From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10207212337.AA24453@clio.rice.edu>
Subject: Re: Emacs CVS and Windows NT 4
To: djgpp-workers AT delorie DOT com
Date: Sun, 21 Jul 2002 18:37:53 -0500 (CDT)
In-Reply-To: <3D3B0240.762A31E8@phekda.freeserve.co.uk> from "Richard Dawe" at Jul 21, 2002 07:49:36 PM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> Charles Sandmann wrote:
> > If the cli fix doesn't work then that means good old NT 4 doesn't even
> > virtualize the cli, it just ignores it ...
> > 
> > Does this version crash?  It disables the keyboard hooking.  This might
> > be a long term fix.  But in the meantime, it seems that unixy sbrk under
> > NT 4.0 is just badly broken, thanks to MS bugz
> 
> No, the new version doesn't crash. 

This confirms the problem is due to hardware interrupts during the sbrk().
One fix might be to call the toggle around the actual resize in sbrk() -
but this only fixes our hooks and not any the user might add.  If both
the DPMI interupt disable call and CLI instruction don't stop interrupts 
from being sent to the DPMI application on NT, I don't see how anything 
we do can be a complete fix.