> Also, any idea how we could turn that explicit null-check into an > implicit one? That should happen in PhaseCFG::implicit_null_check(). So something's not going the way it should there. Roland.