Patch to fix powerpc-ibm-aix4.3.3.0 bootstrap

[David Edelsohn]

	I don't have any problem applying John's patch, but the patch does
seem to paper over the problem of GCC hashing based on an address.  This
is at least the second time that someone has uncovered non-deterministic
behavior in GCC because of hashing an address.  Is there any chance that
we can fix the actual problem of the component being hashed?

Thanks, David