apu.c: Kill audio subsystem if there are no playback devices

[MasonT8198]

fixes issue #354

@GXTX 's initial hypothesis was correct, simply killing the audio subsystem if there are no available audio devices does work...

in the event the dummy audio device does fail to load, assert and kill the program, however in my testing this shouldn't occur

[antangelo]

I tried testing this on linux by stopping pipewire, but the assertion still trips. SDL_GetNumAudioDevices(0) reports value 2.

As an aside, what happens if there is no audio device on loading xemu, but then one is connected after the fact? Is xemu able to output to the newly connected device?

[MasonT8198]

I tried testing this on linux by stopping pipewire, but the assertion still trips. SDL_GetNumAudioDevices(0) reports value 2.

As an aside, what happens if there is no audio device on loading xemu, but then one is connected after the fact? Is xemu able to output to the newly connected device?

I've been unable to repro on my machine as to that specific assertion, even so, I'm not sure why'd it report 2... unless it was grabbing input devices that it couldn't access.

I'm unsure as to how SDL handles that, in most games that use SDL for audio, basically always has to restart to detect new devices.. could be wrong.

As an aside, it's been a minute since I looked at the xemu UI, Using SDL for an audio device selector would be nice.

[mborgerson]

I tried testing this on linux by stopping pipewire, but the assertion still trips. SDL_GetNumAudioDevices(0) reports value 2.

In this case, how about we detect the failed open, then fall back to re-open on the dummy driver

As an aside, what happens if there is no audio device on loading xemu, but then one is connected after the fact? Is xemu able to output to the newly connected device?

This would be nice to have, but not necessarily in this PR