However, there is really no need to throw an exception at this point that callers of rclpy.spin would need to handle; we could instead just return a value indicating why the spin stopped. This would make it so users need less error-handling code, make this act more like rclcpp (which doesn't throw an exception here), and still provide information in case somebody wants to know why the spin stopped.
Further, we can easily stop throwing an exception here with no impact to downstream consumers. If they aren't handling the exception today, this will just make it a bit more robust. If they are handling ExternalShutdownException today, then that effectively becomes dead code but nothing really breaks.
This is an issue split out of https://github.com/ros2/examples/pull/379
Currently the rclpy executor can throw an
ExternalShutdownException: https://github.com/ros2/rclpy/blob/99f1ce933ddd1677fda8fa602af3684b93f814eb/rclpy/rclpy/executors.py#L689However, there is really no need to throw an exception at this point that callers of
rclpy.spinwould need to handle; we could instead just return a value indicating why the spin stopped. This would make it so users need less error-handling code, make this act more like rclcpp (which doesn't throw an exception here), and still provide information in case somebody wants to know why the spin stopped.Further, we can easily stop throwing an exception here with no impact to downstream consumers. If they aren't handling the exception today, this will just make it a bit more robust. If they are handling
ExternalShutdownExceptiontoday, then that effectively becomes dead code but nothing really breaks.