Experimental firmware for people helping to chase specific problems. If nobody sent you here, you want the stable release instead.
0.4.6 switched the basemap to OpenStreetMap tiles, which take several times longer to decode than the old CARTO ones. A full-screen map render (the screensaver above all) decoded every tile back to back and starved the watchdog - shaking on 800x480 boards, a crash on the 7B where the render pass is longest. Round 2 adds the second half of the fix: tiles now decode in their native palette format (a quarter of the memory, measurably faster), which also cures missing map blocks on the route view. Verified clean on an 800x480 board; the 7B is the board that needs the confirmation.
OTA alternative: esp32flight-0.4.7-7b-ota.bin
through the web panel. Command line: merged image,
esptool.py --chip esp32s3 --port PORT write_flash 0x0 esp32flight-0.4.7-7b-full.bin.
When reporting, include:
Panel revision V1 is confirmed working and lives on the stable flasher now. The V2 glass (rear-case batch number 2628 and higher) uses different MIPI timings and its own init table, and no V2 owner has reported back yet. This build carries everything the V1 saga fixed (factory touch firmware table, bring-up retries, sensor grid scaling) on the V2 display path.
Command line: merged image, then
esptool.py --chip esp32p4 --port PORT write_flash 0x0 esp32flight-0.4.7-jc10-v2-full.bin
(back up first: read_flash 0 0x1000000 stock-backup.bin).
When reporting, include: