> There seems to be some problem with something (vague, aren't I?) messing
> around with some location(s) in low memory.  Lots of folks have looked,
> but so far noone has found it.  The problem has existed since at least
> NetBSD/Mac68k v1.0, and possibly even longer.
> 
> Basically, whenever the size of the kernel changes, the behaviour of the
> sound driver is subject to change.  If you're rolling your own kernels,
> you can frequently get the "right" sound behaviour by adding or removing
> superfluous stuff from your kernel.  (For example, if you only have
> internal video, you can add/remove nubus video support.)  If you depend on
> someone else's kernels, you're stuck waiting for them to build a newer,
> different-sized one.

I'm having the same problem with SCSI and the intvid kernels. However, I
think the ASC code now seems to be out of the "trouble zone" since my
IIvx makes a nice metallic "clink" when beeping. I think the tons of video
code I am adding are finally pushing even the SCSI code out of the trouble
zone since I haven't had an error on a new intvid kernel in a while.
Unfortunately, this may mean that my code is moving slowly into "voodoo
memory". :)