Milter-greylist

As I mentioned a few days ago, I'm getting apparently random crashes
without core dumps.

Nothing I did made the crash generate a core dump, so I gave up trying
and ran milter-greylist inside gdb instead. I thought this would be
far simpler than trying to chase down a signal trap that libmilter
might be installing or a chdir() hiding somewhere.

Anyway, the milter crashed eventually, but what gdb showed me did not
look like a valid stack; partly because the stacktrace had only one frame
and partly because what it showed me was the reason for the crash seems
impossible as far as I understand the milter code.

Worse still, even though gdb showed me what threads there were at the
time of the crash, I wasn't able to switch threads to inspect what each
one was doing.

Are there any usual suspects for this kind of behaviour?

Any hints for tracking down what *might* be stack corruption? I'm not
even sure that's it, since it's always possible gdb might have bugs in
its thread handling.

FWIW:

NetBSD 3.1 NetBSD 3.1 (ATLAS) #1: Fri Dec  8 23:24:43 EST 2006  joel@...:/usr/src/sys/arch/sparc/compile/ATLAS sparc

Thanks,

	- Joel