Fix UDP audio frame timestamp handling

gumble/gumble/audio.go

@@ -86,8 +86,8 @@ type AudioPacket struct {
 	Sender *User
 	Target *VoiceTarget
 
-	// Sequence is the UDP audio packet sequence number, used by the
-	// jitter buffer to reorder packets.
+	// Sequence is the UDP audio frame timestamp, used by the jitter buffer to
+	// reorder packets.
 	Sequence int64
 
 	AudioBuffer

gumble/gumble/client.go

@@ -232,11 +232,15 @@ func (c *Client) AudioOutgoing() chan<- AudioBuffer {
 	ch := make(chan AudioBuffer)
 	go func() {
 		var seq int64
+		frameStep := int64(c.Config.AudioFrameSize() / AudioDefaultFrameSize)
+		if frameStep < 1 {
+			frameStep = 1
+		}
 		previous := <-ch
 		for p := range ch {
 			previous.writeAudio(c, seq, false)
 			previous = p
-			seq = (seq + 1) % math.MaxInt32
+			seq = (seq + frameStep) % math.MaxInt32
 		}
 		if previous != nil {
 			previous.writeAudio(c, seq, true)

gumble/gumble/integration_audio_test.go

@@ -106,10 +106,11 @@ func testLocalMumbleAudioRoundTrip(t *testing.T, disableUDP bool, interval time.
 	// Discard the first few decoder warm-up frames before measuring pitch.
 	var packet *gumble.AudioPacket
 	var previousSequence int64
+	frameStep := int64(sendConfig.AudioFrameSize() / gumble.AudioDefaultFrameSize)
 	for i := 0; i < 3; i++ {
 		select {
 		case packet = <-listener.packets:
-			if i > 0 && packet.Sequence != previousSequence+1 {
+			if i > 0 && packet.Sequence != previousSequence+frameStep {
 				t.Fatalf("choppy relay: sequence gap %d -> %d", previousSequence, packet.Sequence)
 			}
 			previousSequence = packet.Sequence
@@ -119,7 +120,7 @@ func testLocalMumbleAudioRoundTrip(t *testing.T, disableUDP bool, interval time.
 	}
 	select {
 	case packet = <-listener.packets:
-		if packet.Sequence != previousSequence+1 {
+		if packet.Sequence != previousSequence+frameStep {
 			t.Fatalf("choppy relay: sequence gap %d -> %d", previousSequence, packet.Sequence)
 		}
 		frequency, purity, peak := audioQuality(packet.AudioBuffer)

gumble/gumble/udp15.go

@@ -733,6 +733,8 @@ func (c *Client) dispatchOpus15(pktNum uint64, session uint32, frameNum int64, o
 
 	if terminator && len(opusData) == 0 {
 		decoder.Reset()
+		user.audioSequenceValid = false
+		user.audioFrameStep = 0
 		log.Info("UDP15 #%d: terminator for %s, decoder reset", pktNum, user.Name)
 		return
 	}
@@ -791,14 +793,23 @@ func (c *Client) handleLegacyUDPVoice(pktNum uint64, data []byte) {
 
 // decodeAndDispatch decodes an Opus frame and dispatches PCM to audio listeners.
 func (c *Client) decodeAndDispatch(pktNum uint64, user *User, decoder AudioDecoder, frameNum int64, opusData []byte, context uint32, position *[3]float32, volumeAdjustment float32) {
-	// Detect sequence gaps.
+	// Frame numbers are timestamps in 10 ms units, not packet counters. For
+	// example, a standard 20 ms Opus packet advances its frame number by two.
+	// Only generate PLC for complete missing packets; treating every timestamp
+	// unit as a packet doubles playout and eventually exhausts OpenAL buffers.
 	if user.audioSequenceValid {
 		gap := frameNum - user.audioSequence
-		if gap > 1 && gap < 100 {
-			log.Info("UDP15 #%d: seq gap for %s: %d -> %d (loss=%d), generating PLC",
-				pktNum, user.Name, user.audioSequence, frameNum, gap-1)
-			for i := int64(1); i < gap; i++ {
-				c.dispatchPLC15(user, decoder)
+		frameStep := user.audioFrameStep
+		if frameStep < 1 {
+			frameStep = 1
+		}
+		if gap > frameStep && gap < 100 {
+			if missing := missingAudioPackets(gap, frameStep); missing > 0 {
+				log.Info("UDP15 #%d: audio gap for %s: %d -> %d (loss=%d), generating PLC",
+					pktNum, user.Name, user.audioSequence, frameNum, missing)
+				for i := int64(1); i <= missing; i++ {
+					c.dispatchPLC15(user, decoder, user.audioSequence+i*frameStep)
+				}
 			}
 		} else if gap < 0 && gap > -100 {
 			log.Info("UDP15 #%d: seq reorder for %s: %d -> %d, resetting decoder",
@@ -809,8 +820,6 @@ func (c *Client) decodeAndDispatch(pktNum uint64, user *User, decoder AudioDecod
 			return
 		}
 	}
-	user.audioSequence = frameNum
-	user.audioSequenceValid = true
 
 	pcm, err := decoder.Decode(opusData, AudioMaximumFrameSize)
 	if err != nil {
@@ -820,6 +829,9 @@ func (c *Client) decodeAndDispatch(pktNum uint64, user *User, decoder AudioDecod
 	}
 
 	log.Debug("UDP15 #%d: Opus OK for %s, pcm_samples=%d", pktNum, user.Name, len(pcm))
+	user.audioSequence = frameNum
+	user.audioSequenceValid = true
+	user.audioFrameStep = audioFrameStep(len(pcm))
 
 	event := AudioPacket{
 		Client:           c,
@@ -836,20 +848,38 @@ func (c *Client) decodeAndDispatch(pktNum uint64, user *User, decoder AudioDecod
 	c.dispatchAudio(user, &event)
 }
 
+// missingAudioPackets returns the number of whole packets absent from a
+// timestamp gap. A non-integral gap cannot reliably identify a missing packet.
+func missingAudioPackets(gap, frameStep int64) int64 {
+	if frameStep < 1 || gap <= frameStep || gap%frameStep != 0 {
+		return 0
+	}
+	return gap/frameStep - 1
+}
+
+// audioFrameStep converts interleaved stereo PCM length to Mumble's 10 ms
+// frame-number units.
+func audioFrameStep(samples int) int64 {
+	frames := samples / AudioChannels
+	step := int64(frames / AudioDefaultFrameSize)
+	if step < 1 {
+		return 1
+	}
+	return step
+}
+
 // dispatchPLC15 generates a Packet Loss Concealment frame for 1.5 UDP.
-func (c *Client) dispatchPLC15(user *User, decoder AudioDecoder) {
+func (c *Client) dispatchPLC15(user *User, decoder AudioDecoder, sequence int64) {
 	pcm, err := decoder.Decode(nil, AudioMaximumFrameSize)
 	if err != nil {
 		decoder.Reset()
 		return
 	}
-	seq := user.audioSequence + 1
-	user.audioSequence = seq
 	event := AudioPacket{
 		Client:      c,
 		Sender:      user,
 		Target:      &VoiceTarget{ID: 0},
-		Sequence:    seq,
+		Sequence:    sequence,
 		AudioBuffer: AudioBuffer(pcm),
 	}
 	c.dispatchAudio(user, &event)

gumble/gumble/udp15_test.go

@@ -223,6 +223,31 @@ func TestUDPAudioProtobufReferenceVector(t *testing.T) {
 	}
 }
 
+func TestAudioFrameTimestampGaps(t *testing.T) {
+	tests := []struct {
+		name      string
+		gap, step int64
+		want      int64
+	}{
+		{"consecutive 10 ms packets", 1, 1, 0},
+		{"consecutive 20 ms packets", 2, 2, 0},
+		{"one missing 20 ms packet", 4, 2, 1},
+		{"two missing 20 ms packets", 6, 2, 2},
+		{"non-integral timestamp gap", 3, 2, 0},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := missingAudioPackets(tt.gap, tt.step); got != tt.want {
+				t.Fatalf("missingAudioPackets(%d, %d) = %d, want %d", tt.gap, tt.step, got, tt.want)
+			}
+		})
+	}
+
+	if got := audioFrameStep(1920); got != 2 {
+		t.Fatalf("audioFrameStep(1920) = %d, want 2", got)
+	}
+}
+
 func TestUDPAudioProtobufIncomingFields(t *testing.T) {
 	var packet bytes.Buffer
 	writeVarint := func(v uint64) {

gumble/gumble/user.go

@@ -52,17 +52,18 @@ type User struct {
 	client  *Client
 	decoder AudioDecoder
 
-	// audioSequence tracks the last UDP audio packet sequence number for
-	// this user, used to detect packet loss and reset the Opus decoder.
+	// audioSequence tracks the last UDP audio frame timestamp for this user,
+	// used to detect packet loss and reset the Opus decoder.
 	audioSequence      int64
 	audioSequenceValid bool
+	audioFrameStep     int64
 
 	// audioMu protects audio-related fields accessed from both the
 	// audio processing goroutine (OnAudioStream) and the UI goroutine.
-	audioMu     sync.Mutex
-	audioSource *openal.Source
-	boost       uint16
-	volume      float32
+	audioMu      sync.Mutex
+	audioSource  *openal.Source
+	boost        uint16
+	volume       float32
 	locallyMuted bool
 }