🧑‍🌾 rclcpp.test_executors failing in Rolling and Jazzy CycloneDDS

[Crola1702]

Bug report

Required Info:

• Operating System:
  • Ubuntu Noble
• Installation type:
  • Source
• Version or commit hash:
  • Rolling and Jammy
• DDS implementation:
  • CycloneDDS
• Client library (if applicable):
  • rclcpp

Steps to reproduce issue

1. Run a build in Rolling Cyclone CI
2. Seeprojectroot.test.rclcpp.test_executors fail.

Additional information

Reference build: https://build.ros2.org/view/Rci/job/Rci__nightly-cyclonedds_ubuntu_noble_amd64/41/#showFailuresLink

Test regressions:

• projectroot.test.rclcpp.test_executors
• rclcpp.TestBusyWaiting/SingleThreadedExecutor.test_spin_all
• rclcpp.TestBusyWaiting/SingleThreadedExecutor.test_spin_some
• rclcpp.TestBusyWaiting/MultiThreadedExecutor.test_spin_all
• rclcpp.TestBusyWaiting/MultiThreadedExecutor.test_spin_some
• rclcpp.TestBusyWaiting/StaticSingleThreadedExecutor.test_spin_some
• rclcpp.TestExecutors/SingleThreadedExecutor.spin_some
• rclcpp.TestExecutors/SingleThreadedExecutor.spin_some_max_duration
• rclcpp.TestExecutors/MultiThreadedExecutor.spin_some
• rclcpp.TestExecutors/MultiThreadedExecutor.spin_some_max_duration
• rclcpp.TestExecutors/StaticSingleThreadedExecutor.spin_some

Log output TextExecutors spin_some:

/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:572
Expected equality of these values:
  my_waitable1->get_count()
    Which is: 1
  original_my_waitable1_count + 1
    Which is: 2
spin_some() failed to execute a waitable that was triggered

spin_some tests is failing since #2465 in CycloneDDS

Log output TestExecutors spin_some_max_duration:

/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:642
Expected: (to_nanoseconds_helper(spin_some_run_time)) > (to_nanoseconds_helper(waitable_callback_duration / 2)), actual: 56575 vs 100000000
spin_some() took less than half the expected time to execute a single waitable, which implies it did not actually execute one when it was expected to


/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:656
Expected equality of these values:
  number_of_waitables_executed
    Which is: 0
  1u
    Which is: 1
expected exactly one of the two waitables to be executed, but my_waitable1->get_count(): 0 and my_waitable2->get_count(): 0

Log output TestBusyWaiting spin_all:

/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:952
Expected: (waitable->get_is_ready_call_count()) < (10u * this->waitable->get_count()), actual: 1 vs 0


/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:975
Expected equality of these values:
  this->waitable->get_count()
    Which is: 0
  2u
    Which is: 2

Log output TestBusyWaiting spin_some :

/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:952
Expected: (waitable->get_is_ready_call_count()) < (10u * this->waitable->get_count()), actual: 1 vs 0


/tmp/ws/src/ros2/rclcpp/rclcpp/test/rclcpp/executors/test_executors.cpp:986
Expected equality of these values:
  this->waitable->get_count()
    Which is: 0
  1u
    Which is: 1

[Crola1702]

I updated the information on this issue with new jobs and test regressions

[Crola1702]

Flaky ratio today (15 days):

job_name	last_fail	first_fail	build_count	failure_count	failure_percentage
Rci__nightly-cyclonedds_ubuntu_noble_amd64	2024-07-02	2024-06-19	11	11	100.0
Jci__nightly-cyclonedds_ubuntu_noble_amd64	2024-07-02	2024-06-20	5	5	100.0