Name resolution is confusing when crates re-exports std items

[Huliiiiii]

Summary

snafu re-exports std::error::Error, which makes clippy show snafu::Error in diagnostics instead of std::error::Error.

cargo check will convert it into StdError.

Reproducer

Code:

trait JustTrait {}

#[derive(Debug)]
struct JustError;

impl std::fmt::Display for JustError {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        todo!()
    }
}

impl std::error::Error for JustError {}

fn foo() -> impl JustTrait {
    Box::new(JustError) as Box<dyn std::error::Error>
}

Current output:

error[E0277]: the trait bound `std::boxed::Box<dyn snafu::Error>: JustTrait` is not satisfied
   --> src/main.rs:125:13
    |
125 | fn foo() -> impl JustTrait {
    |             ^^^^^^^^^^^^^^ the trait `JustTrait` is not implemented for `std::boxed::Box<dyn snafu::Error>`
126 |     Box::new(JustError) as Box<dyn std::error::Error>
    |     ------------------------------------------------- return type was inferred to be `std::boxed::Box<dyn snafu::Error>` here
    |
help: this trait has no implementations, consider adding one
   --> src/main.rs:112:1
    |
112 | trait JustTrait {}
    | ^^^^^^^^^^^^^^^

Desired output:
cargo check output:

error[E0277]: the trait bound `Box<dyn StdError>: JustTrait` is not satisfied
   --> src/main.rs:125:13
    |
125 | fn foo() -> impl JustTrait {
    |             ^^^^^^^^^^^^^^ the trait `JustTrait` is not implemented for `Box<dyn StdError>`
126 |     Box::new(JustError) as Box<dyn std::error::Error>
    |     ------------------------------------------------- return type was inferred to be `Box<dyn StdError>` here
    |
help: this trait has no implementations, consider adding one
   --> src/main.rs:112:1
    |
112 | trait JustTrait {}
    | ^^^^^^^^^^^^^^^

Version

rustc 1.95.0-nightly (3a70d0349 2026-02-27)
binary: rustc
commit-hash: 3a70d0349fa378a10c3748f1a48742e61505020f
commit-date: 2026-02-27
host: x86_64-unknown-linux-gnu
release: 1.95.0-nightly
LLVM version: 22.1.0

Additional Labels

No response

[samueltardieu]

Your example lacks a use snafu as _; or similar at the top, right? Without it, snafu::Error doesn't appear in error messages as the compiler doesn't seem to consider this crate.

With it, both cargo build or cargo check show snafu::Error in the error message as well:

$ cargo build
[…]
error[E0277]: the trait bound `Box<dyn snafu::Error>: JustTrait` is not satisfied
  --> src/lib.rs:16:13
   |
16 | fn foo() -> impl JustTrait {
   |             ^^^^^^^^^^^^^^ the trait `JustTrait` is not implemented for `Box<dyn snafu::Error>`
17 |     Box::new(JustError) as Box<dyn std::error::Error>

cargo check will convert it into StdError.

I get the very same error message with cargo check.

This is not specific to Clippy. This is also not recent (Rust 1.87 also refers to snafu::Error in its error message). Transferring to the rust-lang/rust repository.

@rustbot transfer rust

[eggyal]

snafu::Error is a (doc-hidden) public reexport of core::error::Error, so the diagnostic isn't incorrect but we probably should ignore doc-hidden aliases and/or prefer alias targets (if public) when choosing which is the canonical name of an item to be rendered in diagnostics.

@rustbot label +A-diagnostics +A-visibility +C-bug +D-papercut +T-compiler

[arferreira]

@rustbot claim

[arferreira]

I've traced this to visible_parent_map construction, BFS add_child closure checks if the parent module is #[doc(hidden)] but doesn't check if the re-export itself is. The ModChild.reexport_chain carries the DefId of the use item, so we can call is_doc_hidden on it and send it to the fallback_map, same treatment as doc-hidden parents and underscore re-exports. There's also an existing FIXME test tracking the same class of bug.

[eggyal]

There's also an existing FIXME test tracking the same class of bug.

Is there an existing issue that this is a duplicate of, then?

[arferreira]

Not a duplicate, this is the unhandled half of PR #99698. That one fixed the case where the parent module is #[doc(hidden)], but explicitly left the case where the re-export itself is #[doc(hidden)] as a FIXME.