https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93393
--- Comment #5 from Segher Boessenkool <segher at gcc dot gnu.org> --- It is a backend issue that we cannot solve properly without solving much bigger generic problems first. See the BoF at last year’s cauldron. Without fixing that first, doing signaling floats regresses code quality for us (as it also does on other targets, afaics). So yeah just xfail the new test for us, just like the existing tests with the same issue?
