Fix flaky smoother test #4372
Fix flaky smoother test #4372
Conversation
|
This is ultimately caused by the new analytic expansion technique checking until the end point. Thus, the last point and the goal node can be within floating point error of each other which causes an additional segment to be added when overlapping. This last duplicate point is removed in the planner plugins by checking if there is close overlap but that is not performed in the test. Thus, we can accept either 2 or 3 and either is valid as long as the planner plugins are properly interpreting that duplicate result. Because they're so close, it can be interpreted as 2 or 3 depending on the run and I assume something so low-level in C++ that it is really not worth trying to fix given there is no where that this would functionally appear for users except for in this location bypassing over checks that happen at the plugin level at runtime Definitely a subtle one, good thing we have all these unit tests! |
* fix flaky smoother tets * fix test * fix
- Apply the same fix as in ros-navigation/navigation2#4372
* humble-3.17: automatic update on 20240701-002215
* Update patch for tf2
- Disable copyright and uncrustify tests
* Add workaround-gcc-segv patch for rclcpp_lifecycle
* Add patch to fix flaky smoother test in nav2-smac-planner
- Apply the same fix as in
ros-navigation/navigation2#4372
---------
Co-authored-by: Alpine ROS aports update bot <[email protected]>
Co-authored-by: f-fl0 <[email protected]>
* fix flaky smoother tets * fix test * fix
Removes tolerance so it should now work deterministically if I understand the issue correctly