The document click handler registered on open could catch the same
click event that opened the dropdown, immediately closing it.
Using mousedown avoids the race since it fires before click.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The chevron was too narrow (px-1.5 → px-2.5) and clicks were
caught by the outside-click handler before toggling the dropdown.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Split button with default "Export GPX" (track only) and dropdown:
- Export Route: clean track for use in any app
- Export Plan: track + waypoints + no-go areas in GPX extensions
(trails:planning namespace) for reimporting into the planner
Also:
- Add no-go area serialization to generateGpx (GPX extensions)
- Parse no-go areas from GPX extensions on import
- Initialize no-go areas in Yjs session from imported GPX
- Save to Journal now exports track only (consistent with Export Route)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The button was rendered as a React component, not a Leaflet control,
so clicks propagated to the map. Use L.DomEvent.disableClickPropagation
on the container — same approach Leaflet's built-in controls use.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
BRouter supports a `polygons` parameter with full vertex coordinates.
We were approximating polygons as circles (centroid + max radius),
which poorly represented elongated or concave shapes.
Now passes polygon vertices directly — no approximation, exact
no-go boundaries.
Also removes unused haversineMeters helper and updates the spec.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Vite's dev server can re-evaluate modules, causing prom-client
"already registered" errors for all metrics, not just default ones.
Use getOrCreate pattern to reuse existing metrics from the registry.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Importing metrics.server.ts from yjs-server.ts caused
collectDefaultMetrics() to re-register during HMR, crashing
the planner dev server. Check if metrics already exist first.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
plannerActiveSessions and plannerConnectedClients were registered
but never updated — they always reported 0. Now incremented on
WebSocket connect and decremented on close.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Both new.tsx and api.sessions.ts had duplicated track-segment
waypoint extraction logic. Moved to packages/gpx as
extractWaypoints(gpxData) — uses <wpt> elements when present,
falls back to start of each track segment + end of last.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Same fix as new.tsx — the journal→planner handoff also only used
<wpt> elements from GPX. Now both code paths extract waypoints
from track segments when no explicit waypoints exist.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The planner's GPX import only used <wpt> elements, which many GPX
files don't have. Now falls back to the first and last track point,
giving the planner two endpoints to route between.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
parseGpx (sync) needs browser DOMParser which doesn't exist on the
server — it silently failed in every server-side caller. Removed the
export and migrated all callers to parseGpxAsync. Updated tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Distance was only stored in the elevation profile array, which was
empty when track points had no <ele> elements. Now computed via
haversine independently and exposed as gpxData.distance.
Tested: berlin-dresden-radweg-2021.gpx (5660 points, no elevation)
now correctly reports 248.8 km.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The initial approach (#150) didn't work because:
1. routes.server.ts used parseGpx (sync, needs browser DOMParser)
instead of parseGpxAsync (uses linkedom on Node.js) — GPX parsing
silently failed, so stats AND geom were never populated
2. Drizzle's customType doesn't pass SQL expressions through toDriver
Fixes:
- Switch to parseGpxAsync for all server-side GPX parsing
- Use db.execute() with raw ST_GeomFromGeoJSON SQL after insert
- Remove unused lineStringFromCoords helper from schema
- Share setGeomFromGpx between routes and activities
- Add unit tests for GPX→GeoJSON coordinate extraction
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Extract coordinates from parsed GPX tracks and store as PostGIS
LineString geometry via ST_GeomFromGeoJSON. Applied to:
- createRoute / updateRoute (routes.server.ts)
- createActivity / createRouteFromActivity (activities.server.ts)
Adds lineStringFromCoords() helper to the journal schema that
builds the SQL expression from [lon, lat] coordinate pairs.
Unblocks spatial queries (e.g. route discovery via ST_Intersects).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
invalidateSize() + requestAnimationFrame ensures Leaflet knows
the map's actual dimensions after the elevation chart renders.
Also removes unnecessary filter on already-typed coordinates.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Show error toasts when route calculation fails, with distinct messages:
- "No route found" for BRouter 4xx (unroutable waypoints, missing tiles)
- "Route calculation failed" for server errors / network failures
- "Too many requests" for rate limiting
- Rollback all waypoints to last known good state on route failure
- Guard against restore→recompute loop via restoringRef flag
- Extract reusable useToasts hook from awareness toast logic
- Add BRouterError class to distinguish client vs server errors
- API returns 422 for unroutable requests instead of blanket 502
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The ghost marker (route split indicator) was interfering with waypoint
dragging because it sat above waypoints (zIndexOffset 1000) and stayed
active during drag operations.
Adopt brouter-web's suspension pattern:
- Suspend route interaction on waypoint mouseover (not just dragstart),
unsuspend on mouseout — ghost marker hides before any click/drag
- Lower ghost marker zIndexOffset to -100 so waypoints always receive
clicks first
- Keep drag guards as a safety net
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Now that both server and client use the same locale from context,
the post-hydration useEffect that re-formatted with the browser's
OS locale was causing a visible flicker. Just render with the
context locale directly.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add detectLocale() that extracts the full locale tag (e.g. "de-DE")
from the request, not just the language. The journal root loader passes
it via LocaleContext so ClientDate renders with the correct locale on
both server and client — no more hardcoded "en-US" fallback.
Also fixes settings page hydration mismatch by using ClientDate for
passkey creation dates instead of inline toLocaleDateString().
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Split initI18n() into separate server/client initialization paths
inspired by remix-i18next's architecture:
- initI18nServer(lng): no LanguageDetector, per-request language from
Accept-Language header
- initI18nClient(): LanguageDetector reads <html lang> first to match
server, then localStorage and navigator
- detectLanguage(request): parses Accept-Language for best match
Fixes rehydration errors caused by v26 removing initImmediate option
(replaced with initAsync) and LanguageDetector running on the server
where browser APIs don't exist.
Also makes <html lang> dynamic instead of hardcoded "en".
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The prompt now only appears when coming from signup/login with
?add-passkey=1 AND the user has zero passkeys. Passkey count display
removed from home — that info lives in /settings now.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add `purpose` column (default: 'login') to distinguish login tokens
from email-change tokens
- verifyMagicToken only matches purpose='login'
- verifyEmailChange only matches purpose='email-change'
- Re-check email availability at verification time, not just initiation
— prevents race where someone registers with the new email between
request and verification
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- /settings with profile, security, and account sections
- Profile: edit display name and bio
- Security: list passkeys with device labels, add/delete with
last-passkey warning
- Account: email change with magic link verification, account
deletion with username confirmation
- Settings link in nav for authenticated users
- E2E tests: auth guard, profile editing, passkey add/delete,
last-passkey warning, account deletion, nav visibility
- i18n: full en + de translations for all settings UI
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Login: hide passkey button when browser lacks WebAuthn, default to
magic link mode, hide "back to passkey" link
- Register: detect passkey support and offer magic link registration as
fallback. New server-side registerWithMagicLink() creates account
without credential and sends verification email.
- Home: show passkey count for logged-in users, show amber unsupported
message instead of hiding the add-passkey prompt entirely
- i18n: add passkeyNotSupported, passkeyNotSupportedRegister,
registerWithMagicLink, passkeyStatus keys (en + de)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bug: @simplewebauthn v13 changed credential.id from Uint8Array to
base64url string. Buffer.from(string) without encoding stored the
ASCII text instead of decoded bytes. Authentication then failed to
match because it correctly decoded with "base64url".
Fix: Buffer.from(credential.id, "base64url") in both registration
and add-passkey flows.
Also: catch WebAuthn "not allowed" errors and show a friendly message
instead of the raw browser error.
Existing passkeys must be re-registered after deploy.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@trails-cool/db doesn't export a db instance — only createDb and withDb.
The health check was importing a non-existent export, always failing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Removes replayIntegration and replay sample rates from both apps.
Quota was already exhausted.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Returns SENTRY_RELEASE (set at build time) as the version field,
making it easy to verify which commit is deployed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The route files existed but weren't registered in the explicit routes.ts,
so React Router never compiled them into the build.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
node:25-slim doesn't include CA certificates, causing sentry-cli
SSL errors: "unable to get local issuer certificate". Also removes
debug logging from previous troubleshooting.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The token file may have a trailing newline that causes sentry-cli
API requests to fail with "API request failed".
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Disable Sentry Vite plugin telemetry (telemetry: false)
- Pass SENTRY_RELEASE as build-arg (not secret) since it's just the git sha
- Keep only SENTRY_AUTH_TOKEN as Docker secret
The release version was showing as *** because Docker secrets get masked
by GitHub Actions. Git sha is not sensitive — safe as a build arg.
The sourcemap warning from react-router is a known upstream issue (ships
without sourcemaps) and is harmless.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
pnpm approve-builds in Dockerfile ran after install, so the approval
wasn't picked up. Use onlyBuiltDependencies in package.json instead —
this is read during install and allows build scripts to run.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Approve @sentry/cli and esbuild build scripts so postinstall runs
- Use --mount=type=secret instead of ARG for SENTRY_AUTH_TOKEN
(fixes Docker warning about secrets in build args)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Health endpoints:
- /api/health (Journal) and /health (Planner) with DB connectivity check
- Docker healthchecks updated to use app health endpoints
Structured logging:
- Pino with JSON output in production, pretty-print in dev
- Request logging middleware in Planner (method, path, status, duration)
- Replaced console.log/error with structured logger in email and auth flows
Prometheus metrics:
- prom-client with default Node.js metrics + custom histograms/gauges
- /metrics endpoints on both apps
- http_request_duration, planner_active_sessions, brouter_request_duration
Monitoring stack:
- Prometheus, Loki, Grafana containers in docker-compose
- Grafana provisioned with datasources, dashboards, and alert rules
- Caddy access logging (JSON to stdout for Loki)
- grafana.trails.cool with basic auth via Caddy
Dashboards and alerting:
- Overview: request rate, error rate, latency p50/p95/p99
- Planner: active sessions, connected clients, BRouter latency
- Infrastructure: memory, CPU, event loop lag
- Alerts: disk >80%, app down 2min, error rate >5%
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Enrich BRouter response with per-point 3D coordinates and segment boundary
tracking (EnrichedRoute interface)
- ColoredRoute component: plain, elevation gradient (green→yellow→red), and
surface color modes with invisible wide polyline for click targeting
- Click-to-split: click on route polyline inserts waypoint at nearest point,
mapped to correct segment via boundary indices
- MidpointHandles: draggable CircleMarkers at route segment midpoints for
reshaping, hidden below zoom 12, opaque on hover
- Color mode toggle (select) synced via Yjs routeData
- i18n keys for color mode labels (en + de)
- Unit tests for segment boundary tracking (13 tests)
- E2E tests for enriched route response and color mode toggle
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Transactional emails:
- Add nodemailer SMTP email module with dev-mode console logging
- Magic link template and welcome template with HTML + plain text
- Wire sendMagicLink into login flow, sendWelcome into registration
- Update privacy page and deploy docs for SMTP configuration
Planner features:
- No-go areas: draw polygons on map (leaflet-geoman), synced via Yjs,
passed to BRouter as nogos parameter, route recomputes on change
- Session notes: collaborative Y.Text textarea in sidebar tab
- Crash recovery: periodic localStorage save of Yjs state, restore on reconnect
- Rate limit session creation (10/IP/hour) in /new route
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move initI18n() from entry.client.tsx to root.tsx in both Planner and
Journal so i18n is initialized before any component renders. Downgrade
react-i18next to 16.6.1 and i18next to 25.10.4 to avoid hydration
mismatch issues introduced in newer versions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
With custom entry.client.tsx, initI18n() in root.tsx runs too late —
the client hydrates before i18n is initialized, showing raw translation
keys. Move initI18n() to entry.client.tsx where it runs before
startTransition/hydrateRoot.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
