http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49680
--- Comment #7 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-07-08 22:08:31 UTC --- (In reply to comment #6) > I see the problem wrt cris -- we really shouldn't delay the output > of the CFI note that late. I'll fix that and see if I can still > reproduce the irix problem. Thanks. Still, I don't think there's anything wildly wrong with the quoted rtx so I'm going to do s/PREV_INSN/prev_nonnote_nondebug_insn/g and assert that the right insns and code_label is seen in cris_asm_output_case_end anyway. The current PREV_INSNs are a little too presumptious (QED); there are no documented promises broken at the SEGV.