Message-ID: <34165050.3680@bo.dada.it>
Date: Wed, 10 Sep 1997 09:46:24 +0200
From: Diego Zuccato <dz AT bo DOT dada DOT it>
Organization: CyberSpace Software Labs BBS
MIME-Version: 1.0
To: Bill Currie <billc AT blackmagic DOT tait DOT co DOT nz>
CC: djgpp-workers AT delorie DOT com
Subject: Shared memory (Was: Re: fread slowstart)
References: <341541B4 DOT 2462 AT bo DOT dada DOT it> <199709092121 DOT JAA13127 AT teleng1 DOT tait DOT co DOT nz gatekeeper.tait.co.nz>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

Bill Currie wrote:

> DMPI 1.0 has shared memory. It shouldn't be that difficult to detect
> 1.0 and do the appropriate thing.  It also shouldn't cause too much
> bloat.
Well, DPMI 1.0 have shared memory. DPMI 0.9 let us pass descriptors...
It shouldn't be too difficult to handle both cases (shared memory isn't
just a single handle shared between multiple processes ?).

[single tb]
> I would have to agree with this, I think. I can't think off hand what
> sort of programs would have this problem, but I can imagine the
> possible situations.
Well, I think that every app (well, really, every library function) uses
tb just as temporary buffer. They should save tb contents before
exiting, since another library function destroys it. Eg, if I call
write() then read(), tb is overwritten. I think it's overwritten even if
I call exec() or spawn(). I don't see why the library should assume that
tb won't be changed by a call to exec() or spawn(). IMHO assuming it is
quite dangerous !

BYtE,
 Diego.