test_monitoring fails when run multiple times: ValueError: tool 0 is already in use

[vstinner]

The issue broke all Refleaks buildbots :-(

Example:

$ ./python -m test test_monitoring -R 3:3
0:00:00 load avg: 3.50 Run 1 test sequentially
0:00:00 load avg: 3.50 [1/1] test_monitoring
beginning 6 repetitions
123456
.test test_monitoring failed -- Traceback (most recent call last):
  File "/home/vstinner/python/main/Lib/test/test_monitoring.py", line 1723, in test_gh108976
    sys.monitoring.use_tool_id(0, "test")
ValueError: tool 0 is already in use

test_monitoring failed (1 error)

== Tests result: FAILURE ==

1 test failed:
    test_monitoring

Total duration: 454 ms
Total tests: run=63
Total test files: run=1/1 failed=1
Result: FAILURE

The regression was introduced by: commit 4a69301 of PR GH-109131. cc @markshannon

commit 4a69301ea4539da172a00a80e78c07e9b41c1f8e
Author: Mark Shannon <mark@hotpy.org>
Date:   Mon Sep 11 14:37:09 2023 +0100

    GH-108976. Keep monitoring data structures valid during de-optimization during callback. (GH-109131)

Python 3.12 is also affected, the change was backported to 3.12: cc @Yhg1s

Linked PRs

• gh-109357: Fix test_monitoring.test_gh108976() #109358
• [3.12] gh-109357: Fix test_monitoring.test_gh108976() (GH-109358) #109359

[vstinner]

Oh wait, it's simple to fix this issue in fact: I wrote PR #109358.