search

ros2 / rclcpp

rclcpp (ROS Client Library for C++)
Apache License 2.0
562 stars 425 forks source link

symbol lookup error: undefined symbol _ZN6rclcpp* #2049

Closed tommy-erko closed 2 years ago

tommy-erko commented 2 years ago

Bug report

Required Info:

Steps to reproduce issue

Not included now as the project is internal. I can try to create simpler example that will behave similarly.

Expected behavior

Several rclcpp nodes start up and are ready to control a simulated robot (in Ignition Gazebo).

Actual behavior

Since the latest ROS2 rolling update, once we run multiple rclcpp nodes, we get a "symbol lookup error" from each of them. There are these types:

[node_name_a-5] /opt/ros/rolling/lib/package_a/node_name_a: symbol lookup error: /opt/ros/rolling/lib/package_a/node_name_a: undefined symbol: _ZN6rclcpp11ServiceBaseD2Ev

[node_name_b-14] /opt/ros/rolling/lib/package_b/node_name_b: symbol lookup error: /opt/ros/rolling/lib/package_b/node_name_b: undefined symbol: _ZN6rclcpp10ClientBaseD2Ev

[node_name_c-17] /opt/ros/rolling/lib/package_c/node_name_c: symbol lookup error: /opt/ros/rolling/lib/package_c/node_name_c: undefined symbol: _ZN6rclcpp12experimental28SubscriptionIntraProcessBaseD2Ev

In a simplified case, we launch only one our node and Navigation2 stack and the node dies the same. (with the _ZN6rclcpp10ClientBaseD2Ev)

Additional information

Our project runs on ros-rolling. This issues started to appear with the latest rolling update (since 19.11.22). If we start the

I know the information is rough, but I am not able to disclose any of the code. I would firstly like to get some feedback where this might come from, whether it might be our issue, or inform you that something like this started to happen. Either is appreciated.

clalancette commented 2 years ago

Have you rebuilt your stack since the update? Since we don't guarantee ABI compatibility on Rolling, it is possible that the ABI changed.

tommy-erko commented 2 years ago

Thank you for your reply. We have done so, but if it seems like a likely source, I will do that again and with closer attention.

tommy-erko commented 2 years ago

Sorry for false alarm, it seems our stack rebuild did not work as supposed to. So if anyone encounters something similar, this was caused by part of the components built against older ROS2.