Confusing 'type mismatch' error when using function pointer as fn trait

[Aaron1011]

The following code:

struct MyType;

fn parse_header<T, F: FnOnce(&MyType) -> T>(parse: F) -> T {
    panic!()
}

fn make_it(val: &MyType) -> &MyType { panic!() }

pub fn from_url<'a>() -> &'a MyType {
    parse_header(make_it)
}

fn main() {}

gives this error:

error[E0271]: type mismatch resolving `for<'r> <for<'s> fn(&'s MyType) -> &'s MyType {make_it} as std::ops::FnOnce<(&'r MyType,)>>::Output == _`
  --> src/main.rs:10:5
   |
3  | fn parse_header<T, F: FnOnce(&MyType) -> T>(parse: F) -> T {
   |                                          - required by this bound in `parse_header`
...
10 |     parse_header(make_it)
   |     ^^^^^^^^^^^^ expected bound lifetime parameter, found concrete lifetime

There are several issues with this error:

• The 'type mismatch' message contains several pieces of 'internal' rust syntax (multiple consecutive for<> binders, the weird {make_it} syntax for function pointers, and an inference variable).
• We call the return type T a 'bound', even though it's clearly not a bound.
• We give an confusing 'expected bound lifetime parameter, found concrete lifetime' message, which doesn't really explain what's going on
• We don't point at make_it, even though it's part of the issue.

This was minimized from code posted by @rvolosatovs.

[estebank]

• The 'type mismatch' message contains several pieces of 'internal' rust syntax (multiple consecutive for<> binders, the weird {make_it} syntax for function pointers, and an inference variable).

We can probably improve this error quite dramatically by addressing cases involving closures specifically.

• We call the return type T a 'bound', even though it's clearly not a bound.

It is a bound, it's the Output bound in Fn(Output=T), which is sugared as Fn() -> T. We do need to make that clearer (by maybe not talking about it being a bound).

• We don't point at make_it, even though it's part of the issue.

In E0308 we point at the parameter when possible using some hacky inference that could also be leveraged here.

[Aaron1011]

We can probably improve this error quite dramatically by addressing cases involving closures specifically.

What makes this case tricky is that we never actually create a closure - the function pointer is used directly as F

[estebank]

Current output:

error: implementation of `std::ops::FnOnce` is not general enough
   --> src/main.rs:10:5
    |
10  |       parse_header(make_it)
    |       ^^^^^^^^^^^^ implementation of `std::ops::FnOnce` is not general enough
    |
    = note: `std::ops::FnOnce<(&'0 MyType,)>` would have to be implemented for the type `for<'r> fn(&'r MyType) -> &'r MyType {make_it}`, for some specific lifetime `'0`...
    = note: ...but `std::ops::FnOnce<(&MyType,)>` is actually implemented for the type `for<'r> fn(&'r MyType) -> &'r MyType {make_it}`

[estebank]

Current output:

error[E0308]: mismatched types
  --> src/main.rs:10:5
   |
10 |     parse_header(make_it)
   |     ^^^^^^^^^^^^^^^^^^^^^ one type is more general than the other
   |
   = note: expected reference `&MyType`
              found reference `&MyType`
note: the lifetime requirement is introduced here
  --> src/main.rs:3:42
   |
 3 | fn parse_header<T, F: FnOnce(&MyType) -> T>(parse: F) -> T {
   |                                          ^