> 	What options did you give to crashme - I tried the suggested
> 	options as in the manpage and my sparc5 clone ran for a few
> 	minutes before it paniced with a trap. (I'm running the above
> 	patches plus anything else scheduled for 1.4.1)


Historically, unhandled user generated traps would end up in panic()
on the sparc.  Last time I was near that code, I made that behaviour
optional (but left the default as was) by introducing the patchable
variable `ignore_bogus_traps'.

At that time, I ran `crashme' continuously for a week on a Sparc Classic,
with no ill effects except slowing down my kernel builds.

I guess it's time to invert the default value of `ignore_bogus_traps'.

The sun4 crash problem on the other hand, is very real and appears
to be caused by pointers going astray. I'm looking into this.

-pk