I looked at the IIsi shutdown problem last night.

The code that's there (via_shutdown() in via.c) is clearing the correct bit in
the IIsi VIA2 emulator, if we assume that the via2_reg macro from via.h is
actually giving us the correct bits to play with.

My final conclusion was that either clearing that bit didn't shut off the power
on a IIsi, or that the bit we were clearing was actually not in the correct
byte, and that we were actually just randomly twiddling bits somewhere else in
the address space.

Sadly, since I discovered this, there have been two messages confirming both of
my suspicions.

Given my choice, I'd blame the via.[ch]/machdep.c/pmap_bootstrap.c problem
described by Olivier Galibert/Sarayan, as I had pretty much come to suppose the
same, I just wasn't sure enough to post about it yet.

I just can't imagine Apple implementing a new way of doing something, and not
keeping the old way working. They seem to go to so much trouble to make things
look the same when they implement new hardware that does an old job. I'd
suspect that even though the power management was handed off to the custom ADB
chip in the IIsi, and that there is an ADB like interface, that the VIA2 like
object would still be able to trigger the same effect, albeit by a different
mechanism.

Bernard.