ROS 2 pytest publisher "Destroyable" error

edit

Ubuntu 22.04, ROS 2 Humble, (Cyclone DDS to avoid Nav2 issue)

I am working on a unit test for python class which creates a publisher and publishes a message. I'm running the test via colcon test from a Python unittest.TestCasederived class

I am creating an rclpy context, executor, and node as part of the test setUp, then creating the publisher inside the python code being tested. I verify the node is created properly:

[INFO] [1686855160.909732000] [core_test_7]:  set up core test 7 (1)... 
[WARN] [1686855160.912112918] [core_test_7]:     Node:{<rclpy.node.Node object at 0x7f5da52b56f0>}
[WARN] [1686855160.912704474] [core_test_7]:     Context :{<rclpy.context.Context object at 0x7f5da506b9a0>} (1)

But on attempting to publish, I get an exception cannot use Destroyable because ... :

[WARN] [1686855163.543058631] [core_test_7]: Something went wrong when publishing t!
<class 'rclpy._rclpy_pybind11.InvalidHandle'>: cannot use Destroyable because destruction was requested
[WARN] [1686855163.544163586] [core_test_7]:     Publisher:{<rclpy.publisher.Publisher object at 0x7f5da52b4b20>}
[WARN] [1686855163.545204206] [core_test_7]:     Node:{<rclpy.node.Node object at 0x7f5da52b56f0>} 
[WARN] [1686855163.546243544] [core_test_7]:     Context:{<rclpy.context.Context object at 0x7f5da506b9a0>} (1)

The (1) at end is the output of context.ok() method.

There is no explicit call to destroy the node or publisher before calling publish on the publisher instance.

When I tearDown the node, context, and executor seem OK

INFO] [1686855183.718462381] [core_test_7]:  shutting down core test 7 ... 
[WARN] [1686855183.719658690] [core_test_7]:     Node:{<rclpy.node.Node object at 0x7f5da52b56f0>}
[WARN] [1686855183.720627979] [core_test_7]:     Context :{<rclpy.context.Context object at 0x7f5da506b9a0>} (1)

I've tried ensuring shutdown, even killing the ros2 daemon thinking maybe issue there because the same topic name is used in another test. But nothing has helped.

def setUp(self):
    subprocess.run('ros2 daemon stop', shell=True)  # Make sure we are starting fresh!
    time.sleep(0.1)
    TestCore.test += 1
    self.context = rclpy.context.Context()
    rclpy.init(context=self.context)
    self.executor = MultiThreadedExecutor(context=self.context)
    self.node = rclpy.create_node("core_test_" + str(self.test), context=self.context)
    self.node.get_logger().info(" set up core test %d (%d)... " % (self.test, self.context.ok()))
    self.executor.add_node(self.node)
    self.node.get_logger().warning("    Node:{%s}" % (str(self.node)))
    self.node.get_logger().warning("    Context :{%s} (%d)" % (str(self.context), self.context.ok()))
    time.sleep(0.1)

  def tearDown(self):
        self.node.get_logger().info(" shutting down core test %d ... " % (self.test))
        self.node.get_logger().warning("    Node:{%s}" % (str(self.node)))
        self.node.get_logger().warning("    Context :{%s} (%d)" % (str(self.context), self.context.ok()))
        rclpy.spin_once(self.node, executor=self.executor, timeout_sec=0.1)
        self.node.destroy_node()
        time.sleep(0.1)
        self.executor.shutdown()
        time.sleep(0.1)

        # Kill it with fire to make sure not stray published topics are available
        rclpy.shutdown(context=self.context) 


   time.sleep(0.2)
    subprocess.run('ros2 daemon stop', shell=True)  # Make sure we are starting fresh!
    time.sleep(0.2)

Any suggestions on fixing or avoiding the Destroyable error is appreciated.