chardet 5.0 KeyError with Python 3.10 on Windows

[cas--]

Yesterday I encountered a strange CI failure for our Windows GitHub CI workflows which had been running fine until then. The Python 3.7 job passed fine but the Python 3.10 job failed.

https://github.com/deluge-torrent/deluge/actions/workflows/ci.yml?query=branch%3Adevelop

The only difference I could find from a diff of the logs was the new chardet 5.0.0 being pulled in. So I pinned chardet to 4.0.0 and CI is passing again.

GitHub Actions Environment:

Virtual Environment: windows-2022 (20220626.1)
Python 3.10.5

Just to note that I also tested same error occurs with windows-2019.

The traceback is rather cryptic since it comes from pytest but this is all there is from the job:

INTERNALERROR> Traceback (most recent call last):
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\_pytest\main.py", line 264, in wrap_session
INTERNALERROR>     config._do_configure()
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\_pytest\config\__init__.py", line 995, in _do_configure
INTERNALERROR>     self.hook.pytest_configure.call_historic(kwargs=dict(config=self))
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\pluggy\_hooks.py", line 277, in call_historic
INTERNALERROR>     res = self._hookexec(self.name, self.get_hookimpls(), kwargs, False)
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\pluggy\_manager.py", line 80, in _hookexec
INTERNALERROR>     return self._inner_hookexec(hook_name, methods, kwargs, firstresult)
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\pluggy\_callers.py", line 60, in _multicall
INTERNALERROR>     return outcome.get_result()
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\pluggy\_result.py", line 60, in get_result
INTERNALERROR>     raise ex[1].with_traceback(ex[2])
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\pluggy\_callers.py", line 39, in _multicall
INTERNALERROR>     res = hook_impl.function(*args)
INTERNALERROR>   File "C:\hostedtoolcache\windows\Python\3.10.5\x64\lib\site-packages\_pytest\faulthandler.py", line 27, in pytest_configure
INTERNALERROR>     import faulthandler
INTERNALERROR>   File "<frozen importlib._bootstrap>", line 1024, in _find_and_load
INTERNALERROR>   File "<frozen importlib._bootstrap>", line 171, in __enter__
INTERNALERROR>   File "<frozen importlib._bootstrap>", line 123, in acquire
INTERNALERROR> KeyError: 1832

[dan-blanchard]

Well, that's a baffling one. The traceback doesn't seem to be related to chardet at all

[cas--]

It is rather confounding but I have narrowed it down to this commit: 3222295

And looks to be related to importing the module rather than usage, as suspected from pytest traceback, and I'll continue to investigate when I have a spare moment

Testing with 3.8 and 3.9 didn't show any issues, just 3.10 affected

As an aside the KeyError is a result of the pytest faulthandler calls rather than originating error, as mentioned here with pdb: https://bugs.python.org/issue34609#msg326674

[jayaddison]

(:wave: I landed here as part of reading the state of any bugs related to chardet v5.0.0)

Although I'm not 100% certain, it looks like this could be related to the issue described in python/cpython#/91351 -- the lock-acquisition methods named in the stacktrace are the same, and so is the exception class raised.

[dan-blanchard]

I added Mac and Windows testing to our build matrix in #267 and I can't reproduce this issue.

[jayaddison]

The (likely intermittent) issue that appeared here has, I think, been resolved by gh-91351 in v3.12a5 of Python. It's still a while until the scheduled release of Py3.12 final -- to mention nothing of CI config for chardet -- so maybe it's a distraction to even mention - but I get the sense that this has been resolved.