This seemed safe, but name table init includes ptrtab initialization, and *that* involves doing ctf_type_to_index() on the type the pointer references -- which can be a forward-reference. So ctf_typemax must have its final valid value at all stages in this loop or we can't get an index back out reliably. Playing with ctf_typemax here doesn't really buy us anything, only stopping us from doing name lookups on types we can't look up names for anyway because we haven't yet put them in the name table hashes.