The microphone index given in `arg1` is 1-indexed, so we have to
subtract 1 from the index to access the proper value.
This wasn't caught up before since we were doing direct access of the
array instead of using `::at32` which wouldn't check that the value is
within bounds.
This allows the original SingStar game to boot up again, as it would
otherwise crash then trying to set the volume to the second channel due
to the index being out of bounds.
It would technically never overflow.. but this add minimal protection against adding beyond [PPU thread-count + entries count]
This does not induce an atomic loop which limits until entries.size() but is slower.
When using the emulated SingStar microphone, the device is only
registered when initializing `cellMic`. However, the `mic_context`
thread is only woken up when calling `register_device`. The registration
happens before initializing `cellMic`, so the thread is never woken up
after registering the device.
Add call to `wake_up` in `load_config_and_init` inside the emulated
SingStar microphone specific code to fix this issue, allowing the thread
to detect the device and update the microphone data.
This allows SingStar to receive the microphone data when using the
emulated SingStar microphone that it previously wasn't receiving.
-Process the mouse buttons even when x/y_max aren't yet determined.
-Allows to start the calibration without shaking the mouse before.
-Extend support to 4 players. The games can't use more than 2, but it
allows more flexibility to mix DS3 and GC3.
-Avoid OOB for unsupported Pads
This was a flawed concept due to risk of deadlocks.
Worst case some we will focus o performanc regressins indvidually as we should and not accept this bug.
