Documentation on re-exports is ignored if the item is defined in the same crate

[jyn514]

I tried this code:

mod inner_mod {
    /// docs from inner module
    pub fn g() {}
}

/// docs from outer module
pub use inner_mod::g;

I expected to see this happen: The documentation is added, like it is for cross-crate re-exports:

/// Some docs from outer crate
pub use inner::f;

Instead, this happened: The documentation is ignored:

Meta

rustdoc --version: rustdoc 1.53.0-nightly (392ba2b 2021-04-17)

Originally posted by @jsgf in #84597 (comment)

[Lonami]

Concatenating both doc-comments seems slightly strange to me, at least for the same crate. My expectation is that the re-export would have its own and, if you navigate to the re-exported name, you see the original documentation.

[jyn514]

@Lonami concatenating them is what rustdoc does for cross crate links, it should do the same thing for intra crate links. We shouldn't change it now.

[jyn514]

Well, here's why:

rust/src/librustdoc/clean/inline.rs

Lines 49 to 51 in 6c2dd25

	if did.is_local() {
	return None;
	}

but removing that block breaks rustdoc quite a lot:

failures:
    [rustdoc] rustdoc/foreigntype-reexport.rs
    [rustdoc] rustdoc/inline_local/please_inline.rs
    [rustdoc] rustdoc/intra-doc/field.rs
    [rustdoc] rustdoc/proc-macro.rs
    [rustdoc] rustdoc/reexport-stability-tags-deprecated-and-portability.rs
    [rustdoc] rustdoc/reexport-stability-tags-unstable-and-portability.rs

[Lonami]

I'm not familiar with rustdoc tests, but looks like we should just update the expected results for the tests (assuming we do want to do this change, of course). As in, I don't think things are "breaking" as signaled by the failing tests with this change, but rather it just needs updating.

[jyn514]

@Lonami no, there are larger changes, it's not just documentation being added. I don't have the exact output on hand, it was a lot.

[lambinoo]

I think I'll try to tackle this one! Maybe this will also fix #89020
@rustbot claim