Detect metaclass conflicts, refs #13563

[sobolevn]

Recreate of #13565

[sobolevn]

@AlexWaygood do you have any suggestions about the error message? 🤔

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

[AlexWaygood]

Maybe you could do something like this, to improve the error message:

Suggested change

	metaclasses = [
	entry.metaclass_type
	for entry in typ.mro[1:-1]
	if entry.metaclass_type
	and not is_named_instance(entry.metaclass_type, "builtins.type")
	]
	if not metaclasses:
	return
	if typ.metaclass_type is not None and all(
	is_subtype(typ.metaclass_type, meta) for meta in metaclasses
	):
	return
	self.fail(
	"Metaclass conflict: the metaclass of a derived class must be "
	"a (non-strict) subclass of the metaclasses of all its bases",
	typ,
	)
	subclass_metaclass = typ.metaclass_type
	if subclass_metaclass is None:
	return
	for superclass in typ.mro[1:-1]:
	superclass_metaclass = superclass.metaclass_type
	if superclass_metaclass is None or is_named_instance(
	superclass_metaclass, "builtins.type"
	):
	continue
	if not is_subtype(subclass_metaclass, superclass_metaclass):
	self.fail(
	(
	f"Metaclass conflict: metaclass {subclass_metaclass.fullname!r} "
	f"is not a subclass of {superclass_metaclass.fullname!r}, "
	f"the metaclass of base class {superclass.fullname!r}"
	),
	typ
	)

[sobolevn]

I've tried it and I didn't like it. Why?
Because it either reports several errors for several bases with incompatible metaclasses or reports just one (if break is added) which is potentially confusing or we need even more complex code to accumulate all problematic base types and then raise a single one.

Let's keep it as-is for now?

If someone complains, then we can easily make this more readable and more complex.

(This is why I love coping CPython's error messages 😆 )

[AlexWaygood]

Let's keep it as-is for now?

Sure, the PR looks pretty good right now! It was just an idea :)

[sobolevn]

Thanks everyone! 👏