private bool ProcessPacket(ReadOnlySpan <byte> data, ref Memory <byte> opus, ref Memory <byte> pcm, IList <ReadOnlyMemory <byte> > pcmPackets, out AudioSender voiceSender, out AudioFormat outputFormat)
{
voiceSender = null;
outputFormat = default;
if (!this.Rtp.IsRtpHeader(data))
{
return(false);
}
this.Rtp.DecodeHeader(data, out var sequence, out var timestamp, out var ssrc, out var hasExtension);
if (!this.TransmittingSSRCs.TryGetValue(ssrc, out var vtx))
{
var decoder = Opus.CreateDecoder();
vtx = new AudioSender(ssrc, decoder)
{
// user isn't present as we haven't received a speaking event yet.
User = null
};
}
voiceSender = vtx;
if (sequence <= vtx.LastSequence) // out-of-order packet; discard
{
return(false);
}
var gap = vtx.LastSequence != 0 ? sequence - 1 - vtx.LastSequence : 0;
if (gap >= 5)
{
this.Discord.Logger.LogWarning(VoiceNextEvents.VoiceReceiveFailure, "5 or more voice packets were dropped when receiving");
}
Span <byte> nonce = stackalloc byte[Sodium.NonceSize];
this.Sodium.GetNonce(data, nonce, this.SelectedEncryptionMode);
this.Rtp.GetDataFromPacket(data, out var encryptedOpus, this.SelectedEncryptionMode);
var opusSize = Sodium.CalculateSourceSize(encryptedOpus);
opus = opus.Slice(0, opusSize);
var opusSpan = opus.Span;
try
{
this.Sodium.Decrypt(encryptedOpus, opusSpan, nonce);
// Strip extensions, if any
if (hasExtension)
{
// RFC 5285, 4.2 One-Byte header
// http://www.rfcreader.com/#rfc5285_line186
if (opusSpan[0] == 0xBE && opusSpan[1] == 0xDE)
{
var headerLen = opusSpan[2] << 8 | opusSpan[3];
var i = 4;
for (; i < headerLen + 4; i++)
{
var @byte = opusSpan[i];
// ID is currently unused since we skip it anyway
//var id = (byte)(@byte >> 4);
var length = (byte)(@byte & 0x0F) + 1;
i += length;
}
// Strip extension padding too
while (opusSpan[i] == 0)
{
i++;
}
opusSpan = opusSpan.Slice(i);
}
// TODO: consider implementing RFC 5285, 4.3. Two-Byte Header
}
if (opusSpan[0] == 0x90)
{
// I'm not 100% sure what this header is/does, however removing the data causes no
// real issues, and has the added benefit of removing a lot of noise.
opusSpan = opusSpan.Slice(2);
}
if (gap == 1)
{
var lastSampleCount = this.Opus.GetLastPacketSampleCount(vtx.Decoder);
var fecpcm = new byte[this.AudioFormat.SampleCountToSampleSize(lastSampleCount)];
var fecpcmMem = fecpcm.AsSpan();
this.Opus.Decode(vtx.Decoder, opusSpan, ref fecpcmMem, true, out _);
pcmPackets.Add(fecpcm.AsMemory(0, fecpcmMem.Length));
}
else if (gap > 1)
{
var lastSampleCount = this.Opus.GetLastPacketSampleCount(vtx.Decoder);
for (var i = 0; i < gap; i++)
{
var fecpcm = new byte[this.AudioFormat.SampleCountToSampleSize(lastSampleCount)];
var fecpcmMem = fecpcm.AsSpan();
this.Opus.ProcessPacketLoss(vtx.Decoder, lastSampleCount, ref fecpcmMem);
pcmPackets.Add(fecpcm.AsMemory(0, fecpcmMem.Length));
}
}
var pcmSpan = pcm.Span;
this.Opus.Decode(vtx.Decoder, opusSpan, ref pcmSpan, false, out outputFormat);
pcm = pcm.Slice(0, pcmSpan.Length);
}
finally
{
vtx.LastSequence = sequence;
}
return(true);
}
private bool ProcessPacket(ReadOnlySpan <byte> data, ref Memory <byte> opus, ref Memory <byte> pcm, IList <ReadOnlyMemory <byte> > pcmPackets, out AudioSender voiceSender, out AudioFormat outputFormat)
{
voiceSender = null;
outputFormat = default;
if (!this.Rtp.IsRtpHeader(data))
{
return(false);
}
this.Rtp.DecodeHeader(data, out var sequence, out var timestamp, out var ssrc, out var hasExtension);
var vtx = this.TransmittingSSRCs[ssrc];
voiceSender = vtx;
if (sequence <= vtx.LastSequence) // out-of-order packet; discard
{
return(false);
}
var gap = vtx.LastSequence != 0 ? sequence - 1 - vtx.LastSequence : 0;
if (gap >= 5)
{
this.Discord.DebugLogger.LogMessage(LogLevel.Warning, "VNext RX", "5 or more voice packets were dropped when receiving", DateTime.Now);
}
Span <byte> nonce = stackalloc byte[Sodium.NonceSize];
this.Sodium.GetNonce(data, nonce, this.SelectedEncryptionMode);
this.Rtp.GetDataFromPacket(data, out var encryptedOpus, this.SelectedEncryptionMode);
var opusSize = Sodium.CalculateSourceSize(encryptedOpus);
opus = opus.Slice(0, opusSize);
var opusSpan = opus.Span;
try
{
this.Sodium.Decrypt(encryptedOpus, opusSpan, nonce);
// Strip extensions, if any
if (hasExtension)
{
// RFC 5285, 4.2 One-Byte header
// http://www.rfcreader.com/#rfc5285_line186
if (opusSpan[0] == 0xBE && opusSpan[1] == 0xDE)
{
var headerLen = opusSpan[2] << 8 | opusSpan[3];
var i = 4;
for (; i < headerLen + 4; i++)
{
var @byte = opusSpan[i];
// ID is currently unused since we skip it anyway
//var id = (byte)(@byte >> 4);
var length = (byte)(@byte & 0x0F) + 1;
i += length;
}
// Strip extension padding too
while (opusSpan[i] == 0)
{
i++;
}
opusSpan = opusSpan.Slice(i);
}
// TODO: consider implementing RFC 5285, 4.3. Two-Byte Header
}
if (gap == 1)
{
var lastSampleCount = this.Opus.GetLastPacketSampleCount(vtx.Decoder);
var fecpcm = new byte[this.AudioFormat.SampleCountToSampleSize(lastSampleCount)];
var fecpcmMem = fecpcm.AsSpan();
this.Opus.Decode(vtx.Decoder, opusSpan, ref fecpcmMem, true, out _);
pcmPackets.Add(fecpcm.AsMemory(0, fecpcmMem.Length));
}
else if (gap > 1)
{
var lastSampleCount = this.Opus.GetLastPacketSampleCount(vtx.Decoder);
for (var i = 0; i < gap; i++)
{
var fecpcm = new byte[this.AudioFormat.SampleCountToSampleSize(lastSampleCount)];
var fecpcmMem = fecpcm.AsSpan();
this.Opus.ProcessPacketLoss(vtx.Decoder, lastSampleCount, ref fecpcmMem);
pcmPackets.Add(fecpcm.AsMemory(0, fecpcmMem.Length));
}
}
var pcmSpan = pcm.Span;
this.Opus.Decode(vtx.Decoder, opusSpan, ref pcmSpan, false, out outputFormat);
pcm = pcm.Slice(0, pcmSpan.Length);
}
finally
{
vtx.LastSequence = sequence;
}
return(true);
}
