Get builds going on packet.net about ci

[gerkey]

Summary so far:

• With a few changes (9e3afcb), the docker-based build-and-test job seems to work, and there were only a handful of test failures.
• I'm running it again with the usual args (--retest-until-pass 10, --parallel, etc.) and with a warm ccache to get a sense for performance.

[gerkey]

With a warm ccache, the standard CI build and test took 1h16m and produced 3 test failures. Not too shabby.

$ time docker run --privileged -e UID=1001 -e GID=1001 -e CI_ARGS="--do-venv --force-ansi-color --workspace-path /root/code/ci --fastrtps --isolated --cmake-build-type Debug --ament-build-args --parallel -- --ament-test-args --retest-until-pass 10 --" -e CCACHE_DIR=/home/rosbuild/.ccache -i -v /root/code/ci:/home/rosbuild/ci_scripts -v /root/.ccache:/home/rosbuild/.ccache ros2_batch_ci_aarch64
real	76m30.242s
user	0m5.504s
sys	0m2.636s

$ ./ws/src/ament/ament_tools/scripts/ament.py test_results
Skipping 'ws/src/ros2/robot_model/kdl_parser/test/pr2_desc_vector.xml': File "<string>", line None, xml.etree.ElementTree.ParseError: mismatched tag: line 22, column 2
Skipping 'ws/src/ros2/robot_model/kdl_parser/test/pr2_desc_bracket.xml': File "<string>", line None, xml.etree.ElementTree.ParseError: mismatched tag: line 21, column 2
Skipping 'ws/src/ros2/robot_model/kdl_parser/test/pr2_desc_bracket2.xml': File "<string>", line None, xml.etree.ElementTree.ParseError: mismatched tag: line 21, column 2
ws/build/rosidl_generator_c/test_results/rosidl_generator_c/test_interfaces_c.xml: 1 tests, 0 skipped, 0 errors, 1 failures
ws/build/test_communication/test_results/test_communication/test_requester_replier__Empty__rclcpp__rclpy__rmw_fastrtps_cpp.xunit.xml: 1 tests, 0 skipped, 0 errors, 1 failures
ws/build/test_rclcpp/test_results/test_rclcpp/gtest_multithreaded__rmw_fastrtps_cpp.gtest.xml: 1 tests, 0 skipped, 0 errors, 1 failures
Summary: 8382 tests, 0 errors, 3 failures, 4 skipped

[mikaelarguedas]

that's awesome ! 🎉
Some of these tests are known for their flakiness so it's not very surprising.

[gerkey]

Here's the first real job running on the farm:

http://ci.ros2.org/job/ci_linux-aarch64/27

(That's a cold cache, because it's running as jenkins, instead of root.)

[wjwwood]

@gerkey what about the nightly job? should it also be made to run on the linux_aarch64 label?

[wjwwood]

Since we're down a Linux machine again (slave on master), I went ahead and changed the aarch64 nightly to run on packet.net, and canceled the job that was blocking the rest of the nightly queue. I figure this is safe to do since the packet.net machine is up and running and because we have some nightly history on the qemu nightlies already.

[wjwwood]

actually that didn't seem to work so I changed it back to the normal label and requeued the job:

http://ci.ros2.org/job/nightly_linux-aarch64_debug/37/ (this is the one that failed)

[gerkey]

Definitely we'll want to move the CI and the nightly job. First I wanted to check that it would work, and http://ci.ros2.org/job/ci_linux-aarch64/27/ gives a good result. Now I need to tweak and merge these changes: https://github.com/ros2/ci/compare/packet_aarch64 (your build probably failed because you weren't pulling from that branch).

[gerkey]

Fixed by #47.