webrtc_m130/modules/video_coding/packet_buffer_unittest.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1036 lines
38 KiB
C++
Raw Normal View History

/*
* Copyright (c) 2016 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 "modules/video_coding/packet_buffer.h"
#include <cstring>
#include <limits>
#include <ostream>
#include <string>
#include <utility>
#include "api/array_view.h"
#include "common_video/h264/h264_common.h"
#include "modules/video_coding/frame_object.h"
#include "rtc_base/random.h"
#include "system_wrappers/include/clock.h"
#include "test/field_trial.h"
#include "test/gmock.h"
#include "test/gtest.h"
namespace webrtc {
namespace video_coding {
namespace {
using ::testing::ElementsAre;
using ::testing::ElementsAreArray;
using ::testing::IsEmpty;
using ::testing::Matches;
using ::testing::Pointee;
using ::testing::SizeIs;
constexpr int kStartSize = 16;
constexpr int kMaxSize = 64;
void IgnoreResult(PacketBuffer::InsertResult /*result*/) {}
std::vector<uint16_t> StartSeqNums(
rtc::ArrayView<const std::unique_ptr<RtpFrameObject>> frames) {
std::vector<uint16_t> result;
for (const auto& frame : frames) {
result.push_back(frame->first_seq_num());
}
return result;
}
MATCHER_P(StartSeqNumsAre, seq_num, "") {
return Matches(ElementsAre(seq_num))(StartSeqNums(arg.frames));
}
MATCHER_P2(StartSeqNumsAre, seq_num1, seq_num2, "") {
return Matches(ElementsAre(seq_num1, seq_num2))(StartSeqNums(arg.frames));
}
MATCHER(KeyFrame, "") {
return arg->frame_type() == VideoFrameType::kVideoFrameKey;
}
MATCHER(DeltaFrame, "") {
return arg->frame_type() == VideoFrameType::kVideoFrameDelta;
}
struct PacketBufferInsertResult : public PacketBuffer::InsertResult {
explicit PacketBufferInsertResult(PacketBuffer::InsertResult result)
: InsertResult(std::move(result)) {}
};
void PrintTo(const PacketBufferInsertResult& result, std::ostream* os) {
*os << "frames: { ";
for (size_t i = 0; i < result.frames.size(); ++i) {
const RtpFrameObject& frame = *result.frames[i];
if (i > 0) {
*os << ", ";
}
*os << "{sn: ";
if (frame.first_seq_num() == frame.last_seq_num()) {
*os << frame.first_seq_num();
} else {
*os << "[" << frame.first_seq_num() << "-" << frame.last_seq_num() << "]";
}
*os << "}";
}
*os << " }";
if (result.buffer_cleared) {
*os << ", buffer_cleared";
}
}
class PacketBufferTest : public ::testing::Test {
protected:
explicit PacketBufferTest(std::string field_trials = "")
: scoped_field_trials_(field_trials),
rand_(0x7732213),
clock_(new SimulatedClock(0)),
packet_buffer_(clock_.get(), kStartSize, kMaxSize) {}
uint16_t Rand() { return rand_.Rand<uint16_t>(); }
enum IsKeyFrame { kKeyFrame, kDeltaFrame };
enum IsFirst { kFirst, kNotFirst };
enum IsLast { kLast, kNotLast };
PacketBufferInsertResult Insert(uint16_t seq_num, // packet sequence number
IsKeyFrame keyframe, // is keyframe
IsFirst first, // is first packet of frame
IsLast last, // is last packet of frame
int data_size = 0, // size of data
uint8_t* data = nullptr, // data pointer
uint32_t timestamp = 123u) { // rtp timestamp
PacketBuffer::Packet packet;
packet.video_header.codec = kVideoCodecGeneric;
packet.timestamp = timestamp;
packet.seq_num = seq_num;
packet.video_header.frame_type = keyframe == kKeyFrame
? VideoFrameType::kVideoFrameKey
: VideoFrameType::kVideoFrameDelta;
packet.video_header.is_first_packet_in_frame = first == kFirst;
packet.video_header.is_last_packet_in_frame = last == kLast;
packet.size_bytes = data_size;
packet.data = data;
return PacketBufferInsertResult(packet_buffer_.InsertPacket(&packet));
}
const test::ScopedFieldTrials scoped_field_trials_;
Random rand_;
std::unique_ptr<SimulatedClock> clock_;
PacketBuffer packet_buffer_;
};
TEST_F(PacketBufferTest, InsertOnePacket) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kLast).frames, SizeIs(1));
}
TEST_F(PacketBufferTest, InsertMultiplePackets) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kLast).frames, SizeIs(1));
EXPECT_THAT(Insert(seq_num + 1, kKeyFrame, kFirst, kLast).frames, SizeIs(1));
EXPECT_THAT(Insert(seq_num + 2, kKeyFrame, kFirst, kLast).frames, SizeIs(1));
EXPECT_THAT(Insert(seq_num + 3, kKeyFrame, kFirst, kLast).frames, SizeIs(1));
}
TEST_F(PacketBufferTest, InsertDuplicatePacket) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(seq_num + 1, kKeyFrame, kNotFirst, kLast).frames,
SizeIs(1));
}
TEST_F(PacketBufferTest, SeqNumWrapOneFrame) {
Insert(0xFFFF, kKeyFrame, kFirst, kNotLast);
EXPECT_THAT(Insert(0x0, kKeyFrame, kNotFirst, kLast),
StartSeqNumsAre(0xFFFF));
}
TEST_F(PacketBufferTest, SeqNumWrapTwoFrames) {
EXPECT_THAT(Insert(0xFFFF, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(0xFFFF));
EXPECT_THAT(Insert(0x0, kKeyFrame, kFirst, kLast), StartSeqNumsAre(0x0));
}
TEST_F(PacketBufferTest, InsertOldPackets) {
EXPECT_THAT(Insert(100, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(102, kDeltaFrame, kFirst, kLast).frames, SizeIs(1));
EXPECT_THAT(Insert(101, kKeyFrame, kNotFirst, kLast).frames, SizeIs(1));
EXPECT_THAT(Insert(100, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(100, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(102, kDeltaFrame, kFirst, kLast).frames, SizeIs(1));
packet_buffer_.ClearTo(102);
EXPECT_THAT(Insert(102, kDeltaFrame, kFirst, kLast).frames, IsEmpty());
EXPECT_THAT(Insert(103, kDeltaFrame, kFirst, kLast).frames, SizeIs(1));
}
TEST_F(PacketBufferTest, NackCount) {
const uint16_t seq_num = Rand();
PacketBuffer::Packet packet;
packet.video_header.codec = kVideoCodecGeneric;
packet.seq_num = seq_num;
packet.video_header.frame_type = VideoFrameType::kVideoFrameKey;
packet.video_header.is_first_packet_in_frame = true;
packet.video_header.is_last_packet_in_frame = false;
packet.times_nacked = 0;
IgnoreResult(packet_buffer_.InsertPacket(&packet));
packet.seq_num++;
packet.video_header.is_first_packet_in_frame = false;
packet.times_nacked = 1;
IgnoreResult(packet_buffer_.InsertPacket(&packet));
packet.seq_num++;
packet.times_nacked = 3;
IgnoreResult(packet_buffer_.InsertPacket(&packet));
packet.seq_num++;
packet.video_header.is_last_packet_in_frame = true;
packet.times_nacked = 1;
auto frames = packet_buffer_.InsertPacket(&packet).frames;
ASSERT_THAT(frames, SizeIs(1));
EXPECT_EQ(frames.front()->times_nacked(), 3);
}
TEST_F(PacketBufferTest, FrameSize) {
const uint16_t seq_num = Rand();
uint8_t* data1 = new uint8_t[5]();
uint8_t* data2 = new uint8_t[5]();
uint8_t* data3 = new uint8_t[5]();
uint8_t* data4 = new uint8_t[5]();
Insert(seq_num, kKeyFrame, kFirst, kNotLast, 5, data1);
Insert(seq_num + 1, kKeyFrame, kNotFirst, kNotLast, 5, data2);
Insert(seq_num + 2, kKeyFrame, kNotFirst, kNotLast, 5, data3);
EXPECT_THAT(Insert(seq_num + 3, kKeyFrame, kNotFirst, kLast, 5, data4).frames,
ElementsAre(Pointee(SizeIs(20))));
}
TEST_F(PacketBufferTest, ExpandBuffer) {
const uint16_t seq_num = Rand();
Insert(seq_num, kKeyFrame, kFirst, kNotLast);
for (int i = 1; i < kStartSize; ++i)
EXPECT_FALSE(
Insert(seq_num + i, kKeyFrame, kNotFirst, kNotLast).buffer_cleared);
// Already inserted kStartSize number of packets, inserting the last packet
// should increase the buffer size and also result in an assembled frame.
EXPECT_FALSE(
Insert(seq_num + kStartSize, kKeyFrame, kNotFirst, kLast).buffer_cleared);
}
TEST_F(PacketBufferTest, SingleFrameExpandsBuffer) {
const uint16_t seq_num = Rand();
Insert(seq_num, kKeyFrame, kFirst, kNotLast);
for (int i = 1; i < kStartSize; ++i)
Insert(seq_num + i, kKeyFrame, kNotFirst, kNotLast);
EXPECT_THAT(Insert(seq_num + kStartSize, kKeyFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num));
}
TEST_F(PacketBufferTest, ExpandBufferOverflow) {
const uint16_t seq_num = Rand();
EXPECT_FALSE(Insert(seq_num, kKeyFrame, kFirst, kNotLast).buffer_cleared);
for (int i = 1; i < kMaxSize; ++i)
EXPECT_FALSE(
Insert(seq_num + i, kKeyFrame, kNotFirst, kNotLast).buffer_cleared);
// Already inserted kMaxSize number of packets, inserting the last packet
// should overflow the buffer and result in false being returned.
EXPECT_TRUE(
Insert(seq_num + kMaxSize, kKeyFrame, kNotFirst, kLast).buffer_cleared);
}
TEST_F(PacketBufferTest, OnePacketOneFrame) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(seq_num));
}
TEST_F(PacketBufferTest, TwoPacketsTwoFrames) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(seq_num));
EXPECT_THAT(Insert(seq_num + 1, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 1));
}
TEST_F(PacketBufferTest, TwoPacketsOneFrames) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(seq_num + 1, kKeyFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num));
}
TEST_F(PacketBufferTest, ThreePacketReorderingOneFrame) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(seq_num + 2, kKeyFrame, kNotFirst, kLast).frames,
IsEmpty());
EXPECT_THAT(Insert(seq_num + 1, kKeyFrame, kNotFirst, kNotLast),
StartSeqNumsAre(seq_num));
}
TEST_F(PacketBufferTest, Frames) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(seq_num));
EXPECT_THAT(Insert(seq_num + 1, kDeltaFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 1));
EXPECT_THAT(Insert(seq_num + 2, kDeltaFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 2));
EXPECT_THAT(Insert(seq_num + 3, kDeltaFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 3));
}
TEST_F(PacketBufferTest, ClearSinglePacket) {
const uint16_t seq_num = Rand();
for (int i = 0; i < kMaxSize; ++i)
Insert(seq_num + i, kDeltaFrame, kFirst, kLast);
packet_buffer_.ClearTo(seq_num);
EXPECT_FALSE(
Insert(seq_num + kMaxSize, kDeltaFrame, kFirst, kLast).buffer_cleared);
}
TEST_F(PacketBufferTest, ClearFullBuffer) {
for (int i = 0; i < kMaxSize; ++i)
Insert(i, kDeltaFrame, kFirst, kLast);
packet_buffer_.ClearTo(kMaxSize - 1);
for (int i = kMaxSize; i < 2 * kMaxSize; ++i)
EXPECT_FALSE(Insert(i, kDeltaFrame, kFirst, kLast).buffer_cleared);
}
TEST_F(PacketBufferTest, DontClearNewerPacket) {
EXPECT_THAT(Insert(0, kKeyFrame, kFirst, kLast), StartSeqNumsAre(0));
packet_buffer_.ClearTo(0);
EXPECT_THAT(Insert(2 * kStartSize, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(2 * kStartSize));
EXPECT_THAT(Insert(3 * kStartSize + 1, kKeyFrame, kFirst, kNotLast).frames,
IsEmpty());
packet_buffer_.ClearTo(2 * kStartSize);
EXPECT_THAT(Insert(3 * kStartSize + 2, kKeyFrame, kNotFirst, kLast),
StartSeqNumsAre(3 * kStartSize + 1));
}
TEST_F(PacketBufferTest, OneIncompleteFrame) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num, kDeltaFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(seq_num + 1, kDeltaFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num));
EXPECT_THAT(Insert(seq_num - 1, kDeltaFrame, kNotFirst, kLast).frames,
IsEmpty());
}
TEST_F(PacketBufferTest, TwoIncompleteFramesFullBuffer) {
const uint16_t seq_num = Rand();
for (int i = 1; i < kMaxSize - 1; ++i)
Insert(seq_num + i, kDeltaFrame, kNotFirst, kNotLast);
EXPECT_THAT(Insert(seq_num, kDeltaFrame, kFirst, kNotLast).frames, IsEmpty());
EXPECT_THAT(Insert(seq_num - 1, kDeltaFrame, kNotFirst, kLast).frames,
IsEmpty());
}
TEST_F(PacketBufferTest, FramesReordered) {
const uint16_t seq_num = Rand();
EXPECT_THAT(Insert(seq_num + 1, kDeltaFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 1));
EXPECT_THAT(Insert(seq_num, kKeyFrame, kFirst, kLast),
StartSeqNumsAre(seq_num));
EXPECT_THAT(Insert(seq_num + 3, kDeltaFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 3));
EXPECT_THAT(Insert(seq_num + 2, kDeltaFrame, kFirst, kLast),
StartSeqNumsAre(seq_num + 2));
}
TEST_F(PacketBufferTest, GetBitstream) {
// "many bitstream, such data" with null termination.
uint8_t many_data[] = {0x6d, 0x61, 0x6e, 0x79, 0x20};
uint8_t bitstream_data[] = {0x62, 0x69, 0x74, 0x73, 0x74, 0x72,
0x65, 0x61, 0x6d, 0x2c, 0x20};
uint8_t such_data[] = {0x73, 0x75, 0x63, 0x68, 0x20};
uint8_t data_data[] = {0x64, 0x61, 0x74, 0x61, 0x0};
uint8_t* many = new uint8_t[sizeof(many_data)];
uint8_t* bitstream = new uint8_t[sizeof(bitstream_data)];
uint8_t* such = new uint8_t[sizeof(such_data)];
uint8_t* data = new uint8_t[sizeof(data_data)];
memcpy(many, many_data, sizeof(many_data));
memcpy(bitstream, bitstream_data, sizeof(bitstream_data));
memcpy(such, such_data, sizeof(such_data));
memcpy(data, data_data, sizeof(data_data));
const uint16_t seq_num = Rand();
Insert(seq_num, kKeyFrame, kFirst, kNotLast, sizeof(many_data), many);
Insert(seq_num + 1, kDeltaFrame, kNotFirst, kNotLast, sizeof(bitstream_data),
bitstream);
Insert(seq_num + 2, kDeltaFrame, kNotFirst, kNotLast, sizeof(such_data),
such);
auto frames = Insert(seq_num + 3, kDeltaFrame, kNotFirst, kLast,
sizeof(data_data), data)
.frames;
ASSERT_THAT(frames, SizeIs(1));
EXPECT_EQ(frames[0]->first_seq_num(), seq_num);
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray("many bitstream, such data"));
}
TEST_F(PacketBufferTest, GetBitstreamOneFrameOnePacket) {
uint8_t bitstream_data[] = "All the bitstream data for this frame!";
uint8_t* data = new uint8_t[sizeof(bitstream_data)];
memcpy(data, bitstream_data, sizeof(bitstream_data));
auto frames =
Insert(0, kKeyFrame, kFirst, kLast, sizeof(bitstream_data), data).frames;
ASSERT_THAT(StartSeqNums(frames), ElementsAre(0));
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray(bitstream_data));
}
TEST_F(PacketBufferTest, GetBitstreamOneFrameFullBuffer) {
uint8_t* data_arr[kStartSize];
uint8_t expected[kStartSize];
for (uint8_t i = 0; i < kStartSize; ++i) {
data_arr[i] = new uint8_t[1];
data_arr[i][0] = i;
expected[i] = i;
}
Insert(0, kKeyFrame, kFirst, kNotLast, 1, data_arr[0]);
for (uint8_t i = 1; i < kStartSize - 1; ++i)
Insert(i, kKeyFrame, kNotFirst, kNotLast, 1, data_arr[i]);
auto frames = Insert(kStartSize - 1, kKeyFrame, kNotFirst, kLast, 1,
data_arr[kStartSize - 1])
.frames;
ASSERT_THAT(StartSeqNums(frames), ElementsAre(0));
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray(expected));
}
TEST_F(PacketBufferTest, GetBitstreamAv1) {
const uint8_t data1[] = {0b01'01'0000, 0b0'0100'000, 'm', 'a', 'n', 'y', ' '};
const uint8_t data2[] = {0b10'01'0000, 'b', 'i', 't', 's', 0};
uint8_t* new_data1 = new uint8_t[sizeof(data1)];
memcpy(new_data1, data1, sizeof(data1));
uint8_t* new_data2 = new uint8_t[sizeof(data2)];
memcpy(new_data2, data2, sizeof(data2));
PacketBuffer::Packet packet1;
packet1.video_header.codec = kVideoCodecAV1;
packet1.seq_num = 13;
packet1.video_header.is_first_packet_in_frame = true;
packet1.video_header.is_last_packet_in_frame = false;
packet1.size_bytes = sizeof(data1);
packet1.data = new_data1;
auto frames = packet_buffer_.InsertPacket(&packet1).frames;
EXPECT_THAT(frames, IsEmpty());
PacketBuffer::Packet packet2;
packet2.video_header.codec = kVideoCodecAV1;
packet2.seq_num = 14;
packet2.video_header.is_first_packet_in_frame = false;
packet2.video_header.is_last_packet_in_frame = true;
packet2.size_bytes = sizeof(data2);
packet2.data = new_data2;
frames = packet_buffer_.InsertPacket(&packet2).frames;
ASSERT_THAT(frames, SizeIs(1));
EXPECT_EQ(frames[0]->first_seq_num(), 13);
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), 2),
ElementsAre(0b0'0100'010, 10)); // obu_header and obu_size.
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data() + 2, frames[0]->size() - 2),
ElementsAreArray("many bits"));
}
TEST_F(PacketBufferTest, GetBitstreamInvalidAv1) {
// Two av1 payloads that can't be combined into proper frame.
const uint8_t data1[] = {0b01'01'0000, 0b0'0100'000, 'm', 'a', 'n', 'y', ' '};
const uint8_t data2[] = {0b00'01'0000, 'b', 'i', 't', 's', 0};
uint8_t* new_data1 = new uint8_t[sizeof(data1)];
memcpy(new_data1, data1, sizeof(data1));
uint8_t* new_data2 = new uint8_t[sizeof(data2)];
memcpy(new_data2, data2, sizeof(data2));
PacketBuffer::Packet packet1;
packet1.video_header.codec = kVideoCodecAV1;
packet1.seq_num = 13;
packet1.video_header.is_first_packet_in_frame = true;
packet1.video_header.is_last_packet_in_frame = false;
packet1.size_bytes = sizeof(data1);
packet1.data = new_data1;
auto frames = packet_buffer_.InsertPacket(&packet1).frames;
EXPECT_THAT(frames, IsEmpty());
PacketBuffer::Packet packet2;
packet2.video_header.codec = kVideoCodecAV1;
packet2.seq_num = 14;
packet2.video_header.is_first_packet_in_frame = false;
packet2.video_header.is_last_packet_in_frame = true;
packet2.size_bytes = sizeof(data2);
packet2.data = new_data2;
frames = packet_buffer_.InsertPacket(&packet2).frames;
EXPECT_THAT(frames, IsEmpty());
}
TEST_F(PacketBufferTest, InsertPacketAfterSequenceNumberWrapAround) {
uint16_t kFirstSeqNum = 0;
uint32_t kTimestampDelta = 100;
uint32_t timestamp = 10000;
uint16_t seq_num = kFirstSeqNum;
// Loop until seq_num wraps around.
Reland "SeqNumUnwrapper::Unwrap now returns int64_t instead of uint64_t." This reverts commit b5207b488b035eae4d11dfdcca9526d5a70c9c09. Reason for revert: DecodedFramesHistory has now been updated. Original change's description: > Revert "SeqNumUnwrapper::Unwrap now returns int64_t instead of uint64_t." > > This reverts commit b0f968a761b715da4cf81e4b9c3cab0ccd322cf2. > > Reason for revert: Need to update DecodedFramesHistory to manage negative picture IDs. > > Original change's description: > > SeqNumUnwrapper::Unwrap now returns int64_t instead of uint64_t. > > > > Bug: webrtc:10263 > > Change-Id: Idaeae6be01bd4eba0691226c958d70e114161ffd > > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/127295 > > Commit-Queue: Philip Eliasson <philipel@webrtc.org> > > Reviewed-by: Johannes Kron <kron@webrtc.org> > > Reviewed-by: Karl Wiberg <kwiberg@webrtc.org> > > Cr-Commit-Position: refs/heads/master@{#27129} > > TBR=kwiberg@webrtc.org,eladalon@webrtc.org,terelius@webrtc.org,philipel@webrtc.org,kron@webrtc.org > > Change-Id: I529bb0475bd21a80fa244278aff1fd912a85c169 > No-Presubmit: true > No-Tree-Checks: true > No-Try: true > Bug: webrtc:10263 > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/127885 > Reviewed-by: Philip Eliasson <philipel@webrtc.org> > Commit-Queue: Philip Eliasson <philipel@webrtc.org> > Cr-Commit-Position: refs/heads/master@{#27135} TBR=kwiberg@webrtc.org,eladalon@webrtc.org,terelius@webrtc.org,philipel@webrtc.org,kron@webrtc.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: webrtc:10263 Change-Id: Id59e377010b5070dd37a7ece8df79b23af43835a Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/128568 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Commit-Queue: Philip Eliasson <philipel@webrtc.org> Cr-Commit-Position: refs/heads/master@{#27191}
2019-03-19 12:15:00 +00:00
SeqNumUnwrapper<uint16_t> unwrapper;
while (unwrapper.Unwrap(seq_num) < std::numeric_limits<uint16_t>::max()) {
Insert(seq_num++, kKeyFrame, kFirst, kNotLast, 0, nullptr, timestamp);
for (int i = 0; i < 5; ++i) {
Insert(seq_num++, kKeyFrame, kNotFirst, kNotLast, 0, nullptr, timestamp);
}
Insert(seq_num++, kKeyFrame, kNotFirst, kLast, 0, nullptr, timestamp);
timestamp += kTimestampDelta;
}
// Receive frame with overlapping sequence numbers.
Insert(seq_num++, kKeyFrame, kFirst, kNotLast, 0, nullptr, timestamp);
for (int i = 0; i < 5; ++i) {
Insert(seq_num++, kKeyFrame, kNotFirst, kNotLast, 0, nullptr, timestamp);
}
EXPECT_THAT(
Insert(seq_num++, kKeyFrame, kNotFirst, kLast, 0, nullptr, timestamp)
.frames,
SizeIs(1));
}
// If |sps_pps_idr_is_keyframe| is true, we require keyframes to contain
// SPS/PPS/IDR and the keyframes we create as part of the test do contain
// SPS/PPS/IDR. If |sps_pps_idr_is_keyframe| is false, we only require and
// create keyframes containing only IDR.
class PacketBufferH264Test : public PacketBufferTest {
protected:
explicit PacketBufferH264Test(bool sps_pps_idr_is_keyframe)
: PacketBufferTest(sps_pps_idr_is_keyframe
? "WebRTC-SpsPpsIdrIsH264Keyframe/Enabled/"
: ""),
sps_pps_idr_is_keyframe_(sps_pps_idr_is_keyframe) {}
PacketBufferInsertResult InsertH264(
uint16_t seq_num, // packet sequence number
IsKeyFrame keyframe, // is keyframe
IsFirst first, // is first packet of frame
IsLast last, // is last packet of frame
uint32_t timestamp, // rtp timestamp
int data_size = 0, // size of data
uint8_t* data = nullptr, // data pointer
uint32_t width = 0, // width of frame (SPS/IDR)
uint32_t height = 0) { // height of frame (SPS/IDR)
PacketBuffer::Packet packet;
packet.video_header.codec = kVideoCodecH264;
auto& h264_header =
packet.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
packet.seq_num = seq_num;
packet.timestamp = timestamp;
if (keyframe == kKeyFrame) {
if (sps_pps_idr_is_keyframe_) {
h264_header.nalus[0].type = H264::NaluType::kSps;
h264_header.nalus[1].type = H264::NaluType::kPps;
h264_header.nalus[2].type = H264::NaluType::kIdr;
h264_header.nalus_length = 3;
} else {
h264_header.nalus[0].type = H264::NaluType::kIdr;
h264_header.nalus_length = 1;
}
}
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
packet.video_header.width = width;
packet.video_header.height = height;
packet.video_header.is_first_packet_in_frame = first == kFirst;
packet.video_header.is_last_packet_in_frame = last == kLast;
packet.size_bytes = data_size;
packet.data = data;
return PacketBufferInsertResult(packet_buffer_.InsertPacket(&packet));
}
PacketBufferInsertResult InsertH264KeyFrameWithAud(
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
uint16_t seq_num, // packet sequence number
IsKeyFrame keyframe, // is keyframe
IsFirst first, // is first packet of frame
IsLast last, // is last packet of frame
uint32_t timestamp, // rtp timestamp
int data_size = 0, // size of data
uint8_t* data = nullptr, // data pointer
uint32_t width = 0, // width of frame (SPS/IDR)
uint32_t height = 0) { // height of frame (SPS/IDR)
PacketBuffer::Packet packet;
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
packet.video_header.codec = kVideoCodecH264;
auto& h264_header =
packet.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
packet.seq_num = seq_num;
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
packet.timestamp = timestamp;
// this should be the start of frame.
RTC_CHECK(first == kFirst);
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
// Insert a AUD NALU / packet without width/height.
h264_header.nalus[0].type = H264::NaluType::kAud;
h264_header.nalus_length = 1;
packet.video_header.is_first_packet_in_frame = true;
packet.video_header.is_last_packet_in_frame = false;
packet.size_bytes = 0;
packet.data = nullptr;
IgnoreResult(packet_buffer_.InsertPacket(&packet));
// insert IDR
return InsertH264(seq_num + 1, keyframe, kNotFirst, last, timestamp,
data_size, data, width, height);
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
}
const bool sps_pps_idr_is_keyframe_;
};
// This fixture is used to test the general behaviour of the packet buffer
// in both configurations.
class PacketBufferH264ParameterizedTest
: public ::testing::WithParamInterface<bool>,
public PacketBufferH264Test {
protected:
PacketBufferH264ParameterizedTest() : PacketBufferH264Test(GetParam()) {}
};
INSTANTIATE_TEST_SUITE_P(SpsPpsIdrIsKeyframe,
PacketBufferH264ParameterizedTest,
::testing::Bool());
TEST_P(PacketBufferH264ParameterizedTest, DontRemoveMissingPacketOnClearTo) {
InsertH264(0, kKeyFrame, kFirst, kLast, 0);
InsertH264(2, kDeltaFrame, kFirst, kNotLast, 2);
packet_buffer_.ClearTo(0);
// Expect no frame because of missing of packet #1
EXPECT_THAT(InsertH264(3, kDeltaFrame, kNotFirst, kLast, 2).frames,
IsEmpty());
}
TEST_P(PacketBufferH264ParameterizedTest, GetBitstreamOneFrameFullBuffer) {
uint8_t* data_arr[kStartSize];
uint8_t expected[kStartSize];
for (uint8_t i = 0; i < kStartSize; ++i) {
data_arr[i] = new uint8_t[1];
data_arr[i][0] = i;
expected[i] = i;
}
InsertH264(0, kKeyFrame, kFirst, kNotLast, 1, 1, data_arr[0]);
for (uint8_t i = 1; i < kStartSize - 1; ++i) {
InsertH264(i, kKeyFrame, kNotFirst, kNotLast, 1, 1, data_arr[i]);
}
auto frames = InsertH264(kStartSize - 1, kKeyFrame, kNotFirst, kLast, 1, 1,
data_arr[kStartSize - 1])
.frames;
ASSERT_THAT(StartSeqNums(frames), ElementsAre(0));
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray(expected));
}
TEST_P(PacketBufferH264ParameterizedTest, GetBitstreamBufferPadding) {
uint16_t seq_num = Rand();
uint8_t data_data[] = "some plain old data";
uint8_t* data = new uint8_t[sizeof(data_data)];
memcpy(data, data_data, sizeof(data_data));
PacketBuffer::Packet packet;
auto& h264_header =
packet.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus_length = 1;
h264_header.nalus[0].type = H264::NaluType::kIdr;
h264_header.packetization_type = kH264SingleNalu;
packet.seq_num = seq_num;
packet.video_header.codec = kVideoCodecH264;
packet.data = data;
packet.size_bytes = sizeof(data_data);
packet.video_header.is_first_packet_in_frame = true;
packet.video_header.is_last_packet_in_frame = true;
auto frames = packet_buffer_.InsertPacket(&packet).frames;
ASSERT_THAT(frames, SizeIs(1));
EXPECT_EQ(frames[0]->first_seq_num(), seq_num);
EXPECT_EQ(frames[0]->EncodedImage().size(), sizeof(data_data));
EXPECT_EQ(frames[0]->EncodedImage().capacity(), sizeof(data_data));
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray(data_data));
}
TEST_P(PacketBufferH264ParameterizedTest, FrameResolution) {
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
uint16_t seq_num = 100;
uint8_t data_data[] = "some plain old data";
uint8_t* data = new uint8_t[sizeof(data_data)];
memcpy(data, data_data, sizeof(data_data));
uint32_t width = 640;
uint32_t height = 360;
uint32_t timestamp = 1000;
auto frames = InsertH264(seq_num, kKeyFrame, kFirst, kLast, timestamp,
sizeof(data_data), data, width, height)
.frames;
ASSERT_THAT(frames, SizeIs(1));
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray(data_data));
EXPECT_EQ(frames[0]->EncodedImage()._encodedWidth, width);
EXPECT_EQ(frames[0]->EncodedImage()._encodedHeight, height);
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
}
TEST_P(PacketBufferH264ParameterizedTest, FrameResolutionNaluBeforeSPS) {
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
uint16_t seq_num = 100;
uint8_t data_data[] = "some plain old data";
uint8_t* data = new uint8_t[sizeof(data_data)];
memcpy(data, data_data, sizeof(data_data));
uint32_t width = 640;
uint32_t height = 360;
uint32_t timestamp = 1000;
auto frames =
InsertH264KeyFrameWithAud(seq_num, kKeyFrame, kFirst, kLast, timestamp,
sizeof(data_data), data, width, height)
.frames;
ASSERT_THAT(StartSeqNums(frames), ElementsAre(seq_num));
EXPECT_EQ(frames[0]->EncodedImage().size(), sizeof(data_data));
EXPECT_EQ(frames[0]->EncodedImage().capacity(), sizeof(data_data));
EXPECT_EQ(frames[0]->EncodedImage()._encodedWidth, width);
EXPECT_EQ(frames[0]->EncodedImage()._encodedHeight, height);
EXPECT_THAT(rtc::MakeArrayView(frames[0]->data(), frames[0]->size()),
ElementsAreArray(data_data));
Width and Height was not associated and provided to decoder for H264 streams which have Nalus before SPS Summary: There is an issue with WebRTC for handling of certain H.264 bitstreams where the packets forming the H.264 stream has non-zero packets before the packet containing SPS. Typically a IDR (key frame) will have SPS/PPS (if present) or the IDR slice in the first packet. But this is not required in all cases, for example when packetization-mode = 0, you can have each NALU in separate packet. And certain NALUs can exist before SPS, for example SEI, AUD. The way WebRTC associates width/height to encoded frames is by tracking the dependency of IDR slices to SPS/PPS. RTP packets containing SPS/PPS have correct width/height stored in them during parsing of SPS in RtpDepacketizerH264::ProcessStapAOrSingleNalu IDR packets refer to SPS using ppsid, spsid and the width/height fields get transferred from packet containing SPS to IDR packet in H264SpsPpsTracker::CopyAndFixBitstream. When packets are assembled into a single encoded H264 frame in PacketBuffer::FindFrames, the loop goes through all the packets/nalus in backward scan from last RTP packet of IDR to first one. Hence the order of NALUs during this scan is : Last parts of IDR Slice -> Mid parts of IDR Slice RTP packet -> first IDR slice Packet (this should have correct width / height) -> RTP packet containing SPS/PPS (this should have correct width/height) start_index points to the first RTP packet of the frame and its passed into RtpFrameObject's constructor. RtpFrameObject will use the width/height stored in first RTP packet. This works fine as long as the first RTP packet has width/height, which will be the case if first RTP packet is IDR or SPS. In H.264 first RTP packet may be AUD, SEI in those cases, RtpFrameObject will create IDR with width/height = 0 and this causes problem for Android hardware decoders. On Android hardware decoders rely on correct width/height to initialize the hardware decoder. Verified on real scenario that we have. Simulated on AppRTCMobile on IOS Simulator Added unit tests : ninja -C out/Default && ./out/Default/modules_unittests --gtest_filter=*FrameResolution* Bug: webrtc:11025 Change-Id: Ie2273aae5e81fd62497e1add084876a3aa05af4d Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156260 Reviewed-by: Philip Eliasson <philipel@webrtc.org> Reviewed-by: Sergey Silkin <ssilkin@webrtc.org> Commit-Queue: Shyam Sadhwani <shyamsadhwani@fb.com> Cr-Commit-Position: refs/heads/master@{#29515}
2019-10-16 09:13:38 -07:00
}
TEST_F(PacketBufferTest, FreeSlotsOnFrameCreation) {
const uint16_t seq_num = Rand();
Insert(seq_num, kKeyFrame, kFirst, kNotLast);
Insert(seq_num + 1, kDeltaFrame, kNotFirst, kNotLast);
EXPECT_THAT(Insert(seq_num + 2, kDeltaFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num));
// Insert frame that fills the whole buffer.
Insert(seq_num + 3, kKeyFrame, kFirst, kNotLast);
for (int i = 0; i < kMaxSize - 2; ++i)
Insert(seq_num + i + 4, kDeltaFrame, kNotFirst, kNotLast);
EXPECT_THAT(Insert(seq_num + kMaxSize + 2, kKeyFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num + 3));
}
TEST_F(PacketBufferTest, Clear) {
const uint16_t seq_num = Rand();
Insert(seq_num, kKeyFrame, kFirst, kNotLast);
Insert(seq_num + 1, kDeltaFrame, kNotFirst, kNotLast);
EXPECT_THAT(Insert(seq_num + 2, kDeltaFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num));
packet_buffer_.Clear();
Insert(seq_num + kStartSize, kKeyFrame, kFirst, kNotLast);
Insert(seq_num + kStartSize + 1, kDeltaFrame, kNotFirst, kNotLast);
EXPECT_THAT(Insert(seq_num + kStartSize + 2, kDeltaFrame, kNotFirst, kLast),
StartSeqNumsAre(seq_num + kStartSize));
}
TEST_F(PacketBufferTest, FramesAfterClear) {
Insert(9025, kDeltaFrame, kFirst, kLast);
Insert(9024, kKeyFrame, kFirst, kLast);
packet_buffer_.ClearTo(9025);
EXPECT_THAT(Insert(9057, kDeltaFrame, kFirst, kLast).frames, SizeIs(1));
EXPECT_THAT(Insert(9026, kDeltaFrame, kFirst, kLast).frames, SizeIs(1));
}
TEST_F(PacketBufferTest, SameFrameDifferentTimestamps) {
Insert(0, kKeyFrame, kFirst, kNotLast, 0, nullptr, 1000);
EXPECT_THAT(Insert(1, kKeyFrame, kNotFirst, kLast, 0, nullptr, 1001).frames,
IsEmpty());
}
TEST_F(PacketBufferTest, DontLeakPayloadData) {
// NOTE! Any eventual leak is suppose to be detected by valgrind
// or any other similar tool.
uint8_t* data1 = new uint8_t[5];
uint8_t* data2 = new uint8_t[5];
uint8_t* data3 = new uint8_t[5];
uint8_t* data4 = new uint8_t[5];
// Expected to free data1 upon PacketBuffer destruction.
Insert(2, kKeyFrame, kFirst, kNotLast, 5, data1);
// Expect to free data2 upon insertion.
Insert(2, kKeyFrame, kFirst, kNotLast, 5, data2);
// Expect to free data3 upon insertion (old packet).
packet_buffer_.ClearTo(1);
Insert(1, kKeyFrame, kFirst, kNotLast, 5, data3);
// Expect to free data4 upon insertion (packet buffer is full).
Insert(2 + kMaxSize, kKeyFrame, kFirst, kNotLast, 5, data4);
}
TEST_F(PacketBufferTest, ContinuousSeqNumDoubleMarkerBit) {
Insert(2, kKeyFrame, kNotFirst, kNotLast);
Insert(1, kKeyFrame, kFirst, kLast);
EXPECT_THAT(Insert(3, kKeyFrame, kNotFirst, kLast).frames, IsEmpty());
}
TEST_F(PacketBufferTest, PacketTimestamps) {
absl::optional<int64_t> packet_ms;
absl::optional<int64_t> packet_keyframe_ms;
packet_ms = packet_buffer_.LastReceivedPacketMs();
packet_keyframe_ms = packet_buffer_.LastReceivedKeyframePacketMs();
EXPECT_FALSE(packet_ms);
EXPECT_FALSE(packet_keyframe_ms);
int64_t keyframe_ms = clock_->TimeInMilliseconds();
Insert(100, kKeyFrame, kFirst, kLast);
packet_ms = packet_buffer_.LastReceivedPacketMs();
packet_keyframe_ms = packet_buffer_.LastReceivedKeyframePacketMs();
EXPECT_TRUE(packet_ms);
EXPECT_TRUE(packet_keyframe_ms);
EXPECT_EQ(keyframe_ms, *packet_ms);
EXPECT_EQ(keyframe_ms, *packet_keyframe_ms);
clock_->AdvanceTimeMilliseconds(100);
int64_t delta_ms = clock_->TimeInMilliseconds();
Insert(101, kDeltaFrame, kFirst, kLast);
packet_ms = packet_buffer_.LastReceivedPacketMs();
packet_keyframe_ms = packet_buffer_.LastReceivedKeyframePacketMs();
EXPECT_TRUE(packet_ms);
EXPECT_TRUE(packet_keyframe_ms);
EXPECT_EQ(delta_ms, *packet_ms);
EXPECT_EQ(keyframe_ms, *packet_keyframe_ms);
packet_buffer_.Clear();
packet_ms = packet_buffer_.LastReceivedPacketMs();
packet_keyframe_ms = packet_buffer_.LastReceivedKeyframePacketMs();
EXPECT_FALSE(packet_ms);
EXPECT_FALSE(packet_keyframe_ms);
}
TEST_F(PacketBufferTest, IncomingCodecChange) {
PacketBuffer::Packet packet;
packet.video_header.is_first_packet_in_frame = true;
packet.video_header.is_last_packet_in_frame = true;
packet.size_bytes = 0;
packet.data = nullptr;
packet.video_header.codec = kVideoCodecVP8;
packet.video_header.video_type_header.emplace<RTPVideoHeaderVP8>();
packet.timestamp = 1;
packet.seq_num = 1;
packet.video_header.frame_type = VideoFrameType::kVideoFrameKey;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet).frames, SizeIs(1));
packet.video_header.codec = kVideoCodecH264;
auto& h264_header =
packet.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus_length = 1;
packet.timestamp = 3;
packet.seq_num = 3;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet).frames, IsEmpty());
packet.video_header.codec = kVideoCodecVP8;
packet.video_header.video_type_header.emplace<RTPVideoHeaderVP8>();
packet.timestamp = 2;
packet.seq_num = 2;
packet.video_header.frame_type = VideoFrameType::kVideoFrameDelta;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet).frames, SizeIs(2));
}
TEST_F(PacketBufferTest, TooManyNalusInPacket) {
PacketBuffer::Packet packet;
packet.video_header.codec = kVideoCodecH264;
packet.timestamp = 1;
packet.seq_num = 1;
packet.video_header.frame_type = VideoFrameType::kVideoFrameKey;
packet.video_header.is_first_packet_in_frame = true;
packet.video_header.is_last_packet_in_frame = true;
auto& h264_header =
packet.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus_length = kMaxNalusPerPacket;
packet.size_bytes = 0;
packet.data = nullptr;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet).frames, IsEmpty());
}
TEST_P(PacketBufferH264ParameterizedTest, OneFrameFillBuffer) {
InsertH264(0, kKeyFrame, kFirst, kNotLast, 1000);
for (int i = 1; i < kStartSize - 1; ++i)
InsertH264(i, kKeyFrame, kNotFirst, kNotLast, 1000);
EXPECT_THAT(InsertH264(kStartSize - 1, kKeyFrame, kNotFirst, kLast, 1000),
StartSeqNumsAre(0));
}
TEST_P(PacketBufferH264ParameterizedTest, CreateFramesAfterFilledBuffer) {
EXPECT_THAT(InsertH264(kStartSize - 2, kKeyFrame, kFirst, kLast, 0).frames,
SizeIs(1));
InsertH264(kStartSize, kDeltaFrame, kFirst, kNotLast, 2000);
for (int i = 1; i < kStartSize; ++i)
InsertH264(kStartSize + i, kDeltaFrame, kNotFirst, kNotLast, 2000);
EXPECT_THAT(
InsertH264(kStartSize + kStartSize, kDeltaFrame, kNotFirst, kLast, 2000)
.frames,
IsEmpty());
EXPECT_THAT(InsertH264(kStartSize - 1, kKeyFrame, kFirst, kLast, 1000),
StartSeqNumsAre(kStartSize - 1, kStartSize));
}
TEST_P(PacketBufferH264ParameterizedTest, OneFrameMaxSeqNum) {
InsertH264(65534, kKeyFrame, kFirst, kNotLast, 1000);
EXPECT_THAT(InsertH264(65535, kKeyFrame, kNotFirst, kLast, 1000),
StartSeqNumsAre(65534));
}
TEST_P(PacketBufferH264ParameterizedTest, ClearMissingPacketsOnKeyframe) {
InsertH264(0, kKeyFrame, kFirst, kLast, 1000);
InsertH264(2, kKeyFrame, kFirst, kLast, 3000);
InsertH264(3, kDeltaFrame, kFirst, kNotLast, 4000);
InsertH264(4, kDeltaFrame, kNotFirst, kLast, 4000);
EXPECT_THAT(InsertH264(kStartSize + 1, kKeyFrame, kFirst, kLast, 18000),
StartSeqNumsAre(kStartSize + 1));
}
TEST_P(PacketBufferH264ParameterizedTest, FindFramesOnPadding) {
EXPECT_THAT(InsertH264(0, kKeyFrame, kFirst, kLast, 1000),
StartSeqNumsAre(0));
EXPECT_THAT(InsertH264(2, kDeltaFrame, kFirst, kLast, 1000).frames,
IsEmpty());
EXPECT_THAT(packet_buffer_.InsertPadding(1), StartSeqNumsAre(2));
}
class PacketBufferH264XIsKeyframeTest : public PacketBufferH264Test {
protected:
const uint16_t kSeqNum = 5;
explicit PacketBufferH264XIsKeyframeTest(bool sps_pps_idr_is_keyframe)
: PacketBufferH264Test(sps_pps_idr_is_keyframe) {
packet_.video_header.codec = kVideoCodecH264;
packet_.seq_num = kSeqNum;
packet_.video_header.is_first_packet_in_frame = true;
packet_.video_header.is_last_packet_in_frame = true;
}
PacketBuffer::Packet packet_;
};
class PacketBufferH264IdrIsKeyframeTest
: public PacketBufferH264XIsKeyframeTest {
protected:
PacketBufferH264IdrIsKeyframeTest()
: PacketBufferH264XIsKeyframeTest(false) {}
};
TEST_F(PacketBufferH264IdrIsKeyframeTest, IdrIsKeyframe) {
auto& h264_header =
packet_.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus[0].type = H264::NaluType::kIdr;
h264_header.nalus_length = 1;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet_).frames,
ElementsAre(KeyFrame()));
}
TEST_F(PacketBufferH264IdrIsKeyframeTest, SpsPpsIdrIsKeyframe) {
auto& h264_header =
packet_.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus[0].type = H264::NaluType::kSps;
h264_header.nalus[1].type = H264::NaluType::kPps;
h264_header.nalus[2].type = H264::NaluType::kIdr;
h264_header.nalus_length = 3;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet_).frames,
ElementsAre(KeyFrame()));
}
class PacketBufferH264SpsPpsIdrIsKeyframeTest
: public PacketBufferH264XIsKeyframeTest {
protected:
PacketBufferH264SpsPpsIdrIsKeyframeTest()
: PacketBufferH264XIsKeyframeTest(true) {}
};
TEST_F(PacketBufferH264SpsPpsIdrIsKeyframeTest, IdrIsNotKeyframe) {
auto& h264_header =
packet_.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus[0].type = H264::NaluType::kIdr;
h264_header.nalus_length = 1;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet_).frames,
ElementsAre(DeltaFrame()));
}
TEST_F(PacketBufferH264SpsPpsIdrIsKeyframeTest, SpsPpsIsNotKeyframe) {
auto& h264_header =
packet_.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus[0].type = H264::NaluType::kSps;
h264_header.nalus[1].type = H264::NaluType::kPps;
h264_header.nalus_length = 2;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet_).frames,
ElementsAre(DeltaFrame()));
}
TEST_F(PacketBufferH264SpsPpsIdrIsKeyframeTest, SpsPpsIdrIsKeyframe) {
auto& h264_header =
packet_.video_header.video_type_header.emplace<RTPVideoHeaderH264>();
h264_header.nalus[0].type = H264::NaluType::kSps;
h264_header.nalus[1].type = H264::NaluType::kPps;
h264_header.nalus[2].type = H264::NaluType::kIdr;
h264_header.nalus_length = 3;
EXPECT_THAT(packet_buffer_.InsertPacket(&packet_).frames,
ElementsAre(KeyFrame()));
}
} // namespace
} // namespace video_coding
} // namespace webrtc