[analyzer] Deriving a mixin from a class whose superclass isn't Object not detected

[eernstg]

Cf. #45193, Mixed is used to derive a mixin, and Mixed has the superclass Object with N, so it should be flagged as an error.

[scheglov]

It looks reasonable to me that we should mark these Object with X classes when used as mixins.
But there are existing language tests that fail.
For example language/mixin/mixin7_test.dart and language/covariant/setter_test.

[scheglov]

https://dart-review.googlesource.com/c/sdk/+/189300

[eernstg]

Looking at language/mixin/mixin7_test.dart, I see the following:

The superclass of M<T> is Object for any T, so it's OK to derive a mixin from M<...>, and a raw M would have its type argument inferred.

In class A<U, V> = Object with M implements I<V>;, the specification actually makes Object the superclass of A, not Object with M<...>. That makes this <mixinApplicationClass> slightly different from class A<U, V> extends Object with M implements I<V> {}, and it would in particular be OK to derive a mixin from A<...>.

Cf. https://github.com/dart-lang/language/blob/53f3e4606f7efe25679c37698e167734df7f934c/specification/dartLangSpec.tex#L5197 and https://github.com/dart-lang/language/blob/53f3e4606f7efe25679c37698e167734df7f934c/specification/dartLangSpec.tex#L5387.

In class B<T> = Object with A implements J<T>;, the superclass is again Object, so it's again OK to derive a mixin from B<...> and B.

So where does that test fail? Maybe it's because class A = S with M; is actually treated as class A extends S with M {} such that it gets the superclass S with M?

Looking at language/covariant/setter_test.dart, the situation seems to be the same.

[scheglov]

Thank you. My CL was wrong. I will fix it.

[scheglov]

https://dart-review.googlesource.com/c/sdk/+/189300