Bug 556821 - Use Ogg Skeleton presentation time for start of videos - r=cpearce

--HG--
extra : rebase_source : 83020ac5028e9f94cabe6a3c959cf84eeb25449f

content/media/ogg/nsOggCodecState.cpp

@@ -394,6 +394,7 @@ PRInt64 nsVorbisState::Time(vorbis_info* aInfo, PRInt64 aGranulepos)
 nsSkeletonState::nsSkeletonState(ogg_page* aBosPage)
   : nsOggCodecState(aBosPage),
     mVersion(0),
+    mPresentationTime(0),
     mLength(0)
 {
   MOZ_COUNT_CTOR(nsSkeletonState);
@@ -407,7 +408,8 @@ nsSkeletonState::~nsSkeletonState()
 // Support for Ogg Skeleton 4.0, as per specification at:
 // http://wiki.xiph.org/Ogg_Skeleton_4
 
-// Minimum length in bytes of a Skeleton 4.0 header packet.
+// Minimum length in bytes of a Skeleton header packet.
+#define SKELETON_MIN_HEADER_LEN 28
 #define SKELETON_4_0_MIN_HEADER_LEN 80
 
 // Minimum length in bytes of a Skeleton 4.0 index packet.
@@ -421,6 +423,10 @@ nsSkeletonState::~nsSkeletonState()
 #define SKELETON_VERSION_MAJOR_OFFSET 8
 #define SKELETON_VERSION_MINOR_OFFSET 10
 
+// Byte-offsets of the presentation time numerator and denominator
+#define SKELETON_PRESENTATION_TIME_NUMERATOR_OFFSET 12
+#define SKELETON_PRESENTATION_TIME_DENOMINATOR_OFFSET 20
+
 // Byte-offsets of the length of file field in the Skeleton 4.0 header packet.
 #define SKELETON_FILE_LENGTH_OFFSET 64
 
@@ -434,7 +440,7 @@ nsSkeletonState::~nsSkeletonState()
 
 static PRBool IsSkeletonBOS(ogg_packet* aPacket)
 {
-  return aPacket->bytes >= SKELETON_4_0_MIN_HEADER_LEN && 
+  return aPacket->bytes >= SKELETON_MIN_HEADER_LEN && 
          memcmp(reinterpret_cast<char*>(aPacket->packet), "fishead", 8) == 0;
 }
 
@@ -699,6 +705,13 @@ PRBool nsSkeletonState::DecodeHeader(ogg_packet* aPacket)
   if (IsSkeletonBOS(aPacket)) {
     PRUint16 verMajor = LEUint16(aPacket->packet + SKELETON_VERSION_MAJOR_OFFSET);
     PRUint16 verMinor = LEUint16(aPacket->packet + SKELETON_VERSION_MINOR_OFFSET);
+
+    // Read the presentation time. We read this before the version check as the
+    // presentation time exists in all versions.
+    PRInt64 n = LEInt64(aPacket->packet + SKELETON_PRESENTATION_TIME_NUMERATOR_OFFSET);
+    PRInt64 d = LEInt64(aPacket->packet + SKELETON_PRESENTATION_TIME_DENOMINATOR_OFFSET);
+    mPresentationTime = d == 0 ? 0 : (static_cast<float>(n) / static_cast<float>(d)) * 1000;
+
     mVersion = SKELETON_VERSION(verMajor, verMinor);
     if (mVersion < SKELETON_VERSION(4,0) ||
         mVersion >= SKELETON_VERSION(5,0) ||

content/media/ogg/nsOggCodecState.h

@@ -214,6 +214,9 @@ public:
   virtual PRInt64 Time(PRInt64 granulepos) { return -1; }
   virtual PRBool Init() { return PR_TRUE; }
 
+  // Return PR_TRUE if the given time (in milliseconds) is within
+  // the presentation time defined in the skeleton track.
+  PRBool IsPresentable(PRInt64 aTime) { return aTime >= mPresentationTime; }
 
   // Stores the offset of the page on which a keyframe starts,
   // and its presentation time.
@@ -283,6 +286,9 @@ private:
   // Version of the decoded skeleton track, as per the SKELETON_VERSION macro.
   PRUint32 mVersion;
 
+  // Presentation time of the resource in milliseconds
+  PRInt64 mPresentationTime;
+
   // Length of the resource in bytes.
   PRInt64 mLength;
 

content/media/ogg/nsOggReader.cpp

@@ -628,7 +628,7 @@ PRBool nsOggReader::DecodeVideoFrame(PRBool &aKeyframeSkip,
         // packets we read subsequently.
         mTheoraGranulepos = packet.granulepos;
       }
-    
+
       if (DecodeTheora(frames, &packet) == NS_ERROR_OUT_OF_MEMORY) {
         NS_WARNING("Theora decode memory allocation failure!");
         return PR_FALSE;
@@ -741,6 +741,7 @@ PRBool nsOggReader::DecodeVideoFrame(PRBool &aKeyframeSkip,
 
     PRInt64 time = mTheoraState->StartTime(mTheoraGranulepos);
     NS_ASSERTION(packet.granulepos != -1, "Must know packet granulepos");
+
     if (!aKeyframeSkip ||
         (th_packet_iskeyframe(&packet) == 1 && time >= aTimeThreshold))
     {
@@ -754,13 +755,17 @@ PRBool nsOggReader::DecodeVideoFrame(PRBool &aKeyframeSkip,
   // Push decoded data into the video frame queue.
   for (PRUint32 i = 0; i < frames.Length(); i++) {
     nsAutoPtr<VideoData> data(frames[i].forget());
-    if (aKeyframeSkip && data->mKeyframe) {
-      aKeyframeSkip = PR_FALSE;
-    }
-
-    if (!aKeyframeSkip && data->mEndTime >= aTimeThreshold) {
-      mVideoQueue.Push(data.forget());
-      decoded++;
+    // Don't use the frame if it's outside the bounds of the presentation
+    // start time in the skeleton track.
+    if (!mSkeletonState || mSkeletonState->IsPresentable(data->mTime)) {
+      if (aKeyframeSkip && data->mKeyframe) {
+        aKeyframeSkip = PR_FALSE;
+      }
+ 
+      if (!aKeyframeSkip && data->mEndTime >= aTimeThreshold) {
+        mVideoQueue.Push(data.forget());
+        decoded++;
+      }
     }
   }
 

content/media/test/Makefile.in

@@ -201,6 +201,7 @@ _TEST_FILES += \
 		bug523816.ogv \
 		bug533822.ogg \
 		bug557094.ogv \
+		bug556821.ogv \
 		bug580982.webm \
 		bug603918.webm \
 		bug604067.webm \

content/media/test/manifest.js

@@ -74,6 +74,8 @@ var gPlayTests = [
   { name:"bug504613.ogv", type:"video/ogg", duration:Number.NaN },
   // Multiple audio streams.
   { name:"bug516323.ogv", type:"video/ogg", duration:4.208 },
+  // oggz-chop with non-keyframe as first frame
+  { name:"bug556821.ogv", type:"video/ogg", duration:2.551 },
 
   // Encoded with vorbis beta1, includes unusually sized codebooks
   { name:"beta-phrasebook.ogg", type:"audio/ogg", duration:4.01 },