> After fixing thorpej's bug and ifdef'ing out the pmap_botch message (I
> still want to know whether that's a suboptimality condition or a
> possible-error condition, and why it's happening... I can provide more
> data on just what mmap's are being performed if people want), Solaris
> emulation seems to work fine (modulo an AFS misfeature regarding
> directory entries).

I don't know what SunOS 5 executables actually do with mmap() calls,
but obviously they use MMAP_FIXED a little to often.. That said, the
current method of dealing with cache-aligned addresses returned from
mmap()ing is not fool-proof. This is so because the MI VM code doesn't
support the notion well enough at this moment.

After the release is done, I intend to look more at SunOS5 compatibility
(by then, we'll also have our first Solaris 2.4 installed here, so the fun
can start :-)

-pk