👩‍🌾 Regression on TestExecutor.test_create_task_coroutine_wake_from_another_thread

[claraberendsen]

Generated by Generative AI

No

Operating System:

Windows 10, Windows 11

ROS version or commit hash:

Rolling

RMW implementation (if applicable):

None

RMW Configuration (if applicable):

No response

Client library (if applicable):

No response

'ros2 doctor --report' output

No response

Steps to reproduce issue

1. Run a Windows release job on ci.ros2.org
2. See job error

Expected behavior

Test to pass

Actual behavior

Test errors since the response time is below the expected time.

assert 0.1 <= (8170542.875 - 8170542.781)

Failing test: rclpy.rclpy.test.test_executor.TestExecutor.test_create_task_coroutine_wake_from_another_thread

Additional information

Reference build: https://ci.ros2.org/job/nightly_win_rel/3638/
First failure: https://ci.ros2.org/job/nightly_win_rel/3633/

I think the improvements put in #1469 made the assumptions of time of this test not true.

This is affecting both the Windows nightlies and the Windows 2025 canaries.

[Flova]

I have troubleshooted this issue in: #1563

tl;dr: The internal Windows timer used for many timeouts and the time.monotonic() call only has a resolution of ~15ms which introduces slight errors that are not covered by the strict asserts.

[claraberendsen]

Thanks @Flova for taking a look at this and putting a fix in!