test: make common.noop a getter

[aqrln]

Make common.noop a getter that returns a new function object each time
the propery is read, not the same one. The old behavior could possibly
lead to subtle and hard to catch bugs in some cases.

Refs: #12711 (comment)

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

Affected core subsystem(s)

test

[Trott]

LGTM if CI is green. I'd be just as happy (maybe even slightly happier) to get rid of common.noop altogether, but this works for me!

[aqrln]

CI: https://ci.nodejs.org/job/node-test-pull-request/7737/

[aqrln]

Whoops, unintentional style change. I think I had written something there and then deleted too much.

[not-an-aardvark]

I don't have a strong opinion on this change either way, but if common.noop returns a new function on each access, why is it better than just using () => {}? In cases where it does matter, it seems like it might be confusing for two accesses to common.noop to return different function instances. For example, if I was looking at this code and I didn't have additional context, I would assume that all the listeners were strictly equal to each other.

[Fishrock123]

What if we Object.freeze() the function?

[refack]

How about keeping the old noop and add

export.newNoop = () => { return () => {}; };

[refack]

+0.5
Although it's new syntax () => {} can be considered as an idiom meaning noop...

[benjamingr]

I think at this point writing () => {} is shorter than writing common.noop and we should consider just using that instead.

[Fishrock123]

Even better, we should adjust our lint rules to allow unused arrow fn parameters to be _. e.g.

_ => {}

[refack]

@Fishrock123 although _ => {} is shorter, IMHO () => {} is more idiomatic (e.g. MDN)

[jasnell]

get rid of common.noop altogether

It was just added. A key reason I added this was to make it explicit that the function in question is intentionally a non-op. I'm definitely -1 on reverting back to the old approach. I'm also perfectly fine with cases that are better off not using common.noop simply not using it. In the original PR that added it I intentionally left one of the tests unchanged specifically for that reason.

I'm -1 on this PR.

[benjamingr]

I'd just like to point out JavaScript already ships with a well known noop function - Function.prototype, that is guaranteed to take any number of arguments, return undefined and cause no side effects.

I still think a syntactical construct (like () => {}) or a built in (like Function.prototype) are clearer - but honestly all these ways are pretty obvious anyway and I don't think there is very much to gain/lose here.

[aqrln]

@jasnell should we then maybe warn about the difference between common.noop and () => {} in test/README.md?

[not-an-aardvark]

To me, it seems like if we have to document the gotchas of a noop function, it's already more trouble than it's worth.

[aqrln]

I think at this point this PR is superseded by #12822, so I'm going to go ahead and close it to clean up the backlog of PRs a bit :) Reopening is cheap if there ever be a need to do so.