mirror of
https://github.com/oxen-io/session-android.git
synced 2024-11-28 20:45:17 +00:00
d83a3d71bc
Merge in RedPhone // FREEBIE
499 lines
19 KiB
C++
499 lines
19 KiB
C++
/*
|
|
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
|
|
*
|
|
* Use of this source code is governed by a BSD-style license
|
|
* that can be found in the LICENSE file in the root of the source
|
|
* tree. An additional intellectual property rights grant can be found
|
|
* in the file PATENTS. All contributing project authors may
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
*/
|
|
|
|
#include "webrtc/modules/audio_coding/neteq/interface/neteq.h"
|
|
#include "webrtc/modules/audio_coding/neteq/neteq_impl.h"
|
|
|
|
#include "gmock/gmock.h"
|
|
#include "gtest/gtest.h"
|
|
#include "webrtc/modules/audio_coding/neteq/accelerate.h"
|
|
#include "webrtc/modules/audio_coding/neteq/expand.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_audio_decoder.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_buffer_level_filter.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_decoder_database.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_delay_manager.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_delay_peak_detector.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_dtmf_buffer.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_dtmf_tone_generator.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_packet_buffer.h"
|
|
#include "webrtc/modules/audio_coding/neteq/mock/mock_payload_splitter.h"
|
|
#include "webrtc/modules/audio_coding/neteq/preemptive_expand.h"
|
|
#include "webrtc/modules/audio_coding/neteq/sync_buffer.h"
|
|
#include "webrtc/modules/audio_coding/neteq/timestamp_scaler.h"
|
|
|
|
using ::testing::Return;
|
|
using ::testing::ReturnNull;
|
|
using ::testing::_;
|
|
using ::testing::SetArgPointee;
|
|
using ::testing::InSequence;
|
|
using ::testing::Invoke;
|
|
using ::testing::WithArg;
|
|
|
|
namespace webrtc {
|
|
|
|
// This function is called when inserting a packet list into the mock packet
|
|
// buffer. The purpose is to delete all inserted packets properly, to avoid
|
|
// memory leaks in the test.
|
|
int DeletePacketsAndReturnOk(PacketList* packet_list) {
|
|
PacketBuffer::DeleteAllPackets(packet_list);
|
|
return PacketBuffer::kOK;
|
|
}
|
|
|
|
class NetEqImplTest : public ::testing::Test {
|
|
protected:
|
|
NetEqImplTest()
|
|
: neteq_(NULL),
|
|
config_(),
|
|
mock_buffer_level_filter_(NULL),
|
|
buffer_level_filter_(NULL),
|
|
use_mock_buffer_level_filter_(true),
|
|
mock_decoder_database_(NULL),
|
|
decoder_database_(NULL),
|
|
use_mock_decoder_database_(true),
|
|
mock_delay_peak_detector_(NULL),
|
|
delay_peak_detector_(NULL),
|
|
use_mock_delay_peak_detector_(true),
|
|
mock_delay_manager_(NULL),
|
|
delay_manager_(NULL),
|
|
use_mock_delay_manager_(true),
|
|
mock_dtmf_buffer_(NULL),
|
|
dtmf_buffer_(NULL),
|
|
use_mock_dtmf_buffer_(true),
|
|
mock_dtmf_tone_generator_(NULL),
|
|
dtmf_tone_generator_(NULL),
|
|
use_mock_dtmf_tone_generator_(true),
|
|
mock_packet_buffer_(NULL),
|
|
packet_buffer_(NULL),
|
|
use_mock_packet_buffer_(true),
|
|
mock_payload_splitter_(NULL),
|
|
payload_splitter_(NULL),
|
|
use_mock_payload_splitter_(true),
|
|
timestamp_scaler_(NULL) {
|
|
config_.sample_rate_hz = 8000;
|
|
}
|
|
|
|
void CreateInstance() {
|
|
if (use_mock_buffer_level_filter_) {
|
|
mock_buffer_level_filter_ = new MockBufferLevelFilter;
|
|
buffer_level_filter_ = mock_buffer_level_filter_;
|
|
} else {
|
|
buffer_level_filter_ = new BufferLevelFilter;
|
|
}
|
|
if (use_mock_decoder_database_) {
|
|
mock_decoder_database_ = new MockDecoderDatabase;
|
|
EXPECT_CALL(*mock_decoder_database_, GetActiveCngDecoder())
|
|
.WillOnce(ReturnNull());
|
|
decoder_database_ = mock_decoder_database_;
|
|
} else {
|
|
decoder_database_ = new DecoderDatabase;
|
|
}
|
|
if (use_mock_delay_peak_detector_) {
|
|
mock_delay_peak_detector_ = new MockDelayPeakDetector;
|
|
EXPECT_CALL(*mock_delay_peak_detector_, Reset()).Times(1);
|
|
delay_peak_detector_ = mock_delay_peak_detector_;
|
|
} else {
|
|
delay_peak_detector_ = new DelayPeakDetector;
|
|
}
|
|
if (use_mock_delay_manager_) {
|
|
mock_delay_manager_ = new MockDelayManager(config_.max_packets_in_buffer,
|
|
delay_peak_detector_);
|
|
EXPECT_CALL(*mock_delay_manager_, set_streaming_mode(false)).Times(1);
|
|
delay_manager_ = mock_delay_manager_;
|
|
} else {
|
|
delay_manager_ =
|
|
new DelayManager(config_.max_packets_in_buffer, delay_peak_detector_);
|
|
}
|
|
if (use_mock_dtmf_buffer_) {
|
|
mock_dtmf_buffer_ = new MockDtmfBuffer(config_.sample_rate_hz);
|
|
dtmf_buffer_ = mock_dtmf_buffer_;
|
|
} else {
|
|
dtmf_buffer_ = new DtmfBuffer(config_.sample_rate_hz);
|
|
}
|
|
if (use_mock_dtmf_tone_generator_) {
|
|
mock_dtmf_tone_generator_ = new MockDtmfToneGenerator;
|
|
dtmf_tone_generator_ = mock_dtmf_tone_generator_;
|
|
} else {
|
|
dtmf_tone_generator_ = new DtmfToneGenerator;
|
|
}
|
|
if (use_mock_packet_buffer_) {
|
|
mock_packet_buffer_ = new MockPacketBuffer(config_.max_packets_in_buffer);
|
|
packet_buffer_ = mock_packet_buffer_;
|
|
} else {
|
|
packet_buffer_ = new PacketBuffer(config_.max_packets_in_buffer);
|
|
}
|
|
if (use_mock_payload_splitter_) {
|
|
mock_payload_splitter_ = new MockPayloadSplitter;
|
|
payload_splitter_ = mock_payload_splitter_;
|
|
} else {
|
|
payload_splitter_ = new PayloadSplitter;
|
|
}
|
|
timestamp_scaler_ = new TimestampScaler(*decoder_database_);
|
|
AccelerateFactory* accelerate_factory = new AccelerateFactory;
|
|
ExpandFactory* expand_factory = new ExpandFactory;
|
|
PreemptiveExpandFactory* preemptive_expand_factory =
|
|
new PreemptiveExpandFactory;
|
|
|
|
neteq_ = new NetEqImpl(config_,
|
|
buffer_level_filter_,
|
|
decoder_database_,
|
|
delay_manager_,
|
|
delay_peak_detector_,
|
|
dtmf_buffer_,
|
|
dtmf_tone_generator_,
|
|
packet_buffer_,
|
|
payload_splitter_,
|
|
timestamp_scaler_,
|
|
accelerate_factory,
|
|
expand_factory,
|
|
preemptive_expand_factory);
|
|
ASSERT_TRUE(neteq_ != NULL);
|
|
}
|
|
|
|
void UseNoMocks() {
|
|
ASSERT_TRUE(neteq_ == NULL) << "Must call UseNoMocks before CreateInstance";
|
|
use_mock_buffer_level_filter_ = false;
|
|
use_mock_decoder_database_ = false;
|
|
use_mock_delay_peak_detector_ = false;
|
|
use_mock_delay_manager_ = false;
|
|
use_mock_dtmf_buffer_ = false;
|
|
use_mock_dtmf_tone_generator_ = false;
|
|
use_mock_packet_buffer_ = false;
|
|
use_mock_payload_splitter_ = false;
|
|
}
|
|
|
|
virtual ~NetEqImplTest() {
|
|
if (use_mock_buffer_level_filter_) {
|
|
EXPECT_CALL(*mock_buffer_level_filter_, Die()).Times(1);
|
|
}
|
|
if (use_mock_decoder_database_) {
|
|
EXPECT_CALL(*mock_decoder_database_, Die()).Times(1);
|
|
}
|
|
if (use_mock_delay_manager_) {
|
|
EXPECT_CALL(*mock_delay_manager_, Die()).Times(1);
|
|
}
|
|
if (use_mock_delay_peak_detector_) {
|
|
EXPECT_CALL(*mock_delay_peak_detector_, Die()).Times(1);
|
|
}
|
|
if (use_mock_dtmf_buffer_) {
|
|
EXPECT_CALL(*mock_dtmf_buffer_, Die()).Times(1);
|
|
}
|
|
if (use_mock_dtmf_tone_generator_) {
|
|
EXPECT_CALL(*mock_dtmf_tone_generator_, Die()).Times(1);
|
|
}
|
|
if (use_mock_packet_buffer_) {
|
|
EXPECT_CALL(*mock_packet_buffer_, Die()).Times(1);
|
|
}
|
|
delete neteq_;
|
|
}
|
|
|
|
NetEqImpl* neteq_;
|
|
NetEq::Config config_;
|
|
MockBufferLevelFilter* mock_buffer_level_filter_;
|
|
BufferLevelFilter* buffer_level_filter_;
|
|
bool use_mock_buffer_level_filter_;
|
|
MockDecoderDatabase* mock_decoder_database_;
|
|
DecoderDatabase* decoder_database_;
|
|
bool use_mock_decoder_database_;
|
|
MockDelayPeakDetector* mock_delay_peak_detector_;
|
|
DelayPeakDetector* delay_peak_detector_;
|
|
bool use_mock_delay_peak_detector_;
|
|
MockDelayManager* mock_delay_manager_;
|
|
DelayManager* delay_manager_;
|
|
bool use_mock_delay_manager_;
|
|
MockDtmfBuffer* mock_dtmf_buffer_;
|
|
DtmfBuffer* dtmf_buffer_;
|
|
bool use_mock_dtmf_buffer_;
|
|
MockDtmfToneGenerator* mock_dtmf_tone_generator_;
|
|
DtmfToneGenerator* dtmf_tone_generator_;
|
|
bool use_mock_dtmf_tone_generator_;
|
|
MockPacketBuffer* mock_packet_buffer_;
|
|
PacketBuffer* packet_buffer_;
|
|
bool use_mock_packet_buffer_;
|
|
MockPayloadSplitter* mock_payload_splitter_;
|
|
PayloadSplitter* payload_splitter_;
|
|
bool use_mock_payload_splitter_;
|
|
TimestampScaler* timestamp_scaler_;
|
|
};
|
|
|
|
|
|
// This tests the interface class NetEq.
|
|
// TODO(hlundin): Move to separate file?
|
|
TEST(NetEq, CreateAndDestroy) {
|
|
NetEq::Config config;
|
|
NetEq* neteq = NetEq::Create(config);
|
|
delete neteq;
|
|
}
|
|
|
|
TEST_F(NetEqImplTest, RegisterPayloadType) {
|
|
CreateInstance();
|
|
uint8_t rtp_payload_type = 0;
|
|
NetEqDecoder codec_type = kDecoderPCMu;
|
|
EXPECT_CALL(*mock_decoder_database_,
|
|
RegisterPayload(rtp_payload_type, codec_type));
|
|
neteq_->RegisterPayloadType(codec_type, rtp_payload_type);
|
|
}
|
|
|
|
TEST_F(NetEqImplTest, RemovePayloadType) {
|
|
CreateInstance();
|
|
uint8_t rtp_payload_type = 0;
|
|
EXPECT_CALL(*mock_decoder_database_, Remove(rtp_payload_type))
|
|
.WillOnce(Return(DecoderDatabase::kDecoderNotFound));
|
|
// Check that kFail is returned when database returns kDecoderNotFound.
|
|
EXPECT_EQ(NetEq::kFail, neteq_->RemovePayloadType(rtp_payload_type));
|
|
}
|
|
|
|
TEST_F(NetEqImplTest, InsertPacket) {
|
|
CreateInstance();
|
|
const int kPayloadLength = 100;
|
|
const uint8_t kPayloadType = 0;
|
|
const uint16_t kFirstSequenceNumber = 0x1234;
|
|
const uint32_t kFirstTimestamp = 0x12345678;
|
|
const uint32_t kSsrc = 0x87654321;
|
|
const uint32_t kFirstReceiveTime = 17;
|
|
uint8_t payload[kPayloadLength] = {0};
|
|
WebRtcRTPHeader rtp_header;
|
|
rtp_header.header.payloadType = kPayloadType;
|
|
rtp_header.header.sequenceNumber = kFirstSequenceNumber;
|
|
rtp_header.header.timestamp = kFirstTimestamp;
|
|
rtp_header.header.ssrc = kSsrc;
|
|
|
|
// Create a mock decoder object.
|
|
MockAudioDecoder mock_decoder;
|
|
// BWE update function called with first packet.
|
|
EXPECT_CALL(mock_decoder, IncomingPacket(_,
|
|
kPayloadLength,
|
|
kFirstSequenceNumber,
|
|
kFirstTimestamp,
|
|
kFirstReceiveTime));
|
|
// BWE update function called with second packet.
|
|
EXPECT_CALL(mock_decoder, IncomingPacket(_,
|
|
kPayloadLength,
|
|
kFirstSequenceNumber + 1,
|
|
kFirstTimestamp + 160,
|
|
kFirstReceiveTime + 155));
|
|
EXPECT_CALL(mock_decoder, Die()).Times(1); // Called when deleted.
|
|
|
|
// Expectations for decoder database.
|
|
EXPECT_CALL(*mock_decoder_database_, IsRed(kPayloadType))
|
|
.WillRepeatedly(Return(false)); // This is not RED.
|
|
EXPECT_CALL(*mock_decoder_database_, CheckPayloadTypes(_))
|
|
.Times(2)
|
|
.WillRepeatedly(Return(DecoderDatabase::kOK)); // Payload type is valid.
|
|
EXPECT_CALL(*mock_decoder_database_, IsDtmf(kPayloadType))
|
|
.WillRepeatedly(Return(false)); // This is not DTMF.
|
|
EXPECT_CALL(*mock_decoder_database_, GetDecoder(kPayloadType))
|
|
.Times(3)
|
|
.WillRepeatedly(Return(&mock_decoder));
|
|
EXPECT_CALL(*mock_decoder_database_, IsComfortNoise(kPayloadType))
|
|
.WillRepeatedly(Return(false)); // This is not CNG.
|
|
DecoderDatabase::DecoderInfo info;
|
|
info.codec_type = kDecoderPCMu;
|
|
EXPECT_CALL(*mock_decoder_database_, GetDecoderInfo(kPayloadType))
|
|
.WillRepeatedly(Return(&info));
|
|
|
|
// Expectations for packet buffer.
|
|
EXPECT_CALL(*mock_packet_buffer_, NumPacketsInBuffer())
|
|
.WillOnce(Return(0)) // First packet.
|
|
.WillOnce(Return(1)) // Second packet.
|
|
.WillOnce(Return(2)); // Second packet, checking after it was inserted.
|
|
EXPECT_CALL(*mock_packet_buffer_, Empty())
|
|
.WillOnce(Return(false)); // Called once after first packet is inserted.
|
|
EXPECT_CALL(*mock_packet_buffer_, Flush())
|
|
.Times(1);
|
|
EXPECT_CALL(*mock_packet_buffer_, InsertPacketList(_, _, _, _))
|
|
.Times(2)
|
|
.WillRepeatedly(DoAll(SetArgPointee<2>(kPayloadType),
|
|
WithArg<0>(Invoke(DeletePacketsAndReturnOk))));
|
|
// SetArgPointee<2>(kPayloadType) means that the third argument (zero-based
|
|
// index) is a pointer, and the variable pointed to is set to kPayloadType.
|
|
// Also invoke the function DeletePacketsAndReturnOk to properly delete all
|
|
// packets in the list (to avoid memory leaks in the test).
|
|
EXPECT_CALL(*mock_packet_buffer_, NextRtpHeader())
|
|
.Times(1)
|
|
.WillOnce(Return(&rtp_header.header));
|
|
|
|
// Expectations for DTMF buffer.
|
|
EXPECT_CALL(*mock_dtmf_buffer_, Flush())
|
|
.Times(1);
|
|
|
|
// Expectations for delay manager.
|
|
{
|
|
// All expectations within this block must be called in this specific order.
|
|
InSequence sequence; // Dummy variable.
|
|
// Expectations when the first packet is inserted.
|
|
EXPECT_CALL(*mock_delay_manager_, LastDecoderType(kDecoderPCMu))
|
|
.Times(1);
|
|
EXPECT_CALL(*mock_delay_manager_, last_pack_cng_or_dtmf())
|
|
.Times(2)
|
|
.WillRepeatedly(Return(-1));
|
|
EXPECT_CALL(*mock_delay_manager_, set_last_pack_cng_or_dtmf(0))
|
|
.Times(1);
|
|
EXPECT_CALL(*mock_delay_manager_, ResetPacketIatCount()).Times(1);
|
|
// Expectations when the second packet is inserted. Slightly different.
|
|
EXPECT_CALL(*mock_delay_manager_, LastDecoderType(kDecoderPCMu))
|
|
.Times(1);
|
|
EXPECT_CALL(*mock_delay_manager_, last_pack_cng_or_dtmf())
|
|
.WillOnce(Return(0));
|
|
EXPECT_CALL(*mock_delay_manager_, SetPacketAudioLength(30))
|
|
.WillOnce(Return(0));
|
|
}
|
|
|
|
// Expectations for payload splitter.
|
|
EXPECT_CALL(*mock_payload_splitter_, SplitAudio(_, _))
|
|
.Times(2)
|
|
.WillRepeatedly(Return(PayloadSplitter::kOK));
|
|
|
|
// Insert first packet.
|
|
neteq_->InsertPacket(rtp_header, payload, kPayloadLength, kFirstReceiveTime);
|
|
|
|
// Insert second packet.
|
|
rtp_header.header.timestamp += 160;
|
|
rtp_header.header.sequenceNumber += 1;
|
|
neteq_->InsertPacket(rtp_header, payload, kPayloadLength,
|
|
kFirstReceiveTime + 155);
|
|
}
|
|
|
|
TEST_F(NetEqImplTest, InsertPacketsUntilBufferIsFull) {
|
|
UseNoMocks();
|
|
CreateInstance();
|
|
|
|
const int kPayloadLengthSamples = 80;
|
|
const size_t kPayloadLengthBytes = 2 * kPayloadLengthSamples; // PCM 16-bit.
|
|
const uint8_t kPayloadType = 17; // Just an arbitrary number.
|
|
const uint32_t kReceiveTime = 17; // Value doesn't matter for this test.
|
|
uint8_t payload[kPayloadLengthBytes] = {0};
|
|
WebRtcRTPHeader rtp_header;
|
|
rtp_header.header.payloadType = kPayloadType;
|
|
rtp_header.header.sequenceNumber = 0x1234;
|
|
rtp_header.header.timestamp = 0x12345678;
|
|
rtp_header.header.ssrc = 0x87654321;
|
|
|
|
EXPECT_EQ(NetEq::kOK,
|
|
neteq_->RegisterPayloadType(kDecoderPCM16B, kPayloadType));
|
|
|
|
// Insert packets. The buffer should not flush.
|
|
for (int i = 1; i <= config_.max_packets_in_buffer; ++i) {
|
|
EXPECT_EQ(NetEq::kOK,
|
|
neteq_->InsertPacket(
|
|
rtp_header, payload, kPayloadLengthBytes, kReceiveTime));
|
|
rtp_header.header.timestamp += kPayloadLengthSamples;
|
|
rtp_header.header.sequenceNumber += 1;
|
|
EXPECT_EQ(i, packet_buffer_->NumPacketsInBuffer());
|
|
}
|
|
|
|
// Insert one more packet and make sure the buffer got flushed. That is, it
|
|
// should only hold one single packet.
|
|
EXPECT_EQ(NetEq::kOK,
|
|
neteq_->InsertPacket(
|
|
rtp_header, payload, kPayloadLengthBytes, kReceiveTime));
|
|
EXPECT_EQ(1, packet_buffer_->NumPacketsInBuffer());
|
|
const RTPHeader* test_header = packet_buffer_->NextRtpHeader();
|
|
EXPECT_EQ(rtp_header.header.timestamp, test_header->timestamp);
|
|
EXPECT_EQ(rtp_header.header.sequenceNumber, test_header->sequenceNumber);
|
|
}
|
|
|
|
// This test verifies that timestamps propagate from the incoming packets
|
|
// through to the sync buffer and to the playout timestamp.
|
|
TEST_F(NetEqImplTest, VerifyTimestampPropagation) {
|
|
UseNoMocks();
|
|
CreateInstance();
|
|
|
|
const uint8_t kPayloadType = 17; // Just an arbitrary number.
|
|
const uint32_t kReceiveTime = 17; // Value doesn't matter for this test.
|
|
const int kSampleRateHz = 8000;
|
|
const int kPayloadLengthSamples = 10 * kSampleRateHz / 1000; // 10 ms.
|
|
const size_t kPayloadLengthBytes = kPayloadLengthSamples;
|
|
uint8_t payload[kPayloadLengthBytes] = {0};
|
|
WebRtcRTPHeader rtp_header;
|
|
rtp_header.header.payloadType = kPayloadType;
|
|
rtp_header.header.sequenceNumber = 0x1234;
|
|
rtp_header.header.timestamp = 0x12345678;
|
|
rtp_header.header.ssrc = 0x87654321;
|
|
|
|
// This is a dummy decoder that produces as many output samples as the input
|
|
// has bytes. The output is an increasing series, starting at 1 for the first
|
|
// sample, and then increasing by 1 for each sample.
|
|
class CountingSamplesDecoder : public AudioDecoder {
|
|
public:
|
|
explicit CountingSamplesDecoder(enum NetEqDecoder type)
|
|
: AudioDecoder(type), next_value_(1) {}
|
|
|
|
// Produce as many samples as input bytes (|encoded_len|).
|
|
virtual int Decode(const uint8_t* encoded,
|
|
size_t encoded_len,
|
|
int16_t* decoded,
|
|
SpeechType* speech_type) {
|
|
for (size_t i = 0; i < encoded_len; ++i) {
|
|
decoded[i] = next_value_++;
|
|
}
|
|
*speech_type = kSpeech;
|
|
return encoded_len;
|
|
}
|
|
|
|
virtual int Init() {
|
|
next_value_ = 1;
|
|
return 0;
|
|
}
|
|
|
|
uint16_t next_value() const { return next_value_; }
|
|
|
|
private:
|
|
int16_t next_value_;
|
|
} decoder_(kDecoderPCM16B);
|
|
|
|
EXPECT_EQ(NetEq::kOK,
|
|
neteq_->RegisterExternalDecoder(
|
|
&decoder_, kDecoderPCM16B, kPayloadType));
|
|
|
|
// Insert one packet.
|
|
EXPECT_EQ(NetEq::kOK,
|
|
neteq_->InsertPacket(
|
|
rtp_header, payload, kPayloadLengthBytes, kReceiveTime));
|
|
|
|
// Pull audio once.
|
|
const int kMaxOutputSize = 10 * kSampleRateHz / 1000;
|
|
int16_t output[kMaxOutputSize];
|
|
int samples_per_channel;
|
|
int num_channels;
|
|
NetEqOutputType type;
|
|
EXPECT_EQ(
|
|
NetEq::kOK,
|
|
neteq_->GetAudio(
|
|
kMaxOutputSize, output, &samples_per_channel, &num_channels, &type));
|
|
ASSERT_EQ(kMaxOutputSize, samples_per_channel);
|
|
EXPECT_EQ(1, num_channels);
|
|
EXPECT_EQ(kOutputNormal, type);
|
|
|
|
// Start with a simple check that the fake decoder is behaving as expected.
|
|
EXPECT_EQ(kPayloadLengthSamples, decoder_.next_value() - 1);
|
|
|
|
// The value of the last of the output samples is the same as the number of
|
|
// samples played from the decoded packet. Thus, this number + the RTP
|
|
// timestamp should match the playout timestamp.
|
|
uint32_t timestamp = 0;
|
|
EXPECT_TRUE(neteq_->GetPlayoutTimestamp(×tamp));
|
|
EXPECT_EQ(rtp_header.header.timestamp + output[samples_per_channel - 1],
|
|
timestamp);
|
|
|
|
// Check the timestamp for the last value in the sync buffer. This should
|
|
// be one full frame length ahead of the RTP timestamp.
|
|
const SyncBuffer* sync_buffer = neteq_->sync_buffer_for_test();
|
|
ASSERT_TRUE(sync_buffer != NULL);
|
|
EXPECT_EQ(rtp_header.header.timestamp + kPayloadLengthSamples,
|
|
sync_buffer->end_timestamp());
|
|
|
|
// Check that the number of samples still to play from the sync buffer add
|
|
// up with what was already played out.
|
|
EXPECT_EQ(kPayloadLengthSamples - output[samples_per_channel - 1],
|
|
static_cast<int>(sync_buffer->FutureLength()));
|
|
}
|
|
|
|
} // namespace webrtc
|