privacy: error messages could do better at telling me why something is private

[pnkfelix]

Consider this code:

mod a {
    #[cfg(not(fixed))]
    struct Foo;
    #[cfg(fixed)]
    pub struct Foo;

    impl Foo {
        pub fn new() -> Foo {
            Foo
        }
    }
}

fn main() {
    let foo = a::Foo::new();

    println!("foo: {:?}", foo);
}

When I run rustc today:

% rustc --version
/Users/pnkfelix/opt/rust-dbg/bin/rustc 0.10-pre (d79fbba 2014-03-29 18:56:36 -0700)
host: x86_64-apple-darwin
% rustc /tmp/f.rs
/tmp/f.rs:15:15: 15:26 error: static method `new` is inaccessible
/tmp/f.rs:15     let foo = a::Foo::new();
                           ^~~~~~~~~~~
error: aborting due to previous error
% 

The reason that new is inaccessible, AFAICT, is because Foo is private. However, this is not really relayed in the error message; my first instinct is to look at the definition of new itself, which has been declared pub.

(I had thought there was a related bug filed somewhere about saying that if something is private due to an intermediate non-pub mod along a chain, we could do a better job of pointing the user at one or more of the non-pub mods that would need to be made pub. Obviously that's not perfect since an item may be accessible via more than one path, and so any such suggestion coukd be misleading. But I could not find that related bug. Anyway, the situation today is not ideal.)

[pnkfelix]

I had thought there was a related bug filed somewhere about saying that if something is private due to an intermediate non-pub mod along a chain

I was probably thinking of #13422 when I wrote that. (There are probably many places where it is hard to tell what the source of non-pubness is...)

[alexcrichton]

It's true that there's a bunch of cases, but I'd love to know about missing ones like this!