Commit Graph

15 Commits

Author SHA1 Message Date
lilleman a9f25a7692 Remove completed todo.md + html-css-foundation mockups; strip dead §N phase refs from comments/docs (simplify visual E2E to drop the mockup-comparison oracle) 2026-06-23 22:49:28 +02:00
lilleman 1d198acc97 Canonical host via APP_URL: stop login dumping users on /error from a host mismatch
The from-scratch dev login was broken: open the banner's http://localhost:3000, sign
in as the seeded admin, and you landed on http://127.0.0.1:3000/error "Page not found".
Root cause: the banner/APP_URL said localhost but kratos.yml hard-coded 127.0.0.1, and a
host-scoped Kratos CSRF cookie can't cross localhost<->127.0.0.1, so the cross-host login
POST lost it; Kratos redirected to its error sink, which the app had no route for (404).

Make APP_URL the single source of truth for the public host:
- Canonical-host redirect (app.ts): when APP_URL is set, an off-host GET/HEAD visitor is
  308'd to it (path+query kept) before a flow starts, so the browser, the themed forms and
  the cross-origin Kratos POST share one cookie host. After /public/ so static/health
  checks stay host-agnostic; GET/HEAD only so a 308 never replays a cross-host POST.
- Opt-in (no NODE_ENV / no magic default): unset => no redirect, so a prod deploy that
  forgets APP_URL can't bounce real users to a stale default. The dev stack sets it.
- kratos.yml browser URLs default to localhost (match APP_URL's dev value) and derive from
  ${APP_URL} via compose.override.yml; SERVE_PUBLIC_BASE_URL keeps the dev Ory port.
- Real /error page (views/error.ejs) replaces the catch-all 404 for genuine flow errors.

Tests-first: config (opt-in/validated), app (308 on mismatch, no-redirect on match, static
host-agnostic, POST untouched, /error page), updated kratos.test host pins. New devstack
regression (e2e/devstack-login.spec + compose.e2e-devstack.yml) drives the plain
docker-compose-up topology on the host network: login from localhost works and 127.0.0.1 is
canonicalised; wired into scripts/ci.sh. typecheck + 360 units + full ci.sh (visual 10 ·
auth 1 · oauth 2 · full 7 · devstack 2) green.
2026-06-21 21:52:20 +02:00
lilleman bea9a71d6f §9 trace all fetch + ENV service name + leveled logging (todo §9 follow-up); route every outbound fetch through the request logger, make the OTLP service name implementer-configurable, and add proper leveled logging throughout. An AsyncLocalStorage<Log> makes the per-request logger ambient (runWithLog/currentLog), so all outbound fetch traces with no signature churn: tracedFetch (a typeof fetch) routes through the active request log (client span + propagated W3C traceparent) for string/URL inputs, else plain fetch; server.ts wires it under the Ory timeout into every Kratos/Keto/Hydra + JWKS call (timeout still honoured — log.fetch spreads {...init,headers}). RequestContext gained ctx.log (request logger; additive/contract-stable, silent default) so a handler/plugin logs in-trace and ctx.log.fetch(url) traces upstream calls; the reference plugin's createUpstream defaults to tracedFetch and its handlers log via ctx.log; plugin-api.ts exports tracedFetch + the Log class. SERVICE_NAME (config + createLogger({serviceName})) brands the OTLP service.name. Leveled logging: who-did-what audit info lines on every admin write (user/group/role/client create·delete·assign — actor/target, no secrets), info on login (session mint) + logout, warn on missing-role 403 + CSRF rejections + Ory-unreachable, debug on a JWKS kid-miss reload. app.ts's handler body was extracted to handleRequest run inside runWithLog; end() now fires exactly once after BOTH the handler unwinds AND the response closes, so a client abort mid-handler can't end the log out from under a still-running ctx.log/tracedFetch (regression-tested) and the happy-path access line is never dropped. bootstrap.ts wraps main in runWithLog + traces the seed calls. Tests extended (logger: serviceName/runWithLog/currentLog/tracedFetch-continues-trace; config: SERVICE_NAME; context: ctx.log default+passthrough; app: ctx.log in-trace + ctx.log.fetch propagation + the abort race; plugin-api: tracedFetch+Log). Stability-reviewer: APPROVE, no Critical/High (fixed the abort-race end(); green nits addressed). docs/plugin-contract.md (ctx.log/ctx.log.fetch/tracedFetch) + README (config, Observability tracing/serviceName, plugin note, Layout) updated. typecheck + 333 units + the full scripts/ci.sh E2E gate green (326 → 333). 2026-06-20 15:46:48 +02:00
lilleman a9e3dedbb4 §9 structured logging + OTLP observability (todo §9); structured, OTLP-native logging on @larvit/log (2.3.0, pinned; itself zero-dependency — the one new runtime dep). New pure src/logger.ts: createLogger() builds one app Log tagged service.name=plainpages (level/format/OTLP from config, injectable stdout/stderr); requestLogger() clones it per request (own root trace, inheriting level/format/streams/OTLP) into a "request" span, adopting an inbound W3C traceparent so a request continues an upstream proxy's distributed trace (malformed ⇒ fresh trace; clone honours a passed traceparent while dropping the parent's, unlike parentLog). app.ts builds the per-request log at the top of the handler and on res "close" (fires on completion AND abort, unlike "finish") emits one access line (method/path-without-query/status/ms/requestId, guarded) then end()s to flush the span (fire-and-forget .catch — a flaky collector never crashes a served request); the catch-all 500 + Ory-unreachable re-mint now log via reqLog.error/warn; static.ts mid-stream error takes an injected onError. server.ts builds the app logger, logs discovery/listen/shutdown, end()-flushes on SIGTERM/SIGINT (re-entry-guarded). bootstrap.ts events go structured (the human first-run banner stays raw). Config (environment-agnostic, fail-loud): LOG_LEVEL (info), LOG_FORMAT (text; prod compose → json), OTLP_ENDPOINT (unset ⇒ console-only; set ⇒ export logs + spans to an OTel Collector), OTLP_PROTOCOL (http/json|http/protobuf). compose: base sets LOG_FORMAT=json, dev override flips it to text. Tests-first: logger.test.ts (service.name/severity/level-gate/format, OTLP-only-when-endpoint, a stubbed-fetch proof it POSTs /v1/logs, requestLogger context-merge/own-root-trace/traceparent-continue/malformed-ignored), config.test.ts (4 toggles + validation), app.test.ts (live request emits the JSON access line), compose.test.ts (prod json / dev text). Stability-reviewer: APPROVE, no Critical/High (addressed both yellow nits — guarded access line + "finish"→"close" so aborted requests log; shutdown re-entry guard — and the green ones). README (config table, new Observability section, Status, Layout, runtime-deps) + AGENTS (deps) updated. typecheck + 326 units green (317 → 326). 2026-06-20 02:11:10 +02:00
lilleman a8a018f3e5 §9 optional revocation denylist (todo §9); closes the documented ~10m role/session lag for security-critical revoke, off by default (REVOCATION_DENYLIST, zero hot-path cost + zero behaviour change when off). New pure src/denylist.ts (createDenylist({ttlSec})): an in-memory, auto-evicting Map<sub, revokedAt> — revoke(sub) records now, isRevoked(sub, iat) rejects a subject's tokens minted at/before the revoke (iat <= revokedAt; missing iat fails closed), so a fresh re-login (iat after the revoke) passes while a downgrade lands immediately. Entries self-evict after REVOCATION_TTL_SEC (default 900 ≥ the 10m tokenizer TTL + skew), so it stays a bounded cache like JWKS — no database, Keto stays off the hot path. Wired: jwt-middleware.ts takes the denylist in VerifyOptions and throws TokenError(expired) on a revoked sub, so resolveSession routes it through the existing §4 re-mint (live session → fresh post-revoke JWT with current Keto roles; dead/deactivated → cleared cookie). app.ts merges it into authOptions (the same resolveSession hot-path call) and hands a bound revoke to the Users + Roles admin deps; admin-users.ts revokes on deactivate/delete, admin-roles.ts revokes a direct user: member on assign/unassign (a group:/whole-role change is transitive → left to lag, documented). server.ts builds it only when the toggle is on. Tests-first: denylist.test.ts (iat semantics, cutoff-advance, TTL eviction), jwt-middleware.test.ts (revoked→expired→re-mint, fresh passes), config.test.ts (toggle + posint TTL), app.test.ts (hot-path reject + fresh-login pass; admin deactivate/role-assign/unassign record the revoke). Stability-reviewer on the diff: APPROVE, no Critical/High/Medium (addressed its one Low). Per the §9 security-headers precedent, covered by unit + app-HTTP integration (no new browser E2E — no new user-facing page). README (Auth trade-off + new "Instant revoke" subsection, config table, Layout) updated. typecheck + 317 units green. 2026-06-20 01:38:20 +02:00
lilleman a20f3507e0 §8 review convergence (todo §8); re-ran the architecture + product reviewers to convergence — 5 rounds, until both returned zero new actionable findings. Fixed across rounds 1-4 (tests-first): bounded every outbound Ory fetch with a timeout (src/fetch-timeout.ts withTimeout + ORY_TIMEOUT_SEC default 5, incl. the http JWKS fetch) so a hung Ory can't park a request handler; anonymous on a permission-gated plugin route now 303→/login (was a dead-end 403; signed-in-without-role still 403); an already-signed-in user is sent home from /login + /registration; the onRequest hook short-circuit now sets the fresh CSRF cookie; admin-users malformed :id → 404 (was 500) via safeDecode; parseJwks validates key element shape (fails loud at load); removed the dead COOKIE_SECRET (loaded + enforced + documented but never read); documented HYDRA_ADMIN_URL; admin recovery shows the code + links to the public /recovery instead of the browser-unreachable admin-API link; reference-plugin breadcrumb-label + pagination/datetime README notes; corrected the contract doc to not over-promise a post-login "retry". Declined: unconditional base-ctx chrome (would build the menu per request, regressing the lazy hot path). Deferred → §9: return_to-preservation for deep-link login. Stability-reviewer on the cumulative diff: APPROVE, no Critical/High (addressed its Low nits). typecheck + 310 units + the full scripts/ci.sh gate (visual 9 · auth 1 · oauth 2 · full 6) green. 2026-06-20 00:42:23 +02:00
lilleman 3c8090e8e3 Built-in OAuth2 login-challenge handler (todo §6); /oauth2/login resolves a Hydra login challenge via the Kratos session — skip→accept(subject), live session→accept(identity id), no session→bounce to /login?return_to back here so Kratos lands on the challenge once signed in. New src/hydra-admin.ts (fetch client: get/accept/reject login request + HydraError, mirrors the kratos/keto clients) + src/oauth-login.ts (pure resolveLoginChallenge); wired in app.ts (the absolute return URL derives from the request Host + the SECURE_COOKIES scheme — a spoofed Host can't escape, Kratos validates return_to against its allow-list; /login now bakes return_to into the flow init), config.hydraAdminUrl (default http://hydra:4445), server builds the client, compose web now gates on hydra healthy (the app consumes it). A stale/invalid/consumed challenge (Hydra 4xx — back button, slow login) degrades to a recoverable 400, not a 500; a genuine Hydra 5xx outage still surfaces as 500. Tests-first: hydra-admin/oauth-login units + app/config/compose HTTP integration + full-stack e2e/oauth-login.spec.ts (compose.e2e-oauth.yml — registers an OAuth2 client, starts an auth flow, asserts the unauthenticated bounce and the authenticated accept; boot-verified then torn down). Stability-reviewer run as a local PR: APPROVE, no Critical/High; addressed its one warning (4xx→400 degrade). Deferred §9: document that prod allowed_return_urls entries must be exact origins with a trailing /. typecheck + 253 units + 8 visual + oauth-login E2E green. Consent handler + client registration are the next §6 items. 2026-06-18 21:45:24 +02:00
lilleman b5af4ba6cd E2E for token timeout + refresh (todo §4); full-stack auth-refresh.spec.ts (real Ory stack): a lapsed session JWT is silently re-minted from the live Kratos session (roles re-read from Keto), and cleared once the session is revoked; ory/kratos/e2e.yml shortens the tokenizer ttl to 8s + adds JWT_CLOCK_SKEW_SEC config so re-mint fires at expiry; scope visual suite to visual.spec.ts 2026-06-18 11:32:23 +02:00
lilleman 4b2173cb84 Secure cookie flags + CSRF for our own POST forms (todo §4); SECURE_COOKIES toggle on session/CSRF cookies; csrf.ts signed double-submit token + body.ts form reader; logout is now a CSRF-guarded POST form 2026-06-18 11:12:32 +02:00
lilleman c8b56b85eb JWT session middleware (todo §4); authenticate(): verify the session cookie via cached JWKS (key by kid) → exp/nbf/iss/aud claims (clock skew) → ctx.user/roles; iss/aud opt-in; fail-closed 2026-06-18 09:53:37 +02:00
lilleman 360449e76b Tighten §3 comments (todo §3); drop stale 'next §3 item' forward-refs, condense compose/Ory/bootstrap headers 2026-06-17 17:00:47 +02:00
lilleman e83cf4da88 Address project-wide review (todo §3); fix JWKS_URL default → tokenizer signing key + read-only web mount, cap bootstrap restart, --no-deps for unit commands 2026-06-17 16:49:37 +02:00
lilleman a070362649 Drop NODE_ENV for explicit config toggles (todo §0.1); app is environment-agnostic 2026-06-15 10:53:33 +02:00
lilleman 1fb6f23805 Tighten code comments + README (todo §0): denser, drop redundant prose; no behavior change 2026-06-15 10:30:06 +02:00
lilleman 0bc7998cfe Add env/config loader (todo §0); validate at boot, wire port into server 2026-06-14 19:46:26 +02:00