Skip to content

npm-website group (#442, astro 6→7 inside): opened before the Website build gate existed #485

Description

@explosivebit

Scoped while assessing v0.37.0. Not taken into that release.

Why "CI is green" doesn't mean what it usually means here

#442 shows 13/13 checks green, and Website build is not one of them. The PR was opened 2026-08-17; the website-build job in ci.yml (npm ci + astro build, catching bad frontmatter / broken content collections / dead internal links) postdates it. This PR's green checkmark predates the one gate that actually exercises what it changes.

Inside the group: astro 6.1.10 → 7.2.9 (major), @astrojs/mdx 5.0.6 → 7.0.8 (two majors), @astrojs/react 5.0.2 → 6.0.4 (major), plus 13 minor/patch bumps.

Why not merge-and-see

This repo already has a concrete precedent for what a blanket npm update does to this site: a prior blanket npm update broke the build via a peer-major mismatch (@tailwindcss/vite vs vite, astro vs vite) — the fix that came out of it was "security fixes go in by name, verified with npm run build, not as a batch" (see the site's own dependency-update convention). A two-major-version Astro jump inside an auto-generated group PR is exactly the shape that convention exists to catch before merge, not after.

What actually needs to happen before this can merge

  1. Rebase chore(deps): bump the npm-website group across 1 directory with 16 updates #442 (or let Dependabot do it) so the current website-build job actually runs against it — right now it's absent from the check list, not passing.
  2. If it goes red: don't try to fix it inside the auto-generated PR. Read the Astro 6→7 and MDX 5→7 migration notes, then apply the upgrade by hand, by name, verified with npm run build locally before pushing — the same discipline already established for this site's dependencies.
  3. If it goes green: still worth a manual npm run build + a look at the built output before merging two major versions of the framework the whole site runs on, rather than trusting the CI gate alone on its first real exercise of this exact scenario.

Why not in v0.37.0 either way

v0.37.0 is a Rust-binary release; the website ships independently of it. No reason to couple a two-major frontend migration to a backend trust-layer release with its own risk already.

Refs: #442

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions