sys_sleep doesnt guard against negative delays

[Ryzee119]

If the frame time is longer than framelen, the delay will be negative. See

SDL2-GNUBoy/src/emu.c

Line 88 in f4427f2

delay = framelen - sys_elapsed(timer);

sys_sleep in nix and windows dont check this and will hang. I encountered this went disabling sound.

Upsteam does this:
https://github.com/rofl0r/gnuboy/blob/1655356c6430c7c68c3c6071e1a44e3bea5066e9/sys/nix/nix.c#L63

Secondly,

pcm_submit always returns 1. This should return 0 if not using sound to fall back to the manual delay

SDL2-GNUBoy/sys/sdl2/sdl-audio.c

Lines 77 to 90 in f4427f2

	int pcm_submit()
	{
	#ifdef SOUND
	if (!pcm.buf)
	return 0;
	if (pcm.pos < pcm.len)
	return 1;
	while (!audio_done)
	SDL_Delay(4);
	audio_done = 0;
	pcm.pos = 0;
	#endif
	return 1;
	}

Both these fixes need to be implemented to allow disabling sound while maintaining accurate timing

[AlexOberhofer]

Do you have recreation steps in how you were getting the application to hang up? It still seems to freeze up when resizing/moving the window when the binary was compiled with the SILENT flag

[AlexOberhofer]

Adding an extra check to the system sleep similar to how it was done in upstream allows me to move/resize without hanging the emulation thread... Are there any other use cases I should try before I merge in that fix? See PR #9

[Ryzee119]

Normally this wouldnt actually happen as the code always did this manual delay and never returned 0 so the alternate delay loop never was called. If you did disable sound the game just ran really fast.

SDL2-GNUBoy/sys/sdl2/sdl-audio.c

Line 85 in f4427f2

SDL_Delay(4);

It occurred if you disabled sound and made that function return 0 like its meant to, which made it fallback to the alternative delay:

SDL2-GNUBoy/src/emu.c

Lines 86 to 91 in f4427f2

	if (!pcm_submit())
	{
	delay = framelen - sys_elapsed(timer);
	sys_sleep(delay);
	sys_elapsed(timer);
	}