IntInf asserts (fwd)

[Stephen Weeks]

> Did you mean that the primitive would have type
> 	int * int -> Int.int
> and that it (the C code) would do the allocation?  The problem then is what
> should it do if it runs out of space?

No the semantics is that it would guarantee that that much space was
available before invokinh the primitive.  The idea is to force the
limit check and the primitive call to be "atomic" so that nothing
(like a tuple allocation) can come between them.