https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109231
--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> --- (In reply to r...@cebitec.uni-bielefeld.de from comment #9) > > --- Comment #8 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot > > Uni-Bielefeld.DE> --- > >> --- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> --- > >> No luck reproducing this using a cross. > > > > Ok, so I'll continue with the reghunt. > > It's finished now and identified your patch > > commit 24c06560a7fa39049911eeb8777325d112e0deb9 > Author: Jakub Jelinek <ja...@redhat.com> > Date: Fri Mar 17 18:59:56 2023 +0100 > > tree-inline: Fix up multiversioning with vector arguments [PR105554] > > as the culprit. That is just weird. Looking at the cross case, I see tree_function_versioning being called 5 times, but in all cases DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_SPECIFIC_OPTIMIZATION of both old_decl and new_decl are NULL, 4 times from that from cgraph_node::create_version_clone_with_body with NULL target_attributes. So I don't much see how it could make a difference. So, do you really see changes with that commit, if so, are they in the -fno-checking or -fchecking=1 case (r13-6738 vs. r13-6739), do any changes appear in -fdump-tree-optimized, or if not, where things differ first with -fdump-noaddr -fdump-rtl-all?
