webrtc/audio/audio_send_stream_unittest.cc
Sebastian Jansson 0b69826ffb Don't inject worker queue into send streams.
This prepares for making AudioSendStream use its own task queue. In the
future more of the functionality that depends on running on the task
queue is planned to be moved directly into RtpTransportControllerSend.

They should instead get it from the transport controller. This affects
the media transport tests which previously assumed that the transport
controller could be missing. However, this is not something that is used
in production, so this is an improvement of the tests as they will
behave more like production code.

Bug: webrtc:9883
Change-Id: Ie32f4c2f6433ec37ac16a08d531ceb690ea9c0b5
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/126000
Reviewed-by: Oskar Sundbom <ossu@webrtc.org>
Reviewed-by: Niels Moller <nisse@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Sebastian Jansson <srte@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#27010}
2019-03-07 09:42:26 +00:00

615 lines
25 KiB
C++

/*
* Copyright (c) 2015 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 <string>
#include <utility>
#include <vector>
#include "absl/memory/memory.h"
#include "api/task_queue/global_task_queue_factory.h"
#include "api/test/mock_frame_encryptor.h"
#include "audio/audio_send_stream.h"
#include "audio/audio_state.h"
#include "audio/conversion.h"
#include "audio/mock_voe_channel_proxy.h"
#include "call/test/mock_rtp_transport_controller_send.h"
#include "logging/rtc_event_log/mock/mock_rtc_event_log.h"
#include "modules/audio_device/include/mock_audio_device.h"
#include "modules/audio_mixer/audio_mixer_impl.h"
#include "modules/audio_processing/include/audio_processing_statistics.h"
#include "modules/audio_processing/include/mock_audio_processing.h"
#include "modules/rtp_rtcp/mocks/mock_rtcp_bandwidth_observer.h"
#include "modules/rtp_rtcp/mocks/mock_rtcp_rtt_stats.h"
#include "modules/rtp_rtcp/mocks/mock_rtp_rtcp.h"
#include "rtc_base/task_queue.h"
#include "system_wrappers/include/clock.h"
#include "test/field_trial.h"
#include "test/gtest.h"
#include "test/mock_audio_encoder.h"
#include "test/mock_audio_encoder_factory.h"
namespace webrtc {
namespace test {
namespace {
using testing::_;
using testing::Eq;
using testing::Ne;
using testing::Field;
using testing::Invoke;
using testing::Return;
using testing::StrEq;
const uint32_t kSsrc = 1234;
const char* kCName = "foo_name";
const int kAudioLevelId = 2;
const int kTransportSequenceNumberId = 4;
const int32_t kEchoDelayMedian = 254;
const int32_t kEchoDelayStdDev = -3;
const double kDivergentFilterFraction = 0.2f;
const double kEchoReturnLoss = -65;
const double kEchoReturnLossEnhancement = 101;
const double kResidualEchoLikelihood = -1.0f;
const double kResidualEchoLikelihoodMax = 23.0f;
const CallSendStatistics kCallStats = {112, 13456, 17890};
const ReportBlock kReportBlock = {456, 780, 123, 567, 890, 132, 143, 13354};
const int kTelephoneEventPayloadType = 123;
const int kTelephoneEventPayloadFrequency = 65432;
const int kTelephoneEventCode = 45;
const int kTelephoneEventDuration = 6789;
constexpr int kIsacPayloadType = 103;
const SdpAudioFormat kIsacFormat = {"isac", 16000, 1};
const SdpAudioFormat kOpusFormat = {"opus", 48000, 2};
const SdpAudioFormat kG722Format = {"g722", 8000, 1};
const AudioCodecSpec kCodecSpecs[] = {
{kIsacFormat, {16000, 1, 32000, 10000, 32000}},
{kOpusFormat, {48000, 1, 32000, 6000, 510000}},
{kG722Format, {16000, 1, 64000}}};
class MockLimitObserver : public BitrateAllocator::LimitObserver {
public:
MOCK_METHOD3(OnAllocationLimitsChanged,
void(uint32_t min_send_bitrate_bps,
uint32_t max_padding_bitrate_bps,
uint32_t total_bitrate_bps));
};
std::unique_ptr<MockAudioEncoder> SetupAudioEncoderMock(
int payload_type,
const SdpAudioFormat& format) {
for (const auto& spec : kCodecSpecs) {
if (format == spec.format) {
std::unique_ptr<MockAudioEncoder> encoder(
new testing::NiceMock<MockAudioEncoder>());
ON_CALL(*encoder.get(), SampleRateHz())
.WillByDefault(Return(spec.info.sample_rate_hz));
ON_CALL(*encoder.get(), NumChannels())
.WillByDefault(Return(spec.info.num_channels));
ON_CALL(*encoder.get(), RtpTimestampRateHz())
.WillByDefault(Return(spec.format.clockrate_hz));
return encoder;
}
}
return nullptr;
}
rtc::scoped_refptr<MockAudioEncoderFactory> SetupEncoderFactoryMock() {
rtc::scoped_refptr<MockAudioEncoderFactory> factory =
new rtc::RefCountedObject<MockAudioEncoderFactory>();
ON_CALL(*factory.get(), GetSupportedEncoders())
.WillByDefault(Return(std::vector<AudioCodecSpec>(
std::begin(kCodecSpecs), std::end(kCodecSpecs))));
ON_CALL(*factory.get(), QueryAudioEncoder(_))
.WillByDefault(Invoke(
[](const SdpAudioFormat& format) -> absl::optional<AudioCodecInfo> {
for (const auto& spec : kCodecSpecs) {
if (format == spec.format) {
return spec.info;
}
}
return absl::nullopt;
}));
ON_CALL(*factory.get(), MakeAudioEncoderMock(_, _, _, _))
.WillByDefault(Invoke([](int payload_type, const SdpAudioFormat& format,
absl::optional<AudioCodecPairId> codec_pair_id,
std::unique_ptr<AudioEncoder>* return_value) {
*return_value = SetupAudioEncoderMock(payload_type, format);
}));
return factory;
}
struct ConfigHelper {
ConfigHelper(bool audio_bwe_enabled, bool expect_set_encoder_call)
: clock_(1000000),
stream_config_(/*send_transport=*/nullptr, /*media_transport=*/nullptr),
audio_processing_(new rtc::RefCountedObject<MockAudioProcessing>()),
bitrate_allocator_(&clock_, &limit_observer_),
worker_queue_("ConfigHelper_worker_queue"),
audio_encoder_(nullptr) {
using testing::Invoke;
AudioState::Config config;
config.audio_mixer = AudioMixerImpl::Create();
config.audio_processing = audio_processing_;
config.audio_device_module =
new rtc::RefCountedObject<MockAudioDeviceModule>();
audio_state_ = AudioState::Create(config);
SetupDefaultChannelSend(audio_bwe_enabled);
SetupMockForSetupSendCodec(expect_set_encoder_call);
// Use ISAC as default codec so as to prevent unnecessary |channel_proxy_|
// calls from the default ctor behavior.
stream_config_.send_codec_spec =
AudioSendStream::Config::SendCodecSpec(kIsacPayloadType, kIsacFormat);
stream_config_.rtp.ssrc = kSsrc;
stream_config_.rtp.c_name = kCName;
stream_config_.rtp.extensions.push_back(
RtpExtension(RtpExtension::kAudioLevelUri, kAudioLevelId));
if (audio_bwe_enabled) {
AddBweToConfig(&stream_config_);
}
stream_config_.encoder_factory = SetupEncoderFactoryMock();
stream_config_.min_bitrate_bps = 10000;
stream_config_.max_bitrate_bps = 65000;
}
std::unique_ptr<internal::AudioSendStream> CreateAudioSendStream() {
EXPECT_CALL(rtp_transport_, GetWorkerQueue())
.WillRepeatedly(Return(&worker_queue_));
return std::unique_ptr<internal::AudioSendStream>(
new internal::AudioSendStream(
Clock::GetRealTimeClock(), stream_config_, audio_state_,
&rtp_transport_, &bitrate_allocator_, &event_log_, &rtcp_rtt_stats_,
absl::nullopt,
std::unique_ptr<voe::ChannelSendInterface>(channel_send_)));
}
AudioSendStream::Config& config() { return stream_config_; }
MockAudioEncoderFactory& mock_encoder_factory() {
return *static_cast<MockAudioEncoderFactory*>(
stream_config_.encoder_factory.get());
}
MockChannelSend* channel_send() { return channel_send_; }
RtpTransportControllerSendInterface* transport() { return &rtp_transport_; }
static void AddBweToConfig(AudioSendStream::Config* config) {
config->rtp.extensions.push_back(RtpExtension(
RtpExtension::kTransportSequenceNumberUri, kTransportSequenceNumberId));
config->send_codec_spec->transport_cc_enabled = true;
}
void SetupDefaultChannelSend(bool audio_bwe_enabled) {
EXPECT_TRUE(channel_send_ == nullptr);
channel_send_ = new testing::StrictMock<MockChannelSend>();
EXPECT_CALL(*channel_send_, GetRtpRtcp()).WillRepeatedly(Invoke([this]() {
return &this->rtp_rtcp_;
}));
EXPECT_CALL(*channel_send_, SetLocalSSRC(kSsrc)).Times(1);
EXPECT_CALL(*channel_send_, SetRTCP_CNAME(StrEq(kCName))).Times(1);
EXPECT_CALL(*channel_send_, SetFrameEncryptor(_)).Times(1);
EXPECT_CALL(*channel_send_, SetExtmapAllowMixed(false)).Times(1);
EXPECT_CALL(*channel_send_,
SetSendAudioLevelIndicationStatus(true, kAudioLevelId))
.Times(1);
EXPECT_CALL(rtp_transport_, GetBandwidthObserver())
.WillRepeatedly(Return(&bandwidth_observer_));
if (audio_bwe_enabled) {
EXPECT_CALL(*channel_send_,
EnableSendTransportSequenceNumber(kTransportSequenceNumberId))
.Times(1);
EXPECT_CALL(*channel_send_,
RegisterSenderCongestionControlObjects(
&rtp_transport_, Eq(&bandwidth_observer_)))
.Times(1);
} else {
EXPECT_CALL(*channel_send_, RegisterSenderCongestionControlObjects(
&rtp_transport_, Eq(nullptr)))
.Times(1);
}
EXPECT_CALL(*channel_send_, ResetSenderCongestionControlObjects()).Times(1);
EXPECT_CALL(*channel_send_, SetRid(std::string(), 0, 0)).Times(1);
}
void SetupMockForSetupSendCodec(bool expect_set_encoder_call) {
if (expect_set_encoder_call) {
EXPECT_CALL(*channel_send_, SetEncoderForMock(_, _))
.WillOnce(Invoke(
[this](int payload_type, std::unique_ptr<AudioEncoder>* encoder) {
this->audio_encoder_ = std::move(*encoder);
return true;
}));
}
}
void SetupMockForCallEncoder() {
// Let ModifyEncoder to invoke mock audio encoder.
EXPECT_CALL(*channel_send_, CallEncoder(_))
.WillRepeatedly(
[this](rtc::FunctionView<void(AudioEncoder*)> modifier) {
if (this->audio_encoder_)
modifier(this->audio_encoder_.get());
});
}
void SetupMockForSendTelephoneEvent() {
EXPECT_TRUE(channel_send_);
EXPECT_CALL(*channel_send_, SetSendTelephoneEventPayloadType(
kTelephoneEventPayloadType,
kTelephoneEventPayloadFrequency));
EXPECT_CALL(
*channel_send_,
SendTelephoneEventOutband(kTelephoneEventCode, kTelephoneEventDuration))
.WillOnce(Return(true));
}
void SetupMockForGetStats() {
using testing::DoAll;
using testing::SetArgPointee;
using testing::SetArgReferee;
std::vector<ReportBlock> report_blocks;
webrtc::ReportBlock block = kReportBlock;
report_blocks.push_back(block); // Has wrong SSRC.
block.source_SSRC = kSsrc;
report_blocks.push_back(block); // Correct block.
block.fraction_lost = 0;
report_blocks.push_back(block); // Duplicate SSRC, bad fraction_lost.
EXPECT_TRUE(channel_send_);
EXPECT_CALL(*channel_send_, GetRTCPStatistics())
.WillRepeatedly(Return(kCallStats));
EXPECT_CALL(*channel_send_, GetRemoteRTCPReportBlocks())
.WillRepeatedly(Return(report_blocks));
EXPECT_CALL(*channel_send_, GetANAStatistics())
.WillRepeatedly(Return(ANAStats()));
EXPECT_CALL(*channel_send_, GetBitrate()).WillRepeatedly(Return(0));
audio_processing_stats_.echo_return_loss = kEchoReturnLoss;
audio_processing_stats_.echo_return_loss_enhancement =
kEchoReturnLossEnhancement;
audio_processing_stats_.delay_median_ms = kEchoDelayMedian;
audio_processing_stats_.delay_standard_deviation_ms = kEchoDelayStdDev;
audio_processing_stats_.divergent_filter_fraction =
kDivergentFilterFraction;
audio_processing_stats_.residual_echo_likelihood = kResidualEchoLikelihood;
audio_processing_stats_.residual_echo_likelihood_recent_max =
kResidualEchoLikelihoodMax;
EXPECT_CALL(*audio_processing_, GetStatistics(true))
.WillRepeatedly(Return(audio_processing_stats_));
}
private:
SimulatedClock clock_;
rtc::scoped_refptr<AudioState> audio_state_;
AudioSendStream::Config stream_config_;
testing::StrictMock<MockChannelSend>* channel_send_ = nullptr;
rtc::scoped_refptr<MockAudioProcessing> audio_processing_;
AudioProcessingStats audio_processing_stats_;
testing::StrictMock<MockRtcpBandwidthObserver> bandwidth_observer_;
testing::NiceMock<MockRtcEventLog> event_log_;
testing::NiceMock<MockRtpTransportControllerSend> rtp_transport_;
testing::NiceMock<MockRtpRtcp> rtp_rtcp_;
MockRtcpRttStats rtcp_rtt_stats_;
testing::NiceMock<MockLimitObserver> limit_observer_;
BitrateAllocator bitrate_allocator_;
// |worker_queue| is defined last to ensure all pending tasks are cancelled
// and deleted before any other members.
rtc::TaskQueue worker_queue_;
std::unique_ptr<AudioEncoder> audio_encoder_;
};
} // namespace
TEST(AudioSendStreamTest, ConfigToString) {
AudioSendStream::Config config(/*send_transport=*/nullptr,
/*media_transport=*/nullptr);
config.rtp.ssrc = kSsrc;
config.rtp.c_name = kCName;
config.min_bitrate_bps = 12000;
config.max_bitrate_bps = 34000;
config.send_codec_spec =
AudioSendStream::Config::SendCodecSpec(kIsacPayloadType, kIsacFormat);
config.send_codec_spec->nack_enabled = true;
config.send_codec_spec->transport_cc_enabled = false;
config.send_codec_spec->cng_payload_type = 42;
config.encoder_factory = MockAudioEncoderFactory::CreateUnusedFactory();
config.rtp.extmap_allow_mixed = true;
config.rtp.extensions.push_back(
RtpExtension(RtpExtension::kAudioLevelUri, kAudioLevelId));
config.rtcp_report_interval_ms = 2500;
EXPECT_EQ(
"{rtp: {ssrc: 1234, extmap-allow-mixed: true, extensions: [{uri: "
"urn:ietf:params:rtp-hdrext:ssrc-audio-level, id: 2}], "
"c_name: foo_name}, rtcp_report_interval_ms: 2500, "
"send_transport: null, media_transport: null, "
"min_bitrate_bps: 12000, max_bitrate_bps: 34000, "
"send_codec_spec: {nack_enabled: true, transport_cc_enabled: false, "
"cng_payload_type: 42, payload_type: 103, "
"format: {name: isac, clockrate_hz: 16000, num_channels: 1, "
"parameters: {}}}}",
config.ToString());
}
TEST(AudioSendStreamTest, ConstructDestruct) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
}
TEST(AudioSendStreamTest, SendTelephoneEvent) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
helper.SetupMockForSendTelephoneEvent();
EXPECT_TRUE(send_stream->SendTelephoneEvent(
kTelephoneEventPayloadType, kTelephoneEventPayloadFrequency,
kTelephoneEventCode, kTelephoneEventDuration));
}
TEST(AudioSendStreamTest, SetMuted) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
EXPECT_CALL(*helper.channel_send(), SetInputMute(true));
send_stream->SetMuted(true);
}
TEST(AudioSendStreamTest, AudioBweCorrectObjectsOnChannelProxy) {
ScopedFieldTrials field_trials("WebRTC-Audio-SendSideBwe/Enabled/");
ConfigHelper helper(true, true);
auto send_stream = helper.CreateAudioSendStream();
}
TEST(AudioSendStreamTest, NoAudioBweCorrectObjectsOnChannelProxy) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
}
TEST(AudioSendStreamTest, GetStats) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
helper.SetupMockForGetStats();
AudioSendStream::Stats stats = send_stream->GetStats(true);
EXPECT_EQ(kSsrc, stats.local_ssrc);
EXPECT_EQ(static_cast<int64_t>(kCallStats.bytesSent), stats.bytes_sent);
EXPECT_EQ(kCallStats.packetsSent, stats.packets_sent);
EXPECT_EQ(kReportBlock.cumulative_num_packets_lost, stats.packets_lost);
EXPECT_EQ(Q8ToFloat(kReportBlock.fraction_lost), stats.fraction_lost);
EXPECT_EQ(kIsacFormat.name, stats.codec_name);
EXPECT_EQ(static_cast<int32_t>(kReportBlock.extended_highest_sequence_number),
stats.ext_seqnum);
EXPECT_EQ(static_cast<int32_t>(kReportBlock.interarrival_jitter /
(kIsacFormat.clockrate_hz / 1000)),
stats.jitter_ms);
EXPECT_EQ(kCallStats.rttMs, stats.rtt_ms);
EXPECT_EQ(0, stats.audio_level);
EXPECT_EQ(0, stats.total_input_energy);
EXPECT_EQ(0, stats.total_input_duration);
EXPECT_EQ(kEchoDelayMedian, stats.apm_statistics.delay_median_ms);
EXPECT_EQ(kEchoDelayStdDev, stats.apm_statistics.delay_standard_deviation_ms);
EXPECT_EQ(kEchoReturnLoss, stats.apm_statistics.echo_return_loss);
EXPECT_EQ(kEchoReturnLossEnhancement,
stats.apm_statistics.echo_return_loss_enhancement);
EXPECT_EQ(kDivergentFilterFraction,
stats.apm_statistics.divergent_filter_fraction);
EXPECT_EQ(kResidualEchoLikelihood,
stats.apm_statistics.residual_echo_likelihood);
EXPECT_EQ(kResidualEchoLikelihoodMax,
stats.apm_statistics.residual_echo_likelihood_recent_max);
EXPECT_FALSE(stats.typing_noise_detected);
}
TEST(AudioSendStreamTest, SendCodecAppliesAudioNetworkAdaptor) {
ConfigHelper helper(false, true);
helper.config().send_codec_spec =
AudioSendStream::Config::SendCodecSpec(0, kOpusFormat);
const std::string kAnaConfigString = "abcde";
const std::string kAnaReconfigString = "12345";
helper.config().audio_network_adaptor_config = kAnaConfigString;
EXPECT_CALL(helper.mock_encoder_factory(), MakeAudioEncoderMock(_, _, _, _))
.WillOnce(Invoke([&kAnaConfigString, &kAnaReconfigString](
int payload_type, const SdpAudioFormat& format,
absl::optional<AudioCodecPairId> codec_pair_id,
std::unique_ptr<AudioEncoder>* return_value) {
auto mock_encoder = SetupAudioEncoderMock(payload_type, format);
EXPECT_CALL(*mock_encoder,
EnableAudioNetworkAdaptor(StrEq(kAnaConfigString), _))
.WillOnce(Return(true));
EXPECT_CALL(*mock_encoder,
EnableAudioNetworkAdaptor(StrEq(kAnaReconfigString), _))
.WillOnce(Return(true));
*return_value = std::move(mock_encoder);
}));
auto send_stream = helper.CreateAudioSendStream();
auto stream_config = helper.config();
stream_config.audio_network_adaptor_config = kAnaReconfigString;
helper.SetupMockForCallEncoder();
send_stream->Reconfigure(stream_config);
}
// VAD is applied when codec is mono and the CNG frequency matches the codec
// clock rate.
TEST(AudioSendStreamTest, SendCodecCanApplyVad) {
ConfigHelper helper(false, false);
helper.config().send_codec_spec =
AudioSendStream::Config::SendCodecSpec(9, kG722Format);
helper.config().send_codec_spec->cng_payload_type = 105;
using ::testing::Invoke;
std::unique_ptr<AudioEncoder> stolen_encoder;
EXPECT_CALL(*helper.channel_send(), SetEncoderForMock(_, _))
.WillOnce(
Invoke([&stolen_encoder](int payload_type,
std::unique_ptr<AudioEncoder>* encoder) {
stolen_encoder = std::move(*encoder);
return true;
}));
EXPECT_CALL(*helper.channel_send(), RegisterCngPayloadType(105, 8000));
auto send_stream = helper.CreateAudioSendStream();
// We cannot truly determine if the encoder created is an AudioEncoderCng. It
// is the only reasonable implementation that will return something from
// ReclaimContainedEncoders, though.
ASSERT_TRUE(stolen_encoder);
EXPECT_FALSE(stolen_encoder->ReclaimContainedEncoders().empty());
}
TEST(AudioSendStreamTest, DoesNotPassHigherBitrateThanMaxBitrate) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
EXPECT_CALL(*helper.channel_send(),
OnBitrateAllocation(
Field(&BitrateAllocationUpdate::target_bitrate,
Eq(DataRate::bps(helper.config().max_bitrate_bps)))));
BitrateAllocationUpdate update;
update.target_bitrate = DataRate::bps(helper.config().max_bitrate_bps + 5000);
update.packet_loss_ratio = 0;
update.round_trip_time = TimeDelta::ms(50);
update.bwe_period = TimeDelta::ms(6000);
send_stream->OnBitrateUpdated(update);
}
TEST(AudioSendStreamTest, ProbingIntervalOnBitrateUpdated) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
EXPECT_CALL(*helper.channel_send(),
OnBitrateAllocation(Field(&BitrateAllocationUpdate::bwe_period,
Eq(TimeDelta::ms(5000)))));
BitrateAllocationUpdate update;
update.target_bitrate = DataRate::bps(helper.config().max_bitrate_bps + 5000);
update.packet_loss_ratio = 0;
update.round_trip_time = TimeDelta::ms(50);
update.bwe_period = TimeDelta::ms(5000);
send_stream->OnBitrateUpdated(update);
}
// Test that AudioSendStream doesn't recreate the encoder unnecessarily.
TEST(AudioSendStreamTest, DontRecreateEncoder) {
ConfigHelper helper(false, false);
// WillOnce is (currently) the default used by ConfigHelper if asked to set an
// expectation for SetEncoder. Since this behavior is essential for this test
// to be correct, it's instead set-up manually here. Otherwise a simple change
// to ConfigHelper (say to WillRepeatedly) would silently make this test
// useless.
EXPECT_CALL(*helper.channel_send(), SetEncoderForMock(_, _))
.WillOnce(Return());
EXPECT_CALL(*helper.channel_send(), RegisterCngPayloadType(105, 8000));
helper.config().send_codec_spec =
AudioSendStream::Config::SendCodecSpec(9, kG722Format);
helper.config().send_codec_spec->cng_payload_type = 105;
auto send_stream = helper.CreateAudioSendStream();
send_stream->Reconfigure(helper.config());
}
TEST(AudioSendStreamTest, ReconfigureTransportCcResetsFirst) {
ScopedFieldTrials field_trials("WebRTC-Audio-SendSideBwe/Enabled/");
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
auto new_config = helper.config();
ConfigHelper::AddBweToConfig(&new_config);
EXPECT_CALL(*helper.channel_send(),
EnableSendTransportSequenceNumber(kTransportSequenceNumberId))
.Times(1);
{
::testing::InSequence seq;
EXPECT_CALL(*helper.channel_send(), ResetSenderCongestionControlObjects())
.Times(1);
EXPECT_CALL(*helper.channel_send(), RegisterSenderCongestionControlObjects(
helper.transport(), Ne(nullptr)))
.Times(1);
}
// CallEncoder will be called to re-set overhead.
EXPECT_CALL(*helper.channel_send(), CallEncoder(testing::_)).Times(1);
send_stream->Reconfigure(new_config);
}
TEST(AudioSendStreamTest, OnTransportOverheadChanged) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
auto new_config = helper.config();
// CallEncoder will be called on overhead change.
EXPECT_CALL(*helper.channel_send(), CallEncoder(testing::_)).Times(1);
const size_t transport_overhead_per_packet_bytes = 333;
send_stream->SetTransportOverhead(transport_overhead_per_packet_bytes);
EXPECT_EQ(transport_overhead_per_packet_bytes,
send_stream->TestOnlyGetPerPacketOverheadBytes());
}
TEST(AudioSendStreamTest, OnAudioOverheadChanged) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
auto new_config = helper.config();
// CallEncoder will be called on overhead change.
EXPECT_CALL(*helper.channel_send(), CallEncoder(testing::_)).Times(1);
const size_t audio_overhead_per_packet_bytes = 555;
send_stream->OnOverheadChanged(audio_overhead_per_packet_bytes);
EXPECT_EQ(audio_overhead_per_packet_bytes,
send_stream->TestOnlyGetPerPacketOverheadBytes());
}
TEST(AudioSendStreamTest, OnAudioAndTransportOverheadChanged) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
auto new_config = helper.config();
// CallEncoder will be called when each of overhead changes.
EXPECT_CALL(*helper.channel_send(), CallEncoder(testing::_)).Times(2);
const size_t transport_overhead_per_packet_bytes = 333;
send_stream->SetTransportOverhead(transport_overhead_per_packet_bytes);
const size_t audio_overhead_per_packet_bytes = 555;
send_stream->OnOverheadChanged(audio_overhead_per_packet_bytes);
EXPECT_EQ(
transport_overhead_per_packet_bytes + audio_overhead_per_packet_bytes,
send_stream->TestOnlyGetPerPacketOverheadBytes());
}
// Validates that reconfiguring the AudioSendStream with a Frame encryptor
// correctly reconfigures on the object without crashing.
TEST(AudioSendStreamTest, ReconfigureWithFrameEncryptor) {
ConfigHelper helper(false, true);
auto send_stream = helper.CreateAudioSendStream();
auto new_config = helper.config();
rtc::scoped_refptr<FrameEncryptorInterface> mock_frame_encryptor_0(
new rtc::RefCountedObject<MockFrameEncryptor>());
new_config.frame_encryptor = mock_frame_encryptor_0;
EXPECT_CALL(*helper.channel_send(), SetFrameEncryptor(Ne(nullptr))).Times(1);
send_stream->Reconfigure(new_config);
// Not updating the frame encryptor shouldn't force it to reconfigure.
EXPECT_CALL(*helper.channel_send(), SetFrameEncryptor(_)).Times(0);
send_stream->Reconfigure(new_config);
// Updating frame encryptor to a new object should force a call to the proxy.
rtc::scoped_refptr<FrameEncryptorInterface> mock_frame_encryptor_1(
new rtc::RefCountedObject<MockFrameEncryptor>());
new_config.frame_encryptor = mock_frame_encryptor_1;
new_config.crypto_options.sframe.require_frame_encryption = true;
EXPECT_CALL(*helper.channel_send(), SetFrameEncryptor(Ne(nullptr))).Times(1);
send_stream->Reconfigure(new_config);
}
} // namespace test
} // namespace webrtc