[blfs-support] gcc-4.5.1 (BLFS) vs. gcc-4.7.0 (LFS)

Ken Moffat zarniwhoop at ntlworld.com
Sun May 6 18:34:43 PDT 2012

On Mon, May 07, 2012 at 01:45:50AM +0100, Andrew Benton wrote:
> On Mon, 07 May 2012 00:49:37 +0100
> alex lupu <alupu01 at gmail.com> wrote:
> > My latest build (make) of the Chrome browser failed with
> > apparently a gcc (actually, g++) error of some sort:
> > 
> > "...
> > Please submit a full bug report, ...
> > See <http://gcc.gnu.org/bugs.html> for instructions. ..."
> > 
 Internal Compiler Error (ICE) ?
> > I've had gcc-4.5.1 (in good standing) for quite a while.
> > Rather than going through the big effort, uncertainty and
> > probably the futility of submitting a bug report,
> > I'm thinking about just upgrading to gcc-4.7.0 (the latest
> > as of this writing), thereby killing as many as two birds
> > (in the ideal outcome).
> > 
> > 1.  Mostly theoretical.  Why this dissimilarity between
> > LFS and BLFS versions?
> I think the version of GCC in BLFS is neglected because none of the
> current editors use any of those compilers. I just need a C and C++
> compiler so the GCC installed in LFS is fine for me. 
> If I were you I'd install GCC-4.7 with the configure switches from
> chapter 6 of current LFS. Unless you need fortran or whatever, in which
> case, do you want to volunteer to update the BLFS page ;)
> Andy

 I agree with what Andy says, but there are two things to be aware
of when moving to a newer version of gcc - and 4.5.1 now counts as
"very old" in LFS terms!) :

1. With every new version, the C++ standards are tightened up.  At
the moment I'm still trying to work out the minimal dependencies for
cinelerra (dunno if I'll find it useful, or not) and going round in
circles trying to track a gcc-4.7 problem  : it's actually in
ilmbase, not openexr which is where I was hitting it - seems fixed
now, with a sed to include <cstring>.

 If all the packages you build are in the BLFS book, these issues
should mostly have been solved.  For everything else, you may have
to google and run up some blind alleys.

2. Newer versions of gcc seem to stress the hardware more.  On my
own new AMD machine, gcc-4.6 was mostly ok but some days gcc-4.7
runs into a spate of ICEs on packages I have previously built
successfully.  On bad days, it is eventually unable to continue
(every attempt at make, even with -j1, fails) until I drop the
caches!  This is my own fault for buying a cheap consumer-grade mobo
(and it's a lot less of a problem since I slowed down the memory in
the bios).  Most people will never hit these problems, but it's
always a good idea to know what *can* happen :)

das eine Mal als Tragödie, das andere Mal als Farce

More information about the blfs-support mailing list