Could you remove the changes to Include/internal/pycore_object.h and Objects/typeobject.c?

Now that #27773 is merged, the tests should pass.

FYI: these tests are designed to run on normal debug builds. So they will fail the normal CI to save us the pain of having to catch this with buildbots.

I think I preferred sys,_get_type_version_tag approach.
We aren't just testing the C API here, sys._clear_type_cache is visible to Python code.

What might be worth having is a testcapi function to set the next version tag. Then we can test what happens when version tag overflows. No need to do that for this PR, though.

Sure. I threw it into testcapi because that's what the ma_version_tag tests used, but I'm fine with both sides for/against placing it in sys.

@markshannon: "I think I preferred sys,_get_type_version_tag approach."

Ah, you prefer to add a private function in the sys module? What would be the usage outside unit tests? If it's only for testing/debugging purpose, IMO _testcapi is more appropriate.

Looks good. Thanks.

When you close/reopen a PR, can you please mention which tests of which CI failed please? If some tests are failing randomly, they should be fixed. Everyone is annoyed by such annoying tests.

On the merged commit, I see that test_tools_script_run_tests() of test_regrtest failed on Windows (x64). This CI is no longer mandatory?

When you close/reopen a PR, can you please mention which tests of which CI failed please? If some tests are failing randomly, they should be fixed. Everyone is annoyed by such annoying tests.

Got it, I'll do so in the future. I didn't open a bug on bpo this time because it's the first time I've seen this test fail like this

0:21:19 load avg: 0.02 [428/428/3] test_regrtest crashed (Exit code 1)
Timeout (0:20:00)!
Thread 0x00000d2c (most recent call first):
  File "D:\a\cpython\cpython\lib\subprocess.py", line 1493 in _readerthread
  File "D:\a\cpython\cpython\lib\threading.py", line 946 in run
  File "D:\a\cpython\cpython\lib\threading.py", line 1009 in _bootstrap_inner
  File "D:\a\cpython\cpython\lib\threading.py", line 966 in _bootstrap

so I thought it's just a rare one-off event (unlike test_asyncio, test_concurrent_futures etc.). If I see it a second time, I'll open a bug.

On the merged commit, I see that test_tools_script_run_tests() of test_regrtest failed on Windows (x64). This CI is no longer mandatory?

Sorry, could you please give me the link? I'm not sure which one you're referring to. I see the Windows buildbot is green, and so is the GH actions

When I check on [Show details] on the merged commit line, I see that Windows (x64) failed:
https://github.com/python/cpython/runs/3342514542

I created:

• https://bugs.python.org/issue44936: test_concurrent_futures: test_cancel_futures_wait_false() and test_interpreter_shutdown() failed on GHA Windows x64
• https://bugs.python.org/issue44937: test_regrest: test_tools_script_run_tests() failed on GHA Windows x64

The test_asyncio failure is known: https://bugs.python.org/issue41682#msg399770