Switch Cyclone DDS to 0.8.x release branch

[eboasson]

I think it makes sense to switch to the 0.8.x release branch for Humble. That'll also make the changes needed for ros2/rmw_cyclonedds#256 available.

[clalancette]

Here's full CI for this change:

• Linux
• Linux-aarch64
• macOS
• Windows
• Windows

[eboasson]

Thanks for doing a full CI run @clalancette!

The macOS failure looks like it is fine until the build system gives up.

The Linux build warnings (there'll be one for macOS as well) are caused by correcting a function name in the API and the incorrect name is now marked deprecated. The fix needed in the RMW layer is obvious, the only matter is that it can't be done until this PR gets merged, or, alternatively, until the same fix is pushed to the iceoryx branch of Cyclone first. Happy to do the latter, but I don't really see the point if we decide to switch to the 0.8.x branch and never look back.

This https://ci.ros2.org/job/ci_windows/15325/testReport/junit/projectroot.src.ddsrt/tests/CUnit_ddsrt_library_dlopen_path/ one we need to look into (I guess I forgot to include cyclonedds among the packages to test when I did a quick partial ROS CI run before creating this PR ...). I don't know yet why that succeeds in our own CI but fails in this environment. @k0ekk0ek, if you have a good idea without first having to reproduce it locally, feel free to let us know 🙂

The other failures in https://ci.ros2.org/job/ci_windows/15326/testReport/ I don't know yet.

[clalancette]

The macOS failure looks like it is fine until the build system gives up.

Yeah, that looks like an infrastructure flake. I'll run it again.

The Linux build warnings (there'll be one for macOS as well) are caused by correcting a function name in the API and the incorrect name is now marked deprecated. The fix needed in the RMW layer is obvious, the only matter is that it can't be done until this PR gets merged

OK, that makes sense. Can you open the PR to make that change to the RMW layer as well? Then we can run CI against both at the same time and merge them together.

This https://ci.ros2.org/job/ci_windows/15325/testReport/junit/projectroot.src.ddsrt/tests/CUnit_ddsrt_library_dlopen_path/ one we need to look into

All right, sounds good. That one is definitely new, and doesn't show up in our current nightlies when using the iceoryx branch. I'll wait to hear back from you on this one.

The other failures in https://ci.ros2.org/job/ci_windows/15326/testReport/ I don't know yet.

The failures on Windows Debug are in the nightlies as well (https://ci.ros2.org/view/nightly/job/nightly_win_deb/2098/#showFailuresLink), so aren't caused by this PR.

[eboasson]

@clalancette it took a while to figure out CMake, but it should be done now (also "backported" to releases/0.8.x). I've also made a PR for changing the name in rmw_cyclonedds: ros2/rmw_cyclonedds#336 should take care of the build warning.

A CI run up to rclcpp and testing cyclonedds and rclcpp looks good, so I think it is worth another run of the full CI.

[clalancette]

Another full CI:

• Linux
• Linux-aarch64
• macOS
• Windows
• Windows Debug

[eboasson]

That looks pretty good to me 🙂 at least I don't see how the two failures in Windows Debug could be related.

[clalancette]

That looks pretty good to me slightly_smiling_face at least I don't see how the two failures in Windows Debug could be related.

That is definitely a different problem and fixed by ros-perception/laser_geometry#82 . So totally unrelated to this PR.