Diagnostics for blanket impl of closure traits are suboptimal

[huonw]

fn direct<F: Fn()>(_: F) {}

trait Foo { fn method(&self) {} }

impl<F: Fn()> Foo for F {}

fn blanket<F: Foo>(_: F) {}

struct Moves;
fn main() {
    let x = Moves;   direct(|| drop(x));

    // let y = Moves;   blanket(|| drop(y));
}

The x line gives

<anon>:11:37: 11:38 error: cannot move out of captured outer variable in an `Fn` closure
<anon>:11     let x = Moves;   direct(|| drop(x));
                                              ^

while the y line gives:

<anon>:13:22: 13:29 error: the trait `core::ops::Fn<()>` is not implemented for the type `[closure <anon>:13:30: 13:40]` [E0277]
<anon>:13     let y = Moves;   blanket(|| drop(y));
                               ^~~~~~~

The latter doesn't really give much information about the problem (moving out of y) and, AFAIK, the most effective way to debug is to find the offending statement via binary search commenting out chunks of code. :(

[nikomatsakis]

triage: P-high (1.0)

I consider this a nice to have, but it's prob not too hard to really improve this error message.

[ftxqxd]

You don’t even need a blanket trait impl to get a poor error message; using an intermediate variable binding also gives no hint as to what stops the closure from implementing Fn:

fn foo<T: Fn()>(_: T) {}

fn main() {
    let a = Box::new(1);
    let x = || {
        a;
    };
    foo(x);
}

gives

<anon>:8:5: 8:8 error: the trait `core::ops::Fn<()>` is not implemented for the type `[closure <anon>:5:13: 7:6]` [E0277]
<anon>:8     foo(x);
             ^~~
error: aborting due to previous error

[blaenk]

hahahaha yes, I'm glad I'm not the only one that thinks this. I think this is the weakest area I've witnessed in Rust's error diagnostics and I think it will be a major pain-point for newcomers, which will probably lead them to see Rust closures as very complex and difficult to understand. I had a lot of trouble with this myself until I ran into it multiple times and @p1start helped me track down the problem, so I think I know what to look for now.

I'm happy to see that this is a high priority!

[alexcrichton]

triage: P-low

Note a P-medium issue by today's standards, and also hasn't gone fixed for quite awhile! Also cc @jonathandturner, issue about diagnostics. Also removing the assignee of @nikomatsakis as I believe this isn't actively being worked on

[Mark-Simulacrum]

@arielb1 Unassigning you since this isn't being worked on as far as I know.

with x:

error[E0507]: cannot move out of captured outer variable in an `Fn` closure
  --> test.rs:11:37
   |
11 |     let x = Moves;   direct(|| drop(x));
   |         - captured outer variable   ^ cannot move out of captured outer variable in an `Fn` closure

error: aborting due to previous error

with y:

error[E0525]: expected a closure that implements the `Fn` trait, but this closure only implements `FnOnce`
  --> test.rs:13:30
   |
13 |     let y = Moves;   blanket(|| drop(y));
   |                              ^^^^^^^^^^
   |
note: the requirement to implement `Fn` derives from here
  --> test.rs:13:22
   |
13 |     let y = Moves;   blanket(|| drop(y));
   |                      ^^^^^^^

error: aborting due to previous error

[Mark-Simulacrum]

Today, the y case is fixed. There appear to be UI tests for this already (src/test/ui/closure_context/issue-26046-fn-once.stderr) and an fn-mut partner to it so closing.

error[E0525]: expected a closure that implements the `Fn` trait, but this closure only implements `FnOnce`
  --> test.rs:12:28
   |
12 |     let y = Moves; blanket(|| drop(y));
   |                    ------- ^^^^^^^^^^
   |                    |
   |                    the requirement to implement `Fn` derives from here
   |
note: closure is `FnOnce` because it moves the variable `y` out of its environment
  --> test.rs:12:36
   |
12 |     let y = Moves; blanket(|| drop(y));
   |                                    ^

error: aborting due to previous error(s)