You're absolutely right, holidays aren't us...

I think your suggestion is probably the way out.
We can make the 'actualbufer' larger and fill it
in ksmp blocks. That's what I was thinking 
yesterday. But then, should we make it uniform
across the systems, using the same method for
windows? The trouble is that we don't want to
break something that is working.

We should also be thinking of using the callback
method with linux, instead of the blocking calls
as implemented presently. Wouldn't it work?

Victor

> 
> Victor, you disappoint me!  I thought you were an academic
> and academics do snot have holidays, just days when the
> students have vacated the university....
> 
> Thanks for working that out.  I had got as far as
> realising that the pa-blocking stuff was synchronous with
> krate, but then could not see how it worked on any other
> platform.
> 
> Suppose the callback thread waited on a mutex until there
> was enough audio before releasing it, and the play code
> could release the mutex when there was enough.  Would be
> easier if the audio buffers were a multiple of the krate,
> but not essential.  Would that work?
> 
> ==John ffitch
> 

I have most of the code for the "improved" pa_blocking written, but
testing is causing me problems -- the mac seems to have decided that
it it on vacation -- or at least not trying!
   More later.

==John ffitch