til: add preliminary audio backend

This is an early implementation of something resembling an audio
backend for rototiller/libtil.

The assumption for now is that everything will use signed 16-bit
native-endian stereo output @ 44.1khz.

1 files changed, 204 insertions, 0 deletions

diff --git a/src/sdl_audio.c b/src/sdl_audio.c
new file mode 100644
index 0000000..90fea35
--- /dev/null
+++ b/src/sdl_audio.c
@@ -0,0 +1,204 @@
+#define SDL_MAIN_HANDLED
+#include <assert.h>
+#include <SDL.h>
+#include <stdlib.h>
+#include <errno.h>
+
+#include "til_audio.h"
+#include "til_audio_context.h"
+#include "til_settings.h"
+#include "til_setup.h"
+
+/* sdl audio backend */
+
+#define SDL_AUDIO_DEFAULT_SAMPLES	1024
+
+typedef struct sdl_audio_setup_t {
+	til_setup_t	til_setup;
+
+	unsigned	frames;
+} sdl_audio_setup_t;
+
+typedef struct sdl_audio_t {
+	til_audio_context_t	til_audio_context;
+
+	SDL_AudioDeviceID	dev;
+} sdl_audio_t;
+
+
+static int sdl_audio_init(til_setup_t *setup, til_audio_context_t **res_context);
+static void sdl_audio_shutdown(til_audio_context_t *context);
+static int sdl_audio_queue(til_audio_context_t *context, int16_t *frames, int n_frames);
+static unsigned sdl_audio_n_queued(til_audio_context_t *context);
+static void sdl_audio_drop(til_audio_context_t *context);
+static void sdl_audio_pause(til_audio_context_t *context);
+static void sdl_audio_unpause(til_audio_context_t *context);
+static int sdl_audio_setup(const til_settings_t *settings, til_setting_t **res_setting, const til_setting_desc_t **res_desc, til_setup_t **res_setup);
+
+
+til_audio_ops_t sdl_audio_ops = {
+	.init = sdl_audio_init,
+	.shutdown = sdl_audio_shutdown,
+	.drop = sdl_audio_drop,
+	.pause = sdl_audio_pause,
+	.unpause = sdl_audio_unpause,
+	.queue = sdl_audio_queue,
+	.n_queued = sdl_audio_n_queued,
+	.setup = sdl_audio_setup,
+};
+
+
+static int sdl_err_to_errno(int err)
+{
+	switch (err) {
+	case SDL_ENOMEM:
+		return ENOMEM;
+	case SDL_EFREAD:
+	case SDL_EFWRITE:
+	case SDL_EFSEEK:
+		return EIO;
+	case SDL_UNSUPPORTED:
+		return ENOTSUP;
+	default:
+		return EINVAL;
+	}
+}
+
+
+static int sdl_audio_init(til_setup_t *setup, til_audio_context_t **res_context)
+{
+	sdl_audio_setup_t	*s = (sdl_audio_setup_t *)setup;
+	sdl_audio_t		*c;
+	int			r;
+
+	assert(setup);
+	assert(res_context);
+
+	c = til_audio_context_new(&sdl_audio_ops, sizeof(sdl_audio_t), setup);
+	if (!c)
+		return -ENOMEM;
+
+	SDL_SetMainReady(); /* is it a problem (or necessary) for both sdl_fb and sdl_audio to do this? */
+	r = SDL_InitSubSystem(SDL_INIT_AUDIO);
+	if (r < 0) {
+		free(c);
+		return -sdl_err_to_errno(r);
+	}
+
+	{
+		SDL_AudioSpec	aspec = {
+			.freq = 44100,
+			.format = AUDIO_S16,
+			.channels = 2,
+			.samples = s->frames,
+		};
+
+		c->dev = SDL_OpenAudioDevice(NULL, 0, &aspec, NULL, 0);
+		if (!c->dev) {
+			/* SDL_GetError only returns strings, we need an errno FIXME */
+			free(c);
+			return -EPERM;
+		}
+	}
+
+	*res_context = &c->til_audio_context;
+
+	return 0;
+}
+
+
+static void sdl_audio_shutdown(til_audio_context_t *context)
+{
+	sdl_audio_t	*c = (sdl_audio_t *)context;
+
+	SDL_CloseAudioDevice(c->dev);
+}
+
+
+static void sdl_audio_drop(til_audio_context_t *context)
+{
+	sdl_audio_t	*c = (sdl_audio_t *)context;
+
+	SDL_ClearQueuedAudio(c->dev);
+}
+
+
+static void sdl_audio_pause(til_audio_context_t *context)
+{
+	sdl_audio_t	*c = (sdl_audio_t *)context;
+
+	SDL_PauseAudioDevice(c->dev, 1);
+}
+
+
+static void sdl_audio_unpause(til_audio_context_t *context)
+{
+	sdl_audio_t	*c = (sdl_audio_t *)context;
+
+	SDL_PauseAudioDevice(c->dev, 0);
+}
+
+
+static int sdl_audio_queue(til_audio_context_t *context, int16_t *frames, int n_frames)
+{
+	sdl_audio_t	*c = (sdl_audio_t *)context;
+
+	/* TODO FIXME: SDL_audio.h says this returns a negative error code on
+	 * error, but that's not the same as a -errno, so directly returning
+	 * from here isn't exactly appropriate.
+	 */
+	return SDL_QueueAudio(c->dev, frames, n_frames * (sizeof(*frames) * 2));
+}
+
+
+static unsigned sdl_audio_n_queued(til_audio_context_t *context)
+{
+	sdl_audio_t	*c = (sdl_audio_t *)context;
+
+	return SDL_GetQueuedAudioSize(c->dev) / (sizeof(int16_t) * 2);
+}
+
+
+static int sdl_audio_setup(const til_settings_t *settings, til_setting_t **res_setting, const til_setting_desc_t **res_desc, til_setup_t **res_setup)
+{
+	til_setting_t	*frames;
+	const char	*frames_values[] = {
+				"512",
+				"1024",
+				"2048",
+				"4096",
+				"8192",
+				NULL
+			};
+	int		r;
+
+	r = til_settings_get_and_describe_setting(settings,
+						&(til_setting_spec_t){
+							.name = "Audio frames buffered",
+							.key = "frames",
+							.regex = "[0-9]+",
+							.preferred = TIL_SETTINGS_STR(SDL_AUDIO_DEFAULT_SAMPLES),
+							.values = frames_values,
+							.annotations = NULL
+						},
+						&frames,
+						res_setting,
+						res_desc);
+	if (r)
+		return r;
+
+	if (res_setup) {
+		sdl_audio_setup_t	*setup;
+
+		setup = til_setup_new(settings, sizeof(*setup), NULL, &sdl_audio_ops);
+		if (!setup)
+			return -ENOMEM;
+
+		if (sscanf(frames->value, "%u", &setup->frames) != 1)
+			return til_setup_free_with_failed_setting_ret_err(&setup->til_setup, frames, res_setting, -EINVAL);
+
+		*res_setup = &setup->til_setup;
+	}
+
+	return 0;
+}