http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54792
--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-10-03 09:59:05 UTC --- Created attachment 28340 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28340 gcc48-pr54792.patch Untested fix. The problem appears to be that find_modifiable_mems wasn't scanning the TAIL insn, without -g on the testcase there was a load from memory at the end of a basic block and it wasn't scanned, with -g there were some DEBUG_INSNs after it, thus tail was a DEBUG_INSN and the load was scanned and triggered the optimization.
