The pages carry the CMS markers (data block, list regions, JSON-LD price,
inline price); PackagesProvider reads the host's data block after hydration
and asks the API only when there is none; the site address comes from the
block. A stand-in for the host (config/hostInjector.ts, --cms-data in the
preview server) and browser tests check the behaviour with different data.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
With contactForm.enabled=true the e2e and Lighthouse runs asked a token route
that CI does not have (404, console error, run 666). The preview server gets an
opt-in --contact-stub (token 200 without captcha, submit 202); the e2e server
helper and the Lighthouse start command switch it on. Tests that need another
answer still replace it per page with page.route.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The preview server must behave like the real server, or Lighthouse and the
browser tests prove the wrong thing. The tests also cover the new URLs in a real
browser: one redirect, one title, a clean hydration.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The prerendered 404 page is served for other URLs, so its dehydrated match id
differs and the router renders the catch-all's pending component first. Without
one that render was empty, which gave React error 418 and duplicate head tags.
The tests run the built output in a real browser under the production CSP.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>