GeezerVison

Review requested:

• @nodejs/web-infra

CI: https://ci.nodejs.org/job/node-test-pull-request/55258/

Separated from this (which I think should land ASAP), I wonder if we should just skip doc generation/tests in without-intl builds

Fast-track has been requested by @joyeecheung. Please 👍 to approve.

Separated from this (which I think should land ASAP), I wonder if we should just skip doc generation/tests in without-intl builds

That was a suggestion in #35942 (comment), which references #41091 which I'm now confused over as it suggests we shouldn't be attempting to generate docs without intl available? Maybe either something was missed there or we've subsequently changed something in how the docs are generated that means they're no longer being skipped? 🤔
(My recommendation would be to land this to unbreak the CI and then investigate separately.)

CI: https://ci.nodejs.org/job/node-test-pull-request/55260/

Landed in 67b1383

Maybe either something was missed there or we've subsequently changed something in how the docs are generated that means they're no longer being skipped?

Do we have other Node.js installations available in the system? #41091 checks the availability of intl using whatever Node.js binary available in the system (out/Release/node just takes precedence). If somehow there's a node executable with Intl on the path, that check could be invalid

hmm, I think this is caused by a missing dependency check in test/addons/.docbuildstamp

So it seems the cause is that:

1. We parse addons.md to generate addon tests
2. The parser used to parse addons.md depends on i18n to work (via remark-gfm?)

Not sure how this can be addressed easily. I think we do still want to test the addons in addons.md in without-intl builds. Perhaps we can just use some dumb marker in addons.md to denote where the code examples are and write a custom parser instead of using remark.