Skip to content

refactor(adapter-vite): narrow the public build interface#464

Draft
SisyphusZheng wants to merge 3 commits into
mainfrom
codex/deep-refactor
Draft

refactor(adapter-vite): narrow the public build interface#464
SisyphusZheng wants to merge 3 commits into
mainfrom
codex/deep-refactor

Conversation

@SisyphusZheng

Copy link
Copy Markdown
Member

What changed

  • refresh the Mastodon dogfood lockfile to the current five-package 0.41.0-alpha.15 graph
  • reduce @openelement/adapter-vite from 14 exports to the documented root, nitro-mount, and cli/build interfaces
  • remove the unused sitemap re-export from generated SSR entries so runtime bundles do not load the Vite adapter root
  • synchronize package docs, migration notes, resolver generation, and the public-interface snapshot
  • make export-files:check compare generated content instead of rejecting every legitimate PR diff

Why

The documented package surface and published export map had drifted apart. Internal plugin factories and individual build phases were still importable, leaving a shallow public interface. The generator check also treated any tracked change as stale, preventing intentional interface changes from passing CI.

Impact

This is an intentional Alpha breaking change. Consumers of removed adapter subpaths must import supported helpers from the package root or use the generated @openelement/create lifecycle. The five-package product topology is unchanged.

Validation

  • frozen dependency install
  • format, lint, typecheck, architecture, package graph, type safety, Deno API, signal seam, docs truth, package surface, interface snapshot, and generated export checks
  • 846 Deno tests and coverage thresholds
  • website SSG build and artifact truth
  • npm pack dry-run, package artifacts, packed starter, Deno/Node consumers, third-party Web Components
  • Nitro Node and Workers proofs
  • Chromium: 146/146 E2E
  • Firefox: 146/146 E2E
  • WebKit: 35 tests completed in the combined run before it was stopped for abnormal duration; one page.goto() timeout was flaky and passed 1/1 when rerun alone

Remaining risk

The complete WebKit suite was not allowed to run to completion after 13.5 minutes because navigation became abnormally slow. The sole observed timeout reproduced as passing in isolation and did not exercise the changed package-interface path.

@github-actions

Copy link
Copy Markdown
Contributor

APIError: Insufficient Balance

opencode session  |  github run

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant