Re: Checking GCC version (Re: [Libreoffice-commits] core.git: configure.ac)

On Fri, Aug 09, 2013 at 01:18:30PM +0200, Lubos Lunak wrote:
> On Friday 09 of August 2013, Lionel Elie Mamane wrote:
> > New commits:
> > commit 91ec774c9fff46af6800e75315561e86167fe5d1
> > Author: Lionel Elie Mamane <lionel@mamane.lu>
> > Date:   Fri Aug 9 10:28:51 2013 +0200

> >     $CC --version is too unstructured

> >     Give up and revert to "$CC -dumpversion".
> >     Since by now no test now refers to patchlevel, make it unavailable in
> > GCCVER so that nobody tries to use it.

>  Seeing all these commits about checking GCC version, does it
> actually make sense to do such checks?

It seems that most usages of such checks were eradicated, latest by

 commit 1e0feb5cf916fada5dc2db66a358649624ece578
 Author: Luboš Luňák <l.lunak@suse.cz>
 Date:   Fri Aug 2 14:18:22 2013 +0200

    do not base feature checks on gcc version

I had the first patch of these series in my local tree "for a long
time" (before your commit above), because I-don't-remember-what-bad
happened because my gcc was detected as being 4.7.0 instead of
4.7.2. Today I fatefully decided to push it...

Concretely, we have only *two* places where that information is used:

 - check that version is at least 4.0.0 (line 2936)

 - test for thread-safe statics

I guess we can drop the test for >= 4.0.0 altogether. So, if "someone"
rewrites the thread-safe statics test to not use GCCVER, we can bin
this whole "GCC version" definition.

-- 
Lionel