[PATCH] Teach DeadArgElimination not to eliminate return values of functions with 'returned' arguments

[Stephen Lin]

> Uh, what?  While I'm a big fan of the "returned" attribute (in that it can model some cool optimizations) I didn't "ok" charging ahead against the feeling of the community.
>
> -Chris

As Nick mentioned off-list, my e-mail describing my patch to the list
may not have made clear that a new IR attribute was being added and
what the semantics of it was, so I think it it my fault that it
attracted less discussion at the time than it should have before being
committed.

In any case, I am glad we are debating the merits of this approach now.

Stephen