In muc.lists.netbsd.port.sparc Rick Kelly wrote:

>Michael-John Turner said:

>>Haven't really had a chance to investigate further, but now that you
>>mention threads, I'm wondering if that doesn't have something to do with
>>it, although I thought only sparc64 had thread issues with 2.0?

>The problem, as I understand it, is that if a threaded application swaps
>out then the system dies. If this is still true, then I might as well keep
>running NetBSD 1.5.4 on my nameserver. It just works and never goes down.

2.0.1 on Sparc 64 still has the problem.  While I was running
clamav-milter with native threads, the system crashed about once a day.
Since I have recompiled clamav-milter with pkgsrc "pth", the machine is
rock solid, not a single crash.

There is code in -current to fix this, and I've read reports that it's
fixed for good in -current on Sparc, but as far as I know, Sparc64 is
still not fully fixed.

So if you want to run BIND 9 on 2.0, compile it with pth.  Be prepared 
for some system include file (/usr/include/) mangling, as 2.0 doesn't 
like you to use non-native threads libraries.  At least I haven't found 
a clean way yet to include <signal.h> (which has prototypes using the 
pthread_t data type) without using the system <pthread.h> - which, of 
course, is incompatible to pth pthread.h...

gert
-- 
gert@greenie.muc.de   fax: +49-89-35655025   http://alpha.greenie.net/mgetty/

"Never try to catch two frogs with one hand."
		-- Chinese Proverb