On 04.08.2009, at 21:51, David Corking wrote:

> Bert Freudenberg wrote:
>
>> Only if someone can think of a way to reliably detect the problem.
> ...
>> (PulseAudio) is what appears to break Squeak sound, because it works
>> fine on "real ALSA" but breaks on "emulated ALSA".
>
> Oh, I had hoped the workaround would be an effective fallback for
> everyone on Debian derivatives.  I bet the lack of native ALSA is far
> from trivial to detect in a shell script :(

Yes, but detecting active pulseaudio is simpler. See my comment on

http://tracker.squeakland.org/browse/SQ-214

>> OSS is obsolete. It's not supported out-of-the-box on most modern  
>> Linux
>> distros anymore. If at all, it is provided by an emulation layer like
>> alsa-oss which adds considerable delay. We would get Squeak talking  
>> to OSS
>> talking to emulated ALSA talking to PulseAudio talking to real  
>> ALSA. No
>> wonder there is delay in that.
>
> I haven't done any testing - is the delay noticeable?

We tried it in SoaS on an Intel Classmate in Mexico today. First using  
the alsa-oss wrapper which had an awful delay. Then using the padsp  
wrapper, much better, but still with a slight delay.

> The OSS API
> (via emulation layers) is not obsolete, and OSS is still the standard
> on BSD.

You are right - but since in many Linux distros it is not pre- 
installed anymore (not even the emulation wrapper) we cannot rely on  
it. Likewise, it is not on the OLPC XO (which actually prompted doing  
the ALSA driver).

> 10 years ago, ALSA was the next big thing.  Now, ALSA is fading away,
> and OSS and /dev/dsp , with all their flaws, are still here.

ALSA seems to still be the low-level interface of choice. It's just  
that PulseAudio is trying to virtualize it, breaking some code.

>> Here is an overview of the various Linux sound APIs:
>>
>> http://0pointer.de/blog/projects/guide-to-sound-apis.html
>
> (great guide - thanks -  and the author admits his bias too!  I also
> enjoyed the linked article:  http://lwn.net/Articles/299211/  and
> comments)
>
> Given that so many distros have already gone to PulseAudio, Lennart
> makes a pretty convincing case for the Linux VM to move towards a
> "PulseAudio-safe" subset of the ALSA API.  Thanks for your patience
> with me.


Well, the code is here if you want to try:

http://squeakvm.org/svn/squeak/trunk/platforms/unix/vm-sound-ALSA/sqUnixSoundALSA.c

- Bert -