Fix C/Capture swap, read factory stick calibration, hardening (v1.3.0)

[mlstr0m]

Protocol

• C and Capture were swapped in byte 4 of the input report: Capture is 0x02, C is 0x10. Three independent sources agree: ndeadly hid_reports.md, espp PR #765 switch2_pro_report.hpp, and the hardware capture in Findings from a Pro 2 unit: C/Capture swapped, stick calibration, and USB does work #14. Also fixes Capture → DSU Touch.
• Factory stick calibration is read at connect over the command channel (649d4ac9-…f005, replies on c765a961-…), then player LED 1 is lit. Frame format, SPI addresses 0x130A8/0x130E8 and block layout are adapted from kennethreitz's fork @ 975f329 (MIT, Findings from a Pro 2 unit: C/Capture swapped, stick calibration, and USB does work #14), which is based on BlueRetro #1249; the addresses match SDL's Switch 2 driver. This is best effort: if the controller stays silent, the bridge keeps the nominal 2048 range, as before.
• README: …c0f8 is notify-only, so the command channel and the input report layout are now documented.

Hardening

• mappings.json is written atomically (temp file + os.replace).
• "dsu.enabled" must be a boolean; bool("false") used to evaluate to True.
• A warning is shown when the DSU server binds a non-loopback address.
• DSU socket: dropped SO_REUSEADDR. It is useless for UDP, and on Linux it let a second server silently share the port.
• Worker → UI messages are now taken atomically, and the stick hysteresis read happens under the lock.
• Uses get_running_loop(); logging is set up at startup instead of at import.

Build

• Python 3.10+ (bleak 3 requires it), with upper bounds on major versions.
• py2app moved to requirements-build.txt.
• CI matrix: 3.10, 3.12, 3.13.
• The DMG README explains the macOS 15+ Gatekeeper "Open Anyway" flow.

Tests

• New tests cover the byte-4 bits, calibration decoding (using the real blocks published in the fork's PROTOCOL.md), command frames, an end-to-end session with and without a responding controller, and the config hardening.
• Reverting the C/Capture fix makes 3 tests fail.
• The tests were run on Linux with stubbed rumps/pynput. Not yet verified on real hardware: C/Capture, the "sticks calibrated" status and the player LED.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SGsFjHBorX2mfsanZhFyzs

---

Generated by Claude Code