Re: fix kennerism

[kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)]

    Or if "make check" was run, the results weren't examined.

I saw failures, but they looked pre-existing to me.  I really have to figure
out why I have failures that other systems don't seem to have: it makes it
much harder to spot the real ones.  But I thought we don't use integrate.c
for C anymore and instead use the tree inliner?  That belief was part of
the reason I thought the errors were unrelated.

    The issue is that if the old and new address are the same
    replace_equiv_address_nv will return the original MEM.

Oops.  Right.

Thanks.