[guohui85]

Summary

The Wi-Fi status worker in cp0_lvgl_network.cpp starts two nmcli processes every 3 s for the launcher's whole lifetime. It does this even when nobody can see the status: when the screen is locked, or when another app owns the display. With this PR it refreshes every 3 s only while the status is visible, and every 60 s otherwise.

Change

• poll_loop() still wakes every 3 s. It now calls update_status_cache() only when one of these holds:
  • this is the first poll;
  • the launcher owns the screen (LVGL_RUN_FLAGE == 1) and /sys/class/backlight/backlight/brightness is above 0;
  • 60 s have passed since the last refresh.
• Since the worker still wakes every 3 s, the status catches up within 3 s after the screen comes back.
• test_wifi_connection.cpp defines LVGL_RUN_FLAGE, as the keyboard tests do, because the network source now reads it.

Measurements

One polling round (dev status plus dev wifi list --rescan no) costs about 0.18 s of CPU on the device (user + sys, averaged over 5 rounds). NetworkManager and dbus-daemon also do work for every call.

These were taken on a CardputerZero (0.6.105 plus the system-sound idle fix from the companion PR). Conditions: screen locked, on battery, 60 s of 2 s samples.

	mean current	median	context switches
before	212 mA	203 mA	818 /s
after	204 mA	202 mA	516 /s

No nmcli process showed up in 30 one-second samples while locked. The saving in current is small. Most of the effect is fewer CPU bursts and less work in NetworkManager and the system bus.

Testing

• sh ext_components/cp0_lvgl/tests/run_tests.sh passes.
• On the device:
  • the Wi-Fi icon updates within a few seconds after waking;
  • Settings → Wi-Fi still lists networks and connects.

Note

I prepared this with AI assistance (Claude) and tested it on my own device. The backlight path is the same one cp0_lvgl_settings.cpp uses. If you would rather not read sysfs here, the poll could instead be driven by a "status visible" flag set from the UI side.