Python 3.6 does not support asyncio.get_event_loop().add_signal_handler and async_client.py spams about it on every connect attempt

[bialix]

engineio/asyncio_client.py has the following code at the very start of connect:

        global async_signal_handler_set
        if self.handle_sigint and not async_signal_handler_set and \
                threading.current_thread() == threading.main_thread():
            try:
                asyncio.get_event_loop().add_signal_handler(
                    signal.SIGINT, async_signal_handler)
                async_signal_handler_set = True
            except NotImplementedError:  # pragma: no cover
                self.logger.warning('Signal handler is unsupported')

The problem here is that variable async_signal_handler_set will never be set to True in this case and logger is spamming my logs with useless warning.

May I suggest to rename async_signal_handler_set to async_signal_handler_once and set it regardless of exception raised to avoid it triggered next time? I can make PR if you agree.

[miguelgrinberg]

Yes, your suggestion makes sense. Thanks. If you create the PR I'll review and merge.

[bialix]

The same problem with Python 3.10 @ Windows and I'm not sure why. Maybe it's a asyncio bug on Windows.

[miguelgrinberg]

@bialix I believe there is no support for signal handlers in the Windows event loop, so there is currently no way to exit cleanly on that platform.

[bialix]

Yep, I think you're right.

So, should I compose a patch or you do it yourself?

[bialix]

OK, that's definitely problem with your signal_handler.
Once you set it - your handler overrides my own SIGINT handler. And I'm not sure it's desired side-effect.

[miguelgrinberg]

I'm going to close this issue, since you have now moved the discussion to #277.

[bialix]

#277 is about disconnect method. This issue is about spaming with warnings. I don't understand, but it's up to you to close issues, that's your project.

[miguelgrinberg]

I settled the discussion on the multiple attempts to set a signal in the 2nd message in this issue:

Yes, your suggestion makes sense. Thanks. If you create the PR I'll review and merge.

Everything else in this issue is unrelated to that problem. In any case, I just submitted the fix for the original issue you reported here. Now the signal will be set the first time you connect only.