Media Controls: Fix seek to end when max attr is rounded up

third_party/WebKit/Source/core/html/shadow/MediaControlElements.cpp

 /*
  * Copyright (C) 2008, 2009, 2010, 2011 Apple Inc. All rights reserved.
  * Copyright (C) 2012 Google 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.
@@ skipping 752 matching lines @@
     mediaControls().endScrubbing();
   }
 
   MediaControlInputElement::defaultEventHandler(event);
 
   if (event->type() != EventTypeNames::input)
     return;
 
   double time = value().toDouble();
 
+  double duration = mediaElement().duration();
+  double maxAttribute = getFloatingPointAttribute(maxAttr);
+  // Workaround for floating point error - it's possible for this element's max
+  // attribute to be rounded to a value slightly higher than the duration. If
+  // this happens and scrubber is dragged near the max, seek to duration.
+  if (duration < time && time <= maxAttribute && maxAttribute <= duration + 0.5)
+    time = duration;

[mlamouri (slow - plz ping), 2017/02/27 17:48:12]

I do not understand why this isn't:
```
if (duration < time)
  time = duration;
```
or said otherwise: `time = std:min(time, duration);`

I bet there is a reason. I can actually guess why we do `time <= maxAttribute`
even though that sounds like we could drop the check. However, I don't
understand why `+0.5`.

[johnme, 2017/02/28 18:02:55]

Done.

I was guarding against the case where mediaElement().duration() and and
getFloatingPointAttribute(maxAttr) are significantly out of sync (not just by
rounding error), and trying not to change behaviour in that case. It seems that
can only happen if the webpage calls event.stopImmediatePropagation() in a
capturing durationchange event listener, and since that's such an edge case,
setting time to duration in that case is probably fine too (in fact, it probably
makes more sense).

(The 0.5 was to allow the seconds value to have been rounded, which increases it
by at most 0.5 seconds; though FWIW it seems that it's the milliseconds that get
rounded)

+
   // FIXME: This will need to take the timeline offset into consideration
   // once that concept is supported, see https://crbug.com/312699
   if (mediaElement().seekable()->contain(time))
     mediaElement().setCurrentTime(time);
 
   LayoutSliderItem slider = LayoutSliderItem(toLayoutSlider(layoutObject()));
   if (!slider.isNull() && slider.inDragMode())
     mediaControls().updateCurrentTimeDisplay();
 }
 
@@ skipping 263 matching lines @@
 MediaControlCurrentTimeDisplayElement*
 MediaControlCurrentTimeDisplayElement::create(MediaControls& mediaControls) {
   MediaControlCurrentTimeDisplayElement* element =
       new MediaControlCurrentTimeDisplayElement(mediaControls);
   element->setShadowPseudoId(
       AtomicString("-webkit-media-controls-current-time-display"));
   return element;
 }
 
 }  // namespace blink