MSE: Fix logic bugs with high precision duration

third_party/WebKit/Source/core/html/HTMLMediaElement.cpp

 /*
  * Copyright (C) 2007, 2008, 2009, 2010, 2011, 2012, 2013 Apple Inc. All rights
  * reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
@@ skipping 1988 matching lines @@
   seek(time);
 }
 
 double HTMLMediaElement::duration() const {
   // FIXME: remove m_webMediaPlayer check once we figure out how
   // m_webMediaPlayer is going out of sync with readystate.
   // m_webMediaPlayer is cleared but readystate is not set to kHaveNothing.
   if (!m_webMediaPlayer || m_readyState < kHaveMetadata)
     return std::numeric_limits<double>::quiet_NaN();
 
-  // FIXME: Refactor so m_duration is kept current (in both MSE and
-  // non-MSE cases) once we have transitioned from kHaveNothing ->
-  // kHaveMetadata. Currently, m_duration may be out of date for at least MSE
-  // case because MediaSource and SourceBuffer do not notify the element
-  // directly upon duration changes caused by endOfStream, remove, or append
-  // operations; rather the notification is triggered by the WebMediaPlayer
-  // implementation observing that the underlying engine has updated duration
-  // and notifying the element to consult its MediaSource for current
-  // duration. See http://crbug.com/266644

[chcunningham, 2016/12/15 01:24:26]

Matt, I think this comment is stale (or I misunderstand). I checked all cases of
duration changing (decrease for eos, increase for append, user specified, init
segment) and found they all notify the pipeline which bubbles to WMPI and
ultimately HTMLMediaElement. If you agree we should close that old bug. 

[wolenetz, 2016/12/16 00:56:52]

These do eventually bubble, but in the interim time, m_duration is out of date
(though ::duration() with this change, will return current MSE duration). The
problem is that ::duration() is not always used; there's an m_duration which is
still not necessarily in agreement with WMPI, even with this change. At this
time, it could be a simple fix. WDYT?

[chcunningham, 2017/01/03 17:48:06]

At this point m_duration is only used in OnDurationChanged as a guard against
spamming that event. Having said that, I have taken this opportunity to clean
this up. 

HTMLMediaElement::duration() will now simply return m_duration. 

m_duration will be explicitly set to NaN during
invokeResourceSelectionAlgorithm() 

m_duration will be explicitly set to WMPI->duration() during transition to
HAVE_METADATA. 

m_duration will be updated (already is) whenever duration changes occur from the
lower layer

-
   if (m_mediaSource)
     return m_mediaSource->duration();
 
   return webMediaPlayer()->duration();
 }
 
 bool HTMLMediaElement::paused() const {
   return m_paused;
 }
 
@@ skipping 2065 matching lines @@
       kMostlyFillViewportBecomeStableSeconds, BLINK_FROM_HERE);
 }
 
 void HTMLMediaElement::viewportFillDebouncerTimerFired(TimerBase*) {
   m_mostlyFillingViewport = true;
   if (m_webMediaPlayer)
     m_webMediaPlayer->becameDominantVisibleContent(m_mostlyFillingViewport);
 }
 
 }  // namespace blink