This looks reasonable, but it will definitely require changes in rclcpp and rclpy.

What kind of changes? There are compatibility methods in rcl, so you can still call it the old way.

There is no "nice" way to get this information in C++, we'd need new functions and callback types for rclcpp Clients/Services. It is completely inaccessible in Python atm, AFAICT.

There is no "nice" way to get this information in C++, we'd need new functions and callback types for rclcpp Clients/Services.

The "user", as in a person writing a ROS node, shouldn't need it. AFAICT, the user cannot currently get the message_info_t either, right? At least I could not find a way to create a subscription callback receiving it.

For the executor, the rcl-functions should suffice, shouldn't they?

I mean, I'm happy to add a better way to access this, but currently I don't understand the use-case.

Without python code to convert the new struct into a pyobj the executor, which is in python, cannot use it either. Either way it needs to be in python so user could use it if they wanted.

Got it. That’s in the works already, I’ll push after dinner.

Sorry you were speaking about c++, still the user should have the option of looking at the time stamps in think. It would require new callback signatures to expose that to them. Also for the case of using it without an executor the user needs all this information to make decisions about when to execute it.

CI with all rmw's, building and testing --packages-up-to demo_nodes_cpp demo_nodes_py test_rclcpp:

• Linux
• Linux-aarch64
• macOS
• Windows

This includes these pull requests:

• Service timestamps #217
• Adapt interfaces for service timestamps rmw_implementation#96
• Implement service info structure with timestamps. rcl#627
• Feature/services timestamps rclpy#545
• Feature/services timestamps rmw_fastrtps#369
• Switch take_response and take_request to rmw_service_info_t rmw_connext#412
• Switch take_response and take_request to rmw_service_info_t rosidl_typesupport_connext#53
• implement with_info version of take rmw_cyclonedds#161

New CI after linter fixes:

• Linux:
  • this failure looks new to me, but has nothing to do with the changes in rcl or rmw_cyclone for this pr:
    • https://ci.ros2.org/job/ci_linux/10480/testReport/rcl/TestGuardConditionFixture__rmw_cyclonedds_cpp/test_rcl_guard_condition_accessors/
  • the rest of the failures look to be known flaky

This pull request has been mentioned on ROS Discourse. There might be relevant details there:

https://discourse.ros.org/t/ros2-middleware-change-proposal/15863/7