Monitor previously only closed AP on external WiFi STA; late wired uplink
while in AP never fired network-ready, so WS stayed down and AP display
kept flashing. Poll hasInternet() in AP state and call _emitNetworkReady().
Made-with: Cursor
- Monitor: every tick in ap state led.off() so play never stays on with hotspot
- hasLanCableCarrier: _ifacePhysicalLinkUp (carrier=1 and operstate not down)
- When end0/eth0 exist but link down, return false (no fallback to hasWiredCarrier)
- LAN poll 500ms
Made-with: Cursor
wlan0 cannot scan while in AP mode (hardware limitation).
Now scan nearby networks before starting hotspot and cache
the results for the captive portal page.
Made-with: Cursor
Killing all dnsmasq processes caused NetworkManager to detect its
hotspot dnsmasq died and tear down the hotspot (AP appears briefly
then disappears).
Now leverage NM's built-in dnsmasq-shared.d config directory:
write DNS hijack config before starting AP so NM's own dnsmasq
picks it up. No separate dnsmasq process needed.
Made-with: Cursor
After WiFi is configured and device reboots, NetworkManager needs
a few seconds to auto-connect to saved WiFi. Without waiting,
AP starts immediately and occupies wlan0, preventing NM reconnect.
- Add hasSavedWifiConnection() to check for saved WiFi profiles
- Wait up to 20s for NM auto-connect before falling back to AP
- First boot (no saved WiFi) still starts AP immediately
Made-with: Cursor
Redesign provisioning from one-shot blocking to persistent background manager:
- AP hotspot starts at boot regardless of eth0 status
- Captive portal runs alongside AP for WiFi configuration
- AP automatically shuts down only when WiFi STA connects
- WiFi drops at runtime -> AP auto-restarts
- WiFi connect fails -> AP auto-restarts for retry
- client.js no longer blocks on network; connects WS when ready
Made-with: Cursor
