Sender: nate AT cartsys DOT com
Message-ID: <3783E2C9.951F6654@cartsys.com>
Date: Wed, 07 Jul 1999 16:29:13 -0700
From: Nate Eldredge <nate AT cartsys DOT com>
X-Mailer: Mozilla 4.08 [en] (X11; I; Linux 2.2.10 i586)
MIME-Version: 1.0
To: djgpp-workers AT delorie DOT com
Subject: Re: gcc-crash - and a possible solution
References: <9907061440 DOT AA15007 AT clio DOT rice DOT edu>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

Charles Sandmann wrote:
> 
> > If you mean two DOS boxes with *any* DJGPP programs, not necessarily GCC,
> > then I always do that.  On my office machine, running Windows 95, there's
> > always one DOS box with Emacs, another with Info, and I use two other for
> > command-line tasks, including compiling programs.  I never had any
> > crashes in CC1 like what you see.
> 
> To get the 3.9Gb addresses you must exit an application run earlier (so it
> frees it's memory which is typically lower addresses) and then have an
> application still running allocate more new memory.
> 
> The sorts of things that broke some applications:
> 
>  0-1Mb space, then sbrk returns 3.9Gb block (broke signed/unsigned)
>  0-1Mb space, sbrk returns 3.9Gb address, then sbrk returns 1Mb address
>        (this broke some assumptions in some code about sbrk always
>         increasing)
> 
> It may be the second case which is biting malloc and/or CC1.

Maybe it would be interesting to sadistically patch CWSDPMI to return
blocks all over the address space.  This would make it easier to
reproduce such problems (though I have no idea how easy it would be to
implement).

Or can it be done in sbrk?
-- 

Nate Eldredge
nate AT cartsys DOT com