Python3.8 and above changed the behavior of the multiprocessing module on macOS, resulting in variable scoping being tighter than on other platforms or versions.

As a simplified case, this code crashes on such a machine:

import multiprocessing

def worker(local):
	print(v)

if __name__ == "__main__":
	v = "can i access this?"
	with multiprocessing.Pool() as p:
		devices = p.map(worker, [None])

This patch passes in the relevant variable into the worker in a different style to fix. I've tested this on linux and Mac platforms for regressions.