DCE eliminating valid statement for ACATS c34007p

[Richard Kenner]

    > Yeah, that's what I realized in working on this problem.  Is it worth
    > adding a flag to show that this is a gimple-generated variable so we
    > can have an assert that we're not trying to mark it as addressable?

    What has gimple-generated got to do with it?

Well, if it's gimple-generated, then we know it was assumed to not be
addressable, but I guess you're right: setting *any* variable that would
potentially have already been used in gimple as an operand to be addressable
is a potential bug.  I'm wondering why verify_ssa doesn't detect this.  After
all, it means we have an operand that's not a gimple reg.