regression: parameter type may not live long enough

[Mark-Simulacrum]

See https://crater-reports.s3.amazonaws.com/beta-1.62-1/beta-2022-05-20/gh/FSMaxB.rust-genealogy/log.txt:

[INFO] [stdout] error[E0311]: the parameter type `Extracted` may not live long enough
[INFO] [stdout]   --> genealogy-java-apis/src/comparator.rs:24:13
[INFO] [stdout]    |
[INFO] [stdout] 24 |             compare: Box::new(move |left, right| match (self.compare)(left, right) {
[INFO] [stdout]    |                      ^^^^^^^^ ...so that the type `Extracted` will meet its required lifetime bounds...
[INFO] [stdout]    |
[INFO] [stdout] note: ...that is required by this bound
[INFO] [stdout]   --> genealogy-java-apis/src/comparator.rs:21:28
[INFO] [stdout]    |
[INFO] [stdout] 21 |         for<'a> Extracted: Ord + 'a,
[INFO] [stdout]    |                                  ^^
[INFO] [stdout] help: consider adding an explicit lifetime bound...
[INFO] [stdout]    |
[INFO] [stdout] 21 |         for<'a> Extracted: Ord + 'a + 'a,
[INFO] [stdout]    |                                     ++++
[INFO] [stdout] 
[INFO] [stdout] 
[INFO] [stdout] error: aborting due to previous error

This is a 1.62 regression, it seems, based on Crater. At the very least, it's a diagnostics issue that we're suggesting adding an already existing lifetime bound.

[compiler-errors]

Bisected

searched nightlies: from nightly-2022-04-01 to nightly-2022-05-31
regressed nightly: nightly-2022-05-14
searched commit range: a5ad0d2...f1f721e
regressed commit: a7d6408

bisected with cargo-bisect-rustc v0.6.3

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

cargo bisect-rustc --start=2022-04-01 --end=2022-05-31 

#96899, cc @oli-obk

Minimized

https://play.rust-lang.org/?version=beta&mode=debug&edition=2021&gist=8bcb86adc0a98d57aecbe3e2b80229c0

fn test<T, F, U>(f: F) -> Box<dyn Fn(T) -> U + 'static>
where
    F: 'static + Fn(T) -> U,
    for<'a> U: 'a, // < This is the problematic line -- remove it, and it passes.
{
    Box::new(move |t| f(t))
}

[compiler-errors]

The issue is that for<'a> U: 'a bound is causing the new wf check introduced in #96899 to fail.

Regarding the diagnostics being weird, the issue is probably because we don't handle the case where a higher-ranked lifetime doesn't outlive itself. We just see a borrowck error, see that it's due to the binder's 'a lifetime, and suggest adding it. Not sure if that needs fixed, personally think it's not worth it.

[apiraino]

Assigning priority as discussed in the Zulip thread of the Prioritization Working Group.

@rustbot label -I-prioritize +P-medium

[oli-obk]

This got fixed on nightly by enabling nll by default

[Rageking8]

@rustbot claim
For the test