This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: genconstants failure on darwin 1.4
Stan Shebs wrote:
>
> Andreas Tobler wrote:
> >
> > 2001-10-01 Loren J. Rittle <ljrittle@acm.org>
> >
> > * Makefile.in (STAGE2_FLAGS_TO_PASS): Propagate HOST_CC.
> > (fixinc.sh): Map CC, CFLAGS and LDFLAGS to HOST_* versions for
> > later recursive make invocation.
> >
> > The STAGE2_FLAGS_TO_PASS breaks the genconstants. I'm investigating
> > further, but maybe anyone of you has a quick idea.
>
> A quick scan of the make log shows that cc is compiling both the
> stage 1 and stage 2 gengenrtl.c, because they get compiled with
> HOST_CC listed explicitly, while other files get compiled with CC.
>
> While in theory the Apple 2.95-based compiler and 3.x are binary
> compatible, I'm sure there are enough flags set differently to
> cause chokage. Anyway, it seems pretty clearly wrong to be using
> the host compiler to compile *anything* in stage 2 - perhaps the
> explicit compile commands using HOST_CC need to use CC instead.
What I did is to delete the following lines in gcc/gcc/Makefile.in (2960~):
HOST_CC="$(HOST_CC)" \
HOST_CFLAGS="$(HOST_CFLAGS)" \
HOST_LDFLAGS="$(HOST_LDFLAGS)" \
HOST_CPPFLAGS="$(HOST_CPPFLAGS)" \
Then I was able to bootstrap the fsf mainline. (synced this evening 19 MEST)
Any light is highly appreciated!
Andreas