This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Minimum GCC version for building glibc
- From: Joseph Myers <joseph at codesourcery dot com>
- To: Mike Frysinger <vapier at gentoo dot org>
- Cc: <libc-alpha at sourceware dot org>
- Date: Tue, 4 Nov 2014 17:38:30 +0000
- Subject: Re: Minimum GCC version for building glibc
- Authentication-results: sourceware.org; auth=none
- References: <Pine dot LNX dot 4 dot 64 dot 1410311352200 dot 4263 at digraph dot polyomino dot org dot uk> <20141104165815 dot GK5402 at vapier dot wh0rd dot info>
On Tue, 4 Nov 2014, Mike Frysinger wrote:
> > * Should 4.6 or 4.7 be the minimum?
>
> i tend to prefer to maintain older compatibility when possible. gcc-4.7 was
> released in Mar 2012. when we drop support for older versions, i think a
> prerequisite should be: what does dropping the older version gain us ? when i
> look at the body of code using __GNUC_PREREQ, i see very few hits for <4.7.
>
> you mentioned that we might want to use __atomic_* for some arches. so if we
> aren't requiring 4.7, what's the diff between 4.4 and 4.6 ?
If nothing else, 4.6 provides the _Static_assert keyword (and I think
there's been some desire to use static assertions). I'd definitely like
to get glibc using -Werror for default builds, and not having to allow for
lots of different GCC versions helps reduce the number of cases where you
need version-specific warning workarounds / use of -Wno-error= /
equivalent pragmas. It also seems quite plausible some vector libm
functions or associated tests could require GCC versions more recent than
4.4 (e.g. -mavx2 was added in 4.7, but maybe some were added in 4.6 - if
you require a relevant version then that saves on configure tests).
--
Joseph S. Myers
joseph@codesourcery.com