Anonymous regions only sometimes getting related

[alexcrichton]

This code has some interesting results

fn foo(_: &[&str]) {}

fn bad(a: &str, b: &str) {
    foo([a, b]);
}

fn good(a: &str, b: &str) {
    foo([a.as_slice(), b.as_slice()]);
}

fn main() {}

foo.rs:4:12: 4:13 error: mismatched types: expected `&str` but found `&str` (lifetime mismatch)
foo.rs:4     foo([a, b]);
                     ^
foo.rs:3:25: 5:1 note: the anonymous lifetime #2 defined on the block at 3:25...
foo.rs:3 fn bad(a: &str, b: &str) {
foo.rs:4     foo([a, b]);
foo.rs:5 }
foo.rs:3:25: 5:1 note: ...does not necessarily outlive the anonymous lifetime #1 defined on the block at 3:25
foo.rs:3 fn bad(a: &str, b: &str) {
foo.rs:4     foo([a, b]);
foo.rs:5 }
error: aborting due to previous error

I find it odd that if good compiles then why bad doesn't compile...

cc @nikomatsakis

[nikomatsakis]

That is somewhat odd, needs more investigation.

[aturon]

Triage visit. This example is now broken for other reasons. I've tried a few variations, and cannot reproduce the original issue.

[alexcrichton]

This compiles today:

fn foo(_: &[&str]) {}

fn bad(a: &str, b: &str) {
    foo(&[a, b]);
}

fn good(a: &str, b: &str) {
    foo(&[a.as_slice(), b.as_slice()]);
}

fn main() {}

Flagging as needstest