Update voice code to vws V4

- Update internals to be compatible with v4
- Adds multiple encryption mode support.  Previously only `xsalsa20_poly1305` was supported.  Now `xsalsa20_poly1305_suffix` is also supported.
  Note: There is no (nice) way to manually select a mode.  The user needn't worry about this however.
- Fixed speaking state bug.  When you disconnected from a voice channel while a bot was playing, upon reconnect you would be unable to hear the bot.  This was caused by bots not sending their speaking state while transmitting.  Bots will now set their speaking state properly when transmitting.  
  Note: This does not account for sending actual silence, the speaking indicator will still be active.

4 files changed, 86 insertions, 24 deletions

diff --git a/discord/enums.py b/discord/enums.py
index f7df9059..f11f1ed1 100644
--- a/discord/enums.py
+++ b/discord/enums.py
@@ -26,10 +26,10 @@ DEALINGS IN THE SOFTWARE.
 
 from enum import Enum, IntEnum
 
-__all__ = ['ChannelType', 'MessageType', 'VoiceRegion', 'VerificationLevel',
-           'ContentFilter', 'Status', 'DefaultAvatar', 'RelationshipType',
-           'AuditLogAction', 'AuditLogActionCategory', 'UserFlags',
-           'ActivityType', 'HypeSquadHouse', 'NotificationLevel']
+__all__ = ['ChannelType', 'MessageType', 'VoiceRegion', 'SpeakingState',
+           'VerificationLevel', 'ContentFilter', 'Status', 'DefaultAvatar',
+           'RelationshipType', 'AuditLogAction', 'AuditLogActionCategory',
+           'UserFlags', 'ActivityType', 'HypeSquadHouse', 'NotificationLevel']
 
 class ChannelType(Enum):
     text     = 0
@@ -75,6 +75,15 @@ class VoiceRegion(Enum):
     def __str__(self):
         return self.value
 
+class SpeakingState(IntEnum):
+    none       = 0
+    voice      = 1
+    soundshare = 2
+    priority   = 4
+
+    def __str__(self):
+        return self.name
+
 class VerificationLevel(IntEnum):
     none              = 0
     low               = 1
diff --git a/discord/gateway.py b/discord/gateway.py
index eb17c2ef..83699a44 100644
--- a/discord/gateway.py
+++ b/discord/gateway.py
@@ -38,6 +38,7 @@ import websockets
 
 from . import utils
 from .activity import _ActivityTag
+from .enums import SpeakingState
 from .errors import ConnectionClosed, InvalidArgument
 
 log = logging.getLogger(__name__)
@@ -547,6 +548,10 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
         Receive only. Tells you that your websocket connection was acknowledged.
     INVALIDATE_SESSION
         Sent only. Tells you that your RESUME request has failed and to re-IDENTIFY.
+    CLIENT_CONNECT
+        Indicates a user has connected to voice.
+    CLIENT_DISCONNECT
+        Receive only.  Indicates a user has disconnected from voice.
     """
 
     IDENTIFY            = 0
@@ -559,6 +564,8 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
     RESUME              = 7
     HELLO               = 8
     INVALIDATE_SESSION  = 9
+    CLIENT_CONNECT      = 12
+    CLIENT_DISCONNECT   = 13
 
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
@@ -597,7 +604,7 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
     @classmethod
     async def from_client(cls, client, *, resume=False):
         """Creates a voice websocket for the :class:`VoiceClient`."""
-        gateway = 'wss://' + client.endpoint + '/?v=3'
+        gateway = 'wss://' + client.endpoint + '/?v=4'
         ws = await websockets.connect(gateway, loop=client.loop, klass=cls, compression=None)
         ws.gateway = gateway
         ws._connection = client
@@ -610,7 +617,7 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
 
         return ws
 
-    async def select_protocol(self, ip, port):
+    async def select_protocol(self, ip, port, mode):
         payload = {
             'op': self.SELECT_PROTOCOL,
             'd': {
@@ -618,18 +625,28 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
                 'data': {
                     'address': ip,
                     'port': port,
-                    'mode': 'xsalsa20_poly1305'
+                    'mode': mode
                 }
             }
         }
 
         await self.send_as_json(payload)
 
-    async def speak(self, is_speaking=True):
+    async def client_connect(self):
+        payload = {
+            'op': self.CLIENT_CONNECT,
+            'd': {
+                'audio_ssrc': self._connection.ssrc
+            }
+        }
+
+        await self.send_as_json(payload)
+
+    async def speak(self, state=SpeakingState.voice):
         payload = {
             'op': self.SPEAKING,
             'd': {
-                'speaking': is_speaking,
+                'speaking': int(state),
                 'delay': 0
             }
         }
@@ -642,9 +659,6 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
         data = msg.get('d')
 
         if op == self.READY:
-            interval = data['heartbeat_interval'] / 1000.0
-            self._keep_alive = VoiceKeepAliveHandler(ws=self, interval=interval)
-            self._keep_alive.start()
             await self.initial_connection(data)
         elif op == self.HEARTBEAT_ACK:
             self._keep_alive.ack()
@@ -652,7 +666,12 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
             log.info('Voice RESUME failed.')
             await self.identify()
         elif op == self.SESSION_DESCRIPTION:
+            self._connection.mode = data['mode']
             await self.load_secret_key(data)
+        elif op == self.HELLO:
+            interval = data['heartbeat_interval'] / 1000.0
+            self._keep_alive = VoiceKeepAliveHandler(ws=self, interval=interval)
+            self._keep_alive.start()
 
     async def initial_connection(self, data):
         state = self._connection
@@ -673,15 +692,23 @@ class DiscordVoiceWebSocket(websockets.client.WebSocketClientProtocol):
         # the port is a little endian unsigned short in the last two bytes
         # yes, this is different endianness from everything else
         state.port = struct.unpack_from('<H', recv, len(recv) - 2)[0]
-
         log.debug('detected ip: %s port: %s', state.ip, state.port)
-        await self.select_protocol(state.ip, state.port)
-        log.info('selected the voice protocol for use')
+
+        # there *should* always be at least one supported mode (xsalsa20_poly1305)
+        modes = [mode for mode in data['modes'] if mode in self._connection.supported_modes]
+        log.debug('received supported encryption modes: %s', ", ".join(modes))
+
+        mode = modes[0]
+        await self.select_protocol(state.ip, state.port, mode)
+        log.info('selected the voice protocol for use (%s)', mode)
+
+        await self.client_connect()
 
     async def load_secret_key(self, data):
         log.info('received secret key for voice connection')
         self._connection.secret_key = data.get('secret_key')
         await self.speak()
+        await self.speak(False)
 
     async def poll_event(self):
         try:
diff --git a/discord/player.py b/discord/player.py
index 74cd073f..8ee3bd13 100644
--- a/discord/player.py
+++ b/discord/player.py
@@ -27,6 +27,7 @@ DEALINGS IN THE SOFTWARE.
 import threading
 import subprocess
 import audioop
+import asyncio
 import logging
 import shlex
 import time
@@ -261,6 +262,7 @@ class AudioPlayer(threading.Thread):
 
         # getattr lookup speed ups
         play_audio = self.client.send_audio_packet
+        self._speak(True)
 
         while not self._end.is_set():
             # are we paused?
@@ -309,14 +311,19 @@ class AudioPlayer(threading.Thread):
     def stop(self):
         self._end.set()
         self._resumed.set()
+        self._speak(False)
 
-    def pause(self):
+    def pause(self, *, update_speaking=True):
         self._resumed.clear()
+        if update_speaking:
+            self._speak(False)
 
-    def resume(self):
+    def resume(self, *, update_speaking=True):
         self.loops = 0
         self._start = time.time()
         self._resumed.set()
+        if update_speaking:
+            self._speak(True)
 
     def is_playing(self):
         return self._resumed.is_set() and not self._end.is_set()
@@ -326,6 +333,12 @@ class AudioPlayer(threading.Thread):
 
     def _set_source(self, source):
         with self._lock:
-            self.pause()
+            self.pause(update_speaking=False)
             self.source = source
-            self.resume()
+            self.resume(update_speaking=False)
+
+    def _speak(self, speaking):
+        try:
+            asyncio.run_coroutine_threadsafe(self.client.ws.speak(speaking), self.client.loop)
+        except Exception as e:
+            log.info("Speaking call in player failed: %s", e)
diff --git a/discord/voice_client.py b/discord/voice_client.py
index 91480872..6bd293c7 100644
--- a/discord/voice_client.py
+++ b/discord/voice_client.py
@@ -102,6 +102,7 @@ class VoiceClient:
         self._connected = threading.Event()
         self._handshake_complete = asyncio.Event(loop=self.loop)
 
+        self.mode = None
         self._connections = 0
         self.sequence = 0
         self.timestamp = 0
@@ -110,6 +111,10 @@ class VoiceClient:
         self.encoder = opus.Encoder()
 
     warn_nacl = not has_nacl
+    supported_modes = (
+        'xsalsa20_poly1305_suffix',
+        'xsalsa20_poly1305',
+    )
 
     @property
     def guild(self):
@@ -288,22 +293,30 @@ class VoiceClient:
 
     def _get_voice_packet(self, data):
         header = bytearray(12)
-        nonce = bytearray(24)
-        box = nacl.secret.SecretBox(bytes(self.secret_key))
 
-        # Formulate header
+        # Formulate rtp header
         header[0] = 0x80
         header[1] = 0x78
         struct.pack_into('>H', header, 2, self.sequence)
         struct.pack_into('>I', header, 4, self.timestamp)
         struct.pack_into('>I', header, 8, self.ssrc)
 
-        # Copy header to nonce's first 12 bytes
+        encrypt_packet = getattr(self, '_encrypt_' + self.mode)
+        return encrypt_packet(header, data)
+
+    def _encrypt_xsalsa20_poly1305(self, header, data):
+        box = nacl.secret.SecretBox(bytes(self.secret_key))
+        nonce = bytearray(24)
         nonce[:12] = header
 
-        # Encrypt and return the data
         return header + box.encrypt(bytes(data), bytes(nonce)).ciphertext
 
+    def _encrypt_xsalsa20_poly1305_suffix(self, header, data):
+        box = nacl.secret.SecretBox(bytes(self.secret_key))
+        nonce = nacl.utils.random(nacl.secret.SecretBox.NONCE_SIZE)
+
+        return header + box.encrypt(bytes(data), nonce).ciphertext + nonce
+
     def play(self, source, *, after=None):
         """Plays an :class:`AudioSource`.