Region#subst ICE: trait A<'self, T> { fn f() -> &'self T;}

[nielsle]

The following program causes a compiler error

#[crate_type = "lib"];
trait A<'self, T> { fn f() -> &'self T;}

$ rustc --version
/home/niels/local/src/rust/x86_64-unknown-linux-gnu/stage2/bin/rustc 0.6 (28efc23 2013-03-22 23:09:15 -0700)

$ rustc tests.rs
error: internal compiler error: ty::subst: Reference to self region when given substs with no self region, ty = &'self 'a

[graydon]

reproduced today. seems relatively minor, though real: should be caught earlier.

[emberian]

reproduces, error: internal compiler error: ty::Region#subst(): Reference to self region when given substs with no self region: substs(self_r=None, self_ty=Some(U), tps=[T])

[bblum]

Doesn't need the crate type lib. Definitely should just be an error.

[sruggier]

I just hit this yesterday. Using commit 433fbe8 (August 13th), the following program:

#[crate_type = "lib"];

trait A<'self, T> {
    fn f(&mut self) -> &'self mut T;
    fn p() -> T;
}

triggers the same ICE:

error: internal compiler error: ty::Region#subst(): Reference to self region when given substs with no self region: substs(regions=[], self_ty=Some(BUG[1]), tps=[T])

However, if either of those two functions is removed, the remaining code compiles successfully. I realize this is a bit of a different problem than the original report. I can open a new issue, if that's preferred.

[flaper87]

'self is no longer a special lifetime. This can't be reproduced using a normal lifetime:

#[crate_type = "lib"];

trait A<'a, T> {
    fn f(&mut self) -> &'a mut T;
    fn p() -> T;
}

This probably needs a test. Flagging.