Add a border radius property to inherit it from a parent in a fallback content

third_party/WebKit/Source/core/layout/line/InlineFlowBox.cpp

Index: third_party/WebKit/Source/core/layout/line/InlineFlowBox.cpp
diff --git a/third_party/WebKit/Source/core/layout/line/InlineFlowBox.cpp b/third_party/WebKit/Source/core/layout/line/InlineFlowBox.cpp
index 2fec4ee3304a90a8a87b05cee2def5a9a4f896dc..97738e33ecd85f9e5b9576c985b23d00f620aef6 100644
--- a/third_party/WebKit/Source/core/layout/line/InlineFlowBox.cpp
+++ b/third_party/WebKit/Source/core/layout/line/InlineFlowBox.cpp
@@ -1034,9 +1034,19 @@ bool InlineFlowBox::nodeAtPoint(HitTestResult& result, const HitTestLocation& lo
     }
 
     if (lineLayoutItem().style()->hasBorderRadius()) {

[pdr., 2016/03/07 21:15:51]

I agree that the code in InlineBox::nodeAtPoint is wrong, but I don't understand
why the code here is correct.

Why can't we change includeLogicalLeftEdge to look like:
bool includeLogicalLeftEdge() const { return lineLayoutItem().isBox() ?
m_includeLogicalLeftEdge : false; }

I guess it's just not clear why this isBox check is correct, since I don't see
similar calls elsewhere.

[Miyoung Shin(c), 2016/03/09 12:28:07]

<img> has a <div> child with inline-block for a fallback image, and <img>
inlinebox is a rootInlineBox.
Then, we will set the position in RootInlineBox::alignBoxesInBlockDirection and
the position of <img> is updated by the child's maxAscent.
So, <img>, rootInlineBox(inlineflowbox), has the different position from
LayoutBox's position.
If <img> doesn't have any child or the child's size is none, <img> position is
not updated.
As I know, in general <img> doesn't have chidren.
I've seen the only case that <img> has a child when shadow dom is created for a
fallback image.

Considering this case, I thought we should check borderRadius with <img>
LayoutBox's position.

If I'm wrong, could you please correct me ?

[pdr., 2016/03/21 05:26:10]

I think you're getting stuck on the <img> case when the issue here is unrelated
to images. I would recommend getting everything working for the common case of
divs, then see if images need special handling.

Hit testing and painting are very similar problems so one way to approach this
is to look at how rounded rects are painted, and use the same functions here.
Another approach might be to directly call
LayoutBox::hitTestClippedOutByRoundedBorder.

[Miyoung Shin(c), 2016/03/29 22:49:40]

As you suggested, I've looked at how rounded rects are painted. In case of IMG,
we use the overflow rectangle applied with lineTop and lineBottom to paint the
rounded border, but in case of a general inlineflowbox such as text, we use the
frameRect moved by the paint offset in
InlineFlowBoxPainter::paintBoxDecorationBackground. Currently in this function,
we check hitTest with overflowRect but for borderRadius we use frameRect.
So I think it is inconsistent, and LayoutBox::hitTestClippedOutByRoundedBorder
will be reasonable to use here.

-        LayoutRect borderRect = logicalFrameRect();
+        LayoutRect borderRect;
+        bool includeLeftEdge = true;
+        bool includeRightEdge = true;
+        if (lineLayoutItem().isBox()) {
+            borderRect = toLayoutBox(lineLayoutItem())->borderBoxRect();
+        } else {
+            borderRect = logicalFrameRect();
+            includeLeftEdge = includeLogicalLeftEdge();
+            includeRightEdge = includeLogicalRightEdge();
+        }
+
         borderRect.moveBy(accumulatedOffset);
-        FloatRoundedRect border = lineLayoutItem().style()->getRoundedBorderFor(borderRect, includeLogicalLeftEdge(), includeLogicalRightEdge());
+        FloatRoundedRect border = lineLayoutItem().style()->getRoundedBorderFor(borderRect, includeLeftEdge, includeRightEdge);
         if (!locationInContainer.intersects(border))
             return false;
     }