RSSA

[Matthew Fluet]

> > If you do, bear in mind that that will be a new loop where a signal
> > limit check will be inserted.
> 
> No.  This is one loop that can't have a signal limit check -- a
> pointer array must be completely filled in or the gc will not be able
> to trace it.

Well, I thought that would be the case.  The point then is that we need to
somehow distinguish this loop from other loops and suppress limit check
insertion.  (E.g., in the x86-codegen, this just looks like any other
loop; I don't do anything special with it.)