From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <9704230248.AA15664@clio.rice.edu>
Subject: Re: CWSDPMI problem
To: crough45 AT amc DOT de (Chris Croughton)
Date: Tue, 22 Apr 1997 21:48:48 -0600 (CDT)
Cc: djgpp AT delorie DOT com
In-Reply-To: <97Apr22.120429gmt+0100.21894@internet01.amc.de> from "Chris Croughton" at Apr 22, 97 11:01:07 am
Content-Type: text
Content-Length: 1352      
Precedence: bulk

> Why does it need the free [disk] space?  Is this a dynamic thing, 
> to see if the swap file can be extended, or just at the start?

It is determined at each call to get dpmi memory status to indicate
the current virtual memory available (to be reported to the program).

> What happens if the free space changes during the running
> of the program, will this cause it to fall over?

Nope, it handles this gracefully, by either failing a new sbrk/malloc
request (if it would exceed the combined total of free real memory and
free disk space), or in the worst case of overallocated non-committed 
pages, a page fault.

> This could be a big problem running under DesqView, where 
> even if I had it patched to different files another window
> could easily eat up space on the drive.

Under a multi-tasking enviroment, QDPMI is probably a much better idea.
But I think you would find the multi-file idea to work fine.

> (My C: partition is usually close to full, because of so much
> software which wants to live there.  Like 40Mb or so of
> DJGPP binaries...)

A good reason to use CWSPARAM to modify your copy of CWSDPMI.EXE to
point to the partition you want to page on.

> (Oh well, hopefully I'll be installing Linux soon and no
> worries about DOS limits...)

You still must tell Linux how much swap to use and where to put it -
no magic there.