Problem
The render-review artifact may not represent the actual publication container. In the heat-pump test, the interactive review still was captured at a 900×560 browser viewport, while the accepted destination was article web. The screenshot omitted the source footer, yet review still proceeded.
A visually clean screenshot at an arbitrary size does not prove that the deliverable is legible or complete at the destination size.
@rmdms
Suggested solution
Make review capture destination-aware and provenance-check the viewport.
- Resolve the primary review dimensions from the accepted deliverable/channel and the newsroom CMS/embed profile.
- For responsive interactives, test the configured article container plus documented narrow and wide breakpoints.
- Capture the complete component root, including title, unit, plot, source, credit, interaction affordance, and other required furniture.
- Record CSS viewport, device scale factor, captured root dimensions, destination/deliverable ID, and artifact hash alongside every review image.
- Add deterministic furniture visibility and overflow checks; a footer outside the component contract should fail rather than disappear from the still.
- Keep interaction tests separate, but run them at the same destination breakpoints where practical.
- Make
review-gate refuse a still whose recorded target does not match the accepted delivery profile.
Avoid assuming one universal “article web” rectangle: the newsroom delivery profile should supply its real embed width/height or responsive contract.
Acceptance criteria
- The primary review image matches the configured publication container.
- Required furniture is included and visible in the captured component.
- Responsive outputs are checked at primary, narrow, and wide target breakpoints.
- Review metadata records viewport, DPR, root dimensions, destination, and artifact hash.
- A viewport/destination mismatch blocks review approval.
- Tests reproduce the 900×560/source-footer failure and prove it is caught.
Problem
The render-review artifact may not represent the actual publication container. In the heat-pump test, the interactive review still was captured at a 900×560 browser viewport, while the accepted destination was article web. The screenshot omitted the source footer, yet review still proceeded.
A visually clean screenshot at an arbitrary size does not prove that the deliverable is legible or complete at the destination size.
@rmdms
Suggested solution
Make review capture destination-aware and provenance-check the viewport.
review-gaterefuse a still whose recorded target does not match the accepted delivery profile.Avoid assuming one universal “article web” rectangle: the newsroom delivery profile should supply its real embed width/height or responsive contract.
Acceptance criteria