From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10004251440.AA17526@clio.rice.edu>
Subject: Re: The new cwsdpmi
To: ams AT ludd DOT luth DOT se (Martin Str|mberg)
Date: Tue, 25 Apr 2000 09:40:54 -0500 (CDT)
Cc: djgpp-workers AT delorie DOT com (DJGPP-WORKERS)
In-Reply-To: <200004250447.GAA02932@father.ludd.luth.se> from "Martin Str|mberg" at Apr 25, 2000 06:47:40 AM
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

> > > Once I had an error message saying "FPU overrun", I think.
> > 
> > Very strange - this means the FPU had an exception (only found on 386
> > chips).  I've never seen one.
> 
> Note that this is a 386 without any math coprocessor.

Even more bizzare - I would think impossible ...

> > my diagnostic routine.  CWSDPMI with diagnostics is ugly - there is a 
> Put it one the ftp site!

Well, I have to build it.  Then you have to be very careful using it since
it allows some reentrantcy where there shouldn't be any - which can make 
bugs appear where there normally wouldn't be any if there are any hardware 
interrupts (which means you can't touch the keyboard since DJGPP hooks the 
keyboard interrupt).  For hardware interrupt debugging the only reliable
way to do it is to have a second monochrome monitor which has the frame
buffer directly written to. Since there are so many pitfalls using the 
debug stuff and it's almost impossible to interpret - I usually don't 
have anyone else look at it, or will only enable key debug statements
to verify badness.

> Let us know when you've updated it. Perhaps an announcement on c.o.m.d
> would be good too (for other people not reading djgpp-workers).

It appears the only change was in the mswitch code to double check for
A20 being enabled (a bug fix from r4).  If you were running in raw mode
with VDISK.SYS - it would disable A20 on a VDISK access and you would 
see all sorts of bizzare behavior.