document.documentElement.scrollTop/Left is zero

Source/core/html/HTMLBodyElement.cpp

Index: Source/core/html/HTMLBodyElement.cpp
diff --git a/Source/core/html/HTMLBodyElement.cpp b/Source/core/html/HTMLBodyElement.cpp
index ecf39231e90c8bee182b61938d351ffe0d36ffd0..d8c9fb2eb4d8426cc3cc6610a9366bb1e9363d77 100644
--- a/Source/core/html/HTMLBodyElement.cpp
+++ b/Source/core/html/HTMLBodyElement.cpp
@@ -253,6 +253,9 @@ static int adjustForZoom(int value, Document* document)
 
 int HTMLBodyElement::scrollLeft()
 {
+    if (!document().inQuirksMode())
+        return 0;
+
     // Update the document's layout.
     Document& document = this->document();
     document.updateLayoutIgnorePendingStylesheets();
@@ -275,6 +278,9 @@ void HTMLBodyElement::setScrollLeft(int scrollLeft)
 
 int HTMLBodyElement::scrollTop()
 {
+    if (!document().inQuirksMode())
+        return 0;

[Julien - ping for review, 2013/09/30 23:39:57]

Is this totally correct?

Step 4 of scrollTop says: "Return the y-coordinate of the scrolling area at the
alignment point with the top of the padding edge of the element."

[tonikitoo_, 2013/10/01 14:42:17]

This is a valid observation! I believe the specification is not clear about
this, or it lacks a "return zero and terminate these steps" advice.

The way I implemented it makes Blink to match Firefox, IE, Safari nightly and
Opera (pre-blink) for this case, as it can be seen on
http://13thparallel.com/archive/viewport/example6.htm . Firefox implementation
is explicit about it:
http://lxr.mozilla.org/mozilla-central/source/content/base/src/Element.cpp#510 .

We could maybe chose to match other engines, and clarify with Hixie the spec.

[Julien - ping for review, 2013/10/01 15:19:51]

Clarifying the specification is definitely the way to go. It shouldn't block
this change though (you can put a comment linking to the thread or just a
FIXME).

+
     // Update the document's layout.
     Document& document = this->document();
     document.updateLayoutIgnorePendingStylesheets();