Self-contradictory error message about Fn types on beta/nightly

[jeremysalwen]

https://play.rust-lang.org/?version=nightly&mode=debug&edition=2018&gist=e973951415f5092af183ea13d11177cf

struct Foo<T, F: Fn(&T, &T) -> T> {
    value: T,
    func: F
}

fn main() {
    let lambda = |&x, &y| x + y;
    let foo = Foo {
        value: 5 as i32,
        func: lambda
    };
}

   Compiling playground v0.0.1 (/playground)
error[E0308]: mismatched types
 --> src/main.rs:8:15
  |
8 |     let foo = Foo {
  |               ^^^ one type is more general than the other
  |
  = note: expected type `std::ops::FnOnce<(&i32, &i32)>`
             found type `std::ops::FnOnce<(&i32, &i32)>`

error: aborting due to previous error

For more information about this error, try `rustc --explain E0308`.
error: Could not compile `playground`.

To learn more, run the command again with --verbose.

See also https://stackoverflow.com/questions/54341465/rust-expected-type-error-prints-mismatched-types-that-are-exactly-the-same?noredirect=1

[estebank]

This is a diagnostics regression given the current stable output (even though it wasn't that good to begin with):

error[E0631]: type mismatch in closure arguments
 --> src/main.rs:8:15
  |
7 |     let lambda = |&x, &y| x + y;
  |                  -------------- found signature of `fn(&_, &_) -> _`
8 |     let foo = Foo {
  |               ^^^ expected signature of `for<'r, 's> fn(&'r i32, &'s i32) -> _`
  |
note: required by `Foo`
 --> src/main.rs:1:1
  |
1 | struct Foo<T, F: Fn(&T, &T) -> T> {
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

error[E0271]: type mismatch resolving `for<'r, 's> <[closure@src/main.rs:7:18: 7:32] as std::ops::FnOnce<(&'r i32, &'s i32)>>::Output == i32`
 --> src/main.rs:8:15
  |
8 |     let foo = Foo {
  |               ^^^ expected bound lifetime parameter, found concrete lifetime
  |
note: required by `Foo`
 --> src/main.rs:1:1
  |
1 | struct Foo<T, F: Fn(&T, &T) -> T> {
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

but needs consideration on wether the code as written should be accepted instead. Why are we evaluating to FnOnce in the first place? Note that changing F to FnMut doesn't change the types reported.

CC @rust-lang/lang

[ExpHP]

Why are we evaluating to FnOnce in the first place?

Is it not just showing FnOnce because the associated type is defined on FnOnce? This isn't related to the fact that it does not compile.

[ExpHP]

The workaround is to annotate the closure parameter types so that they get assigned higher-ranked regions or whatever you call them.

let lambda = |&x: &_, &y: &_| x + y;

This is a well-known problem that has hit many a rustacean. But I thought the error messages used to be more descriptive, saying mething about "concrete" versus "bound" lifetimes. Even if it was terribly confusing language, it was at least something the user could Google.

[lqd]

This issue is very similar to #57362 on which I have been working (same RegionResolutionError: a SubSupConflict with two placeholder regions and Subtype region origins — but in this case the expected/found values are PolyTraitRefs instead of TraitRefs, bypassing the nice_region_errors into this "mismatched types" fallback which doesn't show the self tys) and I hope to look into this one next.

[kadiwa4]

Current output (rustc 2023-02-25):

error[E0308]: mismatched types
  --> src/main.rs:8:15
   |
8  |       let foo = Foo {
   |  _______________^
9  | |         value: 5 as i32,
10 | |         func: lambda
11 | |     };
   | |_____^ one type is more general than the other
   |
   = note: expected trait `for<'a, 'b> Fn<(&'a i32, &'b i32)>`
              found trait `Fn<(&i32, &i32)>`
note: this closure does not fulfill the lifetime requirements
  --> src/main.rs:7:18
   |
7  |     let lambda = |&x, &y| x + y;
   |                  ^^^^^^^^
note: the lifetime requirement is introduced here
  --> src/main.rs:1:18
   |
1  | struct Foo<T, F: Fn(&T, &T) -> T> {
   |                  ^^^^^^^^^^^^^^^

error: implementation of `FnOnce` is not general enough
  --> src/main.rs:8:15
   |
8  |       let foo = Foo {
   |  _______________^
9  | |         value: 5 as i32,
10 | |         func: lambda
11 | |     };
   | |_____^ implementation of `FnOnce` is not general enough
   |
   = note: closure with signature `fn(&'2 i32, &i32) -> i32` must implement `FnOnce<(&'1 i32, &i32)>`, for any lifetime `'1`...
   = note: ...but it actually implements `FnOnce<(&'2 i32, &i32)>`, for some specific lifetime `'2`

error: implementation of `FnOnce` is not general enough
  --> src/main.rs:8:15
   |
8  |       let foo = Foo {
   |  _______________^
9  | |         value: 5 as i32,
10 | |         func: lambda
11 | |     };
   | |_____^ implementation of `FnOnce` is not general enough
   |
   = note: closure with signature `fn(&i32, &'2 i32) -> i32` must implement `FnOnce<(&i32, &'1 i32)>`, for any lifetime `'1`...
   = note: ...but it actually implements `FnOnce<(&i32, &'2 i32)>`, for some specific lifetime `'2`

very long, but the problem is made clear

[estebank]

Really happy to see progress on this front! The last two errors should not be emitted (as they are redundant), but at the very least it should be only one, so I'll leave the ticket open, but I'm ok with the current output as is.