[Gc] GC 6.2 on IRIX64 6.5

Shiro Kawai shiro at lava.net
Fri Apr 16 01:22:27 PDT 2004


It worked with 6.3alpha5.  Thanks.

--shiro

From: Shiro Kawai <shiro at lava.net>
Subject: Re: [Gc] GC 6.2 on IRIX64 6.5
Date: Thu, 15 Apr 2004 16:37:31 -1000 (HST)

> Thanks, I'll tell him to do so and report the result here.
> 
> --shiro
> 
> From: "Boehm, Hans" <hans.boehm at hp.com>
> Subject: RE: [Gc] GC 6.2 on IRIX64 6.5
> Date: Thu, 15 Apr 2004 14:17:26 -0700
> 
> > Can you ask him to confirm that this still happens with 6.3alpha5?
> > 
> > This SEGV is expected, and should be caught.  My guess is that 6.2 on Irix
> > used longjmp instead of siglongjmp to jump out of the signal handler, and
> > thus the kernel thought the process was still in the signal handler.  Somehow
> > this information was preserved across the exec (possibly a kernel misfeature),
> > and when the child generates its SIGSEGV, the kernel thinks the SIGSEGV was generated
> > inside a SIGSEGV handler, and hence kills the process.
> > 
> > I haven't seen this problem reported before, but the code was changed in the meantime,
> > since it didn't look correct to me.  If my theory is correct, 6.3alpha5 should
> > work.
> > 
> > Hans
> 


More information about the Gc mailing list