test_pickle are leaked on Windows

[Eclips4]

Tried on current main branch.

./python -m test -R 3:3 test_pickle
Running Debug|x64 interpreter...
0:00:00 Run tests sequentially
0:00:00 [1/1] test_pickle
beginning 6 repetitions
123456
......
test_pickle leaked [13, 12, 173] references, sum=198
test_pickle leaked [11, 12, 113] memory blocks, sum=136
test_pickle failed (reference leak) in 4 min 30 sec

== Tests result: FAILURE ==

1 test failed:
    test_pickle

Total duration: 4 min 30 sec
Tests result: FAILURE

Also, I cannot reproduce it in Linux (WSL, Ubuntu 20.04)

Linked PRs

• gh-106176: Fix reference leak in importlib/_bootstrap.py #106207
• GH-106176, GH-104702: Fix reference leak when importing across multiple threads #108497
• [3.12] GH-106176, GH-104702: Fix reference leak when importing across multiple threads (GH-108497) #108612

[Eclips4]

Fantastic... I can't reproduce it anymore. I just try found bad commit using git bisect...
But I've found reference leak in another way.

 ./python -m test -R 3:3 test_pickle -m PyUnpicklerTests
Running Debug|x64 interpreter...
0:00:00 Run tests sequentially
0:00:00 [1/1] test_pickle
beginning 6 repetitions
123456
......
test_pickle leaked [4, 4, 4] references, sum=12
test_pickle leaked [4, 4, 4] memory blocks, sum=12
test_pickle failed (reference leak)

== Tests result: FAILURE ==

1 test failed:
    test_pickle

Total duration: 908 ms
Tests result: FAILURE

[Eclips4]

First bad commit: #94504
Also, I know about some refleaks in test_import. (#102251)
What is the relationship with this issue and test_import issue?
Seems that both introduced in #94504 PR.

[sunmy2019]

Seems all ref leaks come from test_unpickle_module_race

If this test is removed from AbstractUnpickleTests, all ref leaks are gone.