Skip to Content.
Sympa Menu

sm-discuss - Re: [SM-Discuss] merged gcc and texlive

sm-discuss AT lists.ibiblio.org

Subject: Public SourceMage Discussion List

List archive

Chronological Thread  
  • From: Remko van der Vossen <wich AT yuugen.jp>
  • To: Andraž 'ruskie' Levstik <ruskie AT codemages.net>
  • Cc: sm-discuss AT lists.ibiblio.org
  • Subject: Re: [SM-Discuss] merged gcc and texlive
  • Date: Fri, 25 Sep 2009 09:26:31 +0200

On Fri, Sep 25, 2009 at 08:57:50AM +0200, Andraž 'ruskie' Levstik wrote:
> :2009-09-24T13:47:Eric Sandall:
>
> > So long as it works. ;)
>
> It works for me and I believe chp also is running it.
>
> Anyway merge is done. Test it and fix it if you see anything odd.

Nice, I'll try to give it a go this weekend, but with my todo list only
getting longer I can't make any promises.

> Now as for merged-gcc. What's the decission? gmp in or out? mpfr in or
> out?
>
> And of course test which solution is best. I would just include it. As
> for security fixes in gmp and mpfr. How likely are they to affect gcc
> through it?
>
> Questions questions questions... but to little answers. C'mon folks be a
> little more engaged in all of this.

I would still like to go with separate, there shouldn't be a reason to
include them. The way I see it, including them is just the easy way out
instead of the best way out of the problem.

The issue is however that I still have no idea what exactly the problem
is. I've done some experimenting with recompiling gcc after a march
switch from x86_64 to core2, (post iso install.) For me it worked just
to first recompile gmp and mpfr, then gcc and g++, (which would be
merged gcc in the future.) However the forced_depends rules I added to
the current spells don't seem to force the order, only that they are
recast, so it isn't a complete solution yet. Also Daniel Bolton
experimented a bit, also compiling both gmp and mpfr before gcc and g++,
but for him the system still ended up with a borked compiler, while it
didn't for me, I still don't understand why and I can't reproduce it.

To properly fix this we first need to figure out why the march change
fails for Daniel, first step of which would be to try to reproduce it on
other systems. So I'd like to ask anyone with a core2 or k8 64 bit
system to have a crack at this. The recipe is simple, just install the
iso in a vm, (or alternatively use the system.tar.bz2 from the iso as a
base for a chroot,) switch march to core2 or k8, scribe update, sorcery
hold gcc g++, cast -c gmp mpfr, sorcery unhold gcc g++, cast -c gcc g++,
cast -c $random C based spell.

When we solved this issue and found a proper way to upgrade after a
march switch we will also need a way to force casting order of certain
spells, even when a circular dependency exists. Perhaps this can be done
with triggers? Though that might get messy...

Remko van der Vossen.

Attachment: smime.p7s
Description: S/MIME cryptographic signature




Archive powered by MHonArc 2.6.24.

Top of Page