GNU bug report logs - #72451
[PATCH core-updates] gnu: gsl: Fix build on i686.

Previous Next

Full log

Message #11 received at 72451 <at> debbugs.gnu.org (full text, mbox):

From: Ludovic Courtès <ludo <at> gnu.org>
To: Kaelyn <kaelyn.alexi <at> protonmail.com>
Cc: "72451 <at> debbugs.gnu.org" <72451 <at> debbugs.gnu.org>
Subject: Re: fixing the gsl test failures on i686
Date: Tue, 20 Aug 2024 14:42:56 +0200

Hi Kaelyn,

Kaelyn <kaelyn.alexi <at> protonmail.com> skribis:

> I noticed this morning that you recently pushed commit fa8dbbe59d to fix the i686-linux gsl test failures. I'd submitted a similar fix about two weeks ago (along with a couple of other i686-linux fixes that affect building wine64), and I wanted to mention that I don't think your commit does the right thing with CFLAGS. When I figured out the error, I noticed a side-effect of passing CFLAGS=-fPIC as the #:make-flags is that it disables the optimized build (or at least relies on whatever optimizations the compiler defaults to, which AFAIK is little to none). While your patch also cleans up the definition a bit by no longer disabling some tests, I think the correct approach is to include the package's default CFLAGS as part of the #:make-flags as I did in issue 72451 so that optimizations aren't unintentionally disabled.

Oh, apologies for not noticing <https://issues.guix.gnu.org/72451>.

What you did looks correct to me.  I added a FIXME about ‘CFLAGS=-fPIC’
stating roughly what you explained here.

The reason I did not change CFLAGS on other architectures was to avoid a
massive rebuild at this late stage of the branch (‘gsl’ has 2.9K
dependents).

But I agree: we should clean that up for all platforms, and check which
of the tests really need to be skipped on 32-bit platforms once compiled
with optimizations.  We can create a branch with such a patch once
‘core-updates’ is merged.

WDYT?

Thanks,
Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.