Skip to Content.
Sympa Menu

sm-commit - Re: [SM-Commit] GIT changes to devel-xorg-modular grimoire by Treeve Jelbert (5b73fa5b8f2815509ca860e68f013c9a36c02f1a)

sm-commit AT lists.ibiblio.org

Subject: Source Mage code commit list

List archive

Chronological Thread  
  • From: Thomas Orgis <thomas-forum AT orgis.org>
  • To: Treeve Jelbert <scm AT sourcemage.org>
  • Cc: sm-commit AT lists.ibiblio.org
  • Subject: Re: [SM-Commit] GIT changes to devel-xorg-modular grimoire by Treeve Jelbert (5b73fa5b8f2815509ca860e68f013c9a36c02f1a)
  • Date: Fri, 5 Sep 2014 21:59:30 +0200

Am Fri, 5 Sep 2014 10:39:04 -0500
schrieb Treeve Jelbert <scm AT sourcemage.org>:

> Revert "llvm: => 3.5.0"

Thank you very much. I had a little stroke when reading about your
commit. Two points:

1. llvm is decidedly incompatible on version upgrades, up to the point
where I would rather see llvm3.4 and llvm3.5 spells, etc. One needs to
check dependent spells.

2. Of these dependent spells, the major one (in my view, of course) is
mesalib-1x. Apart from mesalib-1x possibly being broken by llvm
upgrade, reversal of my change to enable non-x86 targets is guaranteed
to kill video support for a box using nouveau and a modern radeon chip
(you guessed it, using R600 llvm output). I'm not telling stories: I
actually have a box running using _both_ nouveau and r600 in a triple
head configuration (Zaphod head, though, merged xrandr display being
tetchy).

This is a decidedly wicked situation. LLVM ist becoming a critical part
of desktop infrastructure and it's dangerous to mess with, even alone,
as you whitnessed with breakage of your build with R600, I presume.

At least merging of the devel-xorg-modular branch should be done before
we start breaking things again by isolated upgrades. I emphasize: If
I'm sounding agitated, it's not so much about you accidentally nearly
breaking things again after I spent many days fixing up the branch to
result in a workable system, but more about the GNU/Linux software
landscape becoming so crazily unstable that it is really hard to avoid
such mishap.

Now, llvm upstream indeed recommends packaging to fix up the mess they
cause. The least we can do is to install several versions in parallel
to keep dependees stable. I didn't even succeed in enabling an RTTI
build together with mesalib-1x, which they also recomment. Free
software went nuts, really cracking nuts.


Alrighty then,

Thomas


PS: Special puzzlement goes out to AMD, where support for a not that
brand new AMD 760G board (a platform nearly 7 years old!) using open
source drivers is still very much bleeding edge. Cheers to Mathieu for
possibly whitnessing me bitching again.

Attachment: signature.asc
Description: PGP signature




Archive powered by MHonArc 2.6.24.

Top of Page