Rclcpp: [rclcpp_lifecycle] introduce new service for valid transitions to call

[Karsten1987]

see discourse discussion here: https://answers.ros.org/question/302402/ros2cli-lifecycle-shutdown-transition/

[dirk-thomas]

Beside the three PRs updating ros2lifecycle (list -> nodes and a new list showing the available transitions from the current state) and this looks :top:

[Karsten1987]

CI:

[Karsten1987]

@Myzhar Do you mind having a look at this and let me know if this meets your expectations? I believe you have quite some experiences with the lifecycle implementation by now (given that you have been vocal about it on github and discourse).

The set of patches (branch lifecycle_refactor) implements a new service to get all available transitions for the current state as well as a service to introspect the state machine completely.

[Myzhar]

I will look at it with very pleasure and will let you know

[Myzhar]

@Karsten1987 I need a few suggestion on how to update my workspace with the patches introduced with the lifecycle_refactor branch

[Karsten1987]

so essentially you want to checkout the branches, right?

So in your workspace you can run the following command to do so:

vcs custom --git --args checkout lifecycle_refactor

If not, you should be able to go into each referenced repository and change the git branch manually.
These are:

• demos
• rclcpp
• rcl
• rcl_interfaces
• ros2lifecycle

[Myzhar]

Ok, from Waffle I found the list of repositories that have been patched:

• ros2/ros2cli
• ros2/rcl_interfaces
• ros2/demos
• ros2/rcl
• ros2/rclcpp

I launched a git checkout manually (I'm sure that a better method exists) for all of them and I was able to compile.

[Update: I received your previous comment while I was writing mine]

I modified the code of my node to follow the new returns type and the other modifications and I finally noticed that the command:
$ ros2 lifecycle set /zed/zed_node shutdown
works :+1:

This is also a very useful command:

$ ros2 lifecycle list /zed/zed_node 
- cleanup [2]
    Start: inactive
    Goal: cleaningup
- activate [3]
    Start: inactive
    Goal: activating
- shutdown [51]
    Start: inactive
    Goal: shuttingdown

I tried also to understand what kind of other modifications you have introduced, to see if I can get enhancements at code level, but it's not easy.

[Karsten1987]

We do have the code review, so don't worry too much about that. It was more like to ask whether that gives you a solution to the issue you posted on discourse.
But then I'll take it, that it indeed works for you as well.

[Myzhar]

What I have not understood is if now it is possible to change state internally calling a transition.
I mainly think about going to "error processing" from, for example, a running thread... this is a possible situation reading the design documents, but not yet available until now (as far as I know).

To be more clear I explain my use case:
I'm writing the ROS2 driver for the Stereolabs ZED and ZED mini.
I configure everything and I open the camera in "on_configuring".
I start the grab thread in "on_activating".
But I have not a "lifecycle way" to handle a disconnection. I'd like to deactivate and cleanup to bring the node in a valid state waiting for a new configuration and activation.

This is how I understood the way of working with the Lifecycle node... if I'm right.

[Karsten1987]

So from what I understand is that you want to shutdown the camera either when deactivate or cleanup is called? In both cases you have a callback for this.
Everything which is happening in other threads has to be manually managed. The callbacks are solely a trigger for it if you want.
You can trigger this callbacks programmatically on your lifecycle node by calling deactivate() or cleanup() on it.

Nevertheless, I am going to merge this then as it seems to fulfill the requirements described in your discourse post.