Update xtypes

[MiguelBarro]

• Fix errors associated to peg library improvements
• Provide windows OS support
• Update ci

[russkel]

@russkel I'm afraid I was not able to reproduce the issue in the ci. Would you be so gentle to provide a reproducer?

I just tried to replicate the issue and was unable to. Now that I am thinking about it, it is possible I had a dirty build environment (where old version of xtypes was being used amongst new versions). I will do some more tests with my code base running new xtypes.

[joakimono]

Will this PR be merged into main when approved, or are there more pending work? I'm eager to test the library (from main).

[MiguelBarro]

That's the idea ... whenever we can spare a reviewer. Meanwhile, I will introduce some bugfix from time to time because I'm working on dependent repos.

[joakimono]

If the environment variable is undefined, std::getenv returns a null pointer. The std::string constructor throws a std::logic_error instead of the graceful exit message. Suggest const char* distro = std::getenv("ROS_DISTRO");, check for null pointer and construct distro string in else .. + std::string(distro) +...

Is the xtypes_idl_validator only intended for ROS? The current implementation demands ROS_DISTRO to be defined.

[MiguelBarro]

Got me! I turns out that construct a std::string from nullptr is undefined behavior. On MSVC leads to an empty string and in gcc a std::logic_error is raised. Crossplatform pitfalls 🙄.
We only used xtypes_idl_validator in an Integration-Service ROS2-SH related project. It is interesting for us to fail without a ROS2 overlay. But is a good hint too. Let's turn it into a warning.