Support STYLE and REGION in WebVTT

Note that STYLE and REGION are not supported in mp4 container due to
spec limitation as 14496-30:2014 does not specify a way to signal
styles/regions inside mp4.

Closes #344.

Change-Id: I05c14df916f7b2c7ca4364ee9407e0eda4dc7a3f

docs/source/build_instructions.md

@@ -237,8 +237,8 @@ $ ninja -C out/Release
 Use `apk` command to install dependencies:
 
 ```shell
-$apk add --no-cache bash build-base curl findutils git ninja python \
-                    bsd-compat-headers linux-headers libexecinfo-dev
+$ apk add --no-cache bash build-base curl findutils git ninja python \
+                     bsd-compat-headers linux-headers libexecinfo-dev
 ```
 
 Alpine uses musl which does not have mallinfo defined in malloc.h. It is

packager/app/test/testdata/hls-audio-video-text-with-ad-cues/bear-subtitle-english-text-1.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:00.000 --> 00:00:00.800
 Yup, that's a bear, eh.
 

packager/app/test/testdata/hls-audio-video-text-with-ad-cues/bear-subtitle-english-text-2.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-audio-video-text-with-ad-cues/bear-subtitle-english-text-3.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-audio-video-text-with-ad-cues/bear-subtitle-english-text-4.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-audio-video-text-with-ad-cues/bear-subtitle-english-text-5.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-audio-video-text-with-ad-cues/bear-subtitle-english-text-6.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-segmented-webvtt/bear-subtitle-english-text-1.vtt

@@ -1,6 +1,9 @@
 WEBVTT
 X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:9000
 
+STYLE
+::cue { color:lime }
+
 00:00:00.000 --> 00:00:00.800
 Yup, that's a bear, eh.
 

packager/app/test/testdata/hls-segmented-webvtt/bear-subtitle-english-text-2.vtt

@@ -1,6 +1,9 @@
 WEBVTT
 X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:9000
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-segmented-webvtt/bear-subtitle-english-text-3.vtt

@@ -1,6 +1,9 @@
 WEBVTT
 X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:9000
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-segmented-webvtt/bear-subtitle-english-text-4.vtt

@@ -1,6 +1,9 @@
 WEBVTT
 X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:9000
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/app/test/testdata/hls-segmented-webvtt/bear-subtitle-english-text-5.vtt

@@ -1,6 +1,9 @@
 WEBVTT
 X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:9000
 
+STYLE
+::cue { color:lime }
+
 00:00:01.000 --> 00:00:04.700
 He 's... um... doing bear-like stuff.
 

packager/media/formats/mp4/mp4_muxer.cc

@@ -538,8 +538,19 @@ bool MP4Muxer::GenerateTextTrak(const TextStreamInfo* text_info,
     // Handle WebVTT.
     TextSampleEntry webvtt;
     webvtt.format = FOURCC_wvtt;
-    webvtt.config.config.assign(text_info->codec_config().begin(),
-                                text_info->codec_config().end());
+
+    // 14496-30:2014 7.5 Web Video Text Tracks Sample entry format.
+    // In the sample entry, a WebVTT configuration box must occur, carrying
+    // exactly the lines of the WebVTT file header, i.e. all text lines up to
+    // but excluding the 'two or more line terminators' that end the header.
+    webvtt.config.config = "WEBVTT";
+    // The spec does not define a way to carry STYLE and REGION information in
+    // the mp4 container.
+    if (!text_info->codec_config().empty()) {
+      LOG(INFO) << "Skipping possible style / region configuration as the spec "
+                   "does not define a way to carry them inside ISO-BMFF files.";
+    }
+
     // TODO(rkuroiwa): This should be the source file URI(s). Putting bogus
     // string for now so that the box will be there for samples with overlapping
     // cues.

packager/media/formats/webvtt/webvtt_file_buffer.cc

@@ -18,9 +18,11 @@ const int kTsTimescale = 90000;
 }
 
 WebVttFileBuffer::WebVttFileBuffer(
-    uint32_t transport_stream_timestamp_offset_ms)
+    uint32_t transport_stream_timestamp_offset_ms,
+    const std::string& style_region_config)
     : transport_stream_timestamp_offset_(transport_stream_timestamp_offset_ms *
-                                         kTsTimescale / 1000) {
+                                         kTsTimescale / 1000),
+      style_region_config_(style_region_config) {
   // Make sure we start with the same state that we would end up with if
   // the caller reset our state.
   Reset();
@@ -38,6 +40,10 @@ void WebVttFileBuffer::Reset() {
                         transport_stream_timestamp_offset_);
   }
   buffer_.append("\n");  // end of header.
+  if (!style_region_config_.empty()) {
+    buffer_.append(style_region_config_);
+    buffer_.append("\n\n");
+  }
 }
 
 void WebVttFileBuffer::Append(const TextSample& sample) {

packager/media/formats/webvtt/webvtt_file_buffer.h

@@ -20,7 +20,8 @@ class TextSample;
 // all the formatting requirements for a webvtt file.
 class WebVttFileBuffer {
  public:
-  explicit WebVttFileBuffer(uint32_t transport_stream_timestamp_offset_ms);
+  WebVttFileBuffer(uint32_t transport_stream_timestamp_offset_ms,
+                   const std::string& style_region_config);
   virtual ~WebVttFileBuffer() = default;
 
   void Reset();
@@ -36,6 +37,7 @@ class WebVttFileBuffer {
   WebVttFileBuffer& operator=(const WebVttFileBuffer&) = delete;
 
   const uint32_t transport_stream_timestamp_offset_ = 0;
+  const std::string style_region_config_;
   std::string buffer_;
   size_t sample_count_ = 0;
 };

packager/media/formats/webvtt/webvtt_parser.cc

@@ -14,6 +14,7 @@
 #include "packager/base/strings/string_util.h"
 #include "packager/media/base/text_stream_info.h"
 #include "packager/media/formats/webvtt/webvtt_timestamp.h"
+#include "packager/status_macros.h"
 
 namespace shaka {
 namespace media {
@@ -75,6 +76,13 @@ bool IsLikelyStyle(const std::string& line) {
 bool IsLikelyRegion(const std::string& line) {
   return base::TrimWhitespaceASCII(line, base::TRIM_TRAILING) == "REGION";
 }
+
+void UpdateConfig(const std::vector<std::string>& block, std::string* config) {
+  if (!config->empty())
+    *config += "\n\n";
+  *config += base::JoinString(block, "\n");
+}
+
 }  // namespace
 
 WebVttParser::WebVttParser(std::unique_ptr<FileReader> source,
@@ -121,14 +129,6 @@ bool WebVttParser::Parse() {
     return false;
   }
 
-  const Status send_stream_info_result = DispatchTextStreamInfo();
-
-  if (send_stream_info_result != Status::OK) {
-    LOG(ERROR) << "Failed to send stream info down stream:"
-               << send_stream_info_result.error_message();
-    return false;
-  }
-
   bool saw_cue = false;
 
   while (reader_.Next(&block) && keep_reading_) {
@@ -141,11 +141,10 @@ bool WebVttParser::Parse() {
     // STYLE
     if (IsLikelyStyle(block[0])) {
       if (saw_cue) {
-        LOG(ERROR)
+        LOG(WARNING)
             << "Found style block after seeing cue. Ignoring style block";
       } else {
-        LOG(WARNING) << "Missing support for style blocks. Skipping block:\n"
-                     << BlockToString(block.data(), block.size());
+        UpdateConfig(block, &style_region_config_);
       }
       continue;
     }
@@ -153,11 +152,10 @@ bool WebVttParser::Parse() {
     // REGION
     if (IsLikelyRegion(block[0])) {
       if (saw_cue) {
-        LOG(ERROR)
+        LOG(WARNING)
             << "Found region block after seeing cue. Ignoring region block";
       } else {
-        LOG(WARNING) << "Missing support for region blocks. Skipping block:\n"
-                     << BlockToString(block.data(), block.size());
+        UpdateConfig(block, &style_region_config_);
       }
       continue;
     }
@@ -223,6 +221,9 @@ Status WebVttParser::ParseCue(const std::string& id,
         "Could not parse start time, -->, and end time from " + block[0]);
   }
 
+  if (!stream_info_dispatched_)
+    RETURN_IF_ERROR(DispatchTextStreamInfo());
+
   // According to the WebVTT spec end time must be greater than the start time
   // of the cue. Since we are seeing content with invalid times in the field, we
   // are going to drop the cue instead of failing to package.
@@ -261,6 +262,8 @@ Status WebVttParser::ParseCue(const std::string& id,
 }
 
 Status WebVttParser::DispatchTextStreamInfo() {
+  stream_info_dispatched_ = true;
+
   const int kTrackId = 0;
   // The resolution of timings are in milliseconds.
   const int kTimescale = 1000;
@@ -269,13 +272,12 @@ Status WebVttParser::DispatchTextStreamInfo() {
   // work nicely with the current demuxer.
   const int kDuration = 0;
   const char kWebVttCodecString[] = "wvtt";
-  const char kCodecConfig[] = "";
   const int64_t kNoWidth = 0;
   const int64_t kNoHeight = 0;
 
   std::shared_ptr<StreamInfo> info = std::make_shared<TextStreamInfo>(
       kTrackId, kTimescale, kDuration, kCodecWebVtt, kWebVttCodecString,
-      kCodecConfig, kNoWidth, kNoHeight, language_);
+      style_region_config_, kNoWidth, kNoHeight, language_);
 
   return DispatchStreamInfo(kStreamIndex, std::move(info));
 }

packager/media/formats/webvtt/webvtt_parser.h

@@ -43,6 +43,8 @@ class WebVttParser : public OriginHandler {
 
   BlockReader reader_;
   std::string language_;
+  std::string style_region_config_;
+  bool stream_info_dispatched_ = false;
   bool keep_reading_ = true;
 };
 

packager/media/formats/webvtt/webvtt_parser_unittest.cc

@@ -14,6 +14,7 @@
 #include "packager/status_test_util.h"
 
 using ::testing::_;
+using ::testing::SaveArgPointee;
 
 namespace shaka {
 namespace media {
@@ -28,6 +29,10 @@ const bool kEncrypted = true;
 
 const char* kNoId = "";
 const char* kNoSettings = "";
+
+std::string ToString(const std::vector<uint8_t>& v) {
+  return std::string(v.begin(), v.end());
+}
 }  // namespace
 
 class WebVttParserTest : public MediaHandlerTestBase {
@@ -71,8 +76,7 @@ TEST_F(WebVttParserTest, ParseOnlyHeader) {
 
   {
     testing::InSequence s;
-    EXPECT_CALL(*Output(kOutputIndex),
-                OnProcess(IsStreamInfo(_, kTimeScale, !kEncrypted, kLanguage)));
+    EXPECT_CALL(*Output(kOutputIndex), OnProcess(_)).Times(0);
     EXPECT_CALL(*Output(kOutputIndex), OnFlush(_));
   }
 
@@ -88,8 +92,7 @@ TEST_F(WebVttParserTest, ParseHeaderWithBOM) {
 
   {
     testing::InSequence s;
-    EXPECT_CALL(*Output(kOutputIndex),
-                OnProcess(IsStreamInfo(_, kTimeScale, !kEncrypted, _)));
+    EXPECT_CALL(*Output(kOutputIndex), OnProcess(_)).Times(0);
     EXPECT_CALL(*Output(kOutputIndex), OnFlush(_));
   }
 
@@ -123,57 +126,7 @@ TEST_F(WebVttParserTest, FailToParseHeaderNotOneLine) {
   ASSERT_NE(Status::OK, parser_->Run());
 }
 
-// Right now we don't support region blocks, but for now make sure that we don't
-// die if we see a region block.
-TEST_F(WebVttParserTest, ParserDoesNotDieOnRegionBlock) {
-  const char* text =
-      "WEBVTT\n"
-      "\n"
-      "REGION\n"
-      "id:fred\n"
-      "width:40%\n"
-      "lines:3\n"
-      "regionanchor:0%,100%\n"
-      "viewportanchor:10%,90%\n"
-      "scroll:up";
-
-  ASSERT_NO_FATAL_FAILURE(SetUpAndInitializeGraph(text));
-
-  {
-    testing::InSequence s;
-    EXPECT_CALL(*Output(kOutputIndex),
-                OnProcess(IsStreamInfo(_, kTimeScale, !kEncrypted, _)));
-    EXPECT_CALL(*Output(kOutputIndex), OnFlush(_));
-  }
-
-  ASSERT_OK(parser_->Run());
-}
-
-// Right now we don't support style blocks, but for now make sure that we don't
-// die if we see a style block.
-TEST_F(WebVttParserTest, ParserDoesNotDieOnStyleBlock) {
-  const char* text =
-      "WEBVTT\n"
-      "\n"
-      "STYLE\n"
-      "::cue {\n"
-      "  background-image: linear-gradient(to bottom, dimgray, lightgray);\n"
-      "  color: papayawhip;\n"
-      "}";
-
-  ASSERT_NO_FATAL_FAILURE(SetUpAndInitializeGraph(text));
-
-  {
-    testing::InSequence s;
-    EXPECT_CALL(*Output(kOutputIndex),
-                OnProcess(IsStreamInfo(_, kTimeScale, !kEncrypted, _)));
-    EXPECT_CALL(*Output(kOutputIndex), OnFlush(_));
-  }
-
-  ASSERT_OK(parser_->Run());
-}
-
-TEST_F(WebVttParserTest, IngoresZeroDurationCues) {
+TEST_F(WebVttParserTest, IgnoresZeroDurationCues) {
   const char* text =
       "WEBVTT\n"
       "\n"
@@ -214,6 +167,44 @@ TEST_F(WebVttParserTest, ParseOneCue) {
   ASSERT_OK(parser_->Run());
 }
 
+TEST_F(WebVttParserTest, ParseOneCueWithStyleAndRegion) {
+  const char* text =
+      "WEBVTT\n"
+      "\n"
+      "STYLE\n"
+      "::cue { color:lime }\n"
+      "\n"
+      "REGION\n"
+      "id:scroll\n"
+      "scrol:up\n"
+      "\n"
+      "00:01:00.000 --> 01:00:00.000\n"
+      "subtitle\n";
+
+  ASSERT_NO_FATAL_FAILURE(SetUpAndInitializeGraph(text));
+
+  StreamData stream_data;
+  {
+    testing::InSequence s;
+    EXPECT_CALL(*Output(kOutputIndex),
+                OnProcess(IsStreamInfo(_, kTimeScale, !kEncrypted, _)))
+        .WillOnce(SaveArgPointee<0>(&stream_data));
+    EXPECT_CALL(*Output(kOutputIndex),
+                OnProcess(IsTextSample(_, kNoId, 60000u, 3600000u, kNoSettings,
+                                       "subtitle")));
+    EXPECT_CALL(*Output(kOutputIndex), OnFlush(_));
+  }
+
+  ASSERT_OK(parser_->Run());
+  EXPECT_EQ(ToString(stream_data.stream_info->codec_config()),
+            "STYLE\n"
+            "::cue { color:lime }\n"
+            "\n"
+            "REGION\n"
+            "id:scroll\n"
+            "scrol:up");
+}
+
 TEST_F(WebVttParserTest, ParseOneEmptyCue) {
   const char* text =
       "WEBVTT\n"

packager/media/formats/webvtt/webvtt_text_output_handler.cc

@@ -18,14 +18,17 @@ namespace shaka {
 namespace media {
 namespace {
 double kMillisecondsToSeconds = 1000.0;
+
+std::string ToString(const std::vector<uint8_t>& v) {
+  return std::string(v.begin(), v.end());
+}
 }  // namespace
 
 WebVttTextOutputHandler::WebVttTextOutputHandler(
     const MuxerOptions& muxer_options,
     std::unique_ptr<MuxerListener> muxer_listener)
     : muxer_options_(muxer_options),
-      muxer_listener_(std::move(muxer_listener)),
-      buffer_(muxer_options.transport_stream_timestamp_offset_ms) {}
+      muxer_listener_(std::move(muxer_listener)) {}
 
 Status WebVttTextOutputHandler::InitializeInternal() {
   return Status::OK;
@@ -50,7 +53,13 @@ Status WebVttTextOutputHandler::Process(
 }
 
 Status WebVttTextOutputHandler::OnFlushRequest(size_t input_stream_index) {
-  DCHECK_EQ(buffer_.sample_count(), 0u)
+  if (!buffer_) {
+    LOG(INFO) << "Skip stream '" << muxer_options_.segment_template
+              << "' which does not contain any sample.";
+    return Status::OK;
+  }
+
+  DCHECK_EQ(buffer_->sample_count(), 0u)
       << "There should have been a segment info before flushing that would "
          "have cleared out all the samples.";
 
@@ -64,6 +73,9 @@ Status WebVttTextOutputHandler::OnFlushRequest(size_t input_stream_index) {
 }
 
 Status WebVttTextOutputHandler::OnStreamInfo(const StreamInfo& info) {
+  buffer_.reset(
+      new WebVttFileBuffer(muxer_options_.transport_stream_timestamp_offset_ms,
+                           ToString(info.codec_config())));
   muxer_listener_->OnMediaStart(muxer_options_, info, info.time_scale(),
                                 MuxerListener::kContainerText);
   return Status::OK;
@@ -89,8 +101,8 @@ Status WebVttTextOutputHandler::OnSegmentInfo(const SegmentInfo& info) {
     return Status(error::FILE_FAILURE, "Failed to open " + filename);
   }
 
-  buffer_.WriteTo(file.get());
-  buffer_.Reset();
+  buffer_->WriteTo(file.get());
+  buffer_->Reset();
 
   if (!file.release()->Close()) {
     return Status(error::FILE_FAILURE, "Failed to close " + filename);
@@ -115,7 +127,7 @@ void WebVttTextOutputHandler::OnTextSample(const TextSample& sample) {
   // Skip empty samples. It is normal to see empty samples as earlier in the
   // pipeline we pad the stream to remove gaps.
   if (sample.payload().size()) {
-    buffer_.Append(sample);
+    buffer_->Append(sample);
   }
 }
 }  // namespace media

packager/media/formats/webvtt/webvtt_text_output_handler.h

@@ -49,7 +49,7 @@ class WebVttTextOutputHandler : public MediaHandler {
   uint64_t total_duration_ms_ = 0;
   uint32_t segment_index_ = 0;
 
-  WebVttFileBuffer buffer_;
+  std::unique_ptr<WebVttFileBuffer> buffer_;
 };
 
 }  // namespace media

packager/media/test/data/bear-subtitle-english.vtt

@@ -1,5 +1,8 @@
 WEBVTT
 
+STYLE
+::cue { color:lime }
+
 00:00:00.000 --> 00:00:00.800
 Yup, that's a bear, eh.