This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author terry.reedy
Recipients corona10, ethan.furman, serhiy.storchaka, sobolevn, terry.reedy, zach.ware
Date 2022-01-18.20:31:30
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <>
Nikita, thanks for doing this.

Ethan's suggestion has two advantages. 1. It would fix most backport issues.  By containing a mix of changes, PR-30666 cannot be backported as is.  'unittest.main' should always backport unless there is a context conflict.  But 3.9 does not have os.helper and TESTFN is still in in 3.9 and that change cannot backport.  (I did not check 3.10.)

2. It restricts the knowledge needed to review (and be willing to merge).  I could and would review and merge a patch with only unittest.main changes, but not, without checking the docs to learn about new changes, the mixture you posted.

I suggest you write a script to directly run each test.test_xyz file via subprocess and report to stdout.  The test for this issue is a clean report.  If you do so, I will run it on Windows in a fresh build of each of 3.9, 3.10, and 3.11.

Note that there are two possible commands: 'python -m test test_xyz' and 'python -m test.test_xyz'.  They are subtly different.  So I would add an option to do it each way.

With current 3.11, test_importlib runs fine on my Windows, with 1436 tests.
Date User Action Args
2022-01-18 20:31:30terry.reedysetrecipients: + terry.reedy, ethan.furman, zach.ware, serhiy.storchaka, corona10, sobolevn
2022-01-18 20:31:30terry.reedysetmessageid: <>
2022-01-18 20:31:30terry.reedylinkissue46425 messages
2022-01-18 20:31:30terry.reedycreate