3.11 tests are failing due to python/cpython#94601, which has been fixed and will be included in 3.11b5. Built docs locally without any issues.

So the root cause of the bug might not have been the issue I linked, but the changed behavior when mixing str and Enum in py3.11. See https://docs.python.org/3.11/library/enum.html#notes.

Hence I adjusted the behavior for 3.11. Also noticed that we show the Enum value on __repr__ for IntEnum derived classes. Is that intended behavior?

About allowed failures: https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idcontinue-on-error

IISC that may lead to some unexpected UI stuff (community/community#15452, actions/runner#2347), but whatever …

Regarding repr, format and str: I think I've changed my perspective a bit by now. Consider the following example

def pprint(arg: int | str) -> str:
    return "Value is" + str(arg)  # explicitly cast in case it's an int

Now pprint(ChatType.PRIVATE) would yield "Value is ChatType.PRIVATE" which I'd argue is unexpected. So I'm now leaning towards having StringEnum == str.__str__. As no user has complained about that so far, I'm also okay with keeping it as is.

In any case

• StringEnum and IntEnum should behave the same
• repr should give unambiguous output and hence ChatType.PRIVATE is still derisable here IMO

Waiting for actions/python-versions#183 to be merged so the 3.11 tests run...