Bug 1112438 - Make newCurrentFrameTime more accurate while seeking by checking audio & video timestamp boundary in addition. r=cpearce

dom/media/MediaDecoderStateMachine.cpp

@@ -2592,7 +2592,16 @@ MediaDecoderStateMachine::SeekCompleted()
     newCurrentTime = mAudioStartTime = seekTime;
   } else if (HasAudio()) {
     AudioData* audio = AudioQueue().PeekFront();
-    newCurrentTime = mAudioStartTime = audio ? audio->mTime : seekTime;
+    // Though we adjust the newCurrentTime in audio-based, and supplemented
+    // by video. For better UX, should NOT bind the slide position to
+    // the first audio data timestamp directly.
+    // While seeking to a position where there's only either audio or video, or
+    // seeking to a position lies before audio or video, we need to check if
+    // seekTime is bounded in suitable duration. See Bug 1112438.
+    int64_t videoStart = video ? video->mTime : seekTime;
+    int64_t audioStart = audio ? audio->mTime : seekTime;
+    newCurrentTime = mAudioStartTime =
+        std::min(std::min(audioStart, videoStart), seekTime);
   } else {
     newCurrentTime = video ? video->mTime : seekTime;
   }