Re: GCC 2.8.0 + Libg++ 2.8.0

[Niklas Hallqvist <niklas@petra.appli.se>]

Sorry for the late answer, been snowboarding..
> Date: Mon, 9 Mar 1998 05:09:56 +0100
> From: Marc Espie <Marc.Espie@liafa.jussieu.fr>

> - why is gcc built with -O ?   Any obvious bug we're stumbling upon while
> building -O2 ?  I'm wondering, since I know for sure that most everybody
> uses gcc with -O2 all the time, heck it's even *recommended* in the 
> documentation to use -O2 for stability.

Mostly because -O is our "standard" level of optimization.  For
quality control we like to share the same level all over our tree.
We could change to -O2 but then we should do that globally, and that
would hurt performance.

> - the Makefile.bsd wrapper does look weird, and not too flexible. For
> instance, building the whole gcc/objc/f77/c++ setup *twice* looks like a
> waste of resource, especially when bootstrap targets that bootstrap the
> C compiler, then build the other tools that don't need to be built twice,
> exist.  Maybe they're very difficult to adapt ?

I don't understand.  We don't bootstrap gcc in-tree, we just build it
once.  Or do I misinterpret you?  If we were to go to two-stage (or
even three stage builds) we should do that globally too, as there are
things like sh, sed, yacc etc. involved in building.  However I think
that is overkill.  Instead we build twice or more when changing
versions and announce that on a mailing list, but that is an
exceptional case.

> - Is there any reason why the openbsd config info does *not* appear in
> the gcc original tree, specifically I'm thinking of egcs ?  Even if it is not
> thoroughly tested, giving that to the egcs people would at least give a
> simpler base to work against in the future.

There's two reasons:

A	Many of the changes stem from work other people have done with
	unclear legal status.  I have messed up once when submitting
	patches to FSF containing code from an individual whose
	employer at the time never did sign the FSF papers.  It was
	hell.  Things must be very thoroughly checked before
	submitting.  The FSF is very fuzzy about having a disclaimer
	from employers, and if you cannot get one out of yours,
	you will not be accepted as a submitter.

B	I haven't had time to collect the legal info on the various
	patches we want submitted, thus I have not done this.  If
	someone would like to step in and sort the mess out, or
	rewrite dubious code sections that would be swell.

> - in the same mindset, why is gcc/configure patched for openbsd targets and
> not configure.in ?  I can understand that we may not want to use 
> autoconf/gnu-m4 and keep the source tree leaner, but having configure track
> configure.in makes it easier to work with mainstream sources.

Oversight.  If you want to please sync them and mail me a patch.
Otherwise I will get to it later.  I have also had a diskcrash which
makes all my duties towards OpenBSD  for me even more cumbersome.
So I am very happy for every help I can get.

> On a lighter note, I've managed to configure/build egcs-pre-release-1.0.2
> using m68k-unknown-netbsd as target without any problems (C and C++).

That's great!

Niklas