Use node clock for diagnostic_aggregator and diagnostic_updater

[kenji-miyake]

Overview

I found there is a bug when using diagnostic_aggregator and diagnostic_updater with use_sim_time=true.
To fix this, I changed to use node's clock.

FYI: As another way, I also tried simply to use RCL_ROS_TIME like clock_(new rclcpp::Clock(RCL_ROS_TIME)), but it had no effect.

How to reproduce the problem

Distribution: ROS Galactic

git clone https://github.com/kenji-miyake/test-diagnostics-sim-time.git
cd test-diagnostics-sim-time
source /opt/ros/galactic/setup.bash
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release
source install/setup.bash

# Terminal 1
ros2 launch test_diagnostics test.launch.xml use_sim_time:=true

# Terminal 2
ros2 topic echo /clock

# Terminal 3
ros2 topic echo /diagnostics
# sim time isn't reflected to diagnostic_updater

# Terminal 4
ros2 topic echo /diagnostics_agg --no-arr
# sim time isn't reflected to diagnostic_aggregator

How to test this PR

git clone https://github.com/kenji-miyake/test-diagnostics-sim-time.git
cd test-diagnostics-sim-time
git clone https://github.com/tier4/diagnostics.git -b fix-sim-time-bug
source /opt/ros/galactic/setup.bash
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release
source install/setup.bash

# Terminal 1
ros2 launch test_diagnostics test.launch.xml use_sim_time:=true

# Terminal 2
ros2 topic echo /clock

# Terminal 3
ros2 topic echo /diagnostics
# sim time is reflected to diagnostic_updater

# Terminal 4
ros2 topic echo /diagnostics_agg --no-arr
# sim time is reflected to diagnostic_aggregator

[clalancette]

@kenji-miyake With #209 merged, the PR jobs should now succeed. Can you rebase this onto the latest ros2-devel code?

[kenji-miyake]

@clalancette Thank you, rebased!

[kenji-miyake]

@clalancette Can this be backported to Galactic if I send a PR?

[clalancette]

@clalancette Can this be backported to Galactic if I send a PR?

Technically, this patch breaks API and ABI. But since this isn't a core package, and doesn't have a Quality Declaration, that is typically allowed. Also the code it changes looks to be internal, so it might be OK anyway.

I'd say open the backport PR. The maintainers can assess whether they should take it. I'll be clear that I'm not a maintainer here; I just saw some PRs that looked reasonable and could be merged, so I did that.