c26f3cd6ec
Three related fixes for per-key lighting state being lost on the wire: 1. The per-key canvas models layout gaps as placeholder cells with zone_id -1. Paint strokes sweeping a gap emitted -1 in the delta, which entered the setting's map, was persisted, and then aborted EVERY subsequent full-map repaint at frame pack time (OverflowError: can't convert negative int to unsigned) — the apply was logged and ignored, so keys silently kept stale colors or went dark. Filter gap cells from the delta before it reaches the sink. 2. PerKeyLighting now sanitizes zone keys against the device-reported zone bitmap in update/write/read paths, so an already-poisoned persisted map (saved by a build without fix 1) heals on read instead of permanently breaking repaints. 3. WIRELESS_DEVICE_STATUS power-on reconfig requests went through the ConfigChange-cookie dedup gate. On a G915 TKL the cookie survives a power cycle, so the gate concluded nothing changed and skipped apply_all_settings — but power-on wipes volatile state (RGB buffers, host-mode lighting), leaving the keyboard dark until manual intervention. A powered-on notification now forces the apply. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| hid_parser | ||
| hidapi | ||
| logitech_receiver | ||
| solaar | ||
| test_keysyms | ||
| __init__.py | ||
| conftest.py | ||