Can n better determine the latest version of node when upstream downloads don't exist?

[cviebrock]

Problem

Node.js made some kind of mistake and published 18.20.5 without actually building the release files in the expected directory. See nodejs/node#55829. Any CI pipeline or tooling that uses n to download the latest 18.x version fails because of this. e.g.:

> if [ $DEV_MODE -eq 1 ]; then n --download engine && n exec engine yarn install; else n --download engine && n exec engine yarn install --production; fi
       found : /var/lib/jenkins/workspace/company_project_PR-2172/package.json
        read : ^18.13.0
      target : 18
  installing : node-v18.20.5
curl: (22) The requested URL returned error: 404 

  Error: download preflight failed for '18.20.5' (https://nodejs.org/dist/v18.20.5/node-v18.20.5-linux-x64.tar.xz)

Obivously this isn't an issue with n per-se ... but I'm wondering if there is something that could be done on the n side to not have errors like this affect users. I notice, for example, that https://nodejs.org/dist/latest-v18.x/ redirects/links to https://nodejs.org/dist/v18.20.4/, so maybe the way that n determines the latest version could be adjusted somehow?

Thanks!

Short Version

Is there a way to change n so that upstream issues like missing download files don't break users' pipelines?

Configuration Details

$ n --version
v9.2.3

$ command -v node
/Users/cmv/.n/bin/node

$ node -p process.platform
darwin

[Nxtra]

Same issue here:

[Container] 2024/11/12 23:59:26.994824 Running command n 18
--
26 | installing : node-v18.20.5
27 | curl: (22) The requested URL returned error: 404
28 |  
29 | Error: download preflight failed for '18.20.5' (https://nodejs.org/dist/v18.20.5/node-v18.20.5-linux-x64.tar.xz)

[shadowspawn]

This may have been fixed already on the node side. Node.js just switched over to a new distribution method (using CloudFlare) within the last couple of weeks, and this might have been a teething problem with the new setup.

My usual expectation is that index.tab is the source of truth for what is available.

Thanks for node issue link. I have subscribed there to follow developments.

[shadowspawn]

For historical interest, switched from web page scraping to using index.tab here: #578

[shadowspawn]

This was a transient problem due to a (hopefully) one-off problem with the new Node.js CDN and release process.

In general, n trusts the available meta-data. And almost always, that is accurate!

One general-purpose idea I had was adding an option for a cool-down period to skip over recent releases.

I don't think there is a common enough problem with a good enough solution to act on for this issue.