gif_to_360: --crop-x option for the horizontal crop position

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
mangavalkandClaude Opus 5.5 committed 2026-10-24 17:31:05 +02:00
1 parent b2f6a52902
commit 21012adbec
2 files changed
+8 -6

No files matched your search

+1 -1
View File
@@ -128,7 +128,7 @@ AliExpress "ELM327 Mini V2.1 Bluetooth OBD **Dual Mode** 5.1" — dual mode = Bl
- **LittleFS** works on the `spiffs` partition of `huge_app.csv` (~900 KB); the first mount after flashing formats it (prints `Corrupted dir pair` / `mount failed (-84)` and some `task_wdt ... task not found` once — harmless).
- BLE (NimBLE) in the user's home: ~12 devices visible within 10 s.
- Big fonts: generate with `npx lv_font_conv` (Node 22 installed) from `lvgl/scripts/generators/built_in_font/Montserrat-Medium.ttf`; `-DLV_LVGL_H_INCLUDE_SIMPLE` (in shared ini) makes the generated `#include` work.
- **GIF size after converting**: upscaling with LANCZOS can make a GIF ~3x bigger (7.2 MB source 498x298 -> 24.9 MB), because blended pixels plus per-frame palettes with dithering break the delta compression. Nearest-neighbour scaling gave 7.2 MB and looked as good. `tools/gif_to_360.py` retries with sharp scaling automatically when the result exceeds 11.5 MB.
- **GIF size after converting**: upscaling with LANCZOS can make a GIF ~3x bigger (7.2 MB source 498x298 -> 24.9 MB), because blended pixels plus per-frame palettes with dithering break the delta compression. Nearest-neighbour scaling gave 7.2 MB and looked as good. `tools/gif_to_360.py` retries with sharp scaling automatically when the result exceeds 11.5 MB. `--crop-x 0..1` moves the crop of wide GIFs (0 = left, 0.5 = center); `3d2c218d...gif` is converted with `--crop-x 0.36` (character is mostly left; best visible share inside the circle at 0.3-0.4).
- Binary assets (GIFs): `board_build.embed_files = data/x.gif` → `extern const uint8_t x_start[] asm("_binary_data_x_gif_start")` (no huge C arrays).
- Python edit scripts: never pass C strings containing `\n` through a bash heredoc into Python — it turned into a real line break twice; write the script with the Write tool instead.
- Serial output from the background task (core 0) is often lost, not only right after reset; don't rely on it as proof. Serial output right after a reset can be lost/garbled (USB-CDC re-enumeration, prints from two cores interleave) — don't conclude a task isn't running from a missing early line.