Partial joints

[v-lopez]

Fixes #31
Fixes #37

[codecov-commenter]

Codecov Report

Merging #68 into master will increase coverage by 2.66%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #68      +/-   ##
==========================================
+ Coverage   33.12%   35.78%   +2.66%     
==========================================
  Files          12       14       +2     
  Lines        1123     1358     +235     
  Branches      740      879     +139     
==========================================
+ Hits          372      486     +114     
+ Misses        103       91      -12     
- Partials      648      781     +133     

Flag	Coverage Δ
#unittests	35.78% <ø> (+2.66%)	⬆️

Impacted Files	Coverage Δ
...include/joint_trajectory_controller/trajectory.hpp	46.66% <0.00%> (-20.00%)	⬇️
...int_trajectory_controller/test/test_trajectory.cpp	24.71% <0.00%> (-1.66%)	⬇️
...ers/joint_trajectory_controller/src/trajectory.cpp	54.11% <0.00%> (ø)
.../joint_state_controller/joint_state_controller.hpp	100.00% <0.00%> (ø)
...jectory_controller/joint_trajectory_controller.hpp	100.00% <0.00%> (ø)
...te_controller/test/test_joint_state_controller.cpp	19.65% <0.00%> (ø)
...te_controller/test/test_joint_state_controller.hpp	100.00% <0.00%> (ø)
...ory_controller/test/test_trajectory_controller.cpp	23.48% <0.00%> (+1.47%)	⬆️
...ory_controller/src/joint_trajectory_controller.cpp	45.10% <0.00%> (+3.41%)	⬆️
...ectory_controller/test/test_trajectory_actions.cpp	38.49% <0.00%> (+8.24%)	⬆️
... and 1 more

---

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 d2e940a...188c0e4. Read the comment docs.

[v-lopez]

@ddengster Could I get your take on these changes?

[ddengster]

Thanks for the PR and sorry for the late reply. In general the only bug to fix is the joint->get_position() call.

[ddengster]

This function is called on a separate thread from the main JTC updating loop, you'll get some kind of race condition if you call joint_state->get_position() here.

traj_external_point_ptr_->update(msg); is where we sync up, we could put some kind of coordination there or delay it until the first parsing of the message (this is how I did immediate playbacks of trajectory messages where messages are sent with time 0). Perhaps we could use a boolean per joint to indicate whether to retain the last known joint position.

[v-lopez]

Thanks for noticing this, it is indeed a problem.

Furthermore, On the topic callback we update the trajectory without acquiring the trajectory_mtx_ which is another race condition.

Probably when we receive a trajectory (whether from topic or action) on an "auxiliary thread" we should do some basic checks to ensure that we can accept it, and store it somewhere protected by a mutex.

Then on the next update call on the "main thread", we should process it, fill the partial goals and do whatever needs to be done.

Something along these lines I believe would fix these issues.

[v-lopez]

I've pushed a fix for these issues on this same branch. 188c0e4