Implement SANACLAP (http://bit.ly/sanaclap).

third_party/WebKit/Source/core/style/ComputedStyle.cpp

Index: third_party/WebKit/Source/core/style/ComputedStyle.cpp
diff --git a/third_party/WebKit/Source/core/style/ComputedStyle.cpp b/third_party/WebKit/Source/core/style/ComputedStyle.cpp
index 20681097384728f6e84ccc056967ab3512b3b711..1811e49f20fe9ada08b32baa261a4a0c21aa4fbb 100644
--- a/third_party/WebKit/Source/core/style/ComputedStyle.cpp
+++ b/third_party/WebKit/Source/core/style/ComputedStyle.cpp
@@ -872,6 +872,9 @@ void ComputedStyle::updatePropertySpecificDifferences(const ComputedStyle& other
                 diff.setTextDecorationOrColorChanged();
         }
     }
+
+    if (diff.needsLayout() || diff.transformChanged())

[ojan, 2016/08/16 23:54:32]

A couple thoughts on this.
1. You need to also check diff.needsPositionedMovementLayout.
2. Putting it in this function seems a bit strange to me. Maybe just put it
directly in visualInvalidationDiff?
3. I think this is fine for now for us to test it in the wild, but the thing we
spec/ship won't be able to use the needsLayout bit directly because it encodes
our current logic for what needs a layout, not something about which properties
changed. So, we probably need a TODO to that effect. For example:

-When opacity changes, we only set the needsLayout bit if the element changed
whether it was a stacking context, but probably want to always disable scroll
anchoring? Or do we?
https://cs.chromium.org/chromium/src/third_party/WebKit/Source/core/style/Com....
-When the textAutoSizingMultiplier changes we set this bit, but probably
shouldn't disable scroll anchoring.

I'm sure there are other examples. I just found those as two quick ones staring
at the code.

Bugs/Sasha/Elliott: Do you have opinions on the best way to incorporate this
logic and/or whether we should always disable scroll anchoring on opacity
changes or just when stacking context state changes? I suspect Tab has opinions
on the latter too.

[Bugs Nash, 2016/08/17 02:08:46]

Isn't that already checked via diff.needsLayout()?
Agree it doesn't seem 'property specific' to me
Not sure. How does scroll anchoring currently decide what is a layout affecting
change? Should be the same no?

[skobes, 2016/08/17 18:50:32]

As bugsnash noted, needsLayout includes this.
Done.
Added TODO.  Agree the spec should reference specific properties and not our
needsLayout concept.  Implementation-wise I wonder if it's better to start with
needsLayout() and add exceptions for the cases where we set needsLayout but
don't want to disable anchoring.

+        diff.setLayoutAffectingPropertyChanged();
 }
 
 void ComputedStyle::addPaintImage(StyleImage* image)