TSan lock-order-inversion reproducible via Nav2 test_bond

[hu1097007701]

Generated by Generative AI

No response

Operating System:

Linux hwk-MS-7D99 7.0.0-28-generic #28~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Wed Jul 1 15:50:57 UTC 2026 x86_64 x86_64 x86_64 GNU/Linux

ROS version or commit hash:

jazzy

RMW implementation (if applicable):

rmw_fastrtps_cpp

RMW Configuration (if applicable):

No response

Client library (if applicable):

rclcpp

'ros2 doctor --report' output

ros2 doctor --report

NETWORK CONFIGURATION
error   : Unable to access network interface information: [Errno 1] Operation not permitted

PLATFORM INFORMATION
system           : Linux
platform info    : Linux-7.0.0-28-generic-x86_64-with-glibc2.39
release          : 7.0.0-28-generic
processor        : x86_64

RMW MIDDLEWARE
middleware name    : rmw_fastrtps_cpp

Steps to reproduce issue

1. Build nav2_util and nav2_lifecycle_manager with ThreadSanitizer enabled.
2. Run Nav2's built-in test_bond test.

Example:

source /opt/ros/jazzy/setup.bash
cd /home/hwk/nav2_ws

colcon build \
  --packages-select nav2_util nav2_lifecycle_manager \
  --cmake-clean-cache \
  --parallel-workers 1 \
  --cmake-args \
    -DBUILD_TESTING=ON \
    -DCMAKE_BUILD_TYPE=RelWithDebInfo \
    -DCMAKE_C_FLAGS="-O1 -g -w -Wno-error -fno-omit-frame-pointer -fsanitize=thread" \
    -DCMAKE_CXX_FLAGS="-O1 -g -w -Wno-error -fno-omit-frame-pointer -fsanitize=thread" \
    -DCMAKE_EXE_LINKER_FLAGS="-fsanitize=thread" \
    -DCMAKE_SHARED_LINKER_FLAGS="-fsanitize=thread"

source /home/hwk/nav2_ws/install/setup.bash

export CTEST_OUTPUT_ON_FAILURE=1
export TSAN_OPTIONS="halt_on_error=0 history_size=7 second_deadlock_stack=1 report_signal_unsafe=0"

cd /home/hwk/nav2_ws/build/nav2_lifecycle_manager
ctest -V -R test_bond

Expected behavior

The test should complete without any ThreadSanitizer lock-order-inversion warning.

Actual behavior

The test logic appears to pass, but TSan reports a lock-order-inversion warning and the test run fails because of the sanitizer report.

The warning summary includes:

• rclcpp::Context::add_on_shutdown_callback(std::function<void ()>)
• rclcpp::graph_listener::GraphListener::start_if_not_started()
• rclcpp::graph_listener::GraphListener::__shutdown()

Additional information

The smallest reproducer I found is Nav2's built-in test:

• nav2_lifecycle_manager/test/test_bond.cpp

The call path on the reproducer side includes:

• LifecycleBondTest.POSITIVE
• nav2_lifecycle_manager::LifecycleManagerClient::startup()

I can also provide the full TSan log if that would be helpful.

[mini-1235]

I would be curious to know if you can reproduce the same issue on Lyrical or Rolling. I think there are already some related issues/PRs, but the fix may not have been backported

[hu1097007701]

Thanks for the suggestion. I haven’t tested Lyrical/Rolling yet. I’ll try to retest on Lyrical in the next few days and follow up here.