Rollup merge of #156823 - ajasad25:rustdoc-self-submodule-tests-84827, r=camelid

Add regression test

Fixes #84827.

Author: JonathanBrouwer

tests/rustdoc-html/intra-doc/self-impl-in-submodule-84827.rs

@@ -0,0 +1,43 @@
+// https://github.com/rust-lang/rust/issues/84827
+//
+// Regression test for resolving `Self` in intra-doc links inside an `impl`
+// block in a submodule. The `Self` type comes from the impl, not from names
+// in scope, so:
+//   1. It resolves even when the implemented type is not in scope.
+//   2. It resolves to the actual `Self` type, not to a different type with
+//      the same name that happens to be in scope.
+
+#![crate_name = "foo"]
+#![deny(rustdoc::broken_intra_doc_links)]
+#![allow(unused_imports)]
+
+pub struct Foo {
+    pub foo: i32,
+}
+
+// Case 1: `Foo` is not in scope inside `bar`, but `Self::foo` still resolves
+// to the field on `crate::Foo`.
+pub mod bar {
+    //@ has foo/struct.Foo.html '//a[@href="struct.Foo.html#structfield.foo"]' 'Self::foo'
+    impl crate::Foo {
+        /// Baz the [`Self::foo`].
+        pub fn baz(&self) {}
+    }
+}
+
+// Case 2: A different type named `Foo` is in scope inside `baz`. `Self::foo`
+// must still resolve to `crate::Foo::foo`, not to `crate::other::Foo` (which
+// is a unit struct and has no `foo` field).
+pub mod baz {
+    use crate::other::Foo;
+
+    //@ has foo/struct.Foo.html '//a[@href="struct.Foo.html#structfield.foo"]' 'Self::foo'
+    impl crate::Foo {
+        /// Quux the [`Self::foo`].
+        pub fn quux(&self) {}
+    }
+}
+
+pub mod other {
+    pub struct Foo;
+}