https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67260
--- Comment #21 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Oleg Endo from comment #17) > > > I guess because the recently added "sibcall_value_pcrel_fdpic" pattern also > uses "=k" as the clobber constraint, it might run into the same problem. To > play safe for now, I'd like to commit the following to trunk and backport it > to 5 and possibly 4.9: > > [...] It seems the patch is not good. I see new failures when testing on sh-elf. I will try to see what's going on there.
