normalize some calls that use keyword arguments to use the ** splat

[doudou]

One important tidbit introduced in this PR are the sanitize_keywords_to_array and sanitize_keywords_to_hash helpers. They handle inconsistencies in Ruby 2.7 (compared to previous Rubies) for calls that mix strings and symbols in the last hash, such as

provides MODEL, "some" => "mapping", some: keywords

[g-arjones]

I'm still not sure about how I feel about these changes in Ruby 2.7. I think the devs may have messed things up :(

(Rubocop violations. Aren't you using vscode anymore?)

[doudou]

(Rubocop violations. Aren't you using vscode anymore?)

I am. Now, with the 1000s of errors I get from the C++ plugin, I end up missing some errors :(

[doudou]

Also, it's passing on my install. Not sure what's going on

[doudou]

I think the devs may have messed things up :(

So do I. For reference: https://bugs.ruby-lang.org/issues/16853, https://discuss.rubyonrails.org/t/new-2-7-3-0-keyword-argument-pain-point/74980/13?u=doudou42

[g-arjones]

https://bugs.ruby-lang.org/issues/16853

Interesting. Definitely messed up then (either in 2.0 or in 2.7)

[doudou]

Not sure what's going on

Now I do. rubocop 0.86 has a breaking change on the CyclomaticComplexity cop. I'm going to stick to 0.85.1, which passes.

[g-arjones]

Did you remove that pin from rock-package_set ?

[doudou]

Did you remove that pin from rock-package_set ?

No, but the checks here are from travis, which does not care about Rock.

We have to see how to handle rubocop versions in the rock package set. It's not the first time there are breaking changes - which, to be fair in this case, are not wrong - the methods it flags are rather complicated. Now, it's a bit stressful to get test break because rubocop stops passing suddenly.

[g-arjones]

No, but the checks here are from travis, which does not care about Rock.

Won't that make your internal CI fail?

[g-arjones]

Now, it's a bit stressful to get test break because rubocop stops passing suddenly.

Maybe we should have a separate build stage that's for linting

[doudou]

Maybe we should have a separate build stage that's for linting

Would be good, yes. It's quite a bit of work, so I would have preferred something less intrusive :P

Actually, could use GitHub actions for that... rubocop does not need to have anything installed. Same with yard.

[g-arjones]

Actually, could use GitHub actions for that

Why not just travis?

[doudou]

Why not just travis?

I want to try out Github actions :P

[doudou]

(and Github advertised a ready-to-use "magic linting" action recently)

[g-arjones]

(and Github advertised a ready-to-use "magic linting" action recently)

:+1: