Add support for handling reordered SS data on the receive-side for VP9.

webrtc/modules/video_coding/main/source/jitter_buffer_unittest.cc

 /*
  *  Copyright (c) 2011 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.h>
 
 #include <list>
 
 #include "testing/gtest/include/gtest/gtest.h"
 #include "webrtc/modules/video_coding/main/source/frame_buffer.h"
 #include "webrtc/modules/video_coding/main/source/jitter_buffer.h"
 #include "webrtc/modules/video_coding/main/source/media_opt_util.h"
 #include "webrtc/modules/video_coding/main/source/packet.h"
 #include "webrtc/modules/video_coding/main/source/test/stream_generator.h"
 #include "webrtc/modules/video_coding/main/test/test_util.h"
 #include "webrtc/system_wrappers/interface/clock.h"
 #include "webrtc/system_wrappers/interface/metrics.h"
 #include "webrtc/test/histogram.h"
 
 namespace webrtc {
 
+namespace {
+  const uint32_t kProcessIntervalSec = 60;
+}  // namespace
+
+class Vp9SsMapTest : public ::testing::Test {
+ protected:
+  virtual void SetUp() {
+    uint16_t seq_num = 1234;
+    uint32_t timestamp = 1;
+    int length = 1400;
+    packet_.reset(new VCMPacket(data_, length, seq_num, timestamp, true));
+    packet_->isFirstPacket = true;
+    packet_->markerBit = true;
+    packet_->frameType = kVideoFrameKey;
+    packet_->codec = kVideoCodecVP9;
+    packet_->codecSpecificHeader.codec = kRtpVideoVp9;
+    packet_->codecSpecificHeader.codecHeader.VP9.flexible_mode = false;
+    packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 0;
+    packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx = kNoTemporalIdx;
+    packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch = false;
+    packet_->codecSpecificHeader.codecHeader.VP9.ss_data_available = true;
+    packet_->codecSpecificHeader.codecHeader.VP9.gof.SetGofInfoVP9(
+        kTemporalStructureMode3);  // kTemporalStructureMode3: 0-2-1-2..
+  }
+
+  Vp9SsMap map_;
+  uint8_t data_[1500];
+  rtc::scoped_ptr<VCMPacket> packet_;

[stefan-webrtc, 2015/10/16 12:26:11]

No need for a scoped_ptr

[åsapersson, 2015/10/20 13:32:20]

Done.

+};
+
+TEST_F(Vp9SsMapTest, Insert) {
+  EXPECT_TRUE(map_.Insert(*packet_));
+}
+
+TEST_F(Vp9SsMapTest, Insert_NoSsData) {
+  packet_->codecSpecificHeader.codecHeader.VP9.ss_data_available = false;
+  EXPECT_FALSE(map_.Insert(*packet_));
+}
+
+TEST_F(Vp9SsMapTest, Find) {
+  EXPECT_TRUE(map_.Insert(*packet_));
+  Vp9SsMap::SsMap::const_iterator it;
+  EXPECT_TRUE(map_.Find(packet_->timestamp, &it));
+  EXPECT_EQ(packet_->timestamp, it->first);
+}
+
+TEST_F(Vp9SsMapTest, Find_WithWrap) {
+  const uint32_t kSsTimestamp1 = 0xFFFFFFFF;
+  const uint32_t kSsTimestamp2 = 100;
+  packet_->timestamp = kSsTimestamp1;
+  EXPECT_TRUE(map_.Insert(*packet_));
+  packet_->timestamp = kSsTimestamp2;
+  EXPECT_TRUE(map_.Insert(*packet_));
+  Vp9SsMap::SsMap::const_iterator it;
+  EXPECT_FALSE(map_.Find(kSsTimestamp1 - 1, &it));
+  EXPECT_TRUE(map_.Find(kSsTimestamp1, &it));
+  EXPECT_EQ(kSsTimestamp1, it->first);
+  EXPECT_TRUE(map_.Find(kSsTimestamp1 + 1, &it));
+  EXPECT_EQ(kSsTimestamp1, it->first);
+  EXPECT_TRUE(map_.Find(kSsTimestamp2 - 1, &it));
+  EXPECT_EQ(kSsTimestamp1, it->first);
+  EXPECT_TRUE(map_.Find(kSsTimestamp2, &it));
+  EXPECT_EQ(kSsTimestamp2, it->first);
+  EXPECT_TRUE(map_.Find(kSsTimestamp2 + 1, &it));
+  EXPECT_EQ(kSsTimestamp2, it->first);
+}
+
+TEST_F(Vp9SsMapTest, Reset) {
+  EXPECT_TRUE(map_.Insert(*packet_));
+  Vp9SsMap::SsMap::const_iterator it;
+  EXPECT_TRUE(map_.Find(packet_->timestamp, &it));
+  EXPECT_EQ(packet_->timestamp, it->first);
+
+  map_.Reset();
+  EXPECT_FALSE(map_.Find(packet_->timestamp, &it));
+}
+
+TEST_F(Vp9SsMapTest, RemoveOld) {
+  Vp9SsMap::SsMap::const_iterator it;
+  const uint32_t kSsTimestamp1 = 10000;
+  packet_->timestamp = kSsTimestamp1;
+  EXPECT_TRUE(map_.Insert(*packet_));
+
+  const uint32_t kTimestamp = kSsTimestamp1 + kProcessIntervalSec * 90000;
+  map_.RemoveOld(kTimestamp - 1);              // Interval not passed.
+  EXPECT_TRUE(map_.Find(kSsTimestamp1, &it));  // Should not been removed.
+
+  map_.RemoveOld(kTimestamp);
+  EXPECT_FALSE(map_.Find(kSsTimestamp1, &it));
+  EXPECT_TRUE(map_.Find(kTimestamp, &it));
+  EXPECT_EQ(kTimestamp, it->first);
+}
+
+TEST_F(Vp9SsMapTest, RemoveOld_WithWrap) {
+  Vp9SsMap::SsMap::const_iterator it;
+  const uint32_t kSsTimestamp1 = 0xFFFFFFFF - kProcessIntervalSec * 90000;
+  const uint32_t kSsTimestamp2 = 10;
+  const uint32_t kSsTimestamp3 = 1000;
+  packet_->timestamp = kSsTimestamp1;
+  EXPECT_TRUE(map_.Insert(*packet_));
+  packet_->timestamp = kSsTimestamp2;
+  EXPECT_TRUE(map_.Insert(*packet_));
+  packet_->timestamp = kSsTimestamp3;
+  EXPECT_TRUE(map_.Insert(*packet_));
+
+  map_.RemoveOld(kSsTimestamp3);
+  EXPECT_FALSE(map_.Find(kSsTimestamp1, &it));
+  EXPECT_FALSE(map_.Find(kSsTimestamp2, &it));
+  EXPECT_TRUE(map_.Find(kSsTimestamp3, &it));
+}
+
+TEST_F(Vp9SsMapTest, UpdatePacket_NoSsData) {
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 0;
+  EXPECT_FALSE(map_.UpdatePacket(packet_.get()));
+}
+
+TEST_F(Vp9SsMapTest, UpdatePacket_NoGofIdx) {
+  EXPECT_TRUE(map_.Insert(*packet_));
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = kNoGofIdx;
+  EXPECT_FALSE(map_.UpdatePacket(packet_.get()));
+}
+
+TEST_F(Vp9SsMapTest, UpdatePacket_InvalidGofIdx) {
+  EXPECT_TRUE(map_.Insert(*packet_));
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 4;
+  EXPECT_FALSE(map_.UpdatePacket(packet_.get()));
+}
+
+TEST_F(Vp9SsMapTest, UpdatePacket) {
+  EXPECT_TRUE(map_.Insert(*packet_));  // kTemporalStructureMode3: 0-2-1-2..
+
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 0;
+  EXPECT_TRUE(map_.UpdatePacket(packet_.get()));
+  EXPECT_EQ(0, packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx);
+  EXPECT_FALSE(packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch);
+  EXPECT_EQ(1U, packet_->codecSpecificHeader.codecHeader.VP9.num_ref_pics);
+  EXPECT_EQ(4, packet_->codecSpecificHeader.codecHeader.VP9.pid_diff[0]);
+
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 1;
+  EXPECT_TRUE(map_.UpdatePacket(packet_.get()));
+  EXPECT_EQ(2, packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx);
+  EXPECT_TRUE(packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch);
+  EXPECT_EQ(1U, packet_->codecSpecificHeader.codecHeader.VP9.num_ref_pics);
+  EXPECT_EQ(1, packet_->codecSpecificHeader.codecHeader.VP9.pid_diff[0]);
+
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 2;
+  EXPECT_TRUE(map_.UpdatePacket(packet_.get()));
+  EXPECT_EQ(1, packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx);
+  EXPECT_TRUE(packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch);
+  EXPECT_EQ(1U, packet_->codecSpecificHeader.codecHeader.VP9.num_ref_pics);
+  EXPECT_EQ(2, packet_->codecSpecificHeader.codecHeader.VP9.pid_diff[0]);
+
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 3;
+  EXPECT_TRUE(map_.UpdatePacket(packet_.get()));
+  EXPECT_EQ(2, packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx);
+  EXPECT_FALSE(packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch);
+  EXPECT_EQ(2U, packet_->codecSpecificHeader.codecHeader.VP9.num_ref_pics);
+  EXPECT_EQ(1, packet_->codecSpecificHeader.codecHeader.VP9.pid_diff[0]);
+  EXPECT_EQ(2, packet_->codecSpecificHeader.codecHeader.VP9.pid_diff[1]);
+}
+

[stefan-webrtc, 2015/10/16 12:26:11]

Maybe we should have a test with an incoming structure that we don't support on
the send-side? And maybe one with spatial layers? Might be a better fit for a
JitterBufferTest though?

[åsapersson, 2015/10/20 13:32:20]

Added test with spatial layers.

 class TestBasicJitterBuffer : public ::testing::Test {
  protected:
   virtual void SetUp() {
     clock_.reset(new SimulatedClock(0));
     jitter_buffer_.reset(new VCMJitterBuffer(
         clock_.get(),
         rtc::scoped_ptr<EventWrapper>(event_factory_.CreateEvent())));
     jitter_buffer_->Start();
     seq_num_ = 1234;
     timestamp_ = 0;
@@ skipping 662 matching lines @@
   // Should be able to decode 3 delta frames, key frame already decoded.
   for (size_t i = 0; i < 3; ++i) {
     frame_out = DecodeCompleteFrame();
     ASSERT_TRUE(frame_out != NULL);
     CheckOutFrame(frame_out, size_, false);
     EXPECT_EQ(kVideoFrameDelta, frame_out->FrameType());
     jitter_buffer_->ReleaseFrame(frame_out);
   }
 }
 
+TEST_F(TestBasicJitterBuffer, TestSkipForwardVp9) {
+  // Verify that JB skips forward to next base layer frame.
+  //  -------------------------------------------------
+  // | 65485 | 65486 | 65487 | 65488 | 65489 | ...
+  // | pid:5 | pid:6 | pid:7 | pid:8 | pid:9 | ...
+  // | tid:0 | tid:2 | tid:1 | tid:2 | tid:0 | ...
+  // |  ss   |   x   |   x   |   x   |       |
+  //  -------------------------------------------------
+  // |<----------tl0idx:200--------->|<---tl0idx:201---
+
+  bool re = false;
+  packet_->codec = kVideoCodecVP9;
+  packet_->codecSpecificHeader.codec = kRtpVideoVp9;
+  packet_->isFirstPacket = true;
+  packet_->markerBit = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.flexible_mode = false;
+  packet_->codecSpecificHeader.codecHeader.VP9.spatial_idx = 0;
+  packet_->codecSpecificHeader.codecHeader.VP9.beginning_of_frame = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.end_of_frame = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx = kNoTemporalIdx;
+  packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch = false;
+
+  packet_->seqNum = 65485;
+  packet_->timestamp = 1000;
+  packet_->frameType = kVideoFrameKey;
+  packet_->codecSpecificHeader.codecHeader.VP9.picture_id = 5;
+  packet_->codecSpecificHeader.codecHeader.VP9.tl0_pic_idx = 200;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 0;
+  packet_->codecSpecificHeader.codecHeader.VP9.ss_data_available = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof.SetGofInfoVP9(
+      kTemporalStructureMode3);  // kTemporalStructureMode3: 0-2-1-2..
+  EXPECT_EQ(kCompleteSession, jitter_buffer_->InsertPacket(*packet_, &re));
+
+  // Insert next temporal layer 0.
+  packet_->seqNum = 65489;
+  packet_->timestamp = 13000;
+  packet_->frameType = kVideoFrameDelta;
+  packet_->codecSpecificHeader.codecHeader.VP9.picture_id = 9;
+  packet_->codecSpecificHeader.codecHeader.VP9.tl0_pic_idx = 201;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 0;
+  packet_->codecSpecificHeader.codecHeader.VP9.ss_data_available = false;
+  EXPECT_EQ(kCompleteSession, jitter_buffer_->InsertPacket(*packet_, &re));
+
+  VCMEncodedFrame* frame_out = DecodeCompleteFrame();
+  EXPECT_EQ(1000U, frame_out->TimeStamp());
+  EXPECT_EQ(kVideoFrameKey, frame_out->FrameType());
+  jitter_buffer_->ReleaseFrame(frame_out);
+
+  frame_out = DecodeCompleteFrame();
+  EXPECT_EQ(13000U, frame_out->TimeStamp());
+  EXPECT_EQ(kVideoFrameDelta, frame_out->FrameType());
+  jitter_buffer_->ReleaseFrame(frame_out);
+}
+
+TEST_F(TestBasicJitterBuffer, ReorderedVp9SsData) {
+  // Verify that frames are updated with SS data when SS packet is reordered.
+  //  --------------------------------
+  // | 65486 | 65487 | 65485 |...
+  // | pid:6 | pid:7 | pid:5 |...
+  // | tid:2 | tid:1 | tid:0 |...
+  // |       |       |  ss   |
+  //  --------------------------------
+  // |<--------tl0idx:200--------->|
+
+  bool re = false;
+  packet_->codec = kVideoCodecVP9;
+  packet_->codecSpecificHeader.codec = kRtpVideoVp9;
+  packet_->isFirstPacket = true;
+  packet_->markerBit = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.flexible_mode = false;
+  packet_->codecSpecificHeader.codecHeader.VP9.spatial_idx = 0;
+  packet_->codecSpecificHeader.codecHeader.VP9.beginning_of_frame = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.end_of_frame = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.temporal_idx = kNoTemporalIdx;
+  packet_->codecSpecificHeader.codecHeader.VP9.temporal_up_switch = false;
+  packet_->codecSpecificHeader.codecHeader.VP9.tl0_pic_idx = 200;
+
+  packet_->seqNum = 65486;
+  packet_->timestamp = 6000;
+  packet_->frameType = kVideoFrameDelta;
+  packet_->codecSpecificHeader.codecHeader.VP9.picture_id = 6;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 1;
+  EXPECT_EQ(kCompleteSession, jitter_buffer_->InsertPacket(*packet_, &re));
+
+  packet_->seqNum = 65487;
+  packet_->timestamp = 9000;
+  packet_->frameType = kVideoFrameDelta;
+  packet_->codecSpecificHeader.codecHeader.VP9.picture_id = 7;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 2;
+  EXPECT_EQ(kCompleteSession, jitter_buffer_->InsertPacket(*packet_, &re));
+
+  // Insert first frame with SS data.
+  packet_->seqNum = 65485;
+  packet_->timestamp = 3000;
+  packet_->frameType = kVideoFrameKey;
+  packet_->width = 352;
+  packet_->height = 288;
+  packet_->codecSpecificHeader.codecHeader.VP9.picture_id = 5;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof_idx = 0;
+  packet_->codecSpecificHeader.codecHeader.VP9.ss_data_available = true;
+  packet_->codecSpecificHeader.codecHeader.VP9.gof.SetGofInfoVP9(
+      kTemporalStructureMode3);  // kTemporalStructureMode3: 0-2-1-2..
+  EXPECT_EQ(kCompleteSession, jitter_buffer_->InsertPacket(*packet_, &re));
+
+  VCMEncodedFrame* frame_out = DecodeCompleteFrame();
+  EXPECT_EQ(3000U, frame_out->TimeStamp());
+  EXPECT_EQ(kVideoFrameKey, frame_out->FrameType());
+  EXPECT_EQ(0, frame_out->CodecSpecific()->codecSpecific.VP9.temporal_idx);
+  EXPECT_FALSE(
+      frame_out->CodecSpecific()->codecSpecific.VP9.temporal_up_switch);
+  EXPECT_EQ(1U, packet_->codecSpecificHeader.codecHeader.VP9.num_ref_pics);
+  EXPECT_EQ(4, packet_->codecSpecificHeader.codecHeader.VP9.pid_diff[0]);
+  jitter_buffer_->ReleaseFrame(frame_out);
+
+  frame_out = DecodeCompleteFrame();
+  EXPECT_EQ(6000U, frame_out->TimeStamp());
+  EXPECT_EQ(kVideoFrameDelta, frame_out->FrameType());
+  EXPECT_EQ(2, frame_out->CodecSpecific()->codecSpecific.VP9.temporal_idx);
+  EXPECT_TRUE(frame_out->CodecSpecific()->codecSpecific.VP9.temporal_up_switch);
+  jitter_buffer_->ReleaseFrame(frame_out);
+
+  frame_out = DecodeCompleteFrame();
+  EXPECT_EQ(9000U, frame_out->TimeStamp());
+  EXPECT_EQ(kVideoFrameDelta, frame_out->FrameType());
+  EXPECT_EQ(1, frame_out->CodecSpecific()->codecSpecific.VP9.temporal_idx);
+  EXPECT_TRUE(frame_out->CodecSpecific()->codecSpecific.VP9.temporal_up_switch);
+  jitter_buffer_->ReleaseFrame(frame_out);
+}
+
 TEST_F(TestBasicJitterBuffer, H264InsertStartCode) {
   packet_->frameType = kVideoFrameKey;
   packet_->isFirstPacket = true;
   packet_->markerBit = false;
   packet_->seqNum = seq_num_;
   packet_->timestamp = timestamp_;
   packet_->insertStartCode = true;
 
   bool retransmitted = false;
   EXPECT_EQ(kIncomplete, jitter_buffer_->InsertPacket(*packet_,
@@ skipping 1478 matching lines @@
 
   // Stream should be decodable from this point.
   clock_->AdvanceTimeMilliseconds(kDefaultFramePeriodMs);
   InsertFrame(kVideoFrameDelta);
   EXPECT_TRUE(DecodeCompleteFrame());
   nack_list = jitter_buffer_->GetNackList(&extended);
   EXPECT_EQ(0u, nack_list.size());
 }
 
 }  // namespace webrtc