Disable actually failing flaky OMPL test

[vatanaksoytezer]

Disable actually failing flaky OMPL test that causes CI failures all over moveit2. See https://github.com/ros-planning/moveit2/runs/2798018498?check_suite_focus=true for a sample output.

[vatanaksoytezer]

Disables #481

[codecov]

Codecov Report

Merging #495 (75bab8c) into main (9ca817d) will decrease coverage by 0.83%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main     #495      +/-   ##
==========================================
- Coverage   46.01%   45.19%   -0.82%     
==========================================
  Files         170      169       -1     
  Lines       18731    18729       -2     
==========================================
- Hits         8618     8462     -156     
- Misses      10113    10267     +154     

Impacted Files	Coverage Δ
...mpl/ompl_interface/src/detail/ompl_constraints.cpp	0.00% <0.00%> (-77.18%)	⬇️
.../include/moveit/robot_model/revolute_joint_model.h	60.00% <0.00%> (-40.00%)	⬇️
...del/include/moveit/robot_model/joint_model_group.h	84.00% <0.00%> (-5.33%)	⬇️
moveit_core/robot_state/src/robot_state.cpp	28.99% <0.00%> (-2.48%)	⬇️
moveit_core/robot_model/src/joint_model_group.cpp	54.09% <0.00%> (-1.05%)	⬇️
...de/moveit/ompl_interface/detail/ompl_constraints.h

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9ca817d...75bab8c. Read the comment docs.

[davetcoleman]

TODO (vatanaksoytezer): Uncomment once this is fixed

I think this is a pretty bad habit to get into... the chances of you ever re-enabling this flakey test is close to zero. Did a Clubhouse story get created to tackle this issue in the next sprint? Do we have a plan to fix it?

MoveIt already has low test coverage, but disabling existing ones only makes that worse.

Tagging @mamoll as this pertains to OMPL.

[vatanaksoytezer]

I think this is a pretty bad habit to get into... the chances of you ever re-enabling this flakey test is close to zero. Did a Clubhouse story get created to tackle this issue in the next sprint? Do we have a plan to fix it?

Thanks for the heads up @davetcoleman. I have a clubhouse story and an issue here to re-enable this. The reason I suggested to disable this was we didn't know the cause and this flakiness was becoming an issue with a lot of PRs failing in CI in the release sprint. #511 has more information from my debugging session. I think @mamoll maybe the right person to look into this.

For the OMPL and servo tests that we disabled I am trying to look whenever I can and really want to enable them back asap.

[v4hn]

just a side remark: if you have to disable tests, you should usually not disable compiling the whole unit.
Instead you can selectively disable the affected gtests. The code will still be compiled (thus break on API changes) and during execution gtest explicitly mentions the disabled tests.

[vatanaksoytezer]

just a side remark: if you have to disable tests, you should usually not disable compiling the whole unit.
Instead you can selectively disable the affected gtests. The code will still be compiled (thus break on API changes) and during execution gtest explicitly mentions the disabled tests.

Thanks a lot @v4hn. Definitely learned something new today! I will make sure I do it this way if we need to disable any test in the future. I might take my time to change this as well if I can't figure out a fix soon.