Rustdoc handling of public re-export is broken

[hanna-kruppe]

Looking at nightly docs of core::num, the presentation of ParseFloatError is confusing. It looks like this:

It should be rendered like ParseIntError, for example. The fact that it's defined in a private submodule is irrelevant and should not leak to unsuspecting users. The type does have a page, which can be found with the search, and links from other pages (functions, traits) do work. It's just the landing page for core::num that sucks.

FWIW, in std::num it is rendered normally (the source link is broken, but whatever).

[marti1125]

@steveklabnik I would like to take this.

[steveklabnik]

Have at it!

[ghost]

I think I'm looking at the same issue, but I want to give my analysis of it, cause the symptoms are a little different than described by @rkruppe.

I made a library crate named kraid for testing this.

Example 1

lib.rs:

/// MyType is _my_ type
pub struct MyType;

A normal looking page is generated for kraid with MyType listed under Structs. The page generated for MyType also looks good.

Example 2

lib.rs:

mod foo {
    /// MyType is _my_ type
    pub struct MyType;
}

pub use self::foo::MyType;

Same exact output as Example 1; the module foo is not mentioned. There was an instance in a non-minimal example where the MyType page was not rendered at all (file not found), but not this time I guess.

Example 3

lib.rs:

mod foo {
    mod bar {
        /// MyType is _my_ type
        pub struct MyType;
    }

    pub use self::bar::MyType;
}

pub use self::foo::MyType;

pub use self::foo::MyType; is listed under Reexports, and foo is a link to a blank page. kraid/foo/struct.MyType.html is also a blank page, and unreachable by link.

---

If anyone thinks this is a separate issue, I can open one up.

[hanna-kruppe]

As you say, the way to trigger it is different: I described a re-export from a hidden, public module. You described a re-export from a private submodule of a private module. However, I hope that this is caused by the same piece of code and both can be fixed in one small patch. Let's just expand the scope of this issue.

[ghost]

Hey @marti1125, have you made progress on this? I might try fixing it myself.

[marti1125]

@adambadawy you can try it! I don't have any progress =(

[ghost]

@marti1125 👌 I'll have a look. Not sure I have the knowledge to fix it.