To:	eliz AT is DOT elta DOT co DOT il
Cc:	djgpp AT delorie DOT com
From:	lce AT mcs DOT com (Larry Ellis)
Subject:	Re: Windows NT
Date:	Sun, 25 May 1997 11:57:23 -0500
Message-ID:	<pz6YpG9s1K48075yn@mcs.com>
In-Reply-To:	<Pine.SUN.3.91.970525140343.4960X-100000@is>
Lines:	52

> 
> On 24 May 1997, Larry Ellis wrote:
> 
> > Under Windows NT 4.0, DJGPP executables often generate SIGSEV failures
> > if CONTROL-BREAK is used.  After the segment violation occurs, the DOS
> > session is no longer usable and must be terminated via the task manager.
> > Sometimes the segment violation is trapped by DJGPP's internal fault
> > handler, other times not (caught instead by NT).
> 
> I'd guess these are bugs in the NT DPMI host, but please tell more
> about what programs crashed and what were you doing when they did.
>

The original problem occured with a little directory compare utility I
wrote. But it also occurs with the GNU DIFF utility.  

I don't think it is a simple as being a bug in the NT DMPI host, though,
because when I compile my program under Watcom C and use DOS4GW as
distributed w/ Watcom, the problem does not occur.  But I'd prefer to use
DJGPP.

> > Has anyone else experienced this or have any further information?
> 
> Emacs is known to crash on NT, and it has been established that this
> is a bug in NT DPMI host.  I don't know about other programs, but you
> didn't say which programs crashed.
> 
> > p.s. I tried including PMODE11 in the executable as an alternative, but
> > using PMODE11 under NT locks the DOS session with no other visible
> > symptom (no fault report or error message).  Anyone have anything further
> > on this?
> 
> You can't do this on NT, since it has its own DPMI host built-in, and
> even if you were to succeed in installing PMODE, DJGPP programs won't
> use it because DPMI services are already available on NT when a DJGPP
> program starts.
> 

I was using PMODE11 as bound into my DJGPP application.  While I agree
that the native host DPMI should get used in this case, I don't think the
mere presence of PMODE11 should cause such catastrophic events.  My
Watcom executable with DOS4G runs flawlessly, for example, control break
and all.  Perhaps PMODE11 is attempting to install itself as the DPMI
server even though one is already in place, and DOS4G does not.  Don't
know.

Anyway, thanks for responding, and I will post that SIGSEV info as soon
as I get it.



Larry Ellis

- Raw text -