pulse simple API

4 files changed, 65 insertions, 17 deletions

diff --git a/README.md b/README.md
index 11d237a..de93934 100644
--- a/README.md
+++ b/README.md
@@ -12,10 +12,10 @@ It features
 # Building from source
 
 camlet requires meson-build, a C compiler, and the development libraries
-for SDL2, SDL2\_ttf, GL (headers only), v4l2, udev, sodium, jpeglib (from IJG), avcodec, avformat, and fontconfig.
+for SDL2, SDL2\_ttf, GL (headers only), v4l2, udev, sodium, jpeglib (from IJG), avcodec, avformat, pulseaudio, and fontconfig.
 
 These can all be installed on Debian/Ubuntu with
 
 ```sh
-sudo apt install clang meson libv4l-dev libudev-dev libsodium-dev libfontconfig-dev libgl-dev libsdl2-dev libsdl2-ttf-dev libjpeg-dev libavcodec-dev libavformat-dev
+sudo apt install clang meson libv4l-dev libudev-dev libsodium-dev libfontconfig-dev libgl-dev libsdl2-dev libsdl2-ttf-dev libjpeg-dev libpulse-dev libavcodec-dev libavformat-dev
 ```
diff --git a/main.c b/main.c
index d1e91ea..0e049c9 100644
--- a/main.c
+++ b/main.c
@@ -558,7 +558,7 @@ static bool take_picture(State *state) {
 		return false;
 	struct tm *tm = localtime((time_t[1]){time(NULL)});
 	strftime(path + strlen(path), sizeof path - strlen(path), "/%Y-%m-%d-%H-%M-%S", tm);
-	const char *extension = state->mode == MODE_VIDEO ? "mkv" : image_format_extensions[state->image_format];
+	const char *extension = state->mode == MODE_VIDEO ? "mp4" : image_format_extensions[state->image_format];
 	snprintf(path + strlen(path), sizeof path - strlen(path), ".%s", extension);
 	bool success = false;
 	switch (state->mode) {
@@ -611,7 +611,6 @@ int main(void) {
 	if (TTF_Init() < 0) {
 		fatal_error("couldn't initialize SDL2_ttf: %s\n", TTF_GetError());
 	}
-	state->video = video_init();
 	{
 		const char *home = getenv("HOME");
 		if (home) {
@@ -661,12 +660,15 @@ int main(void) {
 	if (!window) {
 		fatal_error("couldn't create window: %s", SDL_GetError());
 	}
+	state->video = video_init();
+
 	static const struct {
 		int maj, min;
 	} gl_versions_to_try[] = {
 		{4, 3},
 		{3, 0}
 	};
+	
 	SDL_GLContext glctx = NULL;
 	for (size_t i = 0; !glctx && i < SDL_arraysize(gl_versions_to_try); i++) {
 		gl.version_major = gl_versions_to_try[i].maj;
diff --git a/meson.build b/meson.build
index 027007f..706a021 100644
--- a/meson.build
+++ b/meson.build
@@ -16,6 +16,8 @@ jpeg = dependency('libjpeg')
 avcodec = dependency('libavcodec')
 avformat = dependency('libavformat')
 avutil = dependency('libavutil')
+pulse = dependency('libpulse')
+pulse_simple = dependency('libpulse-simple')
 m_dep = cc.find_library('m', required: false)
 if m_dep.found()
 	add_project_link_arguments('-lm', language: 'c')
@@ -26,5 +28,5 @@ else
 	debug_def = '-DDEBUG=0'
 endif
 executable('camlet', 'main.c', 'camera.c', 'video.c', '3rd_party/stb_image_write.c',
-	dependencies: [v4l2, sdl2, sdl2_ttf, gl, udev, sodium, fontconfig, jpeg, avcodec, avformat, avutil],
+	dependencies: [v4l2, sdl2, sdl2_ttf, gl, udev, sodium, fontconfig, jpeg, avcodec, avformat, avutil, pulse, pulse_simple],
 	c_args: ['-Wno-unused-function', '-Wno-format-truncation', '-Wshadow', debug_def])
diff --git a/video.c b/video.c
index 536f0dc..d2dd374 100644
--- a/video.c
+++ b/video.c
@@ -1,24 +1,31 @@
 #include "video.h"
-#include "camera.h"
 
+#include <stdatomic.h>
+#include <SDL.h>
 #include <libavcodec/avcodec.h>
 #include <libavformat/avformat.h>
+#include <pulse/pulseaudio.h>
+#include <pulse/simple.h>
 
 #include "util.h"
+#include "camera.h"
 
 struct VideoContext {
 	bool recording;
 	double start_time;
+	double last_frame_time;
 	AVFormatContext *avf_context;
 	AVCodecContext *video_encoder;
 	AVFrame *video_frame;
 	AVPacket *av_packet;
 	AVStream *video_stream;
-	int64_t video_pts;
+	int64_t next_video_pts;
+	pa_simple *pulseaudio;
 };
 
 VideoContext *video_init(void) {
-	return calloc(1, sizeof(VideoContext));
+	VideoContext *ctx = calloc(1, sizeof(VideoContext));
+	return ctx;
 }
 
 bool video_start(VideoContext *ctx, const char *filename, int32_t width, int32_t height, int fps, int quality) {
@@ -93,9 +100,23 @@ bool video_start(VideoContext *ctx, const char *filename, int32_t width, int32_t
 		fprintf(stderr, "error: avformat_write_header: %s\n", av_err2str(err));
 		return false;
 	}
+	pa_sample_spec audio_format = {
+		.format = PA_SAMPLE_S16NE,
+		.channels = 2,
+		.rate = 44100
+	};
+	// NOTE: SDL2 audio capture is currently broken
+	//  https://github.com/libsdl-org/SDL/issues/9706
+	// once SDL3 becomes commonplace enough, we can switch to that for capturing audio.
+	ctx->pulseaudio = pa_simple_new(NULL, "camlet", PA_STREAM_RECORD, NULL,
+		"microphone", &audio_format, NULL, NULL, &err);
+	if (!ctx->pulseaudio) {
+		fprintf(stderr, "%s", pa_strerror(err));
+	}
+	pa_simple_read(ctx->pulseaudio, (char[1]){0}, 1, &err);
 	ctx->recording = true;
-	ctx->video_pts = 0;
-	ctx->start_time = get_time_double();
+	ctx->next_video_pts = 0;
+	ctx->start_time = ctx->last_frame_time = get_time_double();
 	return true;
 }
 
@@ -127,21 +148,40 @@ static bool write_frame(VideoContext *ctx, AVCodecContext *encoder, AVStream *st
 }
 
 bool video_submit_frame(VideoContext *ctx, Camera *camera) {
-	if (!ctx || !camera) return false;	
-	int64_t next_pts = ctx->video_pts;
-	int64_t curr_pts = (int64_t)((get_time_double() - ctx->start_time)
+	if (!ctx || !camera || !ctx->recording) return false;
+	double curr_time = get_time_double();
+	double time_since_start = curr_time - ctx->start_time;
+	double time_since_last_frame = curr_time - ctx->last_frame_time;
+	ctx->last_frame_time = curr_time;
+	int64_t video_pts = (int64_t)(time_since_start
 		* ctx->video_encoder->time_base.den
 		/ ctx->video_encoder->time_base.num);
-	if (curr_pts >= next_pts) {
-		int err = av_frame_make_writable(ctx->video_frame);
+	size_t audio_size = ctx->pulseaudio
+		? (size_t)(time_since_last_frame * 44100 * 2 * sizeof(int16_t))
+		: 0;
+	int err;
+	while (audio_size > 0 && audio_size < SIZE_MAX / 2 /* just in case time goes backwards somehow */) {
+		uint8_t buf[4096];
+		size_t n = audio_size > sizeof buf ? sizeof buf : audio_size;
+		if (pa_simple_read(ctx->pulseaudio, buf, n, &err) < 0) {
+			static atomic_flag warned = ATOMIC_FLAG_INIT;
+			if (!atomic_flag_test_and_set_explicit(&warned, memory_order_relaxed)) {
+				fprintf(stderr, "error: pa_simple_read: %s\n", pa_strerror(err));
+			}
+			break;
+		}
+		audio_size -= n;
+	}
+	if (video_pts >= ctx->next_video_pts) {
+		err = av_frame_make_writable(ctx->video_frame);
 		if (err < 0) {
 			fprintf(stderr, "error: av_frame_make_writable: %s\n", av_err2str(err));
 			return false;
 		}
-		ctx->video_frame->pts = curr_pts;
+		ctx->video_frame->pts = video_pts;
 		camera_copy_to_av_frame(camera, ctx->video_frame);
 		write_frame(ctx, ctx->video_encoder, ctx->video_stream, ctx->video_frame);
-		ctx->video_pts = curr_pts + 1;
+		ctx->next_video_pts = video_pts + 1;
 	}
 	return true;
 }
@@ -154,6 +194,10 @@ bool video_is_recording(VideoContext *ctx) {
 void video_stop(VideoContext *ctx) {
 	if (!ctx) return;
 	if (ctx->recording) {
+		if (ctx->pulseaudio) {
+			pa_simple_free(ctx->pulseaudio);
+			ctx->pulseaudio = NULL;
+		}
 		ctx->recording = false;
 		// flush video encoder
 		write_frame(ctx, ctx->video_encoder, ctx->video_stream, NULL);