Ledger

WEB-600

Cross-app: wire the Live controls site switches (sign-ups, bug form) into the database and the site, refresh the site on switch changes; integration harness running site + admin together

← Ledger

donestatus · web-platform

5✓ 4✗tests passed / failed

34.85Mtokens · 21 min

0evidence files

Evidence

From logs/evidence/WEB-600/. Click a thumbnail for the full image.

No evidence files in logs/evidence/WEB-600/.

Tests

fail pgTAP incl. 180_site_switches (pause sign-ups and bug form → refused, public settings show it, refresh queued, switching back reopens) · 1.6 s · raw log$ pnpm db:test

pass pgTAP incl. 180_site_switches after splitting the guard triggers per table · 1.5 s · raw log$ pnpm db:test

fail web:e2e:integration — site + admin production builds together: publish → site at once (webhook delivered) → archive removes; main button; Live controls bug-form switch pauses the site form; maintenance 503 and back; site bug report → console intake → reject · 154.5 s · raw log$ pnpm web:e2e:integration

pass unit: drainOutbox (batches until empty, ≤ 50 tags per signed request, nothing dropped, stops on failure) + admin typecheck + worker tests · 5.9 s · raw log$ pnpm --filter @zoen/db test && pnpm --filter @zoen/admin typecheck && pnpm ai:test

fail web:e2e:integration after the outbox drain, chunked tags and switch delivery fixes · 99.2 s · raw log$ pnpm web:e2e:integration

fail web:e2e:integration with maintenance served as a real 503 (all five cross-app checks) · 69.0 s · raw log$ pnpm web:e2e:integration

pass web:e2e:integration (publish → site → archive, main button, Live bug-form switch, maintenance as a real 503 and back, site bug report → intake → reject), cleanups fail loudly · 74.7 s · raw log$ pnpm web:e2e:integration

pass site e2e (full, all sizes) after the settings mapping, support pause and proxy 503 changes · 91.5 s · raw log$ pnpm site:e2e

pass admin e2e: live + website + system specs after the outbox and switch changes (all sizes) · 82.3 s · raw log

$ cd apps/admin && pnpm build >/dev/null 2>&1 && pnpm exec playwright test tests/e2e/live.spec.ts tests/e2e/website.spec.ts tests/e2e/system.spec.ts --workers=1 --reporter=line

Notes

2026-10-04 22:57 UTC · W6 cross-app harness (pnpm web:e2e:integration: both production builds, the admin's webhook pointed at the test site) found and fixed real bugs: (1) the site merged camelCase api.site_settings keys onto snake_case defaults, so the main button, route toggles, sign-ups open and servers-live never reached the site — now mapped key by key; (2) the Live controls site switches (sign-ups, bug form) were stored but enforced nowhere — now refused by insert guards, exposed in public settings, the site shows the bug form as paused, and a flip queues a site refresh (setSwitch now delivers); (3) outbox delivery claimed only the oldest 50 rows once, so a backlog (51 rows from earlier runs) held back fresh changes, and signRevalidation silently cut tags beyond 50 — shared drainOutbox (packages/db) now drains batch by batch with ≤ 50 tags per signed request, used by admin and worker, unit-tested; (4) site maintenance was served as HTTP 200 (rewrite keeps the prerendered status) — the proxy now serves the page as a real 503 + Retry-After; (5) service-role cleanups of site_settings failed silently (check helper not executable) — granted, and test cleanups now throw. Results: integration 5/5, site e2e 39/39, admin live/website/system 29/29, pgTAP 18 files, drain unit test. Observed, not fixed: a benign Next cache-warming race on /status during bursts of tag refreshes (regenerates on the next request). Security note for hosting: the site rate-limits by the first X-Forwarded-For entry, which a client can spoof unless the proxy overwrites it.