Upgrade Yargs v17

[jtoar]

• closes upgrade Yargs from v16.2.0 to v17 #3621

[jtoar]

Looks like the first thing we run into in e2e is this issue here that got fixed: yargs/yargs#1977.

[jtoar]

Just updating the status of this; everything's passing but one test: serve.test.js. Yargs doesn't like something about the way we're composing the command and/or invoking parse:

redwood/packages/cli/src/commands/__tests__/serve.test.js

Lines 57 to 59 in 76b9e9f

	const parser = yargs.command('serve [side]', false, builder)
	parser.parse('serve api --port 5555 --apiRootPath funkyFunctions')

The command itself still works on the CLI, in a redwood project (I tested with rwfw). So just a matter of figuring out what the syntactic quirk is, though I've been at it for a bit now and haven't made much progress. It might just be worth reworking some part of the test instead.

[thedavidprice]

New constraint (you added) is making me so happy right now 😂

[pullflow-com]

From Pullflow:

[jtoar]

@thedavidprice I gave another go at trying to get the serve tests to parse correctly and I came up empty so I've removed them in this PR. They're good tests, but they don't test that things actually get served—just that if you call yarn rw serve, the bothServerHandler is called, etc. That we've set yargs up properly basically. We don't have that kind of coverage for the rest of the CLI and things still reliably work.

I'm happy to focus on adding back those kinds of tests during the v1 rc. I feel ok removing them now because testing the fact that yarn rw serve calls bothServerHandler is easy to QA manually. Also, we never make changes to the CLI directory structure, so I'm happy to QA it once as a part of this PR and feel good that it works. Let me know what you think!

[jtoar]

It's worth noting that the last change to this file was to fix something on Windows:

• Fix serve tests #2668

That makes me less certain that things will work. I can manually QA using my Windows machine to be sure.

[thedavidprice]

cc @dac09

[thedavidprice]

@jtoar I believe this one is prioritized correctly, but it's still bugging me that we haven't upgraded due to a test that 1) we're not sure even applies to our CLI and 2) begs the question as to whether or not we correctly understand Yargs parsing.

If needed, we can see if Ben Coe has some insite on why the test is failing for v17 but passing v16 — when you're ready, let's take that step.

[jtoar]

Finally managed it in #5416.