Explicitly mark a hack as a HACK and elaborate its comment

Author: fmease

compiler/rustc_hir_analysis/src/errors/wrong_number_of_generic_args.rs

@@ -1049,13 +1049,14 @@ impl<'a, 'tcx> WrongNumberOfGenericArgs<'a, 'tcx> {
             );
 
             if span.is_empty() {
-                // Avoid ICE when types with the same name with `derive`s are in the same scope:
-                // struct NotSM;
-                // #[derive(PartialEq, Eq)]
-                // struct NotSM<T>(T);
-                // With the above code, the suggestion is to remove the generics of the first
-                // `NotSM`, which doesn't *have* generics, so we're suggesting to remove no code
-                // with no code, which ICEs on nightly due to an `assert!`.
+                // HACK: Avoid ICE when types with the same name with `derive`s are in the same scope:
+                //     struct NotSM;
+                //     #[derive(PartialEq, Eq)]
+                //     struct NotSM<T>(T);
+                // With the above code, the suggestion would be to remove the generics of the first
+                // `NotSM`, which doesn't *have* generics, so we would suggest to remove no code with
+                // no code, which would trigger an `assert!` later. Ideally, we would do something a
+                // bit more principled. See closed PR #109082.
             } else {
                 err.span_suggestion(span, msg, "", Applicability::MaybeIncorrect);
             }

tests/ui/duplicate/multiple-types-with-same-name-and-derive.rs

@@ -2,6 +2,8 @@
 // expansion these `impl`s are associated to the the *other* type. There is a suggestion to remove
 // unneded type parameters, but because we're now point at a type with no type parameters, the
 // suggestion would suggest removing code from an empty span, which would ICE in nightly.
+//
+// issue: rust-lang/rust#108748
 
 struct NotSM;