http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59923
Jakub Jelinek <jakub at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jakub at gcc dot gnu.org --- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Created attachment 31982 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31982&action=edit gcc49-pr59923.patch This isn't a dup of PR59575, here the problem is that ce3 pass creates cond_exec insn to restore an unwind info important register and adjust stack pointer. dwarf2 EH info can't express conditional restore and expressing conditional CFA adjustment would be quite ugly there. Untested fix attached, can anyone with an ARM box bootstrap/regtest it?