From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10305112241.AA23271@clio.rice.edu>
Subject: Re: DXE3 in DJGPP CVS breaks cross-compiling
To: eliz AT elta DOT co DOT il
Date: Sun, 11 May 2003 17:41:26 -0500 (CDT)
Cc: djgpp-workers AT delorie DOT com, dborca AT yahoo DOT com
In-Reply-To: <2593-Sun11May2003173008+0300-eliz@elta.co.il> from "Eli Zaretskii" at May 11, 2003 05:30:08 PM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> > (yes, I know it could be more defensive on buffer sizes, but I've been
> > up for over 24 hours straight right now flying back from europe and I
> > really don't care ;-P
> 
> How about making the buffer where you assemble the command line be 64K
> long?  The transfer buffer can never be larger, not until MWDPMI hits
> the street, anyway ;-)

I've actually got the old dxegen code which mallocs/reallocs if needed,
which was done after the first buffer overflow discovery.  But 64K 
does sound nicely overkill and simple to do, and with our default stack
size leaves lots of room for other stack overkill ...

That does remind me that I need to decide what to do with dxegen.c -
it's still in the source tree but no longer used, and I have some
patches for it which were probably never committed.

> I have one more concern about replacing `spawn*' with `system': they
> behave differently as far as argument quoting is considered.  Thus, if
> some file names include whitespace or other characters that need to be
> quoted (like quote characters themselves), we are up for a bumpy
> ride...

The old dxegen didn't support things requiring quoting, and I'd rather
document this as a restriction than fix it.  (People who use characters
which require quoting in file names deserve some pain).

> To remedy this, we need to quote every arg that needs it (or maybe
> just quote all of them).

I'd prefer not.  The old code didn't.  If we really require that, then
maybe we don't need to have cross compile support and should leave it
the way it is.