SVG with zero intrinsic size should be respected always

third_party/WebKit/Source/core/layout/LayoutReplaced.cpp

Index: third_party/WebKit/Source/core/layout/LayoutReplaced.cpp
diff --git a/third_party/WebKit/Source/core/layout/LayoutReplaced.cpp b/third_party/WebKit/Source/core/layout/LayoutReplaced.cpp
index 3e42a90b09d1e94d92945c587cc63ba4d5902a08..c5f778d14a0a874302fb2c7b3e4df3356fb4507a 100644
--- a/third_party/WebKit/Source/core/layout/LayoutReplaced.cpp
+++ b/third_party/WebKit/Source/core/layout/LayoutReplaced.cpp
@@ -282,6 +282,9 @@ LayoutUnit LayoutReplaced::computeReplacedLogicalWidth(ShouldComputePreferred sh
         bool computedHeightIsAuto = hasAutoHeightOrContainingBlockWithAutoHeight();
         bool hasIntrinsicWidth = constrainedSize.width() > 0;
 
+        if (!hasIntrinsicWidth && contentLayoutObject)
+            hasIntrinsicWidth = contentLayoutObject->style()->logicalWidth().isFixed();

[davve, 2016/01/18 15:52:48]

You rely on the (logical) intrinsic width being collected as
presentation attribute, thus ending up in
style()->logicalWidth(). Then you use style()->logicalWidth() as
indication whether there is an intrinsic (logical) width or not.

A weakness of this trick is that there can be "proper" style on the
SVG that shouldn't indicate intrinsic width (or height respectively).

The example below gives a 300x150 green rect in Firefox and Chrome and
I think that's correct. With your patch we would think there is a
(zero) intrinsic width and render nothing.

html:

<object style="background: green" data="some-svg.svg"></object>

some-svg.svg:

<svg xmlns="http://www.w3.org/2000/svg" style="width: 0"></svg>

I think we need something more elaborate here, preferable by
propagating enough information from
SVGSVGElement::intrinsicWidth/Height to make the correct
hasIntrinsicWidth/Height judgement in
LayoutReplaced::computeReplacedLogicalWidth/Height().

[Shanmuga Pandi, 2016/01/20 06:09:34]

Instead of LogicalWidth()/Height(), getting hasIntrinsicWidth()/Height() to
detect whether SVG has intrisic width/height. 

This solve the above the problem and the bug reported.

+
         // If 'height' and 'width' both have computed values of 'auto' and the element also has an intrinsic width, then that intrinsic width is the used value of 'width'.
         if (computedHeightIsAuto && hasIntrinsicWidth)
             return computeReplacedLogicalWidthRespectingMinMaxWidth(constrainedSize.width(), shouldComputePreferred);
@@ -344,6 +347,9 @@ LayoutUnit LayoutReplaced::computeReplacedLogicalHeight() const
     bool widthIsAuto = style()->logicalWidth().isAuto();
     bool hasIntrinsicHeight = constrainedSize.height() > 0;
 
+    if (!hasIntrinsicHeight && contentLayoutObject)
+        hasIntrinsicHeight = contentLayoutObject->style()->logicalHeight().isFixed();
+
     // If 'height' and 'width' both have computed values of 'auto' and the element also has an intrinsic height, then that intrinsic height is the used value of 'height'.
     if (widthIsAuto && hasIntrinsicHeight)
         return computeReplacedLogicalHeightRespectingMinMaxHeight(constrainedSize.height());