[rustdoc-json] Assertion error on reexporting item from private module with 2 names

[aDotInTheVoid]

mod style {
    pub struct Color;
}

pub use style::Color;
pub use style::Color as Colour;

rustdoc +stage1 rename.rs -w json
thread 'rustc' panicked at 'assertion failed: `(left == right)`

Diff < left / right > :
 Item {
     id: Id(
         "0:2",
     ),
     crate_id: 0,
     name: Some(
<        "Color",
>        "Colour",
     ),
     span: Some(
         Span {
             filename: "rename.rs",
             begin: (
                 5,
                 4,
             ),
             end: (
                 5,
                 21,
             ),
         },
     ),
     visibility: Public,
     docs: None,
     links: {},
     attrs: [],
     deprecation: None,
     inner: Struct(
         Struct {
             struct_type: Unit,
             generics: Generics {
                 params: [],
                 where_predicates: [],
             },
             fields_stripped: false,
             fields: [],
             impls: [],
         },
     ),
 }

', src/librustdoc/json/mod.rs:179:17
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

error: internal compiler error: unexpected panic

error: Unrecognized option: 'w'

It's unlear to me what the correct representation should be, as currently for items pub used from a non pub mod, we add the item to the mod it's being imported to, with the name it's imported under.

FWIW, in HTML, 2 seperate pages are created, each with the same content.

Eg: https://docs.rs/ansi_term/0.12.1/ansi_term/enum.Color.html and https://docs.rs/ansi_term/0.12.1/ansi_term/enum.Color.html

(this uses the same diffing as #83718)

@rustbot modify labels: +A-rustdoc-json +T-rustdoc

[CraftSpider]

Would it be reasonable to say that the 'correct' behavior for JSON would be to document the item as in its actual location, possibly with handling to allow it to indicate 'in some module which is private', and document the re-exports as just that? That feels 'more correct' and allows the end user to determine how or if they want to handle inlining like this.

[CraftSpider]

Note this would likely also interact with #83057, and I'm not sure exactly how explicit inlining should be handled in this case.

[Enselic]

I can reproduce with nightly-2022-05-18 but not with nightly-2022-06-27.

If you can confirm I think we can just close this?

Edit: Even though the ICE is gone we maybe still want to change the output. Not sure.

@rustbot labels +C-bug