From: Gilles Filippini <gilles.filippini@...>
Date: 2003-02-27T06:16:48+09:00
Subject: Re: rb_thread_create / rb_thread_schedule

Yukihiro Matsumoto wrote:
> 
> Hi,
> 
> In message "rb_thread_create / rb_thread_schedule"
>     on 03/02/25, Gilles Filippini <gilles.filippini@free.fr> writes:
> 
> |Could somebody tell me what's wrong with the attached piece of code?
> |A ruby thread is created before entering the interpreter's main loop.
> |It segfault when the interpreter exits.
> 
> This may be fixed by:
> 
> Sat Feb 22 02:45:20 2003  Yukihiro Matsumoto  <matz@ruby-lang.org>
> 
>         * eval.c (rb_thread_create): may called from place higher than
>           rb_gc_stack_start.
> 
>         * gc.c (Init_stack): update rb_gc_stack_start if it is lower (or
>           higher if stack grows down) than the previous value.
> 
> Could you try the latest CVS?
> 
>                                                         matz.

It's OK with latest CVS.
This bug was related to MPI Ruby being broken. I need to do some more
testing but at least it doesn't segfault with one node.

Thanks.
_gilles.