Make FakeDecodeFromFile handle codec-internal CNG

This implementation interprets payloads of size 1 as codec-internal SID
frames, marking the start of a CNG period. Changes were made to other
parts of the test payload chain, since it had to make use of the virtual
payload size in the case of header-only RTP files.

BUG=webrtc:2692

Review-Url: https://codereview.webrtc.org/2275903002
Cr-Commit-Position: refs/heads/master@{#13901}

webrtc/modules/audio_coding/neteq/tools/fake_decode_from_file.cc

@@ -19,14 +19,37 @@ namespace test {
 
 int FakeDecodeFromFile::DecodeInternal(const uint8_t* encoded,
                                        size_t encoded_len,
-                                       int /*sample_rate_hz*/,
+                                       int sample_rate_hz,
                                        int16_t* decoded,
                                        SpeechType* speech_type) {
-  RTC_CHECK_GE(encoded_len, 8u);
+  if (encoded_len == 0) {
+    // Decoder is asked to produce codec-internal comfort noise.
+    RTC_DCHECK(!encoded);  // NetEq always sends nullptr in this case.
+    RTC_DCHECK(cng_mode_);
+    RTC_DCHECK_GT(last_decoded_length_, 0u);
+    std::fill_n(decoded, last_decoded_length_, 0);
+    *speech_type = kComfortNoise;
+    return last_decoded_length_;
+  }
+
+  RTC_CHECK_GE(encoded_len, 12u);
   uint32_t timestamp_to_decode =
       ByteReader<uint32_t>::ReadLittleEndian(encoded);
   uint32_t samples_to_decode =
       ByteReader<uint32_t>::ReadLittleEndian(&encoded[4]);
+  if (samples_to_decode == 0) {
+    // Number of samples in packet is unknown.
+    if (last_decoded_length_ > 0) {
+      // Use length of last decoded packet, but since this is the total for all
+      // channels, we have to divide by 2 in the stereo case.
+      samples_to_decode = rtc::CheckedDivExact(
+          last_decoded_length_, static_cast<size_t>(stereo_ ? 2uL : 1uL));
+    } else {
+      // This is the first packet to decode, and we do not know the length of
+      // it. Set it to 10 ms.
+      samples_to_decode = rtc::CheckedDivExact(sample_rate_hz, 100);
+    }
+  }
 
   if (next_timestamp_from_input_ &&
       timestamp_to_decode != *next_timestamp_from_input_) {
@@ -36,10 +59,23 @@ int FakeDecodeFromFile::DecodeInternal(const uint8_t* encoded,
     RTC_CHECK(input_->Seek(jump));
   }
 
-  RTC_CHECK(input_->Read(static_cast<size_t>(samples_to_decode), decoded));
   next_timestamp_from_input_ =
       rtc::Optional<uint32_t>(timestamp_to_decode + samples_to_decode);
 
+  uint32_t original_payload_size_bytes =
+      ByteReader<uint32_t>::ReadLittleEndian(&encoded[8]);
+  if (original_payload_size_bytes == 1) {
+    // This is a comfort noise payload.
+    RTC_DCHECK_GT(last_decoded_length_, 0u);
+    std::fill_n(decoded, last_decoded_length_, 0);
+    *speech_type = kComfortNoise;
+    cng_mode_ = true;
+    return last_decoded_length_;
+  }
+
+  cng_mode_ = false;
+  RTC_CHECK(input_->Read(static_cast<size_t>(samples_to_decode), decoded));
+
   if (stereo_) {
     InputAudioFile::DuplicateInterleaved(decoded, samples_to_decode, 2,
                                          decoded);
@@ -47,16 +83,19 @@ int FakeDecodeFromFile::DecodeInternal(const uint8_t* encoded,
   }
 
   *speech_type = kSpeech;
-  return samples_to_decode;
+  return last_decoded_length_ = samples_to_decode;
 }
 
 void FakeDecodeFromFile::PrepareEncoded(uint32_t timestamp,
                                         size_t samples,
+                                        size_t original_payload_size_bytes,
                                         rtc::ArrayView<uint8_t> encoded) {
-  RTC_CHECK_GE(encoded.size(), 8u);
+  RTC_CHECK_GE(encoded.size(), 12u);
   ByteWriter<uint32_t>::WriteLittleEndian(&encoded[0], timestamp);
   ByteWriter<uint32_t>::WriteLittleEndian(&encoded[4],
                                           rtc::checked_cast<uint32_t>(samples));
+  ByteWriter<uint32_t>::WriteLittleEndian(
+      &encoded[8], rtc::checked_cast<uint32_t>(original_payload_size_bytes));
 }
 
 }  // namespace test

webrtc/modules/audio_coding/neteq/tools/fake_decode_from_file.h

@@ -50,11 +50,13 @@ class FakeDecodeFromFile : public AudioDecoder {
                      int16_t* decoded,
                      SpeechType* speech_type) override;
 
-  // Helper method. Writes |timestamp| and |samples| to |encoded| in a format
+  // Helper method. Writes |timestamp|, |samples| and
-  // that the FakeDecpdeFromFile decoder will understand. |encoded| must be at
+  // |original_payload_size_bytes| to |encoded| in a format that the
-  // least 8 bytes long.
+  // FakeDecodeFromFile decoder will understand. |encoded| must be at least 12
+  // bytes long.
   static void PrepareEncoded(uint32_t timestamp,
                              size_t samples,
+                             size_t original_payload_size_bytes,
                              rtc::ArrayView<uint8_t> encoded);
 
  private:
@@ -62,6 +64,8 @@ class FakeDecodeFromFile : public AudioDecoder {
   rtc::Optional<uint32_t> next_timestamp_from_input_;
   const int sample_rate_hz_;
   const bool stereo_;
+  size_t last_decoded_length_ = 0;
+  bool cng_mode_ = false;
 };
 
 }  // namespace test

webrtc/modules/audio_coding/neteq/tools/neteq_packet_source_input.cc

@@ -43,8 +43,16 @@ std::unique_ptr<NetEqInput::PacketData> NetEqPacketSourceInput::PopPacket() {
   }
   std::unique_ptr<PacketData> packet_data(new PacketData);
   packet_->ConvertHeader(&packet_data->header);
+  if (packet_->payload_length_bytes() == 0 &&
+      packet_->virtual_payload_length_bytes() > 0) {
+    // This is a header-only "dummy" packet. Set the payload to all zeros, with
+    // length according to the virtual length.
+    packet_data->payload.SetSize(packet_->virtual_payload_length_bytes());
+    std::fill_n(packet_data->payload.data(), packet_data->payload.size(), 0);
+  } else {
     packet_data->payload.SetData(packet_->payload(),
                                  packet_->payload_length_bytes());
+  }
   packet_data->time_ms = packet_->time_ms();
 
   LoadNextPacket();

webrtc/modules/audio_coding/neteq/tools/neteq_replacement_input.cc

@@ -84,11 +84,17 @@ void NetEqReplacementInput::ReplacePacket() {
 
   rtc::Optional<RTPHeader> next_hdr = source_->NextHeader();
   RTC_DCHECK(next_hdr);
-  uint8_t payload[8];
+  uint8_t payload[12];
   uint32_t input_frame_size_timestamps =
       next_hdr->timestamp - packet_->header.header.timestamp;
+  if (next_hdr->sequenceNumber != packet_->header.header.sequenceNumber + 1) {
+    // Gap in packet sequence. Cannot estimate payload length based on timestamp
+    // difference.
+    input_frame_size_timestamps = 0;
+  }
   FakeDecodeFromFile::PrepareEncoded(packet_->header.header.timestamp,
-                                     input_frame_size_timestamps, payload);
+                                     input_frame_size_timestamps,
+                                     packet_->payload.size(), payload);
   packet_->payload.SetData(payload);
   packet_->header.header.payloadType = replacement_payload_type_;
   return;

webrtc/modules/audio_coding/neteq/tools/packet.cc

@@ -14,6 +14,7 @@
 
 #include <memory>
 
+#include "webrtc/base/checks.h"
 #include "webrtc/modules/include/module_common_types.h"
 #include "webrtc/modules/rtp_rtcp/include/rtp_header_parser.h"
 
@@ -142,6 +143,7 @@ bool Packet::ParseHeader(const RtpHeaderParser& parser) {
   payload_ = &payload_memory_[header_.headerLength];
   assert(packet_length_bytes_ >= header_.headerLength);
   payload_length_bytes_ = packet_length_bytes_ - header_.headerLength;
+  RTC_CHECK_GE(virtual_packet_length_bytes_, packet_length_bytes_);
   assert(virtual_packet_length_bytes_ >= header_.headerLength);
   virtual_payload_length_bytes_ =
       virtual_packet_length_bytes_ - header_.headerLength;