GeezerVison

Sounds like a regression on Node.js side

Reported - nodejs/node#60704 let's wait an answer

This also prevents Docusaurus dev server from starting: meta/docusaurus#11545

I tried this patch locally for Docusaurus, and it didn't work under Node 25.2:

×   Error: require is not defined
  │ 
  │   - dev.html.template.ejs:63
  │     /projects/docusaurus/packages/docusaurus/lib/webpack/templates/dev.html.template.ejs:63:1
  │ 

I tried adding back the "require" removed, leading to another error:

×   Error: __webpack_require__ is not defined
  │ 
  │ 
  │   - eval
  │ 
  │   - dev.html.template.ejs:10 data:text/javascript,__webpack_public_path__ = __webpack_base_uri__ = htmlWebpackPluginPublicPath;
  │     /projects/docusaurus/packages/docusaurus/lib/webpack/templates/dev.html.template.ejs:10:1

I tried adding back everything you removed and keeping the spread:

...Object.defineProperties({}, Object.getOwnPropertyDescriptors(global))

×   Error: Value of "this" must be of type nullish or must be the global object
  │ 
  │   - exposed-window-or-worker:114 global.cryptoThisCheck
  │     node:internal/bootstrap/web/exposed-window-or-worker:114:15

No idea how to fix it.

Need to double check, but I also think it fails on Node 24:

×   TypeError: Cannot set property crypto of #<Object> which has only a getter
  │ 
  │   - Object.assign

Repro?:

git clone [email protected]:facebook/docusaurus.git
cd docusaurus
yarn install
yarn start:website

@slorber Yeah, I think we will need to wait Node.js answer here, we need to pass globals here, technically we can filter as a workaround, but I still think it is a Node.js regression, let's wait an answer from them

This change in node is spec-compliant, so it's unlikely they'll revert it.

I've re-worked the fix to handle one problem that turned out to be caused by eval, however I couldn't ensure the test completes locally because eslint kept finding dozens of [objectively present] violations in the existing code that I didn't change and some of test:only suites claimed unexpected values while they looked exactly like the expected one.

The test:only has succeeded after I disabled git's autocrlf for the repo.

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@6ef547a). Learn more about missing BASE report.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@           Coverage Diff           @@
##             main    #1880   +/-   ##
=======================================
  Coverage        ?   92.38%           
=======================================
  Files           ?       18           
  Lines           ?      749           
  Branches        ?      209           
=======================================
  Hits            ?      692           
  Misses          ?       54           
  Partials        ?        3           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@tophf I still think it is a bug and should be reverted, any way I am fine with these changes too, do can fix eslint problem?

fix eslint problem

In this PR? But that code is completely unrelated, so it'd be weird to change it here. What if I just exclude the file from eslint checks altogether and someone will deal with those issues separately?

@tophf I mean fix prettier - https://github.com/jantimon/html-webpack-plugin/actions/runs/19344385065/job/55400556942?pr=1880

Yeah, I saw that. Thing is, it doesn't say what's wrong. If the project is using prettier I would expect it to be applied automatically without me doing anything. Should I run prettier explicitly myself?

Okay, I found and ran fix:prettier which fixed lint:prettier test, but I still can't commit the change due to lint:js which finds problems in this file that aren't mine, so I've just disabled the pre-commit hook and force-pushed the changes anyway.

@tophf Thanks for this PR