rustdoc(hack): make is_doc_subtype_of always ret true for TyAlias

it's worth noting that this function is only used in
the handling of "Methods from Deref", and we were previously
assuming all generic parameters were meaningless,
so this is still an improvment from the status quo.

this change means that we will have strictly less false positives
without adding any new false negitives.

Author: lolbinarycat

src/librustdoc/clean/types.rs

@@ -1533,6 +1533,14 @@ impl Type {
         matches!(self, Type::BorrowedRef { .. })
     }
 
+    fn is_type_alias(&self) -> bool {
+        if let Type::Path { path: Path { res: Res::Def(DefKind::TyAlias, _), .. } } = &self {
+            true
+        } else {
+            false
+        }
+    }
+
     /// Check if two types are "the same" for documentation purposes.
     ///
     /// This is different from `Eq`, because it knows that things like
@@ -1561,6 +1569,16 @@ impl Type {
         } else {
             (self, other)
         };
+
+        // FIXME: `Cache` does not have the data required to unwrap type aliases,
+        // so we just assume they are equal.
+        // This is only remotely acceptable because we were previously
+        // assuming all types were equal when used
+        // as a generic parameter of a type in `Deref::Target`.
+        if self_cleared.is_type_alias() || other_cleared.is_type_alias() {
+            return true;
+        }
+
         match (self_cleared, other_cleared) {
             // Recursive cases.
             (Type::Tuple(a), Type::Tuple(b)) => {