Check current buffer time span instead of number of samples in postpone decoding after expand.

Bug: webrtc:10392
Change-Id: I2ad4d8c7a3f87cab32e2ea097b2e05aa179e0bc0
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/126761
Reviewed-by: Minyue Li <minyue@webrtc.org>
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Commit-Queue: Jakob Ivarsson‎ <jakobi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#27080}

modules/audio_coding/neteq/decision_logic.cc

@@ -116,6 +116,7 @@ Operations DecisionLogic::GetDecision(const SyncBuffer& sync_buffer,
 
   const size_t samples_left =
       sync_buffer.FutureLength() - expand.overlap_length();
+  // TODO(jakobi): Use buffer span instead of num samples.
   const size_t cur_size_samples =
       samples_left + packet_buffer_.NumSamplesInBuffer(decoder_frame_length);
 
@@ -169,11 +170,13 @@ Operations DecisionLogic::GetDecision(const SyncBuffer& sync_buffer,
   // if the mute factor is low enough (otherwise the expansion was short enough
   // to not be noticable).
   // Note that the MuteFactor is in Q14, so a value of 16384 corresponds to 1.
+  size_t current_span =
+      samples_left + packet_buffer_.GetSpanSamples(decoder_frame_length);
   if ((prev_mode == kModeExpand || prev_mode == kModeCodecPlc) &&
       expand.MuteFactor(0) < 16384 / 2 &&
-      cur_size_samples < static_cast<size_t>(
-              delay_manager_->TargetLevel() * packet_length_samples_ *
-              kPostponeDecodingLevel / 100) >> 8 &&
+      current_span < static_cast<size_t>(delay_manager_->TargetLevel() *
+                                         packet_length_samples_ *
+                                         kPostponeDecodingLevel / 100)>> 8 &&
       !packet_buffer_.ContainsDtxOrCngPacket(decoder_database_)) {
     return kExpand;
   }

modules/audio_coding/neteq/packet_buffer.cc

@@ -287,6 +287,20 @@ size_t PacketBuffer::NumSamplesInBuffer(size_t last_decoded_length) const {
   return num_samples;
 }
 
+size_t PacketBuffer::GetSpanSamples(size_t last_decoded_length) const {
+  if (buffer_.size() == 0) {
+    return 0;
+  }
+
+  size_t span = buffer_.back().timestamp - buffer_.front().timestamp;
+  if (buffer_.back().frame && buffer_.back().frame->Duration() > 0) {
+    span += buffer_.back().frame->Duration();
+  } else {
+    span += last_decoded_length;
+  }
+  return span;
+}
+
 bool PacketBuffer::ContainsDtxOrCngPacket(
     const DecoderDatabase* decoder_database) const {
   RTC_DCHECK(decoder_database);

modules/audio_coding/neteq/packet_buffer.h

@@ -121,6 +121,10 @@ class PacketBuffer {
   // duplicate and redundant packets.
   virtual size_t NumSamplesInBuffer(size_t last_decoded_length) const;
 
+  // Returns the total duration in samples that the packets in the buffer spans
+  // across.
+  virtual size_t GetSpanSamples(size_t last_decoded_length) const;
+
   // Returns true if the packet buffer contains any DTX or CNG packets.
   virtual bool ContainsDtxOrCngPacket(
       const DecoderDatabase* decoder_database) const;