Correctly handle associated items of a trait inside a #[doc(hidden)] item

[GuillaumeGomez]

Fixes #111064.

cc @compiler-errors
r? @notriddle

[GuillaumeGomez]

Fixed fmt issue.

[notriddle]

I don't think this entirely fixes the problem.

It fixes the case where you specifically inline a trait from a hidden module, but these two test cases both fail:

Test case for inlining a module that inherits hiddenness

#![crate_name = "foo"]

// @!has 'foo/hidden/index.html'
// @!has 'foo/hidden/inner/index.html'
// @!has 'foo/hidden/inner/trait.Foo.html'
#[doc(hidden)]
pub mod hidden {
    pub mod inner {
        pub trait Foo {
            /// Hello, world!
            fn test();
        }
    }
}

// @has 'foo/visible/index.html'
// @has 'foo/visible/trait.Foo.html'
#[doc(inline)]
pub use hidden::inner as visible;

// @has 'foo/struct.Bar.html'
// @count - '//*[@id="impl-Foo-for-Bar"]' 1
pub struct Bar;

impl visible::Foo for Bar {
    fn test() {}
}

Test case for inlining an item that's two levels deep

#![crate_name = "foo"]

// @!has 'foo/hidden/index.html'
// @!has 'foo/hidden/inner/index.html'
// @!has 'foo/hidden/inner/trait.Foo.html'
#[doc(hidden)]
pub mod hidden {
    pub mod inner {
        pub trait Foo {
            /// Hello, world!
            fn test();
        }
    }
}

// @has 'foo/trait.Foo.html'
#[doc(inline)]
pub use hidden::inner::Foo;

// @has 'foo/struct.Bar.html'
// @count - '//*[@id="impl-Foo-for-Bar"]' 1
pub struct Bar;

impl Foo for Bar {
    fn test() {}
}

[GuillaumeGomez]

@notriddle In the two tests you provided, you're underlining a limitation of this fix (which is great!) and another existing bug. I don't think they have the same origin though but I'll try to fix both here as well. After a lot of investigations, I'm still wondering whether or not we should store all reexports and then when we encounter an item that is inside a #[doc(hidden)] path and is re-exported, to check how it is re-exported. I'm still trying to find another way though because that'd be quite bad performance-wise...

[GuillaumeGomez]

@notriddle The first failing test you gave actually allowed me to uncover another bug: reexported modules were not handled in rustdoc. So that's two bugs fixed with this PR.

For the extra files created, since the bug already exists, I'll fix it in a follow-up PR. I opened the issue for it here: #111249

[notriddle]

This seems better.

inherits_doc_hidden's stop_at parameter seems like a more complete fix. #[doc(hidden)] is recursively applied, so it makes sense that a bug fix for it should impact the way the recursion is done and not just check one level. Thank you!

I have nits, but the approach seems sound.

[notriddle]

Why are name and renamed both needed? It seems like name is completely ignored if renamed is Some.

[GuillaumeGomez]

I wondered if I should pass the "computed" name or if I should do it in the function. In the end I picked this approach because it's not possible to pass the "non-computed" name since you need to pass both name and renamed.

[GuillaumeGomez]

Updated!

[notriddle]

@bors r+ rollup

[bors]

📌 Commit 8de4308 has been approved by notriddle

It is now in the queue for this repository.