Why do you think this should handled instead of delegating to mypy?

Hi, afaik, the related line here is ignoring mypy.
In opentelemetry-api/src/opentelemetry/context/contextvars_context.py#50

self._current_context.reset(token)  # type: ignore

Due to this comment, mypy cannot detect TypeError.

If you are suggesting to uncomment the ignore, I'm ok, the draft PR is just my initial proposal.

Hi, afaik, the related line here is ignoring mypy. In opentelemetry-api/src/opentelemetry/context/contextvars_context.py#50

self._current_context.reset(token)  # type: ignore

Due to this comment, mypy cannot detect TypeError.

If you are suggesting to uncomment the ignore, I'm ok, the draft PR is just my initial proposal.

Missed that, thanks. Yeah, I think it would make sense to add type annotations here if feasible

I added type annotations - could you have a check on the latest PR of this issue?

@xrmx I managed to reproduce a problem with a 'None' token using a partially instrumented celery (sender not instrumented, worker instrumented) as seen above. This causes the context.detach to fail. I'm not sure if the fix is simply to to check if the token is not None before calling detach ?