Tomoya Fujita

https://ci.ros2.org/job/ci_windows/18331/testReport/junit/(root)/projectroot/test_publisher_subscriber_cpp__rmw_fastrtps_cpp__BoundedPlainSequences/ windows failure is unrelated.

@clalancette do we want to backport this to humble?

i think this is bug. https://github.com/ros2/rcl/blob/493ebf3a5d8fe2d861a93006a7d6908fd4e592d9/rcl/src/rcl/node.c#L251-L261 `logger_name` is assigned before `rmw_create_node` above. and if `rmw_create_node` fails, it goes to https://github.com/ros2/rcl/blob/493ebf3a5d8fe2d861a93006a7d6908fd4e592d9/rcl/src/rcl/node.c#L305-L311 and call `rcl_logging_rosout_fini_publisher_for_node` w/o calling `rcl_logging_rosout_init_publisher_for_node`. this leads to call...

https://github.com/ros2/rcl/pull/985 can fix the problem, but according to https://github.com/ros2/rcl/pull/985#pullrequestreview-975525676, it would be nice to refactor the `rcl_node_init` and `rcl_node_fini` process to fix the complication and possible problem in the future.

CC: @ivanpauno @wjwwood

I believe that the latter one makes more sense. (related to https://github.com/ros2/rclcpp/pull/2006#discussion_r973768620, adding `start` / `stop` would be useful to support these features.) @ros2/team friendly ping.

@Voldivh thanks for taking care of this, there is build failure https://build.ros2.org/job/Rpr__rcl__ubuntu_jammy_amd64/350/console

@christophebedard Rpr job is supposed to fail because of https://github.com/ros2/rcl/commit/c43c7abb0a254f4b60c42c4c2993917eb08cb0ec?

@Voldivh never mind the rpr job failure, i wll start the CI to verify.

CI: * Linux [](http://ci.ros2.org/job/ci_linux/18881/) * Linux-aarch64 [](http://ci.ros2.org/job/ci_linux-aarch64/13419/) * Windows [](http://ci.ros2.org/job/ci_windows/19637/) with - https://github.com/ros2/rcl/pull/1004 - https://github.com/ros2/rclcpp/pull/2005 - https://github.com/ros2/rclpy/pull/999 - https://github.com/ros2/system_tests/pull/508