From:	sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id:	<9711251500.AA16606@clio.rice.edu>
Subject:	Re: Possible enhancements on v2.02
To:	eliz AT is DOT elta DOT co DOT il (Eli Zaretskii)
Date:	Tue, 25 Nov 1997 09:00:05 -0600 (CST)
Cc:	djgpp-workers AT delorie DOT com
In-Reply-To:	<Pine.SUN.3.91.971125134013.27505D-100000@is> from "Eli Zaretskii" at Nov 25, 97 01:43:08 pm

> On Mon, 24 Nov 1997, Charles Sandmann wrote:
> 
> > Images should always start as unixy sbrk(), and convert to non-move
> > only as required by out of memory, nearptr_enable, crt0 flag, or maybe
> > environment var.  Non-move sbrk() should use variable block sizes to
> > prevent the out of memory handles problem.
> 
> These seem to me as two independent enhancements.  Am I correct in 
> assuming that the automatic switch to non-move sbrk can be implemented 
> independently of the non-move sbrk enhancements?

Yes, they are separate.  Both are relatively simple.  The variable block
size hack flanges in perfectly (it's a 2-3 line assembly change).

> If so, is there something in the current design that should prevent 
> adding the automatic switch in the most naive way (i.e., in all the cases 
> you named just set the flag)?

The hard part is catching the failure, setting the flag, and trying again,
keeping the registers in crt0.s straight in all the cases.  There is the 
issue of the low memory unixy sbrk code - this can probably be improved 
upon to use the transfer buffer.

- Raw text -