[Qemu-devel] Re: [PATCH 45/46] Remove AUDIO_DRIVERS define

[malc]

On Sun, 26 Jul 2009, Avi Kivity wrote:

> On 07/26/2009 06:09 PM, Jan Kiszka wrote:
> > Avi Kivity wrote:
> >    
> > > On 07/26/2009 04:41 PM, Juan Quintela wrote:
> > >      
> > > > The order in audio_drivers is important, that means you can have a
> > > > different default sound driver.
> > > > 
> > > > After some thinking, my current plan is exporting audio_drv_list
> > > > to the Makefile, and then generating something similar to AUDIO_DRIVERS.
> > > > 
> > > > Notice that the "ordering" problems not only happens with sound drivers,
> > > > we have the same problems with graphics console, which one is the
> > > > default one?  It depends on what you have configured, and changing the
> > > > default order is far from trivial.
> > > > 
> > > >        
> > > You could pass a goodness value to the registration function, reflecting
> > > what we think the best driver is.  Then at runtime select the best
> > > driver that is supported on the system.
> > > 
> > >      
> > 
> > I think the easiest way is to forward the sorted driver list as string
> > to the audio layer and let it be processed there. Given such an
> > extension, we could also allow passing a list of drivers via QEMU_AUDIO_DRV.
> >    
> 
> Yes, that is better.  Decouple the goodness from the driver, and allow the
> user to override it.
> 
> > BTW, I still wonder why we have OSS and only OSS as the default driver
> > list under Linux. Most distros I came across kicked it out or pushed it
> > to the end of their list (as it requires exclusive access). Would be a
> > chance to improve the default. At least we should extend the list by all
> > available drivers.

Nothing has changed since we had our private e-mail exchange on the
subject a while ago. As for 'OSS and only OSS as the default' i'm not
sure what that really means, maybe this is an answer though - OSS is
the only thing that doesn't require additional headers and libraries,
and just works.

> > 
> >    
> 
> I have some secondhand memories of pulse and alsa generating very bad output.

I have some firsthand memories of this.

-- 
mailto:address@hidden