[forward compat?] Support Python 3.10

[webknjaz]

Describe the bug

I've recently integrated proxy.py into the test suite of aiohttp. Now that we're trying to add Python 3.10 to the CI, the tests that interact with an instance of proxy.py get a TLS handshake timeout.

I haven't looked into it but my guess is that proxy.py may be incompatible with Python 3.10.

To Reproduce

Steps to reproduce the behavior:

1. Clone aio-libs/aiohttp
2. Run make vvtest

Expected behavior

It should work.

Version information

• OS: Ubuntu with Python 3.10 activated
• Browser: N/A
• Device: GHA CI/CD
• proxy.py Version 2.3.1

Additional context

aio-libs/aiohttp#5927

Screenshots

N/A

[abhinavsingh]

Thank you @webknjaz for trying it on Python 3.10. I haven't yet got my hands on 3.10. I'll try to jump on it in coming week and see how it goes.

[abhinavsingh]

@webknjaz I updated to 3.10 today. All looks good locally, tests are passing and I am able to post this comment (all via proxy.py running on 3.10). I have also added official support for 3.10 in proxy.py CI.

I looked into the linked issue and IIUC tests at your end are seeing ConnectionAbortedError: SSL handshake is taking longer than 60.0 seconds: aborting the connection error. I am unsure if this is linked to proxy.py, but we'll need to investigate deeply to confirm anything.

I'll give make vvtest a try and see if I can find anything useful.

[webknjaz]

Thanks, that should be helpful! I haven't had time to look into that specific error myself but if I'll be able to check it out, I'll report back here too.

[abhinavsingh]

@webknjaz Should we look back into it within aiohttp or do we need to get into proxy.py deeper. In general, proxy.py seems to work fine. I am now using it for more than a month on Python 3.10. Wdyt?

[webknjaz]

Well, I'll only be able to check it in that CI once you make a release.

[abhinavsingh]

@webknjaz Wondering if it's ok to resolve this one now.

Probably will mark as resolve next time I come around this. We can always reopen iff necessary :)

Once again thank you for all your help/contributions/support as we went around this.

I hope all is well at your end. Best.

[webknjaz]

It seems like this is still happening with v2.4.3. And I'm seeing this problem under older CPython versions on Windows too: https://github.com/aio-libs/aiohttp/runs/8199178923?check_suite_focus=true#step:10:4910.

[abhinavsingh]

Thanks for cross verifying. From first look, seems like issue with sharing fd between processes. I'll give it a try once locally on a Windows system and see if I can reproduce it. Will update back.

[webknjaz]

Thanks! It also appears to look like #1000.

[webknjaz]

@abhinavsingh have you had a chance to take a look? We have the tests marked as expected to fail currently but it'd be nice to turn that expectation to success.

[abhinavsingh]

Hi @webknjaz , Its been a busy week. I'll try to pull up sometime over the coming weekend. Best.