I have a 486-based machine.  When I run a kernel based on a source tree
supped 1999-03-11, it works fine.  But when I use a 1.4-based kernel, I
get "pckbdprobe: reset response 0xfc" and the keyboard doesn't work.

With the working kernel, the keyboard is probed as

pckbc0 at isa0 port 0x60-0x64
pckbd0 at pckbc0 (kbd slot)
pckbc0: using irq 1 for kbd slot
wskbd0 at pckbd0: console keyboard

This is weird largely because I can't see any relevant difference.
Checking the $NetBSD$ lines in dev/pckbc/* dev/isa/pckbd* make it
appear I have identical versions of everything except pckbd.c, and that
difference is 1.16 versus 1.17, and the diffs between them are all in
pckbd_set_xtscancode, which is not called when that "reset response"
error occurs.  Besides, I tried using the 1.16 version of pckbd.c and
it didn't make any difference.

I suspect some kind of timing weirdness, because every once in a while
it does something different.  Once, with pckbd.c 1.17, it worked
completely; once, with pckbd.c 1.16, I got "reset error 60" instead of
"reset response 0xfc", and it worked.  (Here, "worked" means only that
I could type RETURN to the "enter pathname for shell" prompt and have
it do something.  When it doesn't work it locks up at that point.)
These "once" results are as compared to some 10 or 15 failures.

Anyone have any idea where I could usefully look?  I still have both
source trees online, so I can mix-&-match source files or fragments
thereof if that could reasonably help find/fix the problem.

					der Mouse