https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106157
--- Comment #3 from Aldy Hernandez <aldyh at gcc dot gnu.org> --- (In reply to Richard Biener from comment #2) > the reduction probably ended up removing the initialization as that's not > needed to reproduce the ICE Ah. I'm seeing a whole slew of uses before initialization in the IL before DOM gets going. Dereferences of uninitialized pointers, returnless functions, etc etc. This is causing all sorts of things in DOM to be assumed to be equal / etc, and ranger is making other assumptions on top of that. Could I trouble the reporter in submitting a reduced testcase with sane IL? If this is impossible, then the closest thing you can provide.
