Bug report

Using the new buffersize parameter of Pool.imap() (or imap_unordered())
and then calling the documented close()/join() after only partially
consuming the result deadlocks forever:

import multiprocessing as mp

def f(x):
    return x

if __name__ == "__main__":
    p = mp.Pool(2)
    it = p.imap(f, range(1000), buffersize=2)
    next(it)        # consume one; the producer is now throttled
    p.close()
    p.join()        # hangs forever

terminate() works because it releases the buffersize semaphores; close()
does not, so the task-handler thread stays blocked in sema.acquire() inside
_guarded_task_generation and join() waits on it. buffersize is new in
3.16 (gh-64192).

CPython versions tested on:

3.16 (main)

Operating systems tested on:

Linux, macOS

Linked PRs

• gh-155477: Fix multiprocessing.Pool deadlock on close() with a buffersize imap #155478