Borrow checker error does not figure out source of 'static requirement through closure capture

[Noratrieb]

Code

fn main() {
    let targets = String::new();
    let queue = targets.as_str();

    (move || require_static(queue))();
}

fn require_static(_: &'static str) {}

Current output

error[E0597]: `targets` does not live long enough
 --> src/main.rs:3:17
  |
2 |     let targets = String::new();
  |         ------- binding `targets` declared here
3 |     let queue = targets.as_str();
  |                 ^^^^^^^---------
  |                 |
  |                 borrowed value does not live long enough
  |                 argument requires that `targets` is borrowed for `'static`
...
6 | }
  | - `targets` dropped here while still borrowed

For more information about this error, try `rustc --explain E0597`.
error: could not compile `playground` (bin "playground") due to 1 previous error

Desired output

error[E0597]: `targets` does not live long enough
 --> src/main.rs:3:17
  |
2 |     let targets = String::new();
  |         ------- binding `targets` declared here
3 |     let queue = targets.as_str();
  |                 ^^^^^^^ borrowed value does not live long enough
4 |
5 |     (move || require_static(queue))();
  |              --------------------- argument requires that `targets` is borrowed for `'static`
6 | }
  | - `targets` dropped here while still borrowed

For more information about this error, try `rustc --explain E0597`.
error: could not compile `playground` (bin "playground") due to 1 previous error

Rationale and extra context

No response

Other cases

This works on stable, but regressed on nightly.

Rust Version

rustc 1.80.0-nightly (804421dff 2024-06-07)

Anything else?

No response

[Noratrieb]

Regression in rust-lang-ci@4ffe0cc
The PR introducing the regression in this rollup is #122820: Stop using <DefId as Ord> in various diagnostic situations

searched nightlies: from nightly-2024-03-21 to nightly-2024-03-26
regressed nightly: nightly-2024-03-23
searched commit range: 0ad927c...85e449a
regressed commit: 1447f9d

bisected with cargo-bisect-rustc v0.6.8

Host triple: x86_64-unknown-linux-gnu
Reproduce with:

cargo bisect-rustc --access github --start 2024-03-21 --end 2024-03-26 --prompt --preserve -- check 

@estebank @oli-obk I see the diagnostics changes caused test changes and you were aware of those, but it's really unfortunate :(.. i was really confused by this, would be great if it could be restored somehow

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

Thanks for bisecting! cc @oli-obk for #122820

@rustbot label -I-prioritize +P-medium

[lqd]

This will be stable later today.

[oli-obk]

Yea, this is unfortunate, but previously was relying on essentially random information to emit the better diagnostic. We could try to get a similar enough mostly-random value for deciding the order of things and thus getting the diagnostic back, but I didn't do that in the hope of actually finding a reasonable heuristic