Okay, this is functional now, code snippet:

import os
def before():
   print("before")

def parent():
   print("parent")
def child():
   print("Child")

os.register_at_fork(before=before, after_in_child=child, after_in_parent=parent)
os.fork()

in rustpython

>>>>> import os
>>>>> def before():
.....   print("before")
.....
>>>>> def parent():
.....   print("parent")
.....
>>>>> def child():
.....   print("Child")
.....
>>>>> os.register_at_fork(before=before, after_in_child=child, after_in_parent=parent)
>>>>> os.fork()
before
parent
26312
>>>>> Child
0

In cpython ( 3.12.0a7+)

>>> import os
>>> def child():
...     print("child")
...
>>> def parent():
...     print("parent")
...
>>> def before():
...     print("before")
...
>>> os.register_at_fork(before=before, after_in_child=child, after_in_parent=parent)
>>> os.fork()
before
parent
26349
child
>>> 0

wrt to error handling:

rustpython

>>>>> def before():
.....   print(x)
.....
>>>>> import os
>>>>> os.register_at_fork(before=before)
>>>>> os.fork()
Exception ignored in: <function before at 0x558949c1cbe0>
  File "<stdin>", line 1, in <module>
NameError: name 'x' is not defined
33503
>>>>> 0

cpython

>>> import os
>>> def before():
...     print(x)
...
>>> os.register_at_fork(before=before)
>>> os.fork()
Exception ignored in: <function before at 0x7f4dbf33c260>
Traceback (most recent call last):
  File "<stdin>", line 2, in before
NameError: name 'x' is not defined
29846
0

Some obvious review changes are needed for:

1. register_at_fork code
2. Obvious issues
3. Fix for weird ps (>>)
4. Random warning (warning: value assigned to pid is never read, I am using this)

Pending: test cases

1. test_fork_warns_when_non_python_thread_exists needs _testcapi
2. Unexpected successes

UNEXPECTED SUCCESS: test_waitpid (test.test_os.PidTests.test_waitpid)
UNEXPECTED SUCCESS: test_waitstatus_to_exitcode (test.test_os.PidTests.test_waitstatus_to_exitcode)
UNEXPECTED SUCCESS: test_waitstatus_to_exitcode_kill (test.test_os.PidTests.test_waitstatus_to_exitcode_kill)

in test_waitpid, it shows # TODO: RUSTPYTHON (AttributeError: module 'os' has no attribute 'spawnv')
but spawnv exists in os (not sure what I am missing, will check this).

Nice start! Some initial things so we get a better picture.

1. Remember to run cargo fmt and clippy to get some easy formatting/lint issues out of the way.
2. Many of the failures are due to the fact that fork does not exist on that platform. You'll need to restrict it to platforms that do with some cfgs (go for a #[cfg(unix)] for starters)
3. If tests are passing, we'll need to mark them as failing only for the platforms that won't have fork (currently, this means don't run them on windows, I remember there was a helper added for this specifically recently, see expectedFailureIfWindows and Add unittset.expectedFailureIf for RustPython  #4597 for usages of it).

I suspect something in test_httpservers.py might be hanging.

The error is:

Thanks for looking this up. I knew that would be error but couldn't see it and hence couldn't add skip tag.

The error seems to be hidden probably because it is printed in forked process.

Yes, but I don't know why it didn't got captured at output

while True:
                try:
                    chunk = os.read(fd, 3000)
                except OSError:  # Assume EIO
                    break
                if not chunk:
                    break

I even added sleep(60) at child so that I could send all chunk of traceback. Probably, just UI issue on github action as it worked fine for you.

I have made required changes (adding missing XXX RustPython and using #[derive(FromArgs)])
I used match and optional arg but still I couldn't found clean way to raise type error if none of arg are matched (although cpython didn't had test case for that so opened PR python/cpython#103725 so that's win)

Made required changes, optional arg is working fine as well.

os.register_at_fork(a=8, before=print)

working without any issue

Thank you! You made a big step on fork. I wonder how many more features needed for gunicorn.

Thanks a lot for review and helping me with this PR.
As you could have guessed, it now fails with AttributeError: 'RLock' object has no attribute '_at_fork_reinit' and few more errors.
Its would need some more work to get it running.