sm_coretest_poll_shutdown_1

Core test for the signal detector’s polling loop under a concurrent rclcpp shutdown. It makes the shutdown land, in most runs, between the loop’s rclcpp::ok() check and its rclcpp::Rate::sleep(), which is the window in which rclcpp::Clock::sleep_until throws std::runtime_error("context cannot be slept with because it's invalid").

Design

• ClSlowUpdate (in OrSlow) is an ISmaccUpdatable client whose update() blocks for slow_update_ms (default 150). SignalDetector::pollOnce calls it on the polling thread, inside the window.
• The polling period is set with the core parameter signal_detector_loop_freq; the launch file uses 5 Hz. With a 150 ms update in a 200 ms period, 75 % of each period is window.
• The update must stay shorter than the period. rclcpp::Rate::sleep() returns without sleeping when the iteration overran, and then never reaches the throwing call; a run that overruns cannot show the defect and cannot prove its absence.
• StIdle is the only state and has no transitions, so the state machine thread is idle and the polling thread is the only one doing work.

Pass criteria

Send SIGINT at a random moment while StIdle is active and wait for the process to exit.

• Exit status 0, with Polling loop exited, Shutdown complete and Finishing State Machine in the log.
• No terminate called line and no context cannot be slept line.
• Under AddressSanitizer the leak check runs at exit.

The failure signature is terminate called after throwing an instance of 'std::runtime_error' with what(): context cannot be slept with because it's invalid, exit status 134, logged right after the shutdown callback’s Terminating scheduler... line and with no Polling loop exited line. With the launch file’s parameters it appears in about three runs out of four.

Running

Plain:

ros2 launch sm_coretest_poll_shutdown_1 sm_coretest_poll_shutdown_1.launch.py

The launch file tees the node output to /tmp/sm_coretest_poll_shutdown_1_latest.log.

For sanitizer runs, build in a separate build and install base with the sanitizer flags and run the node binary directly, passing the parameters on the command line:

install_asan/sm_coretest_poll_shutdown_1/lib/sm_coretest_poll_shutdown_1/sm_coretest_poll_shutdown_1_node \
  --ros-args --log-level INFO -p signal_detector_loop_freq:=5.0 -p slow_update_ms:=150

ASAN_OPTIONS=handle_abort=1 prints the stack of the aborting thread, which shows where the exception came from.