[3.11]gh-127213: Replace os._exit() to sys.exit(), so that .so will alos be unloaded in child process

[Yi-sir]

#127213

As mentioned in this issue, a child process created by multiprocessing.Process() with method 'fork' works not the same as os.fork() or c fork() on unloading of .so files.

replace os._exit() to sys.exit() to fix it.

• Issue: multiprocessing.Process() with method fork() works not the same with os.fork() on unloading of .so files #127213

[ghost]

All commit authors signed the Contributor License Agreement.

[bedevere-app]

Most changes to Python require a NEWS entry. Add one using the blurb_it web app or the blurb command-line tool.

If this change has little impact on Python users, wait for a maintainer to apply the skip news label instead.

[oskar-skog]

I believe os._exit is the intended behaviour here.

sys.exit will raise a SystemExit that may be caught in code that is only expected to run in the parent.

[Yi-sir]

I believe os._exit is the intended behaviour here.

sys.exit will raise a SystemExit that may be caught in code that is only expected to run in the parent.

But os._exit() will not unload .so in child process, which would cause some error.

[ZeroIntensity]

Sorry, 3.11 doesn't take bugfixes anymore. I'm closing this because trying to switch the target branch to main will be a headache, and we might accidentially trigger a codeowner review from half the core team (that tends to happen when switching the target branch). Please fix the issue for the main branch, and then open a new PR, we'll deal with backporting this to 3.13 and 3.12 automatically after it lands on main.