https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100291
--- Comment #8 from Patrick Palka <ppalka at gcc dot gnu.org> --- Maybe not exactly a dup, since this is an ICE-on-invalid testcase and strictly speaking a regression. But it doesn't seem like a regression worth fixing in release branches. And the underlying cause is the same as in PR106086 so might as well call it a dup.