🧹 Versioning: Bump minimum Node.js version from 18.18.0 to 20.19.0

[JoshuaKGoldberg]

Overview

We're gearing up for a new major version of Mocha, Mocha 12 (#5357). That means we can bump the minimum Node engine version! Hooray! This is great for a couple of reasons:

• It unblocks bumping old package versions, e.g. 📦 Package: Bump supports-color from 8 to 10 #5335
• It means we can use newer JavaScript + platform features

Node.js 18 went EOL April 30 2025. Node.js 20 is still LTS for another year, through April 30 2026. Node.js ^20.19.0 || >=22.12.0 is the minimum version that supports unflagged require(ESM).

So, our current thought for Mocha 12 is to raise the minimum Node.js version to ^20.19.0 || >=22.12.0.

Note: see #5226 -> #5354 for our policy on supported browser versions (still in review at time of posting). Node.js 20.0.0 was released on April 18, 2023, so the equivalent browser versions would be:

• Chrome 112
• Edge 112
• Firefox 112
• Safari 16.4

[voxpelli]

Eg ESLint won't be able to upgrade to this version until their next major.

I think this is too soon for us.

[JoshuaKGoldberg]

Is there a list of exemplar users you'd want to wait for to signal that enough of the ecosystem has bumped? Last time, we'd confirmed that TypeScript is ok with staying on an older version of Mocha: https://github.com/mochajs/mocha/pull/5216/files#r1792135513

IMO there are enough benefits to dropping 18.X that we should jump ahead and do it. ACK that some users who are stuck supporting older versions of Node.js will also be stuck on Mocha 11. But considering the list of improvements directly blocked by it:

• Cleaning up require() logic post feat: use require to load esm #5366
• 📦 Package: Bump glob from 10 to 11 #5331
• 📦 Package: Bump log-symbols from 4 to 7 #5332
• 📦 Package: Bump strip-json-comments from 3 to 5 #5334
• 📦 Package: Bump supports-color from 8 to 10 #5335

...and that's assuming we don't also block Mocha 12 on dropping 18.x. That's quite a lot of goodness to keep pending.

[mark-wiemer]

Hmm, there are obvious upsides to dropping support for an EOL runtime, but if it means downstream users may be limited I agree we should be cautious. Do we know when ESLint plans to drop support for Node 18?

I'll also mention that these package bump issues aren't too exciting to me: most of them mention directly that they're "just good ecosystem cleanup" and I'm not sure we need to rush into those. But I do see value in supporting native TS and other Node 20+ stuff. But I see a lot of other work to focus on while we wait for ESLint and others to drop Node 18 support, like bugfixes, updating the docs site, ensuring we exit with POSIX codes instead of the number of failing tests, etc. None of that requires moving to Node 20 yet

[JoshuaKGoldberg]

Do we know when ESLint plans to drop support for Node 18?

Asked in Discord: https://discord.com/channels/688543509199716507/1085655150691102751/1384534520023220274

That would happen in ESLint v10. While we don't have a set timeline for that, this is the last thing we are waiting for to start planning: eslint/rewrite#223

So, hopefully around September/October would be my guess.

Also, I'd say go ahead with Mocha. There's no reason we can't just stay on the current version until the next major release.

I actually find the timeline note quite encouraging on its own, even without the supportive note. That's only 4 months from now. Presumably we'd want a few months of beta testing for 12 anyway.

other work to focus on ... requires moving to Node 20 yet

Yeah. The part that motivates me to push for dropping <20.19 is that I don't want Mocha to have to wait for an additional major version to do so. Right now we have a nice plan with the post-10 majors:

1. 11: purely bumping the Node.js platform and a package version (📌 Mocha 11 Release Plan #5249)
2. 12: that, as well as ironing out deprecated code (📌 Mocha 12 Release Plan #5357)
3. 13: actual behavior changes such as the exit code (milestone/67 v13.0.0)

Moving the Node.js platform bump from v12 to v13 leaves not very much in v12. The existing issues in milestone/66 v12.0.0 are:

• 🐛 Bug: Pending constructor does not inherit from the Error constructor #2228
• 🐛 Bug: remove util.inherits / convert to classes #5025
• 🚀 Feature: Remove the long-deprecated --compilers option #5242
• 🚀 Feature: Remove the long-deprecated --opts option #5243

...which is really not much. I think we're effectively stuck between 2-3 options:

• Having a relatively small (just those issues) Mocha 12 now, then either...
  • Using Mocha 13 to both drop <20.19 & bump dependencies and include actual behavioral breaking changes
  • Moving the actual behavioral breaking changes even further in the future to Mocha 14
• Taking on the <20.19 drop now for Mocha 12, so Mocha 13 can have the actual behavioral changes without platform changes

It's not terrible if we wait an extra year to get those behavioral changes. Stability for Mocha is nice after all. But I really think being able to have a clean slate of relatively-updated dependencies -especially give how rare the CJS->ESM migration is- is worth it sooner.