"The Good..."

My kernel and toolchain compiled perfectly from 1.2G -> 1.3_ALPHA.
Now for userland...

"The Bad..."

options DDB
options DEBUGGING

don't work.  My kernel panics with a data alignment fault with
something like:

jmpl [%g0 + %o7], 0x8

[or something similar].

All the register values which were referenced were on longword boundaries,
so I don't see the problem.

Further, when it dropped into debug mode (the panic started after it started
the rc fsck), I started getting

spurious interrupt on zs1 ipl 0xc...
[or something]

when I was typing on the keyboard trying to do a trace.

DB then came back with "Crazy interrupts" and gave my db> prompt back
with the same trace:

jmpl 	[%g0 + %o7], 0x8

[or whatever it was].

I removed "options DDB" and "options DEBUGGING" and the kernel works fine.
Not a horribly big deal, but we should fix this.

SPARCstation IPX (4c/50), 64MB RAM (never swaps, even with netscrape
running).

"...and the Ugly...":

Speaking of 64MB RAM, I have my swap space spread across three disks
so as to balance the load, but it's in three 32MB chunks.  This
means that I cannot get a kernel crash dump because it tries to dump
ALL of memory.  We need to get something in place which will either
only dump used pages (a' la SunOS/pyramid OSx) or which can span
multiple dump devices.

I notice also that there is no way to hardwire swap/dump space -- at
least it's not well documented (that I could see).  Now the ability
to automagically determine the root device is nice (I know it's always
been there, but after much consternation of trying to boot a kernel
from an alternate device only to have it use its idea of root (hard
wired!), I stopped doing this...), but I'd really like to see the ability
to tell the kernel where it's supposed to dump.

Or does it automagically dump to the first _configured_ swap device?
That would suck, because that means I have to allocate another 32M
[and rebuild my disk].  PITA.  [it's not just bread anymore.]


				--*greywolf;
--
Woof...