On Tue, 18 Sep 2007 16:06:09 +0200
Markus W Kilbinger <mk@kilbi.de> wrote:

> >>>>> "Nick" == Nick Hudson <nick.hudson@dsl.pipex.com> writes:
> 
>     >> http://mail-index.netbsd.org/current-users/2007/09/10/0018.html
> 
>     >> > > I've been speaking with Nick and the problem is understood,
>     >> > > basically gcc is generating assembly which breaks the
>     >> > > restartable sequences in libpthread. Stay tuned.
>     >> >
>     >> > Ah, that does sound like it could be the problem. I "fixed"
>     >> > this problem on my server by compiling libpthread with
>     >> > COPTS="-O0" for the time being.
>     >>
>     >> Many thanks! this worked for me too (thus far).
>     >>
>     >> I hacked my libpthread/Makefile to add -O0 for now.
> 
>     Nick> Most arches are fixed with pthread_lock.c:1.28. I'm working
>     Nick> on making this more robust.
> 
> With this recent change my cobalt qube2 seems to run fine again (esp.
> pthread related).
> 
Unless I made a mistake in testing this morning, it was *awful* for
me.  I had to copy over a libpthread from an Aug 22 build, because my
mailer was 100% unusable.  (I'm certain about a build from yesterday; I
tried a partial rebuild this morning and may have made some mistake
testing it.)


		--Steve Bellovin, http://www.cs.columbia.edu/~smb