Tunnel thread leakage with Python 3.6

[kkrasovskii]

If we try to establish SSH tunnel with connectivity checking to some service that is unavailable we will get thread leaking.

Here is the thing:

When sshtunnel.start() function is called it creates new thread that uses socketserver package serve_forever function as infinite loop that can be broken by setting special flag - __shutdown_request. This flag is set by sshtunnel module when we call it stop() function. The stop function sets __shutdown_request flag in the case when the tunnel_is_up flag are set. If we check remote service availability and it's not available the tunnel_is_up flag is in False state and early created thread is not closed by stop method. The problem can be easily fixed this way:

    def _stop_transport(self):
        """ Close the underlying transport when nothing more is needed """
        try:
            self._check_is_started()
        except (BaseSSHTunnelForwarderError,
                HandlerSSHTunnelForwarderError) as e:
            self.logger.warning(e)
        for _srv in self._server_list:
            tunnel = _srv.local_address
            if self.tunnel_is_up[tunnel]:
                self.logger.info('Shutting down tunnel {0}'.format(tunnel))
            _srv.shutdown()
            _srv.server_close()
            # clean up the UNIX domain socket if we're using one
            if isinstance(_srv, _UnixStreamForwardServer):
                try:
                    os.unlink(_srv.local_address)
                except Exception as e:
                    self.logger.error('Unable to unlink socket {0}: {1}'
                                      .format(self.local_address, repr(e)))

The problem can be easily reproduced by a code in the file:

thread_leaks_example.txt

[eddie-chiang]

The issue is still happening on v0.1.5, with Python 3.7.7.