https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95980
--- Comment #7 from anlauf at gcc dot gnu.org --- (In reply to Dominique d'Humieres from comment #5) > AFAICT the patch fixes the ICE for z2.f90, but not for z1.f90. You're right. I got lost in trying to work on too many PRs. I attached a brute-force patch that makes gfortran reject z1.f90 without ICEing. Can you confirm this, or is there possibly something left to handle? I have the impression that there's a lot of bad things happening with invalid input that is not always caught on x86_64.