[joint_trajectory_controller] tests fails randomly about 1/3 chance

[7675t]

I have noticed the tests for joint_trajectory_controller sometimes fail. The chance of failure is about 1/3, so I believe you can reproduce it to run catkin run_test several times locally. The branch is kinetic-devel.

$ catkin run_tests joint_trajectory_controller | tee log.txt | grep FAILURES

Warnings << joint_trajectory_controller:make /home/tajima/Codes/ws/logs/
joint_trajectory_controller/build.make.010.log
...............................................................................
 * FAILURES: 0
 * FAILURES: 0
 * FAILURES: 0
 * FAILURES: 0
 * FAILURES: 4

And one of the failures is:

[joint_trajectory_controller.rosunit-joint_trajectory_controller_vel_test/topicSingleTraj][FAILURE]
/home/tajima/Codes/ws/src/ros_controllers/joint_trajectory_controller/test/joint_trajectory_controller_test.cpp:427
The difference between traj.points.back().positions[i] and state->desired.positions[i] is 0.78539816339744828, which exceeds EPS, where
traj.points.back().positions[i] evaluates to -0.78539816339744828,
state->desired.positions[i] evaluates to 0, and
EPS evaluates to 0.01.

@k-okada #336 may fix some of this, but I'm afraid it is not enough.

[7675t]

#336 is merged, but I still get failure randomly.

[joint_trajectory_controller.rosunit-joint_trajectory_controller_vel_test/emptyTopicCancelsActionTrajWithDelayStopZero][FAILURE]
/home/tajima/Codes/aisinaw_ws/src/ros_controllers/joint_trajectory_controller/test/joint_trajectory_controller_test.cpp:896
Value of: waitForState(action_client, SimpleClientGoalState::SUCCEEDED, long_timeout)
  Actual: false
Expected: true