[Python] Python multiprocessing: How to release an unknown lock?

This is a question about multiprocessing execution getting stuck in combindation with a third-party package (suppose we call it package). Let's pretend this is my code

def work(arg):
from package import f
return f(arg)

from multiprocessing import Pool
pool = Pool(8)
args = [ i for i in range(80) ]
results = pool.map(work, args)
for result in results:
result
pool.close()


In my particular case, this code runs fine until I use other functionality from package. If I modify the script like so:

def work(arg):
from package import f
return f(arg)

from package import SomeClass
obj = SomeClass()

from multiprocessing import Pool
pool = Pool(8)
args = [ i for i in range(80) ]
results = pool.map(work, args) # execution gets stuck on this line
for result in results:
result
pool.close()


then it gets stuck on the commented line. Since this is a third-party package, I have imperfect knowledge of its inner workings, but I see no reason (using a reasonable amount of time) that SomeClass should depend on what happens in work. Since I don't need the object obj when I get to the multiprocessing, I want to delete it (or otherwise render ineffective its interaction with f) hoping that the map can execute. But it doesn't help

def work(arg):
from package import f
return f(arg)

from package import SomeClass
obj = SomeClass()
obj = None
del obj
import gc
gc.collect()

from multiprocessing import Pool
pool = Pool(8)
args = [ i for i in range(80) ]
results = pool.map(work, args) # execution still gets stuck on this line
for result in results:
result
pool.close()

• Is there something I can do in reasonable time to find out what is causing map execution to get blocked?
• I want to use brute force to get rid of the lock (or whatever it is in SomeClass preventing execution) after am done with obj. What are my options to achieve that?

Edit: I noticed that I can get it to execute but not in parallel. So this worked:

def work(arg):
from package import f
return f(arg)

from package import SomeClass
obj = SomeClass()
obj = None
del obj
import gc
gc.collect()

from multiprocessing import Pool
pool = Pool(8)
arg = 42
work(arg) # execution work if I don't run it in parallel

Continue reading...