Codecov Report

Base: 51.10% // Head: 51.10% // Increases project coverage by +0.01% 🎉

Coverage data is based on head (322739a) compared to base (7fec51b).
Patch coverage: 88.47% of modified lines in pull request are covered.

❗ Current head 322739a differs from pull request most recent head 0357848. Consider uploading reports for the commit 0357848 to get more accurate results

@@            Coverage Diff             @@
##             main    #1556      +/-   ##
==========================================
+ Coverage   51.10%   51.10%   +0.01%     
==========================================
  Files         380      381       +1     
  Lines       31796    31790       -6     
==========================================
- Hits        16246    16243       -3     
+ Misses      15550    15547       -3     

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

I understand your high-level concern but this real-time control loop is inherently tied to timing, so it might not be suitable for CI. That's why I made a unit test to replace the integration test.

For example, I would say that this integration test code is a little questionable:

  while (!sawTrackedStatus() && iterations++ < test_parameters_->timeout_iterations)
  {
    auto msg = std::make_unique<geometry_msgs::msg::TwistStamped>();
    msg->header.stamp = node_->now();
    msg->twist.linear.x = 0.8;
    pub_twist_cmd_->publish(std::move(msg));
    publish_loop_rate.sleep();
  }
  // Test that we didn't timeout
  EXPECT_LT(iterations, test_parameters_->timeout_iterations);

Nevertheless, I think the steady_clock change might make this test more reliable so we can try restoring the integration test now and see how it goes.