Spurious additional lifetime appearing in printout of type involving HRT lifetimes

[steffahn]

“Related” to #102346, and #101280, as this, too, is about diagnostic printing of types with HRT lifetimes (unlike the former related issue, this issue is not a recent regression):

fn g(f: for<'a> fn(fn(&str, &'a str))) -> bool {
    f
}

error[E0308]: mismatched types
 --> src/lib.rs:2:5
  |
1 | fn g(f: for<'a> fn(fn(&str, &'a str))) -> bool {
  |                                           ---- expected `bool` because of return type
2 |     f
  |     ^ expected `bool`, found fn pointer
  |
  = note:    expected type `bool`
          found fn pointer `for<'a> fn(for<'b, 'a> fn(&'b str, &'a str))`

The way this type is displayed includes an additional superflous/confusing/wrong quantification of 'a in the inner for. Correct would be for<'a> fn(for<'b> fn(&'b str, &'a str)), not for<'a> fn(for<'b, 'a> fn(&'b str, &'a str)).

cc @b-naber who is currently working on the above-mentioned “related” issues.

@rustbot label regression-from-stable-to-stable, E-needs-bisection

Regresses between 1.54 and 1.55:

1.54

  = note:    expected type `bool`
          found fn pointer `for<'a> fn(for<'r> fn(&'r str, &'a str))`

1.55

  = note:    expected type `bool`
          found fn pointer `for<'a> fn(for<'r, 'a> fn(&'r str, &'a str))`

[rustbot]

Error: Label needs-bisection can only be set by Rust team members

Please file an issue on GitHub at triagebot if there's a problem with this bot, or reach out on #t-infra on Zulip.

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

@rustbot label -I-prioritize +P-low

[apiraino]

Bisection seems to indicate this group of PRs in the range --start 2021-07-01 --end 2021-12-01.

********************************************************************************
Regression in nightly-2021-07-23
********************************************************************************

  commit[0] 2021-07-21UTC: Auto merge of #87351 - ehuss:rollup-pga85az, r=ehuss
  commit[1] 2021-07-21UTC: Auto merge of #86946 - Aaron1011:eval-proj-obligation, r=nikomatsakis
  commit[2] 2021-07-22UTC: Auto merge of #87246 - rust-lang:placeholder-pretty, r=nikomatsakis
  commit[3] 2021-07-22UTC: Auto merge of #87250 - robojumper:87199-sized-relaxation, r=nikomatsakis
  commit[4] 2021-07-22UTC: Auto merge of #87265 - Aaron1011:hir-wf-fn, r=estebank
  commit[5] 2021-07-22UTC: Auto merge of #86619 - rylev:incr-hashing-profiling, r=wesleywiser
  commit[6] 2021-07-22UTC: Auto merge of #87366 - GuillaumeGomez:rollup-7muueab, r=GuillaumeGomez
  commit[7] 2021-07-22UTC: Auto merge of #87371 - ehuss:update-cargo, r=ehuss
  commit[8] 2021-07-22UTC: Auto merge of #86212 - pnkfelix:mainline-targetted-revert-81473-warn-write-only-fields, r=simulacrum

but due to the age I could not pinpoint a specific nightly.

Also, as pointed out in the previous comment at some point the diagnostic changed and decided to use another wording (for<'a> fn(for<'b, 'a> fn(&'b str, &'a str))) which I wonder if it's related to one of the mentioned other 2 issues

@rustbot label -E-needs-bisection

[steffahn]

Looking through this list, #87246 seems likely.