WORK-439
ID:WORK-439Status:ready

Visual-regression compare-against-base CI job

Deferred from WORK-409 (its final acceptance criterion). The harness runs end-to-end locally and in the pinned Playwright container; what's left is the CI automation, which needs a browser-backed runner to develop against — hence its own item, unassigned to a milestone until a browser CI environment is in place.

Priority:mediumComplexity:moderateSource:SPEC-094
claude/release-v0.22.0 View source

Criteria completion

Criteria completion: 0 of 3 (0%) checked; tracking started on Jun 15, no incremental history yet0%25%50%75%100%Jun 15Jun 15

Tracking started Jun 15 — check back for trends.

Relationships

Implements 1
Related 2
Branches 1
claude/release-v0.22.0 current ready
main ready
History 1
  1. f27a573
    Created (ready)by bjornolofandersson

Scope

  • A CI job that renders the gallery on the PR's merge-base and on the PR head, diffs the two, and reports the result (using @refrakt-md/gallery-harness).
  • Informational by default: a non-empty diff uploads the diff images / posts a summary ("N cells changed across hint, badge; 1 layout") and passes — visual change is usually legitimate, so the diff is a review aid, not a gate.
  • Opt-in expect-empty mode for refactors that assert zero visual change (e.g. the WORK-438 skeleton/skin re-bucketing, pure token plumbing), where a non-empty diff is a failure.
  • Run in the pinned container (mcr.microsoft.com/playwright:v1.60.0-jammy); the core CI / install path stays browser-free.

Acceptance Criteria

  • A compare-against-base CI job renders gallery on merge-base + head, diffs, and reports (diff images / summary); informational by default (non-empty diff passes).
  • An opt-in expect-empty mode fails on a non-empty diff, usable by inert-refactor PRs.
  • Runs in the pinned Playwright container; no browser binary enters the core CI/install path.

References

  • WORK-409 (the harness + the documented local workflows) · SPEC-094 · packages/gallery-harness/README.md (CI section).