UnboundLocalError: local variable 'self' referenced before assignment

[sentry-for-tribler]

Sentry Issue: TRIBLER-1JA

UnboundLocalError: local variable 'self' referenced before assignment
(1 additional frame(s) were not displayed)
...
  File "ipv8\taskmanager.py", line 131, in done_cb
  File "ipv8\taskmanager.py", line 18, in interval_runner
  File "tribler\core\components\libtorrent\download_manager\download_manager.py", line 828, in _invoke_states_cb
  File "tribler\core\components\libtorrent\download_manager\download_manager.py", line 839, in sesscb_states_callback
  File "tribler\core\components\libtorrent\torrentdef.py", line 236, in get_infohash

[drew2a]

Yesterday, @kozlovsky expressed suspicion that the issue could be related to the old Python version (3.8.4) used on our builder machines.

Perhaps it's a good time to upgrade it to 3.9?

@xoriole, is it easy to change the Python version on our builders? I can manage the change for GitHub actions.

Ref:

• https://stackoverflow.com/questions/51964883/keras-and-tensorflow-unboundlocalerror-local-variable-self-referenced-before

[xoriole]

@drew2a I can try upgrading the builders today to 3.9.
For 3.10, we have a blocker.

[sentry-for-tribler]

Sentry issue: TRIBLER-1JT for 7.13.1

[
Traceback (most recent call last):, 
  File "asyncio\events.py", line 81, in _run, 
  File "asyncio\proactor_events.py", line 162, in _call_connection_lost, 
OSError: [WinError 10038] Сделана попытка выполнить операцию на объекте, не являющемся сокетом
]

[kozlovsky]

The last linked Sentry issue shows another unrelated case of UnboundLocalError:

UnboundLocalError: local variable 'exc_info' referenced before assignment
  File "tribler\core\components\libtorrent\download_manager\download.py", line 248, in process_alert
  File "tribler\core\components\libtorrent\download_manager\download.py", line 325, in on_tracker_error_alert
  File "__init__.py", line 1463, in error
  File "__init__.py", line 1570, in _log

Here, the error happens inside the code of the logging library:

    def _log(self, level, msg, args, exc_info=None, extra=None, stack_info=False,
             stacklevel=1):
        """
        Low-level logging routine which creates a LogRecord and then calls
        all the handlers of this logger to handle the record.
        """
        sinfo = None
        if _srcfile:
            #IronPython doesn't track Python frames, so findCaller raises an
            #exception on some versions of IronPython. We trap it here so that
            #IronPython can use logging.
            try:
                fn, lno, func, sinfo = self.findCaller(stack_info, stacklevel)
            except ValueError: # pragma: no cover
                fn, lno, func = "(unknown file)", 0, "(unknown function)"
        else: # pragma: no cover
            fn, lno, func = "(unknown file)", 0, "(unknown function)"
        if exc_info:  # <-- UnboundLocalError: local variable 'exc_info' referenced before assignment

The only reasonable explanation of how this can happen is memory corruption by a misbehaving C++ code.

Note that the error is inside the process_alert call. That looks suspicious because incorrect handling of ltsession.pop_alerts() is a possible reason for memory corruption (see #7886).

[kozlovsky]

Another error:

SystemError: PyEval_EvalCodeEx: NULL globals
  File "asyncio\proactor_events.py", line 526, in _loop_writing
  File "asyncio\windows_events.py", line 522, in sendto
  File "asyncio\windows_events.py", line 734, in _register

Traceback is different, but the underlying reason should be the same: memory corruption

[kozlovsky]

There is a chance #7945 can fix this problem. If not, then it is most probably libtorrent-related. My hope is that upgrading to Python 3.10, that is now possible, can fix the root reason of memory corruption problems.