Drop the injected AlarmSound Protocol and the wecker.Sound class. Each style
now owns its tone by using pygame directly; the runner owns only the audio
engine lifecycle (mixer init/quit), button sampling, and cleanup. This is the
seam for future styles that handle their own tone — a 'talk' style would just
import pygame and play speech, with no shared interface to extend.
- styles/base.py: AlarmStyle.__init__(set_led, music_file) + start() + update()
- styles/blink.py: start() loads+plays music via pygame.mixer.music
- styles/simple.py: owns its beep — synthesises a square-wave buffer in
module (stdlib array+math) and plays it via pygame.mixer.Sound
- wecker.py: setup() brings up the mixer only; run_alarm constructs the style
and guards start() with a clean log+exit on failure
Tests: a shared tests/conftest.py stubs RPi.GPIO/pygame in sys.modules before
any SUT import (order-independent, removes duplicated inline mocking); the
wecker mock_pygame fixture patches one fresh pygame mock into wecker + both
style modules so assertions see the same calls.
README: tone-ownership and plugin-contract updated.
gurix
3a7b6ac7a3