Make ament_cmake_python symlink for symlink installs only

[hidmic]

Precisely what the title says. Otherwise, #350 hits our Windows users.

[hidmic]

@gavanderhoorn give this one a shot and let me know.

[gavanderhoorn]

Seems to have fixed it.

Tested on Win10 + Python 3.8.x.

[gavanderhoorn]

I'll test it some more, but it seems like this has resolved the immediate issue.

Thanks for the fast turn-around @hidmic 👍

[gavanderhoorn]

Will this be backported to galactic? Seeing as it fixes something which seems like a regression?

[hidmic]

Full CI:

• Linux
• Linux-aarch64
• macOS
• Windows

[hidmic]

Will this be backported to galactic? Seeing as it fixes something which seems like a regression?

I'm onboard with that. @cottsay ?

[gavanderhoorn]

Friendly ping?

[hidmic]

I was out for a few days @gavanderhoorn. Sorry for the delay.

Hmm, those rclpy test failures on Windows look suspicious.

Running Windows CI up to rclpy:

[gavanderhoorn]

might be flaky tests?

[clalancette]

might be flaky tests?

They were fixed elsewhere in the stack in the meantime.

[hidmic]

Alright then. Going in!

[gavanderhoorn]

Thanks for fixing this @hidmic 👍

[gavanderhoorn]

Oh. And the backport to Galactic?

[hidmic]

Oh. And the backport to Galactic?

You can cherry pick the commit and open a backport PR. Otherwise, I'll try to do it later today.

Edit: I wonder if we can fast-forward.

[gavanderhoorn]

I can create the PR (I believe).

Not sure whether ff would work. Are all those commits supposed to end up on galactic as well?