Use windows CI build

[christophfroehlich]

The main purpose is to avoid adding code which does not compile on Windows. Github provides windows-2019 runners, which can be used here. However, https://github.com/nektos/gh-act does not support this and I was only to test the workflow online.

• Binary releases are not published for rolling, so I hardcoded jazzy
• I added it now to be triggered at PRs and pushes, but it might be configured to run on comments/labels only.
• Needs Add reusable workflow for windows CI ros2_control_ci#175

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 73.42%. Comparing base (b28ce79) to head (2bd52ea).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #204   +/-   ##
=======================================
  Coverage   73.42%   73.42%           
=======================================
  Files           7        7           
  Lines         414      414           
  Branches       68       68           
=======================================
  Hits          304      304           
  Misses         68       68           
  Partials       42       42           

Flag	Coverage Δ
unittests	73.42% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

[fmauch]

I cannot draw a clear result from your commit history: Are the ALL_CAPS enum values causing problems with MSVC? Shouldn't this make problems in many other places, as well?

[christophfroehlich]

Sorry for my scarce comments on this. Exactly, the all caps ENUMS make strange errors like "missing braces". I found some comments online that on on windows there are lots of macros automatically included for historical reasons. I was not able to see where they are defined and if there is a add_compile_definitions macro in deactivating it. (again, no local possibility, and debugging this in the github runner is not fun).
I could rename them to different ALL-CAPS, or maybe find out which one exactly is the culprit.

[christophfroehlich]

@SENAI-GilmarCorreia could you have a look maybe, how to fix this properly?

[SENAI-GilmarCorreia]

@christophfroehlich I am trying to understand the problem here. Do you want to revert the enums back to capital letters?

As I have seen, ERROR already exists in Windows systems because this identifier is reserved for error codes in Windows.

You can simply rename it by changing the case, as you did, or you can change the name of the enum:

enum class return_type : std::uint8_t {
  OK = 0,
  FAILURE = 1,
  DEACTIVATE = 2,
};

I'm testing here as well.

[christophfroehlich]

Thanks for reaching out. Exactly, if only the ERROR is a problem then I'd rename it as you suggested. Or maybe there is a compile definition to avoid this, like NOMINMAX ;)

[SENAI-GilmarCorreia]

As I've seen, this definition is really important. You can use #undef ERROR, but it might cause some unexpected behaviors.

[fmauch]

This looks fine with me

[saikishor]

LGTM now

[christophfroehlich]

@Mergifyio backport humble

[mergify]

backport humble

✅ Backports have been created

• #221 Use windows CI build (backport #204) has been created for branch humble