https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107097
--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Perhaps try to look up the implicit conversion using the semantic type (i.e. with EXCESS_PRECISION_EXPR not stripped) and then if it is a standard conversion (which exact?) from EXCESS_PRECISION to arithmetic/enumeral type or so, strip away the excess precision and actually convert from the excess precision?
