Add a debug assertion in codegen that unsize casts of the same princi…

…pal trait def id are truly NOPs

compiler/rustc_codegen_cranelift/src/unsize.rs

@@ -34,7 +34,22 @@ pub(crate) fn unsized_info<'tcx>(
  let old_info =
  old_info.expect("unsized_info: missing old info for trait upcasting coercion");
  if data_a.principal_def_id() == data_b.principal_def_id() {
- // A NOP cast that doesn't actually change anything, should be allowed even with invalid vtables.
+ // Codegen takes advantage of the additional assumption, where if the
+ // principal trait def id of what's being casted doesn't change,
+ // then we don't need to adjust the vtable at all. This
+ // corresponds to the fact that `dyn Tr<A>: Unsize<dyn Tr<B>>`
+ // requires that `A = B`; we don't allow *upcasting* objects
+ // between the same trait with different args. If we, for
+ // some reason, were to relax the `Unsize` trait, it could become
+ // unsound, so let's assert here that the trait refs are *equal*.
+ //
+ // We can use `assert_eq` because the binders should have been anonymized,
+ // and because higher-ranked equality now requires the binders are equal.
+ debug_assert_eq!(
+ data_a.principal(),
+ data_b.principal(),
+ "NOP unsize vtable changed principal trait ref: {data_a} -> {data_b}"
+ );
  return old_info;
  }
 

compiler/rustc_codegen_ssa/src/base.rs

@@ -124,6 +124,23 @@ fn unsized_info<'a, 'tcx, Bx: BuilderMethods<'a, 'tcx>>(
  let old_info =
  old_info.expect("unsized_info: missing old info for trait upcasting coercion");
  if data_a.principal_def_id() == data_b.principal_def_id() {
+ // Codegen takes advantage of the additional assumption, where if the
+ // principal trait def id of what's being casted doesn't change,
+ // then we don't need to adjust the vtable at all. This
+ // corresponds to the fact that `dyn Tr<A>: Unsize<dyn Tr<B>>`
+ // requires that `A = B`; we don't allow *upcasting* objects
+ // between the same trait with different args. If we, for
+ // some reason, were to relax the `Unsize` trait, it could become
+ // unsound, so let's assert here that the trait refs are *equal*.
+ //
+ // We can use `assert_eq` because the binders should have been anonymized,
+ // and because higher-ranked equality now requires the binders are equal.
+ debug_assert_eq!(
+ data_a.principal(),
+ data_b.principal(),
+ "NOP unsize vtable changed principal trait ref: {data_a} -> {data_b}"
+ );
+
  // A NOP cast that doesn't actually change anything, should be allowed even with
  // invalid vtables.
  return old_info;

compiler/rustc_mir_transform/src/validate.rs

@@ -1236,21 +1236,6 @@ impl<'a, 'tcx> Visitor<'tcx> for TypeChecker<'a, 'tcx> {
  )) {
  self.fail(location, format!("Unsize coercion, but `{op_ty}` isn't coercible to `{target_type}`"));
  }
-
- // FIXME: Codegen has an additional assumption, where if the
- // principal trait def id of what's being casted doesn't change,
- // then we don't need to adjust the vtable at all. This
- // corresponds to the fact that `dyn Tr<A>: Unsize<dyn Tr<B>>`
- // requires that `A = B`; we don't allow *upcasting* objects
- // between the same trait with different args. Nothing actually
- // validates this, though. While it's true right now, if we for
- // some reason were to relax the `Unsize` trait, it could become
- // unsound. We should eventually validate that, but it would
- // require peeling `&Box<Struct<.., dyn Tr<A>, ..>>` down to
- // the trait object that's being unsized, and that's rather
- // annoying, and also it would need to be opportunistic since
- // this MIR is not yet fully monomorphized, so we may bottom
- // out in an alias or a projection or something.
  }
  CastKind::IntToInt | CastKind::IntToFloat => {
  let input_valid = op_ty.is_integral() || op_ty.is_char() || op_ty.is_bool();