Children Failed to Join

[ruidan-li]

Hello!

I am currently using the kinesis-python (https://github.com/NerdWalletOSS/kinesis-python) library, which use your offspring library to have multiple shard readers, and it turns out that the terminated children fail to join sometime. My main process will do some stuff and call sys.exit() when received SIGINT or SIGTERM, and based on the log, it shows "Caught signal 15" (https://github.com/borgstrom/offspring/blob/master/src/offspring/process.py#L112). And then self.end() is called and so is sys.exit() in the run(). However, the children sometimes never join, and after tracing, it is stuck at os.waitpid() (in multiprocessing/forking.py). Trying to figure out what is going on, I placed the sys.exit() in the signal_handler and it works. So I am wondering if it is possible to refactor the SubprocessLoop.run and the signal_handler, to place self.end() and sys.exit() inside the signal_hander instead. I would also be happy to hear your thoughts on this issue!

Thanks,
Ruidan

[Sytten]

Hum I spent 4h today to arrive at the exact same conclusion. This library is broken.