This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Libtool problem (Was: Multilib libjava problem)


On Tuesday 01 October 2002 18:23, Michael Matz wrote:
> Hi,
>
> On Tue, 1 Oct 2002, Bo Thorsen wrote:
> > This is a libtool problem. I have come up with two different patches
> > to try to solve this:
> >
> > +x86_64*linux*)
> > +  # Find out which ABI we are using.
> > +  echo 'int i;' > conftest.$ac_ext
> > +  if AC_TRY_EVAL(ac_compile); then
> > +    case "`/usr/bin/file conftest.o`" in
> > +    *32-bit*)
> > +      LD="${LD-ld} -m elf_i386"
> > +      ;;
> > +    esac
> > +  fi
> > +  rm -rf conftest*
> > +  ;;
> > +
> >  *-*-sco3.2v5*)
> >    # On SCO OpenServer 5, we need -belf to get full-featured binaries.
> >    SAVE_CFLAGS="$CFLAGS"
> >
> > I tried this and just running configure && make bootstrap
> > (srcdir!=builddir)  but that didn't do any change. The libtool in
> > 32/libjava didn't have any line with x86_64 in it??
>
> You need to recreate configure (in this case at least the one in libjava/)
> for those changes to take effect. 

Yes, I figured that out and ran the full configure & bootstrap. And it worked! 
I'm sending a proper patch in a moment.

> The above change looks otherwise
> correct, although if you anyway touch this you can also look into KDE's
> libtool, which also sets $sys_lib_*search_path_spec to "/lib64 ..." or
> "/lib ..." depending on the current default ABI (and existance of /lib64).

I see no reason for this since it's handled automatically on x86-64.

> > Index: libtool.m4
> > ===================================================================
> > RCS file: /cvs/gcc/gcc/libtool.m4,v
> > retrieving revision 1.9
> > diff -u -r1.9 libtool.m4
> > --- libtool.m4	14 Aug 2002 02:39:52 -0000	1.9
> > +++ libtool.m4	1 Oct 2002 14:50:41 -0000
> > @@ -622,7 +622,7 @@
> >  # This must be Linux ELF.
> >  linux-gnu*)
> >    case $host_cpu in
> > -  alpha* | hppa* | i*86 | powerpc* | sparc* | ia64* )
> > +  alpha* | hppa* | i*86 | powerpc* | sparc* | ia64* | x86_64* )
>
> KDE also includes "m68* | mips | mipsel | s390* | sh*".  Some of them
> might not be tested that good, though (although in the context of KDE and
> linux except sh*, mipsel and m68* everything builds).  Somewhen when
> adding the second or third of them I got the feeling the logic of that
> case should be inverted, but I never came around to do this ;-)

I checked the cvs version of libtool, and that looks like the ones KDE uses. 
So GCC will get them when it upgrades to a newer libtool.

Bo.

-- 

     Bo Thorsen                 |   Praestevejen 4
     Senior Software Engineer   |   5290 Marslev
     Klarälvdalens Datakonsult  |   Denmark


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]