[ROS2] Port kindr_ros and kindr_msgs

[SivertHavso]

Summary of non-ROS2 specific changes. These could potentially be cherry picked into the ROS1 branch to avoid future merge conflicts:

• Using tf2 instead of tf (f022c41)
• Linted all files in kindr_msgs and kindr_ros with cppcheck, cpplint, lint_cmake, xmllint, and uncrustify (af5667c, 51b18ec)
• Removed what looks to be unnecessary variables that triggered unused-but-set-variable warnings (250f0e0)

ROS2 specific changes:

• Updated the package.xml format
• Replaced catkin with ament_cmake
• Changed geometry_msgs include paths, and namespace to the new ROS2 convention.

All 18 gtests run and pass for me. I've targeted ROS2 Galactic, but I see not reason why it shouldn't work with ROS2 Foxy.

The rviz plugins haven't been ported yet.

Partly resolves #22

[remod]

Very nice, thank you!

[remod]

Is this something you intended to do before merging the MR?

[SivertHavso]

It will have to wait until I've finished porting the elevation_mapping package. I'd say merge it now since it "only" affects quality assurance and not the API or ABI. The tests themself still run and there shouldn't be a notable difference in code coverage between the ROS1 and ROS2 branch for now.

[remod]

Out of curiosity, why is this a dependency of the porting of elevation_mapping? This is a downstream dependency, right?

[remod]

@SivertHavso We internally discussed the matter of formatting of code: As we are using this style for most of our closed source and open source code, we decided that it would make sense to use it also on the ROS2 branches.

Would you mind adopting it after merging this PR?

[SivertHavso]

@remod Sure, in the process of doing it now. On my ROS2 working branch I've got it set to run clang-format, cppcheck, gtest, lint_cmake, and xmllint automatically (through the corresponding ament_cmake packages) as part of testing now, and have disabled the cpplint and uncrustify linters because they expect a different style. I'm getting a lot of failures from clang_format with your style though.

For example clang-format wants:

template <typename PrimType_>
inline static void convertFromRosGeometryMsg(const geometry_msgs::msg::Point& geometryPointMsg, kindr::Position<PrimType_, 3>& position) {
  position.x() = static_cast<PrimType_>(geometryPointMsg.x);
  position.y() = static_cast<PrimType_>(geometryPointMsg.y);
  position.z() = static_cast<PrimType_>(geometryPointMsg.z);
}

instead of the current:

kindr_ros/kindr_ros/include/kindr_ros/RosGeometryMsgPhysicalQuantities.hpp

Lines 84 to 92 in f0bd43c

	template <typename PrimType_>
	inline static void convertFromRosGeometryMsg(
	const geometry_msgs::Point& geometryPointMsg,
	kindr::Position<PrimType_, 3>& position)
	{
	position.x() = static_cast<PrimType_>(geometryPointMsg.x);
	position.y() = static_cast<PrimType_>(geometryPointMsg.y);
	position.z() = static_cast<PrimType_>(geometryPointMsg.z);
	}

I get the same results running ament_clang_format or clang-format directly for both the master branch and my ROS2 working branch, "12 files with 218 code style divergences".

I'm quite sure it's loading your custom configuration because settings such as PointerAlignment: Left to PointerAlignment: Right increases the amount of failures.

Can you confirm whether the master branch is formatted correctly according to your style please?

[remod]

Can you confirm whether the master branch is formatted correctly according to your style please?

No you are right, this package is not formatted according to our guidelines yet. I'll do that today!