Wrong static type on ==

[johnniwinther]

The test co19/LanguageFeatures/Extension-types/static_analysis_member_invocation_A01_t05 assumes that the static type of == can be derived from the return type of the declared operator == method:

class B implements A {
  ...
  @override
  BoolET operator ==(Object? other) => BoolET(other == this);
}
...
extension type ET2(B b) implements ET1, B {}

extension type BoolET(bool b) implements bool {}

void main() {
  var e2 = ET2(B());
  ...
  BoolET b2 = e2 == e2;
}

Here e2.== can be found to have type BoolET Function(Object?) as declared in B.

If the latest spec is to be believed, the static type of an equality expression is always defined to be bool: (§17.27: "The static type of an equality expression is bool.")

cc @eernstg

[eernstg]

Right, an equality expression has the type bool no matter what the declaration of the member named == says. I guess the reason is that an equality expression does more than just call operator == on the left hand operand (it performs a couple of tests for null).

This means that "treated as all other non-extension type members" in the test description is true even for ==. However, an expression of the form e1 == e2 is given an unusual treatment in all cases (when it occurs in the interface of the class B, and also when it occurs in the interface of the extension type ET2).

It would probably be helpful for future readers to add a comment like '// The type of an equality expression is always bool.' just before bool b2 = e1 == e2;.