Date: Thu, 21 Apr 94 12:11:11 EDT
From: peprbv AT cfa0 DOT harvard DOT edu (Bob Babcock)
To: WKKST AT vms DOT cis DOT pitt DOT edu
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: Re: 2 questions about readme.dj
Reply-To: babcock AT cfa DOT harvard DOT edu

> Maybe you are right, as I haven't tried so many compiling under the same OS/2
> dos session.  If so, 32-bit gcc leave some garbage in memory after every
> execution and eat up all dos memory after many executions??

Available memory does not decrease.  The _speculation_ was that selectors
were not being released.  Since the real-mode gcc fixed the problem with no
apparent performance degradation, I saw no reason not to use it.  On the
other hand, if djgpp 2.0 is going to not use a Borland compiler, the problem
may come back.

> If this really happened under DPMI, what about under Windows(non-OS/2) dos
> box?  Do you experience the same thing? (Maybe after a lot many # of execs)

I can't remember for sure whether I tested under Windows or not.  If I did, I
did not get failures.  I suppose I should go back and test again, both with
Windows and with other DPMI providers such as QDPMI.