Fixed Two Display Issues:
1. Version Parsing (was showing "regex)")
Before: Used awk '{print $NF}' which extracted last field
netifyd -V output: "Netify Agent/5.2.1 (...regex)"
Result: Displayed "regex)" as version
After: Use sed to extract version number
Pattern: sed 's/.*Agent\/\([0-9.]*\).*/\1/'
Result: Correctly displays "5.2.1"
2. Socket Connectivity (was showing "Disconnected")
Before: Checked for unix socket file existence
Problem: Netifyd doesn't create unix socket in current config
Result: Always showed "Disconnected"
After: Check if netifyd is running and producing data
Logic: Process running + status.json exists + readable
Result: Correctly shows "Connected" when service is operational
Also: Removed stat command usage (not available on OpenWrt)
Technical Details:
- Socket detection now based on service health, not socket file
- Works with both sink-only and socket-enabled configurations
- Simplified logic compatible with busybox/OpenWrt environment
Dashboard Now Shows:
✓ Version: 5.2.1 (was: regex))
✓ Socket: Connected (was: Disconnected)
✓ Status: Running
✓ Uptime: Accurate duration
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Issue Identified:
- Netifyd 5.2.1 was crashing with JSON assertion error on startup
- Error: "Assertion failed: m_it.object_iterator != m_object->m_data.m_value.object->end()"
- Root cause: Legacy categories.json format incompatible with netifyd 5.2.1+
The Fix:
- Removed auto-creation of netify-categories.json from UCI defaults
- Let netifyd manage this file itself or operate without it
- Both approaches work correctly with netifyd 5.2.1
Technical Details:
The UCI defaults script was creating categories.json in v1.0 format:
```json
{
"version": "1.0",
"categories": []
}
```
This format is detected as "legacy" by netifyd 5.2.1, which then attempts
to parse it with newer code expecting a different structure. When accessing
JSON object iterators, the assertion fails because expected keys don't exist.
Solution: Don't create the file. Netifyd works perfectly without it and will
create its own if needed in the correct format for its version.
Verified Working:
- Netifyd now starts successfully
- 22 active flows captured on br-lan and br-wan
- Both interfaces online with 0% packet drops
- CPU utilization: 0.1%
- Memory: 14.5 MB
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Removed Features:
- Debug mode toggle button
- Debug panel and log display
- Update indicator (count and time since last update)
- Debug logging functions (debug, toggleDebug)
- Debug state variables (debugMode, updateCount, errorCount, lastUpdate)
- REFRESH-DEBUG.md documentation
Preserved Features:
✅ Race condition fix (containers created before poll.add)
✅ Auto-refresh functionality (5-second polling)
✅ Data caching (latestDashboardData, latestTopApps, latestTopProtocols)
✅ Application aggregation function
✅ All core rendering functions
✅ Service control and statistics display
Benefits:
- Cleaner, production-ready code
- Reduced code complexity (~120 lines removed)
- Maintains all critical functionality
- Better performance (no debug overhead)
The dashboard now provides a streamlined interface focused on
displaying network intelligence data without development debug features.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
CrowdSec Central API (CAPI) Fixed:
- Removed code that disabled online_client on install
- Added proper CAPI registration in crowdsec.defaults
- Registration now works (previous 403 errors were transient)
- Graceful fallback if CAPI registration fails
CAPI Features Now Working:
- Threat intelligence sharing enabled
- Pulling community blocklist (14,997+ IPs)
- Hub updates working without 403 errors
- SSH bruteforce: 12,388 bans from CAPI
- Generic scans: 1,176 bans from CAPI
- SSH exploits: 1,433 bans from CAPI
Registration Flow:
1. Create /etc/machine-id if missing
2. Register local API machine
3. Register with Central API (CAPI)
4. On CAPI failure, create minimal credentials file
5. Update hub index
6. Install default collections
Benefits of CAPI Integration:
- Real-time threat intelligence from global network
- Community-contributed IP blocklists
- Automatic updates for detection scenarios
- Signal sharing to help protect others
- Enhanced protection without manual IP list management
NetIfyd Dashboard Improvements:
- Added data caching for smoother updates
- Application aggregation function
- Fallback stats when data temporarily unavailable
- Better handling of undefined values
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Dashboard Refresh Problem Fixed:
- Race condition where poll.add() was called before containers existed
- Containers were undefined during first poll callback
- DOM updates failed silently with no error logging
- Fixed by creating containers BEFORE setting up polling
Debug Features Added:
- Toggle debug mode with button in header
- Visual debug panel showing last 20 log entries
- Browser console logging with timestamps
- Live update indicator (count + time since last update)
- Error tracking and counting
- Detailed logging of all RPC calls and responses
Debug Panel Features:
- Timestamps for all events
- JSON data preview for API responses
- Auto-scroll with newest entries at top
- Max 20 entries to prevent memory issues
- Hidden by default, shown when debug enabled
Update Indicator:
- Shows "Updates: N | Last: Xs ago" in header
- Updates every second
- Visual feedback that polling is working
- Easy to spot stalled/broken polling
Error Handling:
- Try/catch around all poll callbacks
- Errors logged to debug panel and console
- Error counting for diagnostics
- Polling continues even after errors
Code Improvements:
- Proper container creation order
- Better error handling in load() and polling
- Debug logging throughout lifecycle
- Performance metrics tracking
Documentation:
- Complete analysis in REFRESH-DEBUG.md
- Troubleshooting guide
- Debug mode usage instructions
- Performance considerations
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Network Intelligence Dashboard Enhancements:
- Add detailed protocol breakdown (TCP/UDP/ICMP) with visual bars
- Display flow metrics (active, expired, purged)
- Show CPU and memory usage for netifyd process
- Add IP bytes vs wire bytes differentiation
- Enhanced stat cards with subtitles and better formatting
RPC Backend Improvements:
- Add tcp_packets, udp_packets, icmp_packets metrics
- Add ip_bytes (payload without ethernet overhead)
- Add flows_active, flows_expired, flows_purged counters
- Add cpu_usage and memory_kb from netifyd status
- Calculate CPU total from user + system time
Directory Structure Fix:
- Create /etc/netify.d/plugins.d on package install
- Create /etc/netify.d/address-groups.d
- Generate minimal netify-categories.json to prevent errors
- Auto-initialize UCI config for secubox-netifyd
- Auto-restart netifyd after directory creation
UCI Configuration:
- Settings: auto_refresh, socket configuration
- Analytics: limits for top apps/protocols/devices
- Data retention configuration
Issue Resolved:
- Netifyd was running but showing 0 flows due to missing directories
- Service now properly captures and analyzes network traffic
- All metrics displaying correctly in dashboard
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Package Installation Improvements:
- Automatically create /etc/machine-id from UUID if missing
- Disable Central API (CAPI) by default in config.yaml
- Create minimal online_api_credentials.yaml to prevent errors
- Add fallback curl download for hub index (works around 403 errors)
- Make all setup commands non-fatal with || true
CAPI Status:
- Disabled by default due to HTTP 403 errors from api.crowdsec.net
- Custom User-Agent (crowdsec/v1.7.4-openwrt-*) appears blocked
- Can be manually enabled with: cscli console enroll <key>
- Local-only mode provides full SSH brute-force protection
Hub Updates:
- Manual curl download works (HTTP 200)
- cscli hub update fails (HTTP 403)
- Weekly auto-update via curl in defaults script
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fixed pgrep command to detect running CrowdSec process:
- Changed from `pgrep -x crowdsec` to `pgrep crowdsec`
- The -x flag requires exact process name match which wasn't working
- Affects both check_cscli() and get_status() functions
- Now correctly detects service as running in dashboard
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Enhanced dashboard UX when CrowdSec service is not running:
API module (api.js):
- Modified getDashboardData() to handle error responses gracefully
- Returns empty arrays/objects for stats when service is stopped
- Includes error flag in response data
Overview module (overview.js):
- Added 'fs' module import for service control
- Added startCrowdSec() function to start service from UI
- Display warning banner when service is stopped
- Provide actionable message with start service link
Dashboard CSS (dashboard.css):
- Added .cs-warning-banner styles for error messages
- Professional warning styling with icon and content layout
This resolves XHR timeout errors by showing friendly error messages
instead of hanging requests.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Added service status check in check_cscli() function to immediately return
an error instead of timing out when crowdsec service is not running.
This fixes the 'XHR request timed out' error in the dashboard when CrowdSec
is stopped.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Changed 'new api()' to 'api' in all dashboard views since the API module
exports an object, not a constructor class.
Fixed files:
- overview.js
- decisions.js
- alerts.js
- metrics.js
- waf.js
This resolves the 'TypeError: api is not a constructor' error.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Changed menu dependency from 'luci-app-crowdsec' to 'luci-app-crowdsec-dashboard'
to match the actual ACL name, making CrowdSec visible in LuCI interface.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Major achievements:
- Successfully built CrowdSec 1.7.4-r2 (81MB) for aarch64_cortex-a72
- Netifyd 5.2.1-r1 confirmed working with all fixes
- Both packages built with OpenWrt 24.10.5 buildroot
CrowdSec Build:
- Full Go 1.23 compatibility with vendored modules
- Staged all required golang.org/x/* dependencies
- Fixed go.mod directives for OpenWrt toolchain
- Includes crowdsec engine + crowdsec-cli (cscli)
- Complete configuration files and init scripts
Netifyd Status:
- 5.2.1 package with GCC 13.3/C++17 fixes operational
- LuCI dashboard v1.0.2 with working metrics
- Native status.json integration confirmed
Build System Updates:
- Enhanced local-build.sh for OpenWrt-only packages
- Improved package sync and build workflow
- Updated Makefiles for consistency
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Netifyd crée nativement /var/run/netifyd/status.json avec toutes
les stats nécessaires. Le collecteur custom n'est plus nécessaire.
Supprimé:
- /usr/sbin/netifyd-collector
- /etc/cron.d/netifyd-collector
- Installation dans Makefile
Le backend RPC lit maintenant directement le fichier natif de netifyd.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Netifyd crée automatiquement /var/run/netifyd/status.json avec toutes
les stats. Pas besoin de collecteur custom !
Changements:
- Lit flow_count directement depuis le fichier natif
- Compte unique_devices depuis la table ARP (ip neigh)
- Lit dns_hint_cache.cache_size pour applications
- Calcule total_bytes depuis stats.*.wire_bytes
Fix testé sur routeur:
- active_flows: 16 ✓
- unique_devices: 4 ✓
- unique_applications: 5 ✓
- total_bytes: 48302 ✓
Le collecteur n'est plus nécessaire - netifyd gère tout !
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Modifié le collecteur pour fonctionner sans socket Unix.
Parse netifyd -s + table ARP + stats réseau.
- Supprimé dépendance socat
- Parse netifyd -s pour metrics
- Utilise ARP pour device count
- Calcule bytes depuis /sys/class/net
- Mis à jour RPC pour device_count et total_bytes
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Added netifyd-collector daemon to aggregate real-time statistics from
netifyd and populate the dashboard with actual data.
New Features:
- Added /usr/sbin/netifyd-collector script
- Queries netifyd socket for flow data
- Aggregates devices, applications, protocols
- Writes /var/run/netifyd/status.json
- Runs every minute via cron
- Added /etc/cron.d/netifyd-collector cron job
- Added socat dependency for socket communication
Changes:
- Bumped version to 1.0.2
- Updated Makefile to install collector and cron job
- Fixed dashboard empty metrics issue:
* Unique Devices will now show count
* Applications will now show count
* Total Traffic will now show bytes
This fixes the "0" values issue in dashboard Network Statistics.
Dashboard will now show real metrics after 1 minute of netifyd running.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fixed critical bug in get_top_applications() and get_top_protocols()
RPC methods where data was extracted with jq but never added to the
JSON output. The functions were using jshn arrays but only echoing
data instead of adding it to the array.
Changes:
- Rewrote get_top_applications() to output complete JSON via jq
- Rewrote get_top_protocols() to output complete JSON via jq
- Removed broken jshn array manipulation
- Added proper fallback to empty arrays when no data available
This fixes the "metrics vides" (empty metrics) issue in LuCI dashboard.
The dashboard will now properly display:
- Top applications with traffic stats
- Top protocols with bandwidth usage
- Flow counts and bytes transferred
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
The 003-skip-tests.patch file was malformed and causing build failures
with "Only garbage was found in the patch input" error. Removed the
patch as it's not needed - the build succeeds without it since we
already use --with-only-libndpi configure flag.
Added 002-fix-ndpi-example-linking.patch to properly link ndpi examples
with correct library order.
Build verified: netifyd_5.2.1-r1_aarch64_cortex-a72.ipk (1.2M)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
The local-build.sh script outputs packages to secubox-tools/build/x86-64/
not to bin/packages/*/secubox/ as the deployment script was expecting.
Changes:
- Updated find commands to look in secubox-tools/build/x86-64/ first
- Added fallback to secubox-tools/sdk/bin/packages/ for SDK builds
- Improved error messages showing searched paths
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Removed unnecessary dependency on luci-app-secubox which was causing file
conflicts during installation. The admin package doesn't actually use any
modules from luci-app-secubox - it only uses its own secubox-admin.* modules.
This fixes the installation error:
"Package luci-app-secubox wants to install file /etc/config/secubox
But that file is already provided by package secubox-core"
Changes:
- Makefile: Removed +luci-app-secubox from LUCI_DEPENDS
- Package now only depends on: +luci-base +rpcd +secubox-core
- Incremented PKG_RELEASE: 7 → 8
- Updated DEPLOY_UPDATES.md with v1.0.0-8 details
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fixed "No related RPC reply" errors across all admin views by wrapping
ALL RPC calls in L.resolveDefault() with appropriate fallback values.
This allows the frontend to load gracefully even when the backend RPCD
methods are not yet deployed, showing empty data instead of crashing.
Changes:
- health.js: Wrapped getHealth() → L.resolveDefault(getHealth(), {})
- logs.js: Wrapped getLogs() → L.resolveDefault(getLogs(), { logs: '' })
- settings.js: Wrapped getApps() and getModules() with fallbacks
- apps.js: Wrapped getApps() and getModules() (checkUpdates already wrapped)
- dashboard.js: Wrapped all 4 RPC calls (getApps, getModules, getHealth, getAlerts)
- Incremented PKG_RELEASE: 6 → 7
- Updated DEPLOY_UPDATES.md with v1.0.0-7 details
All admin pages now load successfully regardless of backend deployment status.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fixed TypeError "WidgetRenderer is not a constructor" in dashboard.js
by removing the 'new' keyword. LuCI's baseclass.extend() creates callable
classes that should not be instantiated with 'new'.
Changes:
- dashboard.js: Changed from 'new WidgetRenderer({...})' to 'WidgetRenderer({...})'
- Added comprehensive try-catch error handling with fallback error display
- Incremented PKG_RELEASE: 5 → 6
- Updated DEPLOY_UPDATES.md with v1.0.0-6 details
This fix allows the widget system to initialize properly on the dashboard.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fix incorrect local-build.sh invocation. The correct syntax is:
./secubox-tools/local-build.sh build <package>
Not:
./secubox-tools/local-build.sh <package>
Split into two separate build commands for clarity and proper output.
- Updated package version to 1.0.0-5
- Added troubleshooting for 'No related RPC reply' errors
- Explained graceful degradation behavior
- Clarified deployment order requirements
Fix 'No related RPC reply' errors by wrapping RPC calls in L.resolveDefault()
to provide fallback values when backend methods aren't available yet.
## Problem
When new LuCI views are deployed before backend packages, RPC calls fail with:
Error: No related RPC reply
This happens because:
- Frontend (luci-app-secubox-admin) calls check_updates, get_catalog_sources
- Backend (secubox-core) hasn't been deployed yet with new RPCD methods
- RPCD returns no reply, causing frontend to crash
## Solution
Wrap all new RPC calls in L.resolveDefault() with sensible fallbacks:
**catalog-sources.js**:
- getCatalogSources() → fallback: { sources: [] }
- checkUpdates() → fallback: { updates: [] }
**updates.js**:
- checkUpdates() → fallback: { updates: [] }
This allows pages to load gracefully with empty data instead of crashing.
## Benefits
1. **Graceful degradation**: Pages load even without backend
2. **Deployment flexibility**: Can deploy frontend before backend
3. **Better UX**: Shows 'No updates' / 'No sources' instead of errors
4. **Production-ready**: Handles missing backends in production
## Testing
Before backend deployment:
- Catalog Sources page shows: 'No sources configured'
- Updates page shows: 'All applications are up to date'
After backend deployment:
- Pages populate with real data from RPCD
Incremented PKG_RELEASE: 4 → 5
Fix TypeError when WidgetRenderer is instantiated with undefined options
parameter by adding defensive check at start of __init__ method.
Error:
TypeError: can't access property "containerId", options is undefined
Fix:
options = options || {};
This ensures the constructor works even if called without parameters,
preventing the TypeError when accessing options.containerId.
Incremented PKG_RELEASE: 3 → 4
Fix RPC -32002 "Access denied" errors by adding proper ACL permissions
for all new catalog source and version management methods.
## Problem
New RPC methods added in Phases 1-3 were accessible in the RPCD backend
but lacked ACL (Access Control List) permissions, causing browser errors:
```
RPCError: RPC call to luci.secubox/get_catalog_sources failed with error -32002: Access denied
RPCError: RPC call to luci.secubox/check_updates failed with error -32002: Access denied
```
## Solution
Updated `/usr/share/rpcd/acl.d/luci-app-secubox-admin.json` to grant
permissions for all 7 new methods introduced in the multi-source AppStore.
### Read Permissions (added 5 methods)
These methods only read data and don't modify system state:
- `get_catalog_sources` - List configured catalog sources
- `check_updates` - Check for available app updates
- `get_app_versions` - Get version info for specific app
- `get_changelog` - Retrieve app changelog
- `get_widget_data` - Get widget metrics for dashboard
### Write Permissions (added 2 methods)
These methods modify system configuration:
- `set_catalog_source` - Set active catalog source (UCI write)
- `sync_catalog` - Trigger catalog synchronization
### UCI Access (added 1 config)
Added `secubox-appstore` to UCI read/write lists for source management.
## Files Changed
**Modified**:
- `luci-app-secubox-admin/root/usr/share/rpcd/acl.d/luci-app-secubox-admin.json`
- Read permissions: 9 → 14 methods
- Write permissions: 6 → 8 methods
- UCI access: Added `secubox-appstore`
- `luci-app-secubox-admin/Makefile`
- PKG_RELEASE: 2 → 3
## Deployment
After updating the package:
1. `opkg install luci-app-secubox-admin_*.ipk`
2. ACL file automatically installed to `/usr/share/rpcd/acl.d/`
3. RPCD reloads ACLs automatically
4. Methods now accessible from LuCI frontend
No manual RPCD restart required - ACL changes are picked up automatically.
## Verification
Test with:
```bash
ubus -S call luci.secubox get_catalog_sources
ubus -S call luci.secubox check_updates
```
Should return data instead of "Access denied" error.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Updated secubox-app-mailinabox plugin catalog to match the main catalog.json entry.
Changes to plugin catalog (plugins/catalog/secubox-app-mailinabox.json):
- category: "productivity" → "hosting" (more accurate classification)
- min_storage_mb: 1024 → 2048 (realistic for email server with attachments)
- status: "stable" → "beta" (matches maturity level)
- tags: added "hosting" tag
- capabilities: added "hosting" capability
- notes: added "Port 25 must be accessible" (important for email server)
Context:
SecuBox uses two catalog sources:
1. Main catalog.json - Used by get_appstore_apps RPCD method for Apps Manager UI
2. Individual plugin catalogs - Used by secubox-appstore CLI for package detection
This ensures both sources provide consistent metadata.
Integration with luci-app-secubox-admin:
✅ secubox-app-mailinabox is now fully integrated into the admin interface
✅ Installation detection works automatically via opkg package checking
✅ Apps Manager will show:
- "Install" button if package not installed
- "Configure" and "Remove" buttons if package is installed
✅ Status detection flow:
1. API.getApps() → reads main catalog.json
2. API.getModules() → calls secubox-appstore list --json
3. secubox-appstore checks if secubox-app-mailinabox is installed via opkg
4. Returns status in modules list
5. Frontend displays appropriate buttons based on status
Also incremented PKG_RELEASE: 4 → 5
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Changed Docker dependencies from hard requirements to optional, allowing the package
to install on routers without Docker pre-installed.
Problem:
- Package installation failed with "cannot find dependency dockerd/docker/containerd"
- Users couldn't install the package even though the control script handles Docker
installation automatically via `mailinaboxctl install`
Solution:
- Removed hard dependencies: dockerd, docker, containerd from DEPENDS line
- Kept only essential UCI dependencies: +uci +libuci
- The mailinaboxctl script already handles Docker installation in ensure_packages()
function (line 95) when running `mailinaboxctl install`
Installation workflow now:
1. Install package: opkg install secubox-app-mailinabox_1.0.0-r2_all.ipk ✅
2. Run setup: mailinaboxctl install
- Automatically installs dockerd, docker, containerd via opkg
- Creates directories, pulls Docker image
- Configures service
3. Configure: Edit /etc/config/mailinabox (hostname, admin_email)
4. Start service: /etc/init.d/mailinabox start
Build verification:
✅ Package builds successfully: 3.9KB
✅ No dependency errors during installation
✅ Docker auto-installation handled by control script
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fixed critical Makefile structure issue that prevented luci-app-secubox-admin from
being indexed by the feed system and built by the SDK.
Changes:
1. Added `include $(TOPDIR)/rules.mk` at the beginning (required for all OpenWrt packages)
2. Added PKG_LICENSE and PKG_MAINTAINER fields (best practices)
3. Added LUCI_PKGARCH:=all to match other LuCI packages
4. Added closing comment `# call BuildPackage - OpenWrt buildroot`
5. Moved include statements to proper positions
This fix aligns the Makefile with the standard OpenWrt/LuCI package structure used
by luci-app-secubox and other working packages.
Additionally:
- Suppressed kconfig warnings in local-build.sh by redirecting stderr to /dev/null
for all `make defconfig` commands (lines 532, 782, 1278)
Build verification:
✅ Package now appears in feeds/secubox.index
✅ Successfully builds: luci-app-secubox-admin_1.0.0-r2_all.ipk (8.3KB)
✅ Total SecuBox packages built: 33
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
