ROS2 service : spin_until_future_complete blocking when calling a service that calls another service [closed]

edit

Hello,

I'm using ROS2 Bouncy, built from source on Ubuntu 18.04.

I have a node which create a service A. The callback from this service A calls a function which calls another service B from another node. The service B does answer but the service A still get stuck on rclpy.spin_until_future_complete(self, future).

When I use the function to call the service B from the code it works, the problems comes when this function is called from a service. Is it how it's supposed to work ? Any suggestion on how I could call a service from another service without getting stuck ?

You can find the full file here : https://github.com/MarcTestier/kone_o...

You can run the full thing by running : ros2 run kone_open_opc_server_simu kone_simu then ros2 run kone_open_opc_client kone_client and then call the service : ros2 service call /send_lift_to_floor kone_open_opc_interfaces/SendLiftToFloor "site: 'a' location: 'b' group: 'c' lift: 'd' floor: 3 door_side: 1"

Here is just the interesting part of the function called by service A :

    req = AgvCmd.Request()
    req.id = self.agvID
    req.cmd = cmd
    req.value = 0

    self.read_agv_cmd = self.create_client(AgvCmd, 'read_agv_cmd')
    # Wait for the read_agv_cmd service to be running
    while not self.read_agv_cmd.wait_for_service(timeout_sec=1.0):
        self.get_logger().info('Service read_agv_cmd not available, waiting again...')

    future = self.read_agv_cmd.call_async(req)
    self.get_logger().info('OK 1')
    rclpy.spin_until_future_complete(self, future)
    self.get_logger().info('OK 2')

    if future.result() is not None:
        self.get_logger().info("AGV [%s] called the read service for command [%s] and heard [%d]" % (self.agvID, cmd, future.result().value))
        return 1, future.result().value

EDIT : same problem when trying to use :

self.read_agv_cmd = self.create_client(AgvCmd, 'read_agv_cmd')
while not self.read_agv_cmd.wait_for_service(timeout_sec=1.0):
    self.get_logger().info('Service read_agv_cmd not available, waiting again...')

future = self.read_agv_cmd.call_async(req)

while rclpy.ok():
    self.get_logger().info('OK 1')
    rclpy.spin_once(self)
    self.get_logger().info('OK 2')
    if future.done():
        if future.result() is not None:
            self.get_logger().info("AGV [%s] called the read service for command [%s] and heard [%d]" % (self.agvID, cmd, future.result().value))
            return 1, future.result().value

        else:
            self.get_logger().error('AGV [%s] tried to call the read service but failed %r' % (self.agvID, future.exception()))

It blocks on rclpy.spin_once(self).

EDIT 2 : Same thing when using self.read_agv_cmd.call(req)

EDIT 3: I tried to change Service A to a Subscriber and the result is the same, the callback stay stuck on rclpy.spin_until_future_complete(self, future)

EDIT 4: Actually, I get stuck when trying to rclpy.spin_once(self) in the callback function of a subscriber in the node.

EDIT 5: You can find a minimal example of the problem here : https://answers.ros.org/question/3020...