Hello,

I already posted about what appeared to be random kernel hangs when
using M$ Win telnet client (NetBSD 1.6) on my VS3100. I'm using puttytel
now and since i thought i should check out some X11 packages from pkgsrc
that are not avaliable in pre-compiled form i had the machine running
few days now (turned off for few hours during the night), compiling the
likes of perl, graphics libs and so on. 

(i'm sure perl exists in precompiled form too, before anybody mentions
that  ... and if i knew it was going to take until 2 AM for it to finish
compiling i'd go for the binary package, my eyes are like those of
angora rabbit)

Anyway, i thought i solved the problem i attributed to weird telnet
client and i was under the impression that hardware (RAM) can't be
faulty if 40+ hours of compiling went through without a single mishap.
When icewm was finally built (along with some tools to use it) i started
xdm and played for a while from my other box, without any problem
either  .. icewm looks pretty .. and pretty fast too for m38 BTW, i was
expecting a slide-show but it would allow decent work as an office
computer if only i had an X terminal that wasn't way more powerful than
my VS - things like menus open exactly as fast as they do on 1.7 P4
Windoze i used as Xserver :-))) Anyway, back to my story: then i left
puttytel running, with no other login (nothing on serial console) and
forgot about it after a while. No more than 2 hours later the vax
crashed. I hooked up a serial console and i saw it crashed to the >>>
prompt, while doing nothing ?!?!

So ... does anyone have any suggestions ? :) I can't have serial console
hooked up all the time so i couldn't see what kind of HLT that was - is
it possible to check it somehow *after* a crash occurs ?? How comes it
didn't crash under heavy load (top said 98+%) with just about all of
internal RAM (16 MB) used and few MB of swap (~5 MB out of 32 MB) used ,
but it did crash sitting idle, doing nothing, just blocked on select on
virtual terminal ? Once again, i'm using 1.6 because 1.6.1 refuses to
compile. And i'm most certainly not going to unpacking of that gigantic
source tree just to have it fail again because it doesn't seem to like
my compiler (from 1.6; complains something about an option mentioning
PIC not working, yet pkgsrc stuff all compiled with flag looking alot
like that one ... leaky memory on my part but i do recall -f and PIC
part). This time i only had 1 virtual terminal connected, i used to have
2 when it crashed last time before today.

Blaz Antonic