[red-knot] is_subtype_of no longer implies is_assignable_to for all Type::Tuple types

[github-actions]

Failing property-test run listed here: https://github.com/astral-sh/ruff/actions/runs/14265011645

[AlexWaygood]

failures:

---- types::property_tests::stable::all_type_pairs_are_assignable_to_their_union stdout ----

thread 'types::property_tests::stable::all_type_pairs_are_assignable_to_their_union' panicked at /home/runner/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/quickcheck-1.0.3/src/tester.rs:165:28:
[quickcheck] TEST FAILED. Arguments: (AlwaysFalsy, Tuple([]))
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- types::property_tests::stable::subtype_of_implies_assignable_to stdout ----

thread 'types::property_tests::stable::subtype_of_implies_assignable_to' panicked at /home/runner/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/quickcheck-1.0.3/src/tester.rs:165:28:
[quickcheck] TEST FAILED. Arguments: (Tuple([]), AlwaysFalsy)

@sharkdp I think this may be due to the changes we made to Type::is_assignable_to in #17168...?

[sharkdp]

@sharkdp I think this may be due to the changes we made to Type::is_assignable_to in #17168...?

That seems likely, thank you for the ping. I will look into it.

[AlexWaygood]

haven't tested it, but I think this might fix things:

diff --git a/crates/red_knot_python_semantic/src/types.rs b/crates/red_knot_python_semantic/src/types.rs
index 4f89feb57..17a43a519 100644
--- a/crates/red_knot_python_semantic/src/types.rs
+++ b/crates/red_knot_python_semantic/src/types.rs
@@ -1071,9 +1071,12 @@ impl<'db> Type<'db> {
             // This special case is required because the left-hand side tuple might be a
             // gradual type, so we can not rely on subtyping. This allows us to assign e.g.
             // `tuple[Any, int]` to `tuple`.
-            (Type::Tuple(_), _) => KnownClass::Tuple
-                .to_instance(db)
-                .is_assignable_to(db, target),
+            (Type::Tuple(_), _) => {
+                self.is_subtype_of(db, target)
+                    || KnownClass::Tuple
+                        .to_instance(db)
+                        .is_assignable_to(db, target)
+            }
 
             // `type[Any]` is assignable to any `type[...]` type, because `type[Any]` can
             // materialize to any `type[...]` type.

though there may also be a more elegant or more efficient solution.