Hrvoje Niksic writes: This isn't a good reason?! If configure can't build a binary that will work reliably _to the best of our knowledge_ then it should stop. If we want to offer the user a --use-broken-compiler option then that's fine--- how else will the problems get debugged. But using a known broken compiler should not be a default action. Then that is what we should be trying to discover here in the beta group. I don't think we should subject the users to sickening crashes just because we don't yet know exactly why egcs 1.1 is killing us. configure spews a long series of messages. A warning will be lost in all that. I build XEmacs like this ./configure; make and then go do something else while it grinds. If we expect people to pay attention to the warnings, either configure has to generate much less output or configure has to abort, so that the user is forced to read the message and take some action.