binary frame parse

include/dpp/discordvoiceclient.h

@@ -131,6 +131,7 @@ enum voice_websocket_opcode_t : uint8_t {
 	voice_opcode_multiple_clients_connect = 11,
 	voice_opcode_client_connect = 12,
 	voice_opcode_client_disconnect = 13,
+	voice_opcode_media_sink = 15,
 	voice_client_flags = 18,
 	voice_client_platform = 20,
 	voice_client_dave_prepare_transition = 21,
@@ -146,6 +147,22 @@ enum voice_websocket_opcode_t : uint8_t {
 	voice_client_dave_mls_invalid_commit_welcome = 31,
 };
 
+/**
+ * @brief DAVE E2EE Binary frame header
+ */
+#pragma pack(push, 1)
+struct dave_binary_header_t {
+	/**
+	 * @brief Sequence number
+	 */
+	uint16_t seq;
+	/**
+	 * @brief Opcode type
+	 */
+	uint8_t opcode;
+};
+#pragma pack(pop)
+
 /** @brief Implements a discord voice connection.
  * Each discord_voice_client connects to one voice channel and derives from a websocket client.
  */
@@ -456,9 +473,10 @@ class DPP_EXPORT discord_voice_client : public websocket_client
 	/**
 	 * @brief DAVE - Discord Audio Visual Encryption
 	 * Used for E2EE encryption. dave_protocol_none is
-	 * the default before negotiation.
+	 * the default right now.
+	 * @warning DAVE E2EE is an EXPERIMENTAL feature!
 	 */
-	dave_version_t dave_version{dave_version_none};
+	dave_version_t dave_version;
 
 	/**
 	 * @brief Send data to UDP socket immediately.

src/davetest/dave.cpp

@@ -27,9 +27,10 @@
 int main() {
 	using namespace std::chrono_literals;
 	char* t = getenv("DPP_UNIT_TEST_TOKEN");
-	dpp::snowflake TEST_GUILD_ID(std::string(getenv("TEST_GUILD_ID")));
-	dpp::snowflake TEST_VC_ID(std::string(getenv("TEST_VC_ID")));
-	if (t) {
+	if (t != nullptr) {
+		dpp::snowflake TEST_GUILD_ID(std::string(getenv("TEST_GUILD_ID")));
+		dpp::snowflake TEST_VC_ID(std::string(getenv("TEST_VC_ID")));
+		std::cout << "Test Guild ID: " << TEST_GUILD_ID << " Test VC ID: " << TEST_VC_ID << "\n\n";
 		dpp::cluster dave_test(t, dpp::i_default_intents | dpp::i_guild_members);
 		dave_test.set_websocket_protocol(dpp::ws_etf);
 		dave_test.on_log(dpp::utility::cout_logger());
@@ -40,6 +41,6 @@ int main() {
 				s->connect_voice(TEST_GUILD_ID, TEST_VC_ID, false, false, true);
 			}
 		});
-		dave_test.start(dpp::st_wait);
+		dave_test.start(false);
 	}
 }

src/dpp/discordclient.cpp

@@ -713,7 +713,7 @@ voiceconn& voiceconn::connect(snowflake guild_id) {
 		auto t = std::thread([guild_id, this]() {
 			try {
 				this->creator->log(ll_debug, "Connecting voice for guild " + std::to_string(guild_id) + " channel " + std::to_string(this->channel_id));
-				this->voiceclient = new discord_voice_client(creator->creator, this->channel_id, guild_id, this->token, this->session_id, this->websocket_hostname);
+				this->voiceclient = new discord_voice_client(creator->creator, this->channel_id, guild_id, this->token, this->session_id, this->websocket_hostname, this->dave);
 				/* Note: Spawns thread! */
 				this->voiceclient->run();
 			}

src/dpp/discordvoiceclient.cpp

@@ -316,7 +316,7 @@ void discord_voice_client::voice_courier_loop(discord_voice_client& client, cour
 }
 
 discord_voice_client::discord_voice_client(dpp::cluster* _cluster, snowflake _channel_id, snowflake _server_id, const std::string &_token, const std::string &_session_id, const std::string &_host, bool enable_dave)
-	: websocket_client(_host.substr(0, _host.find(":")), _host.substr(_host.find(":") + 1, _host.length()), "/?v=" + std::to_string(voice_protocol_version), OP_TEXT),
+	: websocket_client(_host.substr(0, _host.find(':')), _host.substr(_host.find(':') + 1, _host.length()), "/?v=" + std::to_string(voice_protocol_version), OP_TEXT),
 	runner(nullptr),
 	connect_time(0),
 	mixer(std::make_unique<audio_mixer>()),
@@ -490,6 +490,44 @@ bool discord_voice_client::handle_frame(const std::string &data)
 {
 	log(dpp::ll_trace, std::string("R: ") + data);
 	json j;
+
+	/**
+	 * Because all discord JSON must be valid UTF-8, if we see a packet with the 2nd character
+	 * being less than 32 (' '), then we know it is a binary MLS frame, as all the binary frame
+	 * opcodes are purposefully less than 32. We then try and parse it as MLS binary.
+	 */
+	if (data.size() >= sizeof(dave_binary_header_t) && data[2] <= voice_client_dave_mls_invalid_commit_welcome) {
+
+		/* Debug, remove once this is working */
+		std::cout << dpp::utility::debug_dump((uint8_t*)(data.data()), data.length()) << "\n";
+
+		dave_binary_header_t dave_header{};
+		std::memcpy(&dave_header, data.data(), sizeof(dave_binary_header_t));
+
+		switch (dave_header.opcode) {
+			case voice_client_dave_mls_external_sender: {
+				log(ll_debug, "voice_client_dave_mls_external_sender");
+			}
+				break;
+			case voice_client_dave_mls_proposals: {
+				log(ll_debug, "voice_client_dave_mls_proposals");
+			}
+				break;
+			case voice_client_dave_announce_commit_transaction: {
+				log(ll_debug, "voice_client_dave_announce_commit_transaction");
+			}
+				break;
+			case voice_client_dave_mls_welcome: {
+				log(ll_debug, "voice_client_dave_mls_welcome");
+			}
+				break;
+			default:
+				log(ll_debug, "Unexpected DAVE frame opcode");
+				break;
+		}
+
+		return true;
+	}
 
 	try {
 		j = json::parse(data);
@@ -507,6 +545,7 @@ bool discord_voice_client::handle_frame(const std::string &data)
 			case voice_opcode_connection_heartbeat_ack:
 				/* These opcodes do not require a response or further action */
 			break;
+			case voice_opcode_media_sink:
 			case voice_client_flags: {
 			}
 			break;
@@ -622,8 +661,16 @@ bool discord_voice_client::handle_frame(const std::string &data)
 					}
 				}
 
-				/* This is needed to start voice receiving and make sure that the start of sending isn't cut off */
-				send_silence(20);
+				if (dave_version != dave_version_none) {
+					if (j["d"]["dave_protocol_version"] != static_cast<uint32_t>(dave_version)) {
+						log(ll_error, "We requested DAVE E2EE but didn't receive it from the server, downgrading...");
+						dave_version = dave_version_none;
+						send_silence(20);
+					}
+				} else {
+					/* This is needed to start voice receiving and make sure that the start of sending isn't cut off */
+					send_silence(20);
+				}
 
 				/* Fire on_voice_ready */
 				if (!creator->on_voice_ready.empty()) {
@@ -1166,8 +1213,12 @@ void discord_voice_client::one_second_timer()
 			if (time(nullptr) > last_heartbeat + ((heartbeat_interval / 1000.0) * 0.75)) {
 				queue_message(json({
 					{"op", voice_opcode_connection_heartbeat},
-					{"d", rand()},
-					{"seq_ack", sequence}
+					{
+						"d", {
+							{"t", rand()},
+							{"seq_ack", receive_sequence},
+						}
+					},
 				}).dump(-1, ' ', false, json::error_handler_t::replace), true);
 				last_heartbeat = time(nullptr);
 			}

src/dpp/utility.cpp

@@ -373,7 +373,7 @@ std::string debug_dump(uint8_t* data, size_t length) {
 			ascii.clear();
 		}
 		ascii.push_back(*ptr >= ' ' && *ptr <= '~' ? *ptr : '.');
-		out << to_hex(*ptr);
+		out << to_hex(*ptr) << " ";
 	}
 	out << "    " << ascii;
 	out << "\n";