GeezerVison

Skipping news since this is purely a fix to CPython's own test suite.

@Yhg1s how would you feel about backporting this test-suite-only fix to 3.12, to reduce spurious failures in 3.12 if anyone runs the 3.12 tests with a code watcher active, and reduce conflicts on future backports in these tests?

in what scenario is this test run with other code watchers registered?

in what scenario is this test run with other code watchers registered?

In practice today, probably it isn't. I found this bug when backporting code watchers to Cinder 3.10 so that the Cinder JIT can use them; in that case the Cinder JIT installs a code watcher, and the test suite should pass with the JIT active. In future, in principle, any Python optimizer or debug/monitoring tool that installs a code watcher may want to validate that the Python test suite passes with it active, and I think the tests should work in such cases. (The Cinder JIT for Python 3.12 will certainly be one such case.)

There is the possible edge case in which someone tries to run these tests with more than six code watchers active, in which case they'd fail because there aren't enough code watcher slots left. I doubt that case will ever be encountered unless artificially constructed, but a possible safeguard could be to skip the tests if too many code watchers are already installed.

This makes sense to me to backport to 3.12. I think it's worth realising that the Python test suite does get run in situations different from the default one. (At work we run the whole testsuite with an embedded interpreter and custom importers, for example, and I could totally see us adding something that uses watchers for something or other.)

Thanks @carljm for the PR 🌮🎉.. I'm working now to backport this PR to: 3.12.
🐍🍒⛏🤖 I'm not a witch! I'm not a witch!

GH-107835 is a backport of this pull request to the 3.12 branch.