I don't think this is what we meant on the discussion (at least not what I mean). 🤔

I mean to raise a warning if lifespan is used on Router if not provided from Starlette.

Something like this?

Yes. Are you able to create tests for it?

Something like this? Note, locally some existing tests would still fail with:

+ venv/bin/mypy starlette tests
starlette/routing.py:638: error: Incompatible types in assignment (expression has type "Union[Callable[[Any], AbstractAsyncContextManager[None]], Callable[[Any], AbstractAsyncContextManager[Mapping[str, Any]]], None]", variable has type "Union[Callable[[Any], AbstractAsyncContextManager[None]], Callable[[Any], AbstractAsyncContextManager[Mapping[str, Any]]]]")  [assignment]
Found 1 error in 1 file (checked 67 source files)

But this has nothing to do with my patch!

Just noticed that this is the same reason why the online tests fail too, but I have no idea how to fix those...

Something like this? Note, locally some existing tests would still fail with:

+ venv/bin/mypy starlette tests
starlette/routing.py:638: error: Incompatible types in assignment (expression has type "Union[Callable[[Any], AbstractAsyncContextManager[None]], Callable[[Any], AbstractAsyncContextManager[Mapping[str, Any]]], None]", variable has type "Union[Callable[[Any], AbstractAsyncContextManager[None]], Callable[[Any], AbstractAsyncContextManager[Mapping[str, Any]]]]")  [assignment]
Found 1 error in 1 file (checked 67 source files)

But this has nothing to do with my patch!

It does. You added the conditional in a block of other elseifs.

I am not sure I understand why it still fails. :(

I see your edit of moving my if inside the previous if (makes sense) but still it behaves very weird...

I'm not sure I can help anymore here... :(

I now completely understand why so few projects have 100% test coverage. The amount of time I spend trying to make this trivial fix was close to 10 times more time in test authoring, rather than the actual code change itself (at least!!!).

I am not sure if you really persist and become more proficient how much more efficient this process ever becomes... So as much as I would like to also have 100% coverage in my projects I do not know if I would ever find the time to actually do it.

This is a task that I would not mind if some "testGPT" would ever take over! :)

I now completely understand why so few projects have 100% test coverage. The amount of time I spend trying to make this trivial fix was close to 10 times more time in test authoring, rather than the actual code change itself (at least!!!).

I am not sure if you really persist and become more proficient how much more efficient this process ever becomes... So as much as I would like to also have 100% coverage in my projects I do not know if I would ever find the time to actually do it.

This is a task that I would not mind if some "testGPT" would ever take over! :)

It is worth it, and it actually saves a lot of time. :)

You should have asked for feedback if you were feeling frustrated. I believe the interpretation of the coverage report on your behalf was not accurate.

You should have asked for feedback if you were feeling frustrated. I believe the interpretation of the coverage report on your behalf was not accurate.

I hope you mean the first time only, when I said "I'm not sure I can help anymore here... 😵‍💫 ", because after that I could not accept defeat and took a more serious look and finally changed the tests enough to pass the whole test suite. Give me some credit here! 😄