Not sure what pyright wants here, bug?

Not sure what pyright wants here, bug?

Oh, no idea what's going on there. We're a few pyright releases behind in CI; we could try updating to the latest and see if it still reproduces?

Yeah, seems like the failures persist with the latest pyright

cc @erictraut any idea what's going wrong here?

This is caused by a circular dependency involving the EnumMeta class.

The ReprEnum class derives from Enum which uses the metaclass EnumMeta. The EnumMeta has a __new__ method that describes a keyword parameter boundary of type FlagBoundary. This type derives from StrEnum which derives from ReprEnum, and we're back to the start.

This circularity can be resolved if analysis is done in precisely the right order. Since pyright is a lazy (just-in-time) type analyzer, the standard ordering can easily change. I suspect that the change in this PR perturbed things just enough to change the evaluation ordering. I have logic in place that's designed to detect circularities like this and force them to resolve in a satisfactory order, but the logic is complex, and it appears that it's not handling this case. I'll investigate it further.

I've filed this bug in the pyright issue tracker.

This will be fixed in the next pyright release (1.1.312).

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