Oops, I take it (somewhat) back. As you suggest, the testcase is valid because the cast can be considered to be a reinterpret_cast. However, a reinterpret_cast should do no adjustment, so please update your patch accordingly. Thanks, Jason