patch was done as a prerequisite for the actual owerflow fix. But as you stated, I missed the extra memcpy by mistake.

> 
> -- 
> 雷米‧德尼-库尔蒙
> http://www.remlab.net/
> 
> 
> 
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel


More information about the vlc-devel mailing list