Felix von Drigalski

If I understand correctly, you would have to store the task description along with the solution to know which `stage_id` to look up, which to me is just another opportunity...

> if you directly get the correct id from the task That is assuming that you have access to the task, which you would not if you get the solution...

> But for any kind of interleaved-execution or custom hooks, you would need [access to the Task] anyway. The only other way to support similar functionality is by additional information...

Do you know if the problem exists if you build from source?

If you're running this right now, could you add pictures of the goal positions/configurations and an example of what you consider "strange"?

After changing these to: ``` roslaunch ur_gazebo ur3e_bringup.launch limited:=true roslaunch ur3e_moveit_config ur3e_moveit_planning_execution.launch sim:=true limited:=true ``` this: ``` roslaunch ur_gazebo ur3_joint_limited.launch roslaunch ur3_e_moveit_config ur3_e_moveit_planning_execution.launch sim:=true limited:=true ``` and executing your code...

Right, there was an update that I hadn't pulled. I was on commit 00447a of https://github.com/fmauch/universal_robot. I haven't tested the newest commit yet. I can't look at this further right...

There's no obstacle to using the master branch on Ubuntu 18.04 and Melodic, as long your nodes use Python 2.7. Using Python 3 in Melodic is a bit more complicated,...

That's bad. Did you also test [c4f8fd](https://github.com/ros-planning/moveit/commit/c4f8fd4f5e48a5ddee0a320b3009b63dd54ba21a), or is the regression only in d90acb3?

Can you give an example of the use cases you are envisioning? I don't know how many people would use it, and this runs the risk of duplicating collision checking...