[lfs-support] LFS 7.2 GCC pass 1

Richard Melville richard.melville69 at googlemail.com
Wed Sep 5 02:43:02 PDT 2012


>
> On Tue, Sep 4, 2012 at 4:23 AM, Richard Melville
> <richard.melville69 at googlemail.com> wrote:
> >> I extracted all of these packages from within the GCC-4.7.1 folder
>
> <snip>
> >
> > I'd still be interested to know why we build GMP, MPC, and MPFR inside
> GCC
> > except on the final build where they are built separately.
> >
> > Richard
> >
>
> During pass 1, GCC requires the GMP, MPC and MPFR libraries, but we
> don't want GCC to get these libraries from the host. GCC searches for
> the libraries either via the regular search paths *or* inside it's own
> source tree. We install a temporary copy inside the GCC source tree to
> take advantage of this, and thus allow GCC to not be contaminated with
> host libraries
>
> During pass 2, we are in a protected chroot environment, so we are no
> longer concerned about the host. So other programs in pass 2 can take
> advantage of the GMP, MPC and MPFR libraries later in the build, we
> install them before GCC instead of in the source tree.
>
> --
> -- -
> Steve Crosby
>
>
Thanks for the reply Steve ( and Eleanor earlier).  Picking up on what
Bruce said about the possibility of race conditions relating to building
GCC with MAKEFLAGS set to -j > 1, I'm wondering if there may be a race
condition affecting the GCC build with GMP, MPC, and MPFR building inside
the GCC directory at the same time.  I have no proof for this; it's just a
hypothesis, but I was wondering what others may think.  There is
*definitely* a problem where the GCC build sometimes fails at the same
point each time (checking for MPFR), and then builds OK on a random
attempt.  I'm not aware of the problem ever occurring on the final build of
GCC where GMP, MPC, and MPFR are built outside the GCC directory.

Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfromscratch.org/pipermail/lfs-support/attachments/20120905/69c54f1f/attachment.html>


More information about the lfs-support mailing list