Misleading note for accessing closures within structs.

[ix]

Hi, I encountered an issue the other day wherein I was attempting to call a closure stored within a struct.

The compiler gives a misleading error which doesn't give the user an idea on how to fix the problem.
I've prepared an example which reproduces the misleading text:

struct Example {
    example: Box<Fn(i32) -> i32>
}

fn main() {
    let demo = Example {
        example: Box::new(|x| {
            x + 1
        })
    };

    demo.example(1);
    // (demo.example)(1);
}

The error this yields from the compiler is as follows:

structs.rs:12:10: 12:20 error: no method named `example` found for type `Example` in the current scope
structs.rs:12     demo.example(1);
                       ^~~~~~~~~~
structs.rs:12:10: 12:20 note: did you mean to write `demo.example`?
structs.rs:12     demo.example(1);
                       ^~~~~~~~~~
error: aborting due to previous error

Note that in the source code I've provided, underneath the line which yields this error, I have provided a comment demonstrating the correct way to access and call the example member of the structure, however the compiler's example is identical to the erroneous code and thus doesn't help the user in working out how to properly call the member.

Would it be at all possible to improve this error message?

I've confirmed this is reproducible on rustc 1.7.0 (a5d1e7a59 2016-02-29).

[Aatch]

If I'm right, we should be able to check to see if the field implements one of the Fn* traits and alter the error message based on that.

[durka]

The check already exists, but it only checks for FnOnce and doesn't try deref coercions (which you will need if you want to catch Box<Fn()>).

[Aatch]

@durka as far as I know anything that implements any of the Fn* traits will implement FnOnce, (since if you can call it more than once, you can clearly call just once). So it's just a matter of checking deref too.

[durka]

@Aatch is it feasible to start up the deref machinery just for a note in an error message? I don't really know how that stuff works.

[eddyb]

@durka it's a single call, search for autoderef in typeck.