regrename bug fix

[David Edelsohn]

>>>>> Joern Rennecke writes:

Joern> Unfortunately, I don't have an AIX system handy here to check this.
Joern> Does my regrename patch from today fix the problem?
Joern> The old code could sometimes calculate the right result for the
Joern> wrong reason - i.e. it put the highpart of a value in the list
Joern> of things that are equivalent to the value itself, and then it
Joern> ignored the offset associated with a mode change.
	
	No, your regrename.c:copy_value() patch does not fix the problem.

	Would you please revert your find_oldest_value_reg patch while we
track down this problem?  It is going to be difficult because the patch is
causing the stage2 compiler itself to be miscompiled.

Thanks, David