Date: Sun, 26 Jul 1998 12:07:20 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Nate Eldredge <nate AT cartsys DOT com>
cc: b52g AT usa DOT net, djgpp AT delorie DOT com
Subject: Re: EMM386 limiting memory to 32Mb
In-Reply-To: <35B7BB95.57E7B48C@cartsys.com>
Message-ID: <Pine.SUN.3.91.980726120655.26387J-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Thu, 23 Jul 1998, Nate Eldredge wrote:

> 256MB as of r4.  And CWSDPMI can't use EMS anyway, only XMS and the
> like.  (Yes, pedantic I know.)

Even more nit-picking: under a memory manager such as QEMM or EMM386,
CWSDPMI uses the VCPI interface, which is an extension of EMS.  The
NOEMS parameter on the manager command line disables VCPI together
with EMS (and so is not recommended, unless you have a good reason for
using it).

> > If you use QEMM's DPMI server you won't be constrained by CWSDPMI.
> > I don't know if there are any limits inherent in the compiler libs or
> > not. (Apart from 4GB addressing)
> 
> I seem to have heard about show-stopping bugs in at least some versions
> of QDPMI.

Indeed.  I don't recommend QDPMI except for some very special cases
(e.g., if you *must* run DOS4GW-extended programs under DJGPP
programs).  At least in QEMM v8 and before, QDPMI doesn't detect NULL
pointer dereferences, will crash your program if it generates Int 24h
(e.g. by accessing an empty floppy drive), and aborts the parent
program when you interrupt a child program with Ctrl-C, even if the
parent has installed a handler for SIGINT.

FWIW, I have QDPMI loaded but turned OFF on my system, and don't turn
it ON except when testing a program for any incompatibilities with
QDPMI proper.