Name resolution does not report ambiguities

I tried this code:

mod foo {
    pub struct Qux;
    pub fn f() -> char { 'a' }
}

mod bar {
    pub struct Qux;
    pub fn f() -> i32 { 15 }
}

fn main() {
    use foo::*;
    use bar::*;
    
    let a: i32 = f();
}

I expected to see this happen:

error[E0659]: `f` is ambiguous (glob import vs glob import in the same module)
  --> test.rs:15:18
   |
15 |     let a: i32 = f();
   |                  ^ ambiguous name
   |
note: `f` could refer to the function imported here
  --> test.rs:12:9
   |
12 |     use foo::*;
   |         ^^^^^^
   = help: consider adding an explicit import of `f` to disambiguate
note: `f` could also refer to the function imported here
  --> test.rs:13:9
   |
13 |     use bar::*; // Ok, no name conflict.
   |         ^^^^^^
   = help: consider adding an explicit import of `f` to disambiguate

warning: unused import: `bar::*`
  --> test.rs:13:9
   |
13 |     use bar::*; // Ok, no name conflict.
   |         ^^^^^^
   |
   = note: `#[warn(unused_imports)]` on by default

error[E0308]: mismatched types
  --> test.rs:15:18
   |
15 |     let a: i32 = f();
   |            ---   ^^^ expected `i32`, found `char`
   |            |
   |            expected due to this

Instead, this happened:

test.rs:15:5: error: expected ‘i32’ got ‘char’
   15 |     let a: i32 = f();
      |     ^~~    ~~~   ~

So the name resolver picks the first declaration of f without reporting about the conflict. I added a type mismatch error to see which one was being picked - but if there is no type error, then gccrs does not emit the name resolution error, which is important here.

Meta

• What version of Rust GCC were you using, git sha if possible.

533e1ef3ca4

Reproduce the issue with the Rust snippet in the report and inspect the gccrs name-resolution handling for two glob imports that define the same name. Done means the resolver reports an ambiguity for f, identifies both imports, and does not silently select the first declaration.