Changed es_parser_h264 to wait for initial I-frame to start emitting samples.

Change-Id: I7b9ca4a73a8765501a1c29e467e877e1eee5f9d9

media/formats/mp2t/es_parser_h264.cc

@@ -39,7 +39,8 @@ EsParserH264::EsParserH264(
       next_access_unit_pos_(0),
       stream_converter_(new H264ByteToUnitStreamConverter),
       decoder_config_check_pending_(false),
-      pending_sample_duration_(0) {
+      pending_sample_duration_(0),
+      waiting_for_key_frame_(true) {
 }
 
 EsParserH264::~EsParserH264() {
@@ -102,6 +103,7 @@ void EsParserH264::Reset() {
   decoder_config_check_pending_ = false;
   pending_sample_ = scoped_refptr<MediaSample>();
   pending_sample_duration_ = 0;
+  waiting_for_key_frame_ = true;
 }
 
 bool EsParserH264::FindAUD(int64* stream_pos) {
@@ -233,9 +235,14 @@ bool EsParserH264::ParseInternal() {
     }
   }
 
-  // Emit a frame and move the stream to the next AUD position.
-  RCHECK(EmitFrame(current_access_unit_pos_, access_unit_size,
-                   is_key_frame, pps_id_for_access_unit));
+  if (waiting_for_key_frame_) {
+    waiting_for_key_frame_ = !is_key_frame;
+  }
+  if (!waiting_for_key_frame_) {
+    // Emit a frame and move the stream to the next AUD position.
+    RCHECK(EmitFrame(current_access_unit_pos_, access_unit_size,
+                     is_key_frame, pps_id_for_access_unit));
+  }
   current_access_unit_pos_ = next_access_unit_pos_;
   es_queue_->Trim(current_access_unit_pos_);
 

media/formats/mp2t/es_parser_h264.h

@@ -94,6 +94,9 @@ class EsParserH264 : public EsParser {
   // Frame for which we do not yet have a duration.
   scoped_refptr<MediaSample> pending_sample_;
   uint64 pending_sample_duration_;
+
+  // Indicates whether waiting for first key frame.
+  bool waiting_for_key_frame_;
 };
 
 }  // namespace mp2t

media/formats/mp2t/es_parser_h264_unittest.cc

@@ -117,20 +117,24 @@ void AppendAUD(
 
 class EsParserH264Test : public testing::Test {
  public:
-  EsParserH264Test() : sample_count_(0) {
-  }
+  EsParserH264Test()
+      : sample_count_(0),
+        first_frame_is_key_frame_(false) {}
 
   void LoadStream(const char* filename);
   void ProcessPesPackets(const std::vector<Packet>& pes_packets);
 
   void EmitSample(uint32 pid, scoped_refptr<MediaSample>& sample) {
     sample_count_++;
+    if (sample_count_ == 1)
+      first_frame_is_key_frame_ = sample->is_key_frame();
   }
 
   void NewVideoConfig(scoped_refptr<StreamInfo>& config) {
   }
 
   size_t sample_count() const { return sample_count_; }
+  bool first_frame_is_key_frame() { return first_frame_is_key_frame_; }
 
   // Stream with AUD NALUs.
   std::vector<uint8> stream_;
@@ -140,6 +144,7 @@ class EsParserH264Test : public testing::Test {
 
  protected:
   size_t sample_count_;
+  bool first_frame_is_key_frame_;
 };
 
 void EsParserH264Test::LoadStream(const char* filename) {
@@ -204,7 +209,8 @@ TEST_F(EsParserH264Test, OneAccessUnitPerPes) {
 
   // Process each PES packet.
   ProcessPesPackets(pes_packets);
-  ASSERT_EQ(sample_count(), access_units_.size());
+  EXPECT_EQ(sample_count(), access_units_.size());
+  EXPECT_TRUE(first_frame_is_key_frame());
 }
 
 TEST_F(EsParserH264Test, NonAlignedPesPacket) {
@@ -228,7 +234,8 @@ TEST_F(EsParserH264Test, NonAlignedPesPacket) {
 
   // Process each PES packet.
   ProcessPesPackets(pes_packets);
-  ASSERT_EQ(sample_count(), access_units_.size());
+  EXPECT_EQ(sample_count(), access_units_.size());
+  EXPECT_TRUE(first_frame_is_key_frame());
 }
 
 TEST_F(EsParserH264Test, SeveralPesPerAccessUnit) {
@@ -258,7 +265,21 @@ TEST_F(EsParserH264Test, SeveralPesPerAccessUnit) {
 
   // Process each PES packet.
   ProcessPesPackets(pes_packets);
-  ASSERT_EQ(sample_count(), access_units_.size());
+  EXPECT_EQ(sample_count(), access_units_.size());
+  EXPECT_TRUE(first_frame_is_key_frame());
+}
+
+TEST_F(EsParserH264Test, NonIFrameStart) {
+  LoadStream("bear_no_iframe_start.h264");
+
+  // One to one equivalence between PES packets and access units.
+  std::vector<Packet> pes_packets(access_units_);
+
+  // Process each PES packet.
+  ProcessPesPackets(pes_packets);
+  // Ensure samples were emitted, but fewer than number of AUDs.
+  EXPECT_LT(sample_count(), access_units_.size());
+  EXPECT_TRUE(first_frame_is_key_frame());
 }
 
 }  // namespace mp2t

media/test/data/README

@@ -75,5 +75,8 @@ bear.h264:
   ffmpeg -i bear.mp4 -vcodec copy -vbsf h264_mp4toannexb \
       -an bear.h264
 
+bear_no_i_frame_start.h264
+   Derived from bear.h264. Consists of 29-non-I-frames, followed by a single I-frame, and 29 non-I-frames.
+
 avc-byte-stream-frame.h264 - Single IDR frame extracted from test-25fps.h264 in Annex B byte stream format.
 avc-unit-stream-frame.h264 - Single IDR frame from avc-byte-stream-frame.h264 converted to unit stream format.