Jeremy Huntwork wrote: > Thanks. This was just an oversight. I meant to include the contents of the > pure64 patch for gcc pass2, but overlooked it due to the similarly named > pure64_specs patch. Here's what the pure64 patch would give us (Approximately. > The following is a faux diff due to line wrapping issues on my news client): > > gcc-4.1.2.orig/gcc/config/i386/t-linux64 > gcc-4.1.2/gcc/config/i386/t-linux64 > > MULTILIB_OPTIONS = m64/m32 > MULTILIB_DIRNAMES = 64 32 > -MULTILIB_OSDIRNAMES = ../lib64 ../lib > +MULTILIB_OSDIRNAMES = ../lib ../lib32 > > Since the specs patch is already set up for pure64 and modifies the 64-bit > linker to be in /tools/lib, I'm thinking to merge these small changes above to > the pure64_specs patch and call it done. Any objections to that?
Umm, you appear to have missed the point completely. Please re-read the info I pointed to. MULTILIB_OSDIRNAMES needs to be *non-existent* to work around the surprising (buggy?) GCC behavior I'm talking about. Then again, I haven't tested x86_64 in a while so I might be talking out of my arse. Preferably on an Ubuntu64 host, please post the verbose output of gcc-pass2 so we can what is going on ie: echo 'main(){}' | gcc -xc -o /dev/null -v - Another point - To be honest, I don't see why any "pure64" patch is needed at all. That's what the symlinks are for. All the patch seems to do is diverge the `x86_64 --disable-multilib' build further away from x86 than is necessary. Regards Greg -- http://www.diy-linux.org/ -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page