Fix all test failures on Linux CI Batch Job about ros2

[dirk-thomas]

ros2/rosidl_dds@64a1f36
ros2/rmw_connext@57cf7f5
ros2/rmw_opensplice@cb75a3d

Locally that leaves me to failing tests for:

• copyright
• lint_cmake
• uncrustify

[wjwwood]

Just FYI, there is another CMake linter I found the other day: https://github.com/polysquare/polysquare-cmake-linter

[dirk-thomas]

@esteve @tfoote @wjwwood Please take a close look to the reference PRs which apply the uncrustify formatting rules. Even if it is a lot of white spaces changes we should make sure that the formatting is exactly what we want - and if necessary adjust it before merging them.

[dirk-thomas]

I have filled the following tickets upstream:

• https://github.com/bengardner/uncrustify/issues/362
• https://github.com/bengardner/uncrustify/issues/363

[dirk-thomas]

I have updated the list in the first comment after looking through the currently failing tests.

[wjwwood]

I finally fixed the communication problem within docker, I had to do docker run with the --privileged option and I had to add an ifconfig eth0 multicast as part of the run (not possible in the build part of the docker image and it requires --privileged). After that the userland talker/listener test passes.

Now the only failing tests on this Linux build are uncrustify related:

http://54.183.26.131:8080/job/ros2_batch_ci_linux/13/console

Though it still doesn't find Connext, so there maybe more failures lurking in the connext packages still.

[wjwwood]

This build is green: http://54.183.26.131:8080/job/ros2_batch_ci_linux/

[wjwwood]

🎆 Yay! 🎆

I'll close this now, deferring to #33 and #34 for OS X and Windows respectively.

[dirk-thomas]

The current linux batch job does not use Connext.

[dirk-thomas]

The current linux batch job (http://54.183.26.131:8080/view/ros2/job/ros2_batch_ci_linux/27/) fails the test_publish_subscribe with OpenSplice. It seems that both endpoints can't communicate again?

[wjwwood]

@dirk-thomas maybe it's something else this time. Last time it said in the logs that multicast was not available in the ospl log, but now it doesn't:

http://54.183.26.131:8080/view/ros2/job/ros2_batch_ci_linux/ws/workspace/src/ros2/examples/userland/ospl-info.log

(Btw the tests drop the ospl log in the source space)

The next thing to do would be to log in to the docker instance locally and run the talker and listener in different folders since they some times clobber each other writing to the same ospl log file.