Adding private imports to a module can break downstream crates

[jyn514]

I tried this code: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=7a55c241c30f1231a2711260adc95d22

mod a {
    pub type Result<T> = std::result::Result<T, ()>;
}

mod inner {
    pub use crate::a::*;
    use std::result::Result;
}

fn f() -> inner::Result<i32> {
    Ok(1)
}

I expected to see this happen: The code compiles fine, because inner::Result resolves to a::Result, just as it would without use std::result::Result.

Instead, this happened:

error[E0603]: enum import `Result` is private
   --> src/lib.rs:10:18
    |
10  | fn f() -> inner::Result<i32> {
    |                  ^^^^^^ private enum import
    |
note: the enum import `Result` is defined here...
   --> src/lib.rs:7:9
    |
7   |     use std::result::Result;
    |         ^^^^^^^^^^^^^^^^^^^
note: ...and refers to the enum `Result` which is defined here

Meta

rustc --version --verbose:

1.59.0

I don't know if this is strictly a bug in the sense that it will break things, but it seems like an unfortunate side effect of the resolution rules ... cc @petrochenkov, do you have opinions on this?

[jyn514]

@rustbot label: +A-resolve

[memoryruins]

I've seen this used before as an intentional way to hide items from pub use globs (for lack of a special syntax for hiding from globs) and have suggested it to someone else before (heh), which is something that a crater run might not catch if this was changed.

Curious about how long it's worked this way, before 1.15 it would error with "a type named Result has already been imported in this module", but 1.15 onward, RFC 1560 was stabilized. The current behavior appears to fall under the "Explicit names may shadow implicit names" section. Although the section doesn't feature an example of importing from the same module, we can think of it similarly: the second use is the explicit import that shadows the glob's import of the item (so it uses explicit import's visibility).

In that sense, I do not think this is a bug, though I understand why it seems surprising at first.