returning an unconstrained, bound region should be an error

[nikomatsakis]

If you have a function like this:

fn foo() -> &a.int

we should report an error. The only thing you could legally do in that case is fail or return a static item, so you miight as well write:

fn foo() -> &static.int

This sounds like a good lint warning/error.

[bblum]

My vote is warning rather than error. If you're right that it's always the wrong thing to do, then you shouldn't be able to write non-unsafe code that does this that doesn't error somewhere else.

And maybe you want to try doing it just while hacking / experimenting with the language.

[catamorphism]

Reproduced with d2ad028

fn foo() -> &a/int {
    return &x;
}
const x: int = 5;
fn main() {}

This compiles successfully.

[nikomatsakis]

Not critical for 0.6; removing milestone

[catamorphism]

Reproduced with 64963d6

fn foo<'a>() -> &'a int {  //~ ERROR unconstrained region
    return &x;
}
static x: int = 5;
fn main() {}

Still compiles successfully.

[catamorphism]

Nominating for milestone 5, production-ready

[graydon]

just a bug, removing milestone/nomination.

[pnkfelix]

Visiting for triage email 2013-07-22

reproduced with 8a7b636

[bstrie]

As of lifetime elision (#15767) we now indeed require a 'static bound in that position. Closing.