fix: Fix cache pollution when copying mutable references

[cmanallen]

• Removes manual overrides of copy behavior and leaves it up to the caller.
  • E.g. a future use case may require a non-deepcopy. If we override copy they would have to remove the dunder copy, update every implementation which relies copy, before finally creating their own copy implementation.
• Deepcopies the flag buffer.
  • Though we do not cache mutable references yet we may soon and so this foot gun should be removed from possibility.
• Removes "copy" test coverage from test_lru_cache.py. We're no longer assuming copy usage and leave it up to the caller.
  • The existing test in tests/test_scope.py covers the cache pollution case originally mentioned here.
  • The mutable cache pollution case is not covered because we do not currently cache mutable objects.

In general a generic class should assume as few implementation details as possible. If we leave the existing copy method someone may assume copy semantics and rely on it in a way that is inappropriate.

Closes: #3886

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
13808	1	13807	4140

View the top 1 failed tests by shortest run time

tests.integrations.launchdarkly.test_launchdarkly test_launchdarkly_integration_threaded

Stack Traces | 0.111s run time

.../integrations/launchdarkly/test_launchdarkly.py:90: in test_launchdarkly_integration_threaded
    assert len(events) == 3
E   AssertionError: assert 4 == 3
E    +  where 4 = len([{'breadcrumbs': {'values': []}, 'contexts': {'runtime': {'build': '3.12.8 (main, Dec  4 2024, 06:20:10) [GCC 9.4.0]', 'name': 'CPython', 'version': '3.12.8'}, 'trace': {'parent_span_id': None, 'span_id': '932f498e64d5c9af', 'trace_id': 'be5c9d7f646540af80bd45541a3516e9'}}, 'environment': 'production', 'event_id': 'bcb716664ee741cc94df69e5e8c60dc9', ...}, {'breadcrumbs': {'values': []}, 'contexts': {'flags': {'values': [{'flag': 'hello', 'result': True}, {'flag': 'world', 'result': False}]}, 'runtime': {'build': '3.12.8 (main, Dec  4 2024, 06:20:10) [GCC 9.4.0]', 'name': 'CPython', 'version': '3.12.8'}, 'trace': {'parent_span_id': None, 'span_id': '89ed637b6b19fee0', 'trace_id': '0a88fb77d3444306942349fd938e2d95'}}, 'environment': 'production', 'event_id': '436339a8fc0747d0988122aaaacb2e47', ...}, {'breadcrumbs': {'values': []}, 'contexts': {'flags': {'values': [{'flag': 'hello', 'result': True}, {'flag': 'other', 'result': False}]}, 'runtime': {'build': '3.12.8 (main, Dec  4 2024, 06:20:10) [GCC 9.4.0]', 'name': 'CPython', 'version': '3.12.8'}, 'trace': {'parent_span_id': None, 'span_id': '89ed637b6b19fee0', 'trace_id': '0a88fb77d3444306942349fd938e2d95'}}, 'environment': 'production', 'event_id': '58300f9a104549b2948a66da0deb53c8', ...}, {'breadcrumbs': {'values': []}, 'contexts': {'flags': {'values': [{'flag': 'hello', 'result': True}]}, 'runtime': {'build': '3.12.8 (main, Dec  4 2024, 06:20:10) [GCC 9.4.0]', 'name': 'CPython', 'version': '3.12.8'}, 'trace': {'parent_span_id': None, 'span_id': '89ed637b6b19fee0', 'trace_id': '0a88fb77d3444306942349fd938e2d95'}}, 'environment': 'production', 'event_id': 'c2cdce18f16849b9a58c50c91735d447', ...}, {'breadcrumbs': {'values': []}, 'contexts': {'runtime': {'build': '3.12.8 (main, Dec  4 2024, 06:20:10) [GCC 9.4.0]', 'name': 'CPython', 'version': '3.12.8'}, 'trace': {'parent_span_id': None, 'span_id': '932f498e64d5c9af', 'trace_id': 'be5c9d7f646540af80bd45541a3516e9'}}, 'environment': 'production', 'event_id': '791ff1b6df5f4becb1903d3c0acea0af', ...}])

To view more test analytics, go to the Test Analytics Dashboard
📢 Thoughts on this report? Let us know!

[antonpirker]

Sounds good to me. Not changing the copy behavior and leaving it for the caller to do the right thing makes sense.