chore(deps): update chokidar to v4

[ggodlewski]

PR Checklist

• Addresses an existing open issue: fixes Chore: bump chokidar to v4 #5214
• That issue was marked as status: accepting prs
• Steps in CONTRIBUTING.md were taken

Overview

Current chokidar is bundled with tons of spam deps including braces which trigger some audit alert.

All test are working. It supports globs unlike #5215

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: ggodlewski / name: Grzegorz Godlewski (d214cda, 6893764, f9fa7e8, 90d7cce, 65eee8d)

[JoshuaKGoldberg]

Hey thanks for getting this started! Sorry it took so long to review - since Chokidar didn't end up needing to be bumped with v11, this fell by the wayside. But I'm excited to help push this forward along now! 🚀

Requesting changes primarily on a performance point, as well as a design point. Happy to talk if you want. 🙂

[ggodlewski]

Hey @JoshuaKGoldberg, there is no way I remember the details after a month :). All I know it's been a struggle to get all tests working. In the meantime we switched to ava to resolve our issues, so I can't do much.
I updated the code to fix your issues. If there is anything that I can do fast let me know.

[talentlessguy]

@JoshuaKGoldberg is there any blocker to this PR besides merge conflicts of a lockfile?

[JoshuaKGoldberg]

Just a re-request for review. I'll add it to the queue for review :)

[JoshuaKGoldberg]

Source code looks good, thanks for the refactor!

I finally found time to really dig in and try this in production - and think I found a performance bug. LMK what you think of it.

[JoshuaKGoldberg]

[Bug] I tried this PR out on real-world code and found a performance issue from files being renamed. I found it on the ESLint codebase but also reproduced it in isolation here: https://github.com/JoshuaKGoldberg/repros/tree/mocha-chokidar-v4-pr-watching. Could you please take a look?

tl;dr: in a minimal codebase with two root-level test files, re-watching is reported as taking >200ms

[ggodlewski]

How exactly are you renaming the file? I run mv b.test.js z.test.js and I don't have the state with 0 passing:

$ node ../mocha/bin/mocha.js "*.test.js" --watch
  a
    ✔ passes
  b
    ✔ passes
  2 passing (2ms)

ℹ [mocha] waiting for changes...

  a
    ✔ passes
  b
    ✔ passes

  2 passing (1ms)

ℹ [mocha] waiting for changes...

[JoshuaKGoldberg]

Interesting. I did it through the VS Code UI. This is on an m1 Mac.

[ggodlewski]

I'm running Ubuntu. I have no hardware to reproduce it. It's strange because the tracker.regenerate(); is only triggered on event === 'add'. I don't see how the list of files can be empty during rename of 1 file. Maybe you can add some debug output so we can trace the tracker method calls?

[JoshuaKGoldberg]

👍 added to the README.md. Interestingly, the long delay of ~200ms only happens if an npm install populated node_modules/. If you rm -rf node_modules/ then the delay is only around 20-30ms.

[ggodlewski]

I added one more improvement. Glob should not traverse node_modules. Can you check if it improves anything?

[JoshuaKGoldberg]

💯 that fixed it for me! Back to 0-1ms on the middle waiting.

+1 that it's a generalized solution too, not specific to node_modules. Nice 🙂

[JoshuaKGoldberg]

🔥 This is fantastic, thanks a bunch @ggodlewski! I really appreciate you sticking through & working with me on the reviews.

I've played with this a bunch locally on relatively small and large repros, and haven't found any issues. I'd really prefer to get another review from someone on @mochajs/maintenance-crew just to be safe - but otherwise ✅ on my end.

Nice job!