changed Audio.cpp to not overflow _audioOutput buffer, and 2 other things

added _consecutiveNotMixedCount to prevent premature injector stream
deletion;  made silent-frame drop only occur in dynamic jitter buffer
mode

assignment-client/src/audio/AudioMixerClientData.cpp

@@ -138,11 +138,14 @@ void AudioMixerClientData::pushBuffersAfterFrameSend() {
         // this was a used buffer, push the output pointer forwards
         PositionalAudioRingBuffer* audioBuffer = *i;
 
+        const int INJECTOR_CONSECUTIVE_NOT_MIXED_THRESHOLD = 100;
+
         if (audioBuffer->willBeAddedToMix()) {
             audioBuffer->shiftReadPosition(audioBuffer->getSamplesPerFrame());
             audioBuffer->setWillBeAddedToMix(false);
         } else if (audioBuffer->getType() == PositionalAudioRingBuffer::Injector
-                   && audioBuffer->hasStarted() && audioBuffer->isStarved()) {
+                   && audioBuffer->hasStarted() && audioBuffer->isStarved()
+                   && audioBuffer->getConsecutiveNotMixedCount() > INJECTOR_CONSECUTIVE_NOT_MIXED_THRESHOLD) {
             // this is an empty audio buffer that has starved, safe to delete
             // also delete its sequence number stats
             QUuid streamIdentifier = ((InjectedAudioRingBuffer*)audioBuffer)->getStreamIdentifier();

interface/src/Audio.cpp

@@ -67,7 +67,7 @@ Audio::Audio(int16_t initialJitterBufferSamples, QObject* parent) :
     _proceduralAudioOutput(NULL),
     _proceduralOutputDevice(NULL),
     _inputRingBuffer(0),
-    _ringBuffer(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO),
+    _ringBuffer(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO, false, 100),
     _isStereoInput(false),
     _averagedLatency(0.0),
     _measuredJitter(0),
@@ -869,14 +869,11 @@ void Audio::processReceivedAudio(const QByteArray& audioByteArray) {
         _numFramesDisplayStarve = 10;
     }
     
-    // if there is anything in the ring buffer, decide what to do
-    if (_ringBuffer.samplesAvailable() > 0) {
-        
-        int numNetworkOutputSamples = _ringBuffer.samplesAvailable();
-        int numDeviceOutputSamples = numNetworkOutputSamples / networkOutputToOutputRatio;
-        
-        QByteArray outputBuffer;
-        outputBuffer.resize(numDeviceOutputSamples * sizeof(int16_t));
+    int numSamplesAudioOutputRoomFor = _audioOutput->bytesFree() / sizeof(int16_t);
+    int numNetworkOutputSamples = std::min(_ringBuffer.samplesAvailable(), (int)(numSamplesAudioOutputRoomFor * networkOutputToOutputRatio));
+
+    // if there is data in the ring buffer and room in the audio output, decide what to do
+    if (numNetworkOutputSamples > 0) {
         
         int numSamplesNeededToStartPlayback = std::min(NETWORK_BUFFER_LENGTH_SAMPLES_STEREO + (_jitterBufferSamples * 2),
             _ringBuffer.getSampleCapacity());
@@ -885,6 +882,11 @@ void Audio::processReceivedAudio(const QByteArray& audioByteArray) {
             //  We are still waiting for enough samples to begin playback
             // qDebug() << numNetworkOutputSamples << " samples so far, waiting for " << numSamplesNeededToStartPlayback;
         } else {
+            int numDeviceOutputSamples = numNetworkOutputSamples / networkOutputToOutputRatio;
+
+            QByteArray outputBuffer;
+            outputBuffer.resize(numDeviceOutputSamples * sizeof(int16_t));
+
             //  We are either already playing back, or we have enough audio to start playing back.
             //qDebug() << "pushing " << numNetworkOutputSamples;
             _ringBuffer.setIsStarved(false);

libraries/audio/src/PositionalAudioRingBuffer.cpp

@@ -99,7 +99,8 @@ PositionalAudioRingBuffer::PositionalAudioRingBuffer(PositionalAudioRingBuffer::
     _listenerUnattenuatedZone(NULL),
     _desiredJitterBufferFrames(1),
     _currentJitterBufferFrames(-1),
-    _dynamicJitterBuffers(dynamicJitterBuffers)
+    _dynamicJitterBuffers(dynamicJitterBuffers),
+    _consecutiveNotMixedCount(0)
 {
 }
 
@@ -129,7 +130,7 @@ int PositionalAudioRingBuffer::parseData(const QByteArray& packet) {
         numSilentSamples = getSamplesPerFrame();
         
         if (numSilentSamples > 0) {
-            if (_currentJitterBufferFrames > _desiredJitterBufferFrames) {
+            if (_dynamicJitterBuffers && _currentJitterBufferFrames > _desiredJitterBufferFrames) {
                 // our current jitter buffer size exceeds its desired value, so ignore some silent
                 // frames to get that size as close to desired as possible
                 int samplesPerFrame = getSamplesPerFrame();
@@ -206,11 +207,12 @@ bool PositionalAudioRingBuffer::shouldBeAddedToMix() {
     if (!isNotStarvedOrHasMinimumSamples(samplesPerFrame + desiredJitterBufferSamples)) {
         // if the buffer was starved, allow it to accrue at least the desired number of
         // jitter buffer frames before we start taking frames from it for mixing
-        
+
         if (_shouldOutputStarveDebug) {
             _shouldOutputStarveDebug = false;
         }
 
+        _consecutiveNotMixedCount++;
         return  false;
     } else if (samplesAvailable() < samplesPerFrame) { 
         // if the buffer doesn't have a full frame of samples to take for mixing, it is starved
@@ -222,6 +224,7 @@ bool PositionalAudioRingBuffer::shouldBeAddedToMix() {
         // reset our _shouldOutputStarveDebug to true so the next is printed
         _shouldOutputStarveDebug = true;
 
+        _consecutiveNotMixedCount++;
         return false;
     }
     
@@ -231,6 +234,7 @@ bool PositionalAudioRingBuffer::shouldBeAddedToMix() {
         // minus one (since a frame will be read immediately after this) is the length of the jitter buffer
         _currentJitterBufferFrames = samplesAvailable() / samplesPerFrame - 1;
         _isStarved = false;
+        _consecutiveNotMixedCount = 0;
     }
 
     // since we've read data from ring buffer at least once - we've started

libraries/audio/src/PositionalAudioRingBuffer.h

@@ -83,6 +83,8 @@ public:
     int getDesiredJitterBufferFrames() const { return _desiredJitterBufferFrames; }
     int getCurrentJitterBufferFrames() const { return _currentJitterBufferFrames; }
 
+    int getConsecutiveNotMixedCount() const { return _consecutiveNotMixedCount; }
+
 protected:
     // disallow copying of PositionalAudioRingBuffer objects
     PositionalAudioRingBuffer(const PositionalAudioRingBuffer&);
@@ -107,9 +109,7 @@ protected:
     bool _dynamicJitterBuffers;
 
     // extra stats
-    int _starveCount;
-    int _silentFramesDropped;
-
+    int _consecutiveNotMixedCount;
 };
 
 #endif // hifi_PositionalAudioRingBuffer_h