There was a small typo, so I fixed it and reran CI:

• Linux
• Linux-aarch64
• macOS
• Windows

Ok, now it's failing because of the version of ASIO, I think.

Ok, now it's failing because of the version of ASIO, I think.

Mac builds should have gotten farther than the other two, since they're on asio from homebrew.

Edit: spelling and grammar.

You're right, looks like API breaking changes that rmw_fastrtps_cpp will need to adapt:

14:58:53 /Users/osrf/jenkins-agent/workspace/ci_osx/ws/src/ros2/rmw_fastrtps/rmw_fastrtps_shared_cpp/include/rmw_fastrtps_shared_cpp/custom_subscriber_info.hpp:83:34: error: no member named 'getUnreadCount' in 'eprosima::fastrtps::Subscriber'; did you mean 'get_unread_count'?
14:58:53     uint64_t unread_count = sub->getUnreadCount();
14:58:53                                  ^~~~~~~~~~~~~~
14:58:53                                  get_unread_count
14:58:53 /Users/osrf/jenkins-agent/workspace/ci_osx/ws/install/fastrtps/include/fastrtps/subscriber/Subscriber.h:118:14: note: 'get_unread_count' declared here
14:58:53     uint64_t get_unread_count() const;
14:58:53              ^
14:58:53 In file included from /Users/osrf/jenkins-agent/workspace/ci_osx/ws/src/ros2/rmw_fastrtps/rmw_fastrtps_shared_cpp/src/custom_subscriber_info.cpp:15:
14:58:53 /Users/osrf/jenkins-agent/workspace/ci_osx/ws/src/ros2/rmw_fastrtps/rmw_fastrtps_shared_cpp/include/rmw_fastrtps_shared_cpp/custom_subscriber_info.hpp:143:34: error: no member named 'getUnreadCount' in 'eprosima::fastrtps::Subscriber'; did you mean 'get_unread_count'?
14:58:53     uint64_t unread_count = sub->getUnreadCount();
14:58:53                                  ^~~~~~~~~~~~~~
14:58:53                                  get_unread_count
14:58:53 /Users/osrf/jenkins-agent/workspace/ci_osx/ws/install/fastrtps/include/fastrtps/subscriber/Subscriber.h:118:14: note: 'get_unread_count' declared here
14:58:53     uint64_t get_unread_count() const;
14:58:53              ^
14:58:53 2 errors generated.
14:58:53 make[2]: *** [CMakeFiles/rmw_fastrtps_shared_cpp.dir/src/custom_subscriber_info.cpp.o] Error 1
14:58:53 make[2]: *** Waiting for unfinished jobs....
14:58:53 make[1]: *** [CMakeFiles/rmw_fastrtps_shared_cpp.dir/all] Error 2
14:58:53 make: *** [all] Error 2
14:58:53 ---

I'll try it locally and form a patch.

Our windows packages are on asio 1.12.1 which appears to be sufficient as well (as it happens Fast-RTPS's Windows instructions currently recommend installing our chocolatey packages for tinyxml2 and asio 😶)

Also, it doesn't build (at least on macOS) without explicitly turning on -DSECURITY=ON, see: eProsima/Fast-DDS#610

Another round of CI with my new patch (ros2/rmw_fastrtps#302):

• Linux
• Linux-aarch64
• macOS
• Windows

Still having issues with the ASIO version :(

Yeah, I don't know what we'll do about that. For now we'll stick to the 1.8.1 release. The failures on macOS Windows are unrelated I believe, but I didn't check closely.

eProsima/Fast-DDS#616 should fix the compilation error against Asio 1.10.

Thanks for the quick update @richiware!

CI after a rebase, reverting the version pinning, and also including my patch to rmw_fastrtps_cpp:

• Linux
• Linux-aarch64
• macOS
• Windows

One of the tests is failing on all platforms and I'm not familiar with it from nightly's, so it may be a regression:

https://ci.ros2.org/job/ci_linux/7597/testReport/junit/launch_testing_ros.test/test_examples/test_examples_talker_listener_test_py_/

FastRTPS-only builds also using ros2/rmw_fastrtps#302:

• Linux
• Linux-aarch64
• macOS
• Windows

All test failures are also present on the latest nightly builds. So it looks to me that we can switch back to master for FastRTPS?

We can, though I never figured out what was the cause behind https://ci.ros2.org/job/ci_linux/7597/testReport/junit/launch_testing_ros.test/test_examples/test_examples_talker_listener_test_py_/

Since the test passed in the recent build (https://ci.ros2.org/job/ci_linux/7827/testReport/junit/launch_testing_ros.test/test_examples/test_examples_talker_listener_test_py_/) I would move forward with it. If it happens to reappear in the future we can investigate it.

Also the previously failing test doesn't stand out as being a significant regression - then I would expect many more failures.

Also the previous build used an earlier beta of FastRTPS which has received further patches till the now released version.

This pull request regressed 12 tests in test_communication when communicating between Fast-RTPS and Connext.

The tests will continue to fail in the nightly packaging and Debug jobs until this is reverted or resolved.

See ros2/build_farmer#230 for more info.