Ada bootstrap failure due to new DCE bug

[Richard Kenner]

    PS: The test case doesn't fail in C because the C front end decomposes
    &arr[n] to pointer arithmetic, and that formation somehow does the
    right thing in tree-ssa-alias.c.  Not sure what the difference is,
    exactly.

I'm not sure it has anything to do with decomposition vs. not decomposing
the reference.  I tried to make a smaller test case for this in Ada and
failed because a memory tag wasn't used.  The difficulty in reproducing
it in C may have been party due to that.

    I've had a patch pending for ages to get C and C++ to not decompose
    array_refs so early, but it doesn't work 100%.  This may be the reason.

Indeed.

Thanks for looking into this so quickly and sorry that I was off email
yesterday.

BTW, I found a major bug on Tuesday in the way elaboration was done with in
the Ada front-end C code (gigi) that fixed a bunch of ACATS tests, so we're
down to about 37 failures, some of which may be specific to x86-64.

I think there's still a loop problem pending, since I have it loop
optimization off in my tree, but we're getting very close to being able to
consider Ada "working" again.