From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10210151630.AA21488@clio.rice.edu>
Subject: Re: 2.03 vs 3.2
To: djgpp-workers AT delorie DOT com
Date: Tue, 15 Oct 2002 11:30:46 -0500 (CDT)
In-Reply-To: <30240030515.20021015122801@softhome.net> from "Laurynas Biveinis" at Oct 15, 2002 12:28:01 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

> > Someone mentioned they didn't believe building 2.03 with newer GCC
> > should be considered; in fact the patch for bzero (?) probably breaks
> > building that library function with newer GCC since the code has the 
> > wrong prototype.
> 
> Could somebody elaborate on this, please?

bzero has the wrong prototype in 2.03 (it returns a pointer to the
object, not void).  But newer GCCs notice this and dislike it, since
they have a builtin which conflicts.  For 2.04 we changed the library
function (minor incompatibility).  For 2.03 it was too late to change 
the library when this was brought up; it also would break strict binary
compatibility if a user was actually using the value.  So, the prototype
changes in the header depending on the GCC release (string.h).

At a minimum you would probably need an #ifdef in bzero.c to compile
with GCC 3.x

Issues like this, DXE linker scripts, strict warnings, warning on errors,
all made us decide that building 2.03 with GCC 3.x was an exercise best
left to the user :-)

Not the best use of valuable development time ...

> > I'd rather spend time getting 2.04 released.
> 
> Indeed. However, 2.04 is still a few months off, and 2.03 update could
> be made sooner. I think...

I still haven't finished the last one, and it's been 10 months, so 
2.04 might happen first :-(

> BTW, I think we really should start using branches in future - as soon
> as we don't accept new features into 2.04 we should branch it off. DJ
> once objected, according to him branches are PITA to merge, however
> current situation is even bigger PITA. What do you think?

In the 1.x days we did maintenance releases, and they were a pit.  We
decided to just do real releases more frequently.  This worked OK for
early 2.x releases.  2.03 was around a long time, and too much stuff 
was in CVS/2.04 without incremental releases.  So we branched 2.03
for the refresh (bug fixes only, no new features).  And it has been 
a small time pit.

I don't mind CVS branches, but I wouldn't want to create any without
needing to.  So unless we have a bunch of bug fixes for 2.04, and we
can't manage to get 2.05 out timely, I'd want to avoid it.