--- Comment #9 from margali at imapmail dot org 2010-09-26 14:04:21 UTC ---
So in the end I did the following and it finally compiled:
- uninstall binutils
- apply cumulative fink patch to source directory
- hack configure to include the --with-dwarf2 flag (also did this to but didn't want to install another version of autoconf etc. just
to regenerate configure)
- configure with --disable-libjava-multilib --program-suffix=-4.5

I did not run the testsuite as I don't have the right software installed and
would not know how to interpret the results in any case.

Installed with --prefix=/usr/local/stow/gcc-4.5.1 and then stowed in the usual

gcc-4.5 -v:

Using built-in specs.
Target: powerpc-apple-darwin8.11.0
Configured with: ../gcc-4.5.1/configure --disable-libjava-multilib
Thread model: posix
gcc version 4.5.1 (GCC) 

I'm still having problems but I think these are due to my ignorance rather than
anything else.

Note that this method not only involves steps not listed in the host-specific
installation notes but involves steps which seem to directly contradict the
instructions given there.

For example, I took it that two sections of the notes were relevant to me: the
notes for powerpc-*-* and powerpc-*-darwin*. The former suggests that binutils
is required but in fact installing binutils actually prevents the build from
succeeding. The latter suggested to me that I probably needed to get cctools
but in fact this wasn't the case. (I admit the notes here merely suggested this
to me and I likely misinterpreted them but perhaps it could be a little

In addition, the notes say nothing about the cumulative fink patch or using
--with-dwarf2 or the need to disable multilib support for libjava. But these
are obviously not fink-specific requirements since I'm not using fink. Rather,
they are host-specific and it would be helpful if they were listed with the
host-specific installation notes.

Perhaps the host-specific notes for darwin could be updated and a section
specific to darwin8.11.0/os x 10.4.11 could be added outlining the stages
necessary to compile? That would have been a huge help to me and would have
saved me a great deal of time spent on failed builds. (My g4 takes a *long*
time to build gcc!)

This is not a criticism of the gcc developers since it is not, I know, your
responsibility. But it seems a shame that issues such as the need to disable
multilib for libjava are not flagged up even though they are obviously well
recognised by those packaging gcc for fink. There never seems to be much
feedback to upstream from those working on fink/macports etc. packages even
when issues are clearly platform-specific rather than relevant only to the
particular packaging system. This is obviously not confined to gcc but it does
seem to me a shame that known fixes don't benefit the upstream sources more

Reply via email to