apps/api/src/auth/linkOidc.ts (new):
- OidcLinkConflictError: thrown when iss+sub already belongs to a different user
- linkOidcToUser(userId, iss, sub): preflight SELECT for conflict, then db.transaction
(UPDATE users SET oidc_iss/sub/claimed + DELETE local_credentials); atomic, no email (D-10)
- 88 lines; no email in source (D-10/T-19-08 assertion passes)
apps/api/src/routes/me.ts:
- POST /api/me/link-oidc: resolveUserId (401 if null), sign state JWT
({ linkUserId, nonce, iat, exp } HS256 with LOCAL_SESSION_SECRET, 10-min window)
- Returns { signedState, authorizationUrl } — 19-03 /callback reads linkUserId from state
- authorizationUrl constructed from OIDC env vars when configured, null otherwise
- T-19-09: per-request nonce in state prevents CSRF/replay
- POST /api/me/password: verifyPassword(current) gate before hashPassword(new) update
- 401 on wrong current password, 404 if no local_credentials row, 200 on success
- meNoEchoHook on /password route (T-19-06, never echo submitted password)
- resolveAdminAndSetupStatus extended with hasLocalCredential (AUTH-LOCAL-17)
- GET /api/me response includes hasLocalCredential alongside isAdmin/needsProviderSetup
- POST /api/admin/members: atomic tx (users + local_credentials), 409 on dup username
- POST /api/admin/members/:id/password: admin reset (no current-pwd required), 404 if no local cred
- GET /api/admin/members: LEFT JOIN local_credentials, hasLocalCredential in each member row
- noEchoHook on both POST routes (T-19-06); requireAdmin via router.use('*') remains first statement
- Dup-entry detection via error message string match (Drizzle wraps mysql2 ER_DUP_ENTRY)
- node:crypto scrypt (N=16384, r=8, p=1, 32-byte output) — zero new dependencies (D-08)
- 16-byte random salt per hash; PHC-encoded format: scrypt$N$r$p$salt_b64url$hash_b64url
- timingSafeEqual for constant-time comparison (prevents timing oracle attacks, T-19-01)
- verifyPassword returns false on any error (never throws); passwords never logged
- All 5 unit tests pass (round-trip, wrong-pw, unique-salt, malformed-hash, PHC-shape)
- status returns { setupComplete, dbName } from process.env.DB_NAME (null fallback)
- only the DB name; never DB_HOST/DB_USER/DB_PASSWORD
- SetupStatusResponse carries dbName?: string | null for the PWA read-only field
- read app_config.vapid_public_key and compare to process.env.VAPID_PUBLIC_KEY
- mismatched/absent submitted key → 400 before the structural check
- VAPID_PRIVATE_KEY still env-only, never compared or returned (T-12-06)
- apps/api/src/auth/user.ts: upsertUser step-5 insert now sets claimed=true
for all OIDC-created users. An identity-bound OIDC user is never a pending
wizard bootstrap user; explicit claimed=true prevents ambiguity with the
(oidcIss IS NULL AND claimed=false) sentinel used by the TOCTOU guard and
isSetupLocked. First-login-claims path is unaffected (it updates a
pre-existing oidcIss=null row; this change only touches the fresh insert).
- apps/api/src/routes/setup.ts: TOCTOU guard in POST /credential now queries
WHERE oidc_iss IS NULL AND claimed = false FOR UPDATE, matching the exact
definition of a pending wizard bootstrap user. This provides defense-in-depth
against any future path that could produce claimed=false OIDC rows.
- apps/api/tests/auth/user.test.ts: new WR-01 test asserts that the fresh
OIDC insert sets claimed=true in the values passed to db.insert().
- apps/api/tests/routes/setup.test.ts: new WR-01 integration test seeds an
OIDC user with claimed=false (oidcIss NOT NULL) and verifies POST /credential
still succeeds (guard ignores the OIDC row, only counts local wizard rows).
All 402 API tests, 253 PWA tests, and typecheck pass.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
appExternalUrl is injected as OIDC_AUTH_EXTERNAL_URL (the redirect URI
base); Authelia rejects non-https redirect URIs in production. Added
.refine() guard matching the existing oidcIssuer pattern. Added test
that verifies http:// appExternalUrl is rejected with 400.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
isSetupLocked() now checks for an unclaimed local wizard user
(oidcIss IS NULL, claimed=false) before firing the effective-config
branch. During the credential→complete window, this sentinel prevents
a production container with VAPID env set from blocking POST /complete
with 423. The explicit setup_complete flag (Check 1) still locks
unconditionally once written. Adds regression test that sets VAPID env
explicitly (no beforeEach clearing) to reproduce the production scenario.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Drizzle mysql2 execute() returns [rows, fields] for SELECT queries; the
generic type parameter alone does not correctly type the result. Use
unknown cast pattern consistent with admin.ts to access the count row.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Without a prerequisite check, an operator could call POST /api/setup/complete
directly, setting setup_complete=true with no admin user or credential row,
leaving no recovery path without manual DB surgery.
Add an inner join check for an unclaimed user with an associated credential;
return 422 if absent. Update /complete tests to seed the prerequisite for
the success path and add an explicit 422 regression test.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The catch block previously echoed err.message (which may contain internal
network addresses like ECONNREFUSED 192.168.1.50:9091) to the pre-auth
caller. Log the raw message server-side only and return a generic user-
facing string with no internal network detail.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The oidcConfigFallbackMiddleware permanently mutates process.env on first
request then never re-reads from DB. Log an explicit info message when each
value is written so operators can see when a restart is required to pick up
config changes, and add inline documentation of the single-write semantics
to prevent silent misconfiguration after a re-run of the wizard.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two concurrent POST /api/setup/credential requests could both pass
isSetupLocked(), observe no unclaimed row, and both insert — leaving two
unclaimed admin rows with no recovery path. Wrap the count-check + user
insert in a transaction with SELECT COUNT(*) ... FOR UPDATE to acquire a
row/gap lock, ensuring at most one unclaimed admin row is created.
Returns 409 when a concurrent request already holds an unclaimed row.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After $returningId() insert, if the re-select returns null the handler
returned 503 without deleting the just-inserted user row, leaving an
unclaimed admin row with no credential. Delete before returning 503 to
mirror the cleanup already present in the catch block.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add isNull import from drizzle-orm; add appConfig to schema imports
- After identity lookup, read app_config.setup_complete per call (D-10 freshness)
- When setup_complete='true' and unclaimed user exists (isNull(oidcIss) AND claimed=false):
claim it via db.update() — binds oidcIss/oidcSub, sets claimed=true, preserves is_admin
- shouldBeAdmin gated: flagRow?.value !== 'true' AND admin COUNT === 0 (T-12-11)
- No email keying in claim branch — isNull(oidcIss) AND claimed=false only (D-10/T-12-12)
- 399 tests pass; typecheck clean
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- A2 CONFIRMED: @hono/oidc-auth reads OIDC_ISSUER/OIDC_CLIENT_ID/OIDC_AUTH_EXTERNAL_URL
at per-request call time (env(c) → process.env), NOT at import time — fresh instance
boots cleanly without OIDC env vars
- Implement oidcConfigFallbackMiddleware in auth/middleware.ts: reads OIDC_ISSUER,
OIDC_CLIENT_ID, OIDC_AUTH_EXTERNAL_URL from app_config when process.env is absent,
injects into process.env before oidcAuthMiddleware() reads it (D-02/D-03/Recommendation a)
- Mount oidcConfigFallbackMiddleware before oidcAuthMiddleware() in index.ts so
wizard-configured instances work before a container restart
- Verify /api/setup mount order: line 49 < devAuthBypass line 54 (T-12-09/Pitfall 1)
- Fix push.test.ts vi.doMock for middleware.js: add oidcConfigFallbackMiddleware stub
- 394 tests pass | 5 todo (D-08 RED scaffolds); typecheck clean
- Fill setupRouter: GET /status, POST /config, POST /validate/{db,oidc,vapid},
POST /credential, POST /complete (SETUP-01/02)
- isSetupLocked() is FIRST statement in every handler; returns 423 if locked (SETUP-04/D-10)
- /status uses isSetupLocked() directly: covers both explicit + effective-config branches
- /config: zod-validates {oidcIssuer:https, oidcClientId, vapidPublicKey, appExternalUrl};
upserts oidc_issuer|oidc_client_id|vapid_public_key|app_external_url into app_config
- /validate/db: db.execute(sql`SELECT 1`); 200 ok, 503 on failure
- /validate/oidc: fetches discovery doc with AbortSignal.timeout(5000); reads oidc_issuer
from app_config; 200 ok, 400 on unreachable/non-2xx
- /validate/vapid: webpush.setVapidDetails() structural check; reads ONLY from process.env
(VAPID_PRIVATE_KEY never from app_config, never returned; T-12-06/SC-3)
- /credential: inserts local user (oidcIss=null, claimed=false, isAdmin=true) FIRST
(Pitfall 5 FK), then calls validateEncryptAndStoreCredential(); noEchoHook + error map
- /complete: upserts setup_complete='true'; 200 first call, 423 second (Pitfall 8/D-10)
- Mount setupRouter pre-auth in index.ts BEFORE devAuthBypass() (T-12-09/Pitfall 1)
- All 394 tests pass (5 todo = D-08 RED scaffolds); typecheck clean
- Add apps/api/src/lib/setupGuard.ts exporting isSetupLocked(): Promise<boolean>
(Wave-0 stub returns false; real DB impl ships in Plan 02)
Doc comment enforces D-10: re-evaluate fresh on every call, never module-cache
- Add apps/api/src/routes/setup.ts exporting setupRouter = new Hono()
(empty router; handlers + index.ts mount added in Plan 02)
Doc comment notes pre-auth surface position — before /api/* OIDC chain
- Remove .notNull() from users.oidc_iss and users.oidc_sub (wizard creates
local rows before OIDC identity is known; first-login-claims binds later)
- Add users.claimed boolean (default false NOT NULL) to distinguish pending
wizard rows from OIDC-bound rows (D-07)
- Add Phase 12 app_config key documentation + prohibition comment (D-01/SC-3)
- Generate migration 0002_lethal_millenium_guard.sql via drizzle-kit generate
(MODIFY COLUMN for nullable, ADD COLUMN claimed — no DROP/recreate)
- Append backfill: UPDATE users SET claimed=true WHERE oidc_iss IS NOT NULL
so existing OIDC users cannot be matched by first-login-claims (D-08)
- Apply migration via drizzle-kit migrate — users.claimed column verified in dev DB
These two files (from the WR-01 / IN-03 review fixes) had formatting that
failed `pnpm format:check`. No logic change — whitespace/wrapping only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The UPDATE and CREATE all-day branches each called getHouseholdTimezone(db)
independently, so a drain processing both an all-day create row and an
all-day update row issued two identical app_config SELECTs. Add a lazy
per-cycle TimezoneResolver (mirroring the existing clientCache thread-through)
created in runOutboxDrain and passed into dispatchRow. The read stays lazy —
cycles with no all-day work never touch the DB — but is shared across all
all-day rows in a cycle. Behavior unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The seed handler computed seeded from a pre-flight SELECT then returned
seeded:!alreadySet. Under a genuine concurrent race both requests can
SELECT the empty table, both enter the insert branch, and both return
seeded:true though only one row was actually written. Replace the
SELECT + conditional onDuplicateKeyUpdate with a single INSERT IGNORE
and derive seeded from affectedRows (1 = inserted, 0 = ignored/existing
row preserved, D-03). On MariaDB onDuplicateKeyUpdate(value=value)
reports affectedRows 1 for both insert and no-op, so it cannot
distinguish them; INSERT IGNORE can. timezone is bound via a
parameterized sql template and is already IANA-validated by zod. Adds a
test asserting seeded:false for a directly-pre-inserted row.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The GET /config/timezone handler SELECTed app_config then, on the unset
path, called getHouseholdTimezone(db) which re-issued the identical
SELECT before falling back (IN-01). The fallback decision also lived in
two places (IN-02). Route the handler through the centralized
resolveHouseholdTimezone(row?.value) added for WR-01: no redundant
round-trip, single source for the D-06 policy. Behavior unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The D-06 fallback used row?.value ?? process.env.TZ ?? Intl..., but ??
only short-circuits on null/undefined. A set-but-empty TZ ('' or ' ')
leaked through and yielded an invalid IANA zone that throws inside
Intl.DateTimeFormat({ timeZone }) downstream, silently dropping the
all-day reminder. Extract resolveHouseholdTimezone() which trims and
treats empty/whitespace candidate values (stored value and TZ) as
absent so they fall through to the Intl resolved zone. Adds RED->GREEN
unit tests for empty and whitespace-only TZ.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Import sql from drizzle-orm in admin.ts
- Add onDuplicateKeyUpdate({ set: { value: sql\`value\` } }) to the
conditional INSERT in POST /config/timezone/seed so a concurrent seed
(or seed racing a PUT) cannot 500 on the app_config.key PK constraint
- Existing value is preserved per D-03 no-overwrite (no-op ODKU)
- seeded flag still reflects the pre-flight SELECT (winner: true, loser: false)
- Add tests: 403 access control, seeded:true on first seed, seeded:false
on second seed without throw (WR-02 idempotent race)
- reminderScheduler.ts: add import { getHouseholdTimezone } from '../lib/householdTimezone.js'
and replace bare process.env.TZ ?? Intl... at line 247 with await getHouseholdTimezone(db)
- outboxWorker.ts: add same import and replace BOTH bare tz lookups at the update-branch
(~line 501) and create-branch (~line 607) with await getHouseholdTimezone(db)
- D-05 satisfied: all three all-day sites now read from the single stored accessor
- D-06 satisfied: getHouseholdTimezone falls back to process.env.TZ → Intl when unset;
existing process.env.TZ-pinned tests pass unchanged
- D-07 satisfied: eventDateTime.ts and hydrateEvents.ts are not modified
- outboxWorker.test.ts: update wireMockChain() to handle app_config table with where().limit()
chain returning empty rows (D-06 fallback), so existing CAL-13 all-day test stays green
- reminderScheduler.test.ts: update mockTwoQueries to mock the new third db.select() call
(getHouseholdTimezone) returning no row (D-06 fallback), keeping all 37 existing tests green
- All 76 broker tests pass; tsc --noEmit clean
- Add appConfig + getHouseholdTimezone/isValidIanaTimezone imports to admin.ts
- Add timezoneSchema: z.object({ timezone: z.string().min(1).max(64).refine(isValidIanaTimezone) })
No noEchoHook — timezone strings are non-sensitive (T-18-06)
- GET /api/admin/config/timezone: returns { timezone, isExplicitlySet } using D-06 fallback
- PUT /api/admin/config/timezone: validates via timezoneSchema + upserts via onDuplicateKeyUpdate
- POST /api/admin/config/timezone/seed: SELECT-then-INSERT (no onDuplicateKeyUpdate) to enforce D-03 no-overwrite
- All three routes appended AFTER existing routes so line-41 requireAdmin covers them (T-18-03)
- All 25 admin.test.ts tests pass; 366/366 full API suite green; tsc --noEmit clean
- getHouseholdTimezone(db): selects household_timezone from app_config
- D-06 fallback chain: stored value → process.env.TZ → Intl.DateTimeFormat().resolvedOptions().timeZone
- isValidIanaTimezone: try/catch Intl.DateTimeFormat (no Intl.supportedValuesOf per RESEARCH Pitfall 2)
- Exports match D-05 single-accessor contract for reminderScheduler + outboxWorker
- All 11 unit tests pass; 358/358 total suite green; tsc --noEmit clean
- eventFieldsSchema (events.ts): rejects reminderLeadMinutes > 10080 with 400
- outboxPayloadSchema (outboxWorker.ts): hard-fails row when value exceeds 1-week cap
- 10080 = 1 week in minutes; matches UI select maximum
classifyValarms: check sign of dur.toSeconds() before preset lookup.
Positive value = alarm fires after event (RFC 5545 TRIGGER:+PT15M or
TRIGGER;RELATED=END:PTNm) → return {kind:'custom'} for preserve path.
Compute leadMinutes as -seconds/60 (was Math.abs) for negative triggers.
Prevents alarm direction inversion: +PT15M was being stored as 15-min-before
lead and re-fired at dtstartUtc-15min — the opposite of the original intent.
- humanizeLeadMinutes: add isAllDay=false param; all-day branch returns
"Today" (lead=0), "Tomorrow" (1440), "In 1 week" (10080), "In N days" (other)
- byKey map: store isAllDay flag (false for timed, true for all-day)
- dispatch loop: pass event.isAllDay to humanizeLeadMinutes
All-day same-day reminder push now reads "Today" instead of "Starts in 0 min".
Timed event wording unchanged (isAllDay defaults to false).
- expand.ts: add reminderIsCustom:boolean to CalendarOccurrence interface;
derived from classifyValarms kind==='custom'; propagated to both
non-recurring and recurring occurrence branches
- client.ts: mirror reminderIsCustom on CalendarOccurrence (atomic mirror)
- EventForm.tsx: extend deriveReminderValue to accept isCustom flag;
returns '__custom__' when true, making the existing D-08 preserve branch
live — editing a custom-alarm event now omits reminderLeadMinutes from
the payload so outboxWorker extractValarms keeps the original VALARM
- Fix existing test fixtures (EventForm.test.tsx, EventDetailPopover.test.tsx)
to include reminderIsCustom:false on all CalendarOccurrence literals
Fixes CAL-14 Pitfall 1: Apple Calendar absolute DATE-TIME / multi-VALARM
alarms no longer silently stripped on any edit round-trip from the PWA.
- Add all-day 9 AM tests: 0-lead fires at EDT alert UTC, not midnight
- Add 1440-lead (day-before) and 10080-lead (7-day-before) tests
- Add all-day dedup test: same uid:dtstartMs fires once across ticks
- Fix all-day prune bug: store start-of-next-day as pruneMs instead of
UTC midnight (which was always <= now by fire time, causing immediate prune)
- Separate dtstartMs (dedup key component) from pruneMs (map cleanup value)
- 28/28 tests GREEN; full API suite 314/314; tsc --noEmit clean
- Add reminderLeadMinutes: number | null to CalendarOccurrence interface (D-06)
- Import classifyValarms in expand.ts; derive series-level value once per VEVENT
- Add reminderLeadMinutes to both non-recurring and recurring occurrence construction
- Add reminderLeadMinutes to GET /api/events select for edit-mode pre-population
- Replace fixed 16-min window with per-event variable-lead fire-time check
- Two separate DB queries: timed (allDay=false) + all-day (allDay=true)
- Remove eq(calendars.isShared, true) — personal events now dispatch (NOTIF-05)
- Remove eq(calendarEvents.allDay, false) — all-day handled in separate query
- Add reminder_lead_minutes IS NOT NULL WHERE predicate (NOTIF-05)
- Skip timed events with reminderLeadMinutes===0 in JS (D-06: 0 on timed = None)
- Change dedup key from bare uid to uid:dtstartMs compound key (NOTIF-06)
- Update prune loop to use compound key
- Import computeAlertInstantUtc from vevent.js (Plan 11-01, wave 2 dep)
- Add humanizeLeadMinutes export (Task 2 body formatter, used in dispatch)
- Update test helper mockTwoQueries() to handle two-query dispatch pattern
- All 14 tests GREEN; tsc --noEmit clean; setInterval retained, no node-cron
- Import classifyValarms from vevent.ts
- Derive reminderLeadMinutesValue: preset/offlist → leadMinutes; custom/none → null
- Add reminderLeadMinutes to .values() and .onDuplicateKeyUpdate({ set: {} })
- Scheduler now has ground truth for native-client VALARMs (T-11-07 mitigated)
PUT /api/admin/calendars/:id/shared cleared the current shared calendar then
set the target in two non-transactional UPDATEs without checking the target
exists — a bad/stale id wiped the family shared lane and still returned ok.
Verify the target inside a transaction; return 404 when absent. Adds a
regression test (RED→GREEN).
Lucas Berger