TeaSpeak-Client/native/serverconnection/src/connection/audio/AudioSender.cpp

#include "AudioSender.h"
#include "VoiceConnection.h"
#include "../ServerConnection.h"
#include "../../logger.h"
#include "AudioEventLoop.h"
#include "../../audio/AudioMerger.h"

using namespace std;
using namespace tc;
using namespace tc::audio;
using namespace tc::audio::codec;
using namespace tc::connection;

VoiceSender::VoiceSender(tc::connection::VoiceConnection *handle) : handle(handle) {}

VoiceSender::~VoiceSender() {
	audio::encode_event_loop->cancel(dynamic_pointer_cast<event::EventEntry>(this->_ref.lock()));
	this->clear_buffer(); /* buffer might be accessed within encode_raw_frame, but this could not be trigered while this will be deallocated! */
}

bool VoiceSender::initialize_codec(std::string& error, connection::codec::value codec, size_t channels, size_t rate, bool reset_encoder) {
	auto& data = this->codec[codec];
	bool new_allocated = !data;
	if(new_allocated) {
		data = make_unique<AudioCodec>();
		data->packet_counter = 0;
		data->last_packet = chrono::system_clock::now();
	}

	auto info = codec::get_info(codec);
	if(!info || !info->supported) {
		if(new_allocated)
			log_error(category::voice_connection, tr("Tried to send voice packet  but we dont support the current codec ({})"), codec);
		return false;
	}

	if(!data->converter) {
		data->converter = info->new_converter(error);
		if(!data->converter)
			return false;
	} else if(reset_encoder) {
		data->converter->reset_encoder();
	}

	if(!data->resampler || data->resampler->input_rate() != rate)
		data->resampler = make_shared<AudioResampler>(rate, data->converter->sample_rate(), data->converter->channels());

	if(!data->resampler->valid()) {
		error = "resampler is invalid";
		return false;
	}

	return true;
}

void VoiceSender::set_voice_send_enabled(bool flag) {
	this->voice_send_enabled = flag;
}

void VoiceSender::send_data(const void *data, size_t samples, size_t rate, size_t channels) {
	unique_lock lock(this->_execute_lock);
	if(!this->handle) {
		log_warn(category::voice_connection, tr("Dropping raw audio frame because of an invalid handle."));
		return;
	}
	lock.unlock();

	if(!this->voice_send_enabled) {
		log_warn(category::voice_connection, tr("Dropping raw audio frame because voice sending has been disabled!"));
		return;
	}

	auto frame = make_unique<AudioFrame>();
	frame->sample_rate = rate;
	frame->channels = channels;
	frame->buffer = pipes::buffer{(void*) data, samples * channels * 4};
	frame->timestamp = chrono::system_clock::now();

	{
		lock_guard buffer_lock(this->raw_audio_buffer_lock);
		this->raw_audio_buffers.push_back(move(frame));
	}

	audio::encode_event_loop->schedule(dynamic_pointer_cast<event::EventEntry>(this->_ref.lock()));
}

void VoiceSender::send_stop() {
	unique_lock lock(this->_execute_lock);
	if(!this->handle) {
		log_warn(category::voice_connection, tr("Dropping audio end frame because of an invalid handle."));
		return;
	}
	lock.unlock();


	auto frame = make_unique<AudioFrame>();
	frame->sample_rate = 0;
	frame->channels = 0;
	frame->buffer = pipes::buffer{nullptr, 0};
	frame->timestamp = chrono::system_clock::now();

	{
		lock_guard buffer_lock(this->raw_audio_buffer_lock);
		this->raw_audio_buffers.push_back(move(frame));
	}

	audio::encode_event_loop->schedule(dynamic_pointer_cast<event::EventEntry>(this->_ref.lock()));
}

void VoiceSender::finalize() {
	lock_guard lock(this->_execute_lock);
	this->handle = nullptr;
}

void VoiceSender::event_execute(const std::chrono::system_clock::time_point &point) {
	static auto max_time = chrono::milliseconds(10);

	bool reschedule = false;
	auto now = chrono::system_clock::now();
	while(true) {
		unique_lock buffer_lock(this->raw_audio_buffer_lock);
		if(this->raw_audio_buffers.empty())
			break;

		if(chrono::system_clock::now() - now > max_time) {
			reschedule = true;
			break;
		}

		auto entry = move(this->raw_audio_buffers.front());
		this->raw_audio_buffers.pop_front();
		buffer_lock.unlock();

		//TODO: Drop too old buffers!
		this->encode_raw_frame(entry);
	}

	if(reschedule) {
		log_warn(category::voice_connection, tr("Audio data decode will take longer than {} us. Enqueueing for later"), chrono::duration_cast<chrono::microseconds>(max_time).count());
		audio::decode_event_loop->schedule(dynamic_pointer_cast<event::EventEntry>(this->_ref.lock()));
	}
}

void VoiceSender::encode_raw_frame(const std::unique_ptr<AudioFrame> &frame) {
	auto codec = this->_current_codec;
	auto& codec_data = this->codec[codec];

	bool flag_head = true, flag_reset = true;
	if(codec_data) {
		if(codec_data->last_packet + chrono::seconds(1) < frame->timestamp)
			codec_data->packet_counter = 0;

		flag_head = codec_data->packet_counter < 5;
		flag_reset = codec_data->packet_counter == 0;

		codec_data->packet_counter++;
		codec_data->last_packet = frame->timestamp;
	}

	if(frame->channels == 0 || frame->sample_rate == 0 || frame->buffer.empty()) {
		lock_guard lock(this->_execute_lock);
		if(!this->handle) {
			log_warn(category::voice_connection, tr("Dropping audio end because of an invalid handle."));
			return;
		}

		if(codec_data)
			codec_data->packet_counter = 0;
		auto server = this->handle->handle();
		server->send_voice_data(this->_buffer, 0, codec, flag_head);
		return;
	}

	string error;
	if(flag_reset) {
		log_trace(category::voice_connection, tr("Resetting encoder for voice sender"));
	}
	if(!this->initialize_codec(error, codec, frame->channels, frame->sample_rate, flag_reset)) {
		log_error(category::voice_connection, tr("Failed to initialize codec: {}"), error);
		return;
	}

	auto merged_channel_byte_size = codec_data->converter->channels() * (frame->buffer.length() / frame->channels);
	auto estimated_resampled_byte_size = codec_data->resampler->estimated_output_size(merged_channel_byte_size);
	this->ensure_buffer(max(estimated_resampled_byte_size, merged_channel_byte_size));

	auto codec_channels = codec_data->converter->channels();
	if(!audio::merge::merge_channels_interleaved(this->_buffer, codec_channels, frame->buffer.data_ptr(), frame->channels, frame->buffer.length() / frame->channels / 4)) {
		log_warn(category::voice_connection, tr("Failed to merge channels to output stream channel count! Dropping local voice packet"));
		return;
	}

	auto resampled_samples = codec_data->resampler->process(this->_buffer, this->_buffer, merged_channel_byte_size / codec_channels / 4);
	if(resampled_samples <= 0) {
		log_error(category::voice_connection, tr("Resampler returned {}"), resampled_samples);
		return;
	}

	if(resampled_samples * codec_channels * 4 != codec_data->converter->bytes_per_frame()) {
		log_error(category::voice_connection,
				tr("Could not encode audio frame. Frame length is not equal to code frame length! Codec: {}, Packet: {}"),
				codec_data->converter->bytes_per_frame(), resampled_samples * codec_channels * 4
		);
		return;
	}

	char _packet_buffer[512];
	auto encoded_bytes = codec_data->converter->encode(error, this->_buffer, _packet_buffer, 512);
	if(encoded_bytes <= 0) {
		log_error(category::voice_connection, tr("Failed to encode voice: {}"), error);
		return;
	}

	{
		lock_guard lock(this->_execute_lock);
		if(!this->handle) {
			log_warn(category::voice_connection, tr("Dropping audio frame because of an invalid handle."));
			return;
		}

		auto server = this->handle->handle();
		server->send_voice_data(_packet_buffer, encoded_bytes, codec, flag_head);
	}
}