The problem is that it's certainly also not the case that *each* of the registers in question is being set to the result of the CALL. I don't remember the exact bug, but that was basically what was going wrong. I agree that CLOBBER isn't precisely correct *either*.