Deadlock when limit_pending is reached

[hlecnt]

Based on the example code from the README, the following script leads into a deadlock and never ends:

async def coro(id, timeout):
    print(f"Start coro #{id} with timeout {timeout}", flush=True)
    await asyncio.sleep(timeout)
    print(f"End coro #{id} with timeout {timeout}", flush=True)


async def main():
    scheduler = await aiojobs.create_scheduler(limit=1, pending_limit=1)
    for i in range(3):
        # spawn jobs
        await scheduler.spawn(coro(i, 2))
        print(f"Coro #{i} scheduled.", flush=True)

    print("All coro are scheduled.", flush=True)
    await asyncio.sleep(10.0)
    # not all scheduled jobs are finished at the moment

    # gracefully close spawned jobs
    print("Waiting for pending jobs.", flush=True)
    await scheduler.close()


asyncio.get_event_loop().run_until_complete(main()) 

Running the script gives:

Coro #0 scheduled.
Coro #1 scheduled.
Start coro #0 with timeout 2
End coro #0 with timeout 2

=> stuck here, Coro #3 is never scheduled, Coro #2 is scheduled but not run.

Testing env:

• Python 3.6.12 + aiojobs 0.3.0
• Python 3.812 + aiojobs 1.0.0

[hlecnt]

@asvetlov please find a pull request for this issue:
#297

[pollydrag]

Fixed by #135