Expected python bug using multiprocessing and inheritance

This is a bug report:

import time
import multiprocessing as mp
class A(object):
    def __init__(self, childnodes, **kwargs) -> None:
        self._childnodes = childnodes
        self._kwargs = kwargs

    def process(self, datablock, meta):
        print("Hello")
        time.sleep(1)
        print("world")

class B(A):
    def process(self, datablock, meta):
        print("before")
        mp.Process(target=super().process, args=(self, datablock, meta)).start()
        print("after")

if __name__ == "__main__":
    b = B([1,2,3])
    b.process([4,5,6], {'foo':42})

This seems to run forerever which is not expected. Reason seems to be that that somehow, super().process actually calls B.process instead of A.process in the subprocess. This is not expected behaviour. As additional evidence towards that, the program behaviour can be corrected by changing

mp.Process(target=super().process, args=(self, datablock, meta)).start()`meta)).start()

with

mp.Process(target=A.process, args=(self, datablock, meta)).start()

Machine:
MacBook Pro (16-inch, 2021)
MacOS Monteray 12.5
Chip: Apple M1 Max
Python version: 3.9.12

Commencez par exécuter le reproducteur fourni avec Python 3.9.12 et retracez la manière dont multiprocessing gère la cible liée super().process dans le processus enfant. Comparez cela avec la cible explicite A.process ; le travail est considéré comme terminé lorsque l’appel à la superclasse ne récursive plus, tout en conservant le comportement attendu de multiprocessing.