feat: more straightforward throttling implementation

include/dpp/discordvoiceclient.h

@@ -561,9 +561,9 @@ class DPP_EXPORT discord_voice_client : public websocket_client
 	std::chrono::high_resolution_clock::time_point last_timestamp;
 
 	/**
-	 * @brief Fraction of the sleep that was not executed after the last audio packet was sent
+	 * @brief Last sent packet duration
 	 */
-	std::chrono::nanoseconds last_sleep_remainder{};
+	uint64_t last_duration;
 
 	/**
 	 * @brief Maps receiving ssrc to user id
@@ -799,12 +799,11 @@ class DPP_EXPORT discord_voice_client : public websocket_client
 	 * audio data because Discord does not expect to receive, say, 3 minutes'
 	 * worth of audio data in 1 second.
 	 *
-	 * There are some inaccuracies in the throttling method used by the recorded
+	 * There was some inaccuracies in the throttling method used by the recorded
 	 * audio mode on some systems (mainly Windows) which causes gaps and stutters
 	 * in the resulting audio stream. The overlap audio mode provides a different 
-	 * implementation that fixes the issue. This method is slightly more CPU 
-	 * intensive, and should only be used if you encounter issues with recorded audio 
-	 * on your system.
+	 * implementation that fixes the issue in the past. This method is not used
+	 * anymore and behave the same as the recorded audio mode.
 	 * 
 	 * Use discord_voice_client::set_send_audio_type to change this value as
 	 * it ensures thread safety.

src/dpp/voice/enabled/constructor.cpp

@@ -50,6 +50,7 @@ discord_voice_client::discord_voice_client(dpp::cluster* _cluster, full_reconnec
 	timestamp(0),
 	packet_nonce(1),
 	last_timestamp(std::chrono::high_resolution_clock::now()),
+	last_duration(0),
 	sending(false),
 	tracks(0),
 	dave_version(enable_dave ? dave_version_1 : dave_version_none),

src/dpp/voice/enabled/write_ready.cpp

@@ -31,23 +31,27 @@
 namespace dpp {
 
 void discord_voice_client::write_ready() {
-	bool needs_write = false;
-	{
-		std::lock_guard<std::mutex> lock(this->stream_mutex);
-		const bool needs_stop_frames = this->paused && !this->sent_stop_frames;
-		const bool needs_send_audio = !this->paused && !outbuf.empty();
-		needs_write = needs_stop_frames || needs_send_audio;
-	}
+	std::chrono::nanoseconds latency{0};
+	if (send_audio_type != satype_live_audio) {
+		const auto now = std::chrono::high_resolution_clock::now();
+
+		const auto minimum = std::chrono::nanoseconds(last_duration);
+		const auto elapsed = std::chrono::duration_cast<std::chrono::nanoseconds>(now - last_timestamp);
 
-	if (needs_write) {
-		udp_events.flags = WANT_READ | WANT_WRITE | WANT_ERROR;
-		owner->socketengine->update_socket(udp_events);
+		if (elapsed < minimum) {
+			std::this_thread::sleep_for(minimum - elapsed);
+			udp_events.flags = WANT_READ | WANT_WRITE | WANT_ERROR;
+			owner->socketengine->update_socket(udp_events);
+			return;
+		}
+
+		latency = elapsed - minimum;
+		last_timestamp = now;
 	}
 
 	uint64_t duration = 0;
 	bool track_marker_found = false;
 	uint64_t bufsize = 0;
-	send_audio_type_t type = satype_recorded_audio;
 	{
 		std::lock_guard<std::mutex> lock(this->stream_mutex);
 		if (this->paused) {
@@ -58,7 +62,6 @@ void discord_voice_client::write_ready() {
 
 			/* Fallthrough if paused */
 		} else if (!outbuf.empty()) {
-			type = send_audio_type;
 			if (outbuf[0].packet.size() == sizeof(uint16_t) && (*(reinterpret_cast<uint16_t*>(outbuf[0].packet.data()))) == AUDIO_TRACK_MARKER) {
 				outbuf.erase(outbuf.begin());
 				track_marker_found = true;
@@ -74,40 +77,15 @@ void discord_voice_client::write_ready() {
 					outbuf.erase(outbuf.begin());
 				}
 			}
+			if (!outbuf.empty()) {
+				udp_events.flags = WANT_READ | WANT_WRITE | WANT_ERROR;
+				owner->socketengine->update_socket(udp_events);
+			}
 		}
 	}
 	if (duration) {
-		if (type == satype_recorded_audio) {
-			std::chrono::nanoseconds latency = std::chrono::duration_cast<std::chrono::nanoseconds>(std::chrono::high_resolution_clock::now() - last_timestamp);
-			std::chrono::nanoseconds sleep_time = std::chrono::nanoseconds(duration) - latency;
-			if (sleep_time.count() > 0) {
-				std::this_thread::sleep_for(sleep_time);
-			}
-		}
-		else if (type == satype_overlap_audio) {
-			std::chrono::nanoseconds latency = std::chrono::duration_cast<std::chrono::nanoseconds>(std::chrono::high_resolution_clock::now() - last_timestamp);
-			std::chrono::nanoseconds sleep_time = std::chrono::nanoseconds(duration) + last_sleep_remainder - latency;
-			std::chrono::nanoseconds sleep_increment = (std::chrono::nanoseconds(duration) - latency) / AUDIO_OVERLAP_SLEEP_SAMPLES;
-			if (sleep_time.count() > 0) {
-				uint16_t samples_count = 0;
-				std::chrono::nanoseconds overshoot_accumulator{};
-
-				do {
-					std::chrono::high_resolution_clock::time_point start_sleep = std::chrono::high_resolution_clock::now();
-					std::this_thread::sleep_for(sleep_increment);
-					std::chrono::high_resolution_clock::time_point end_sleep = std::chrono::high_resolution_clock::now();
-
-					samples_count++;
-					overshoot_accumulator += std::chrono::duration_cast<std::chrono::nanoseconds>(end_sleep - start_sleep) - sleep_increment;
-					sleep_time -= std::chrono::duration_cast<std::chrono::nanoseconds>(end_sleep - start_sleep);
-				} while (std::chrono::nanoseconds(overshoot_accumulator.count() / samples_count) + sleep_increment < sleep_time);
-				last_sleep_remainder = sleep_time;
-			} else {
-				last_sleep_remainder = std::chrono::nanoseconds(0);
-			}
-		}
-
-		last_timestamp = std::chrono::high_resolution_clock::now();
+		const auto latcount = latency.count();
+		last_duration = duration > latcount ? duration - latcount : duration;
 		if (!creator->on_voice_buffer_send.empty()) {
 			voice_buffer_send_t snd(owner, 0, "");
 			snd.buffer_size = bufsize;