This happens on amd64 with profile 2005.0 1. emerge gcc 2. Compile a little test program with -lgcc_s 3. Get libgcc_s.so.1 not found error message After emerging gcc, libgcc_s.so.1 (or a symlink) should be put here: /usr/lib/gcc/x86_64-pc-linux-gnu/3.4.3 But libgcc_s.so.1 is put to /usr/lib/gcc/x86_64-pc-linux-gnu/lib64 By placing a symlink and regenerating /etc/ld.so.cache one can get around this.
This bug is present in 3.4.2 as well and prevents 'groff' and a number of things from building. groff is apart of the base system. USE='-multilib' here.
I have no doubt that 3.4.2 is broken... that ebuild is ugly and doesn't work well with multilib, but the bug with recent versions (3.4.3-r1 and 20050110) on 2005.0 should be fixed. Can you please reemerge and test.
Sadly last we attempted 3.4.3 is unable to rebuild itself on hardened systems. So we are limited to a masking of <3.4.3 for the moment. I will try none the less and see if I can build gcc-3.4.3 twice.
Ok, solar said they hardenend doesn't need 3.4.2 any more... but if you change your mind, I have the bump to fix it in my overlay and can be committed any time if you decide you need it. As for these problems in general, toolchain.eclass has been greatly updated to move the libs to the proper dirs, so I'm closing this bug.