uaudio patch, configurable buffer size

[Mathew Kanner]

On Mar 06, Alexander Leidinger wrote:
> On Sun, 6 Mar 2005 12:10:27 -0500
> Mathew Kanner <mat at cnd.mcgill.ca> wrote:
> 
> > > Can you also please have a look at PR usb/78028? It adds some info to
> > > /dev/sndstat.
> > 
> > 	For now I don't object to his first try that always producing
> > verbose output for format discovery.  The second part that adds sndstd
> 
> I don't think we should clutter the boot process with this output. If
> it's interesting for someone, he shouldn't need to look at those
> messages. That's the reason I proposed to add the information to
> sndstat.
> 
> > output I think is wrong as it duplicates sndstd itself (though a
> > little prettier).  
> 
> Do we have a way to extend sndstat? If not, we should add it, since a
> device may have to say something interesting about its capabilities
> (read: it would be nice if every device would tell the use about its
> capabilities in sndstat).
> 
> > 	However in the long term I don't think it's the right thing.
> > The reason (I'm guessing) he wants the verbose output all the time is
> > to be able to choose according the devices capabilities.  The problem
> > is we always reports supporting speeds in between [4000,48000], so for
> 
> Why?

	As I say below, he probably needs this because uaudio (the pcm
bridge of it at least) wrongly reports it's capabilities.  I would be
that his first solution and second solution don't even yield the same
results.

> 
> > example on my device which only supports 48000hz, mplayer will happily
> > try 44100hz and produce no output.  Forcing resample to 48000 (mplayer
> > -af resample=48000 ...) works.  The situation is even worse in regards
> > to reporting soft formats as you might not get a vote in what speed
> > the hardware device gets set at.  (All the same applies to formats as
> > well).
> 
> So either we need a format converter in the kernel (looks ugly to me, it
> belongs IMHO into the userland, but NetBSD seems to have this
> possibility in the kernel), or we need to reject incompatible use.

	It does belong is userland.  We could get the same effect by
having the something probe capabilities via ioctls but we need to fix
the reporting of caps first.
	--Mat

> 
> Bye,
> Alexander.
> 
> -- 
>               To boldly go where I surely don't belong.
> 
> http://www.Leidinger.net                       Alexander @ Leidinger.net
>   GPG fingerprint = C518 BC70 E67F 143F BE91  3365 79E2 9C60 B006 3FE7

--