René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"0.14.1 esp32s3 OTA black-screens on Fri3d badges: drivers.display.st7789 not frozen into the build","articleBody":"## Summary\n\nThe **0.14.1 esp32s3 OTA image black-screens on Fri3d badges**. MicroPython boots fine, but `mpos.main` aborts during board init because the SPI display driver `drivers.display.st7789` is not present in the firmware. Because the app never marks the new image valid, the ESP-IDF bootloader rolls back to the previous version, so users see: download OK, then black screen, then back to the old release.\n\nThis affects **all Fri3d badges** (both `fri3d_2024.py` and `fri3d_2026.py` do `import drivers.display.st7789`), so it is not board-specific.\n\n## Environment\n\n- Firmware: `MicroPythonOS_esp32s3_0.14.1.ota` (from `https://updates.micropythonos.com/releases/esp32s3/`)\n- MicroPython: `1.27.0`, build `78ff170de9-dirty` (2026-07-06)\n- Hardware: Fri3d 2024 badge (see \"Secondary\" below about detection)\n- Prior working release on the same badge: 0.7.x (rolls back to it cleanly)\n\n## Symptom / serial boot log\n\nCaptured over USB serial during the failing boot:\n\n```\nInitializing fri3d_2026 hardware\nError importing mpos.main, sleeping 5 seconds before printing the exception...\nTraceback (most recent call last):\n  File \"main.py\", line 36, in \u003cmodule\u003e\n  File \"lib/mpos/main.py\", line 49, in \u003cmodule\u003e\n  File \"lib/mpos/board/fri3d_2026.py\", line 17, in \u003cmodule\u003e\nImportError: no module named 'st7789'\nMicroPythonOS exiting.\nMicroPython 78ff170de9-dirty on 2026-07-06; Generic ESP32S3 module with Octal-SPIRAM with ESP32S3\nType \"help()\" for more information.\n\u003e\u003e\u003e\n```\n\nThe `no module named 'st7789'` is the missing leaf of `import drivers.display.st7789`.\n\n## Diagnosis (confirmed on the live 0.14.1 REPL)\n\n- `import lcd_bus` -\u003e OK\n- `import lvgl` -\u003e OK\n- `import drivers.display.st7789` -\u003e ImportError\n\nSo the low-level display bus and LVGL are in the build, but the **lvgl_micropython display driver `drivers.display.st7789` (and its `display_driver_framework` dependency) were not frozen into the 0.14.1 esp32s3 image.** These come from `lvgl_micropython/api_drivers/common_api_drivers/display/`. They were present in 0.13.0 and earlier (that is why 0.7.x boots).\n\nRoot cause: a build/freeze regression between 0.13.0 (works) and 0.14.1 (broken), most likely from a `lvgl_micropython` submodule bump or a change in how the display drivers get frozen for the esp32s3 target.\n\n## Step-by-step fix\n\n1. **Reproduce** on any 0.14.1 esp32s3 build: at the REPL, `import drivers.display.st7789` raises ImportError while `import lcd_bus` and `import lvgl` succeed.\n2. **Locate the regression.** Diff the build between the last-good release (0.13.0) and 0.14.1, focusing on:\n   - the `lvgl_micropython` submodule commit (a bump is the most likely culprit),\n   - the esp32s3 `make.py` invocation in `scripts/build_mpos.sh` (whether a display driver is passed / how api_drivers get frozen),\n   - `manifests/manifest.py`.\n   Something in that set stopped freezing `api_drivers/common_api_drivers/display/` (which freezes as `drivers.display.*` plus `display_driver_framework`).\n3. **Restore the freeze.** Ensure `drivers.display.st7789` and `display_driver_framework` are frozen into the esp32s3 image again, exactly as in 0.13.0. Both `fri3d_2024.py` and `fri3d_2026.py` import `drivers.display.st7789`, so both boards need it.\n4. **Rebuild and republish.** Build a fixed esp32s3 OTA (e.g. 0.14.2), publish the `.ota` to `updates.micropythonos.com`, and update the per-hardware `osupdate_\u003chardware_id\u003e.json` manifest to point at it.\n5. **Verify on hardware.** OTA a Fri3d 2024 and a Fri3d 2026 badge and confirm it boots to the UI (no black screen), and that `import drivers.display.st7789` works at the REPL.\n\n## Secondary (not the cause, but worth a look)\n\nMy badge is a **Fri3d 2024**, yet 0.14.1 logged `Initializing fri3d_2026 hardware`. That points to a possible board-misdetection in 0.14.x (`detect_board()` distinguishes the boards by IMU I2C address: 0x6A -\u003e 2026, 0x6B -\u003e 2024). This does not cause the black screen (a correctly-detected 2024 hits the same missing `st7789` import), but the misdetection is worth verifying while the display fix is in.\n\n## How it was captured\n\nBoot log captured over USB serial at 115200 during the OTA attempt, with an auto-reconnecting reader to survive the reboot into the new image.\n","author":{"url":"https://github.com/jnuyens","@type":"Person","name":"jnuyens"},"datePublished":"2026-07-06T21:30:55.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":0},"url":"https://github.com/186/MicroPythonOS/issues/186"}