Your Jan 8th, 2001 rtlanal.c:note_stores change

[Richard Kenner]

    Now, Richard, in your gcc-patches posting when this change went in you
    say "as discussed all over the lists".  From such a description, it
    ought to be easy to find many postings and some involved thread
    discussing the topic.  I spent all night looking for such postings but
    to no avail.

Unfortunately, I can't remember what I was referring to, but it might well
have been discussions that occurred many months before.

    The fact that no record can be found by anyone so far of any
    discussion about your sparc64 problems proportedly fixed by these
    changes is very troubling.  I'm actually quite upset about this.

The changes were motivated by a *specific* problem I ran into, I'm
sure.  The reference to the older discussion was just to tie things together.
   
    If you can't come up with a test case that shows the problem you were
    having on sparc64 or at least show us what inside the compiler fails
    without the CLOBBER note_stores stuff, I'm going to recommend that the
    CLOBBER bits be reverted because they are wrong and I can show a
    testcase (gdb/infcmd.i with -mcpu=ultrasparc) that fails due to the
    bogus CLOBBERs.

I agree the CLOBBERs are wrong, but so is a SET.  It will cause problems to
show each of those registers as being set to *the same thing*, whatever that
thing is.