[css-flexbox] Add test case for definite widths

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

Index: third_party/WebKit/Source/core/layout/LayoutBox.cpp
diff --git a/third_party/WebKit/Source/core/layout/LayoutBox.cpp b/third_party/WebKit/Source/core/layout/LayoutBox.cpp
index 70e2509cb39ec661154035ac38b7e44ec691b6e9..9188b1f89c6d17c18afdc73830f2f91818ea16b2 100644
--- a/third_party/WebKit/Source/core/layout/LayoutBox.cpp
+++ b/third_party/WebKit/Source/core/layout/LayoutBox.cpp
@@ -4180,6 +4180,7 @@ static bool logicalWidthIsResolvable(const LayoutBox& layoutBox)
     const LayoutBox* box = &layoutBox;
     while (!box->isLayoutView() && !box->isFloatingOrOutOfFlowPositioned()
         && (box->style()->logicalWidth().isAuto() || box->isAnonymousBlock())
+        && !isStretchingColumnFlexItem(box)
         && !box->hasOverrideContainingBlockLogicalWidth())
         box = box->containingBlock();
 
@@ -4189,6 +4190,8 @@ static bool logicalWidthIsResolvable(const LayoutBox& layoutBox)
         return true;
     if (box->isOutOfFlowPositioned() && !box->style()->logicalLeft().isAuto() && !box->style()->logicalRight().isAuto())
         return true;
+    if (isStretchingColumnFlexItem(box))

[mstensho (USE GERRIT), 2016/05/03 12:08:52]

logicalWidthIsResolvable(), isn't that about the size being definite or not?
I.e. https://drafts.csswg.org/css-sizing-3/#definite ?

Some flexbox extensions to the definition of "definite size" here:
https://drafts.csswg.org/css-flexbox/#definite-sizes

But all widths and heights of flexboxes and flex items in the test case are
auto, so there's nothing definite here, is there?

Reduced test:

<p>There should be no red or yellow here.</p>
<div id="stf" style="float:left; background:yellow;">
    <div id="outerFlex" style="display:flex; flex-direction:column;">
        <div id="innerFlex" style="display:flex; background:red;">
            <div id="elm" style="width:100%; background:green;">xxxxx</div>
        </div>
    </div>
    <div style="width:20em;"></div>
</div>

Without the patch, I see that #innerFlex gets the correct width. It's a flex
item in #outerFlex, and its width is the cross size, which by default is
stretched. #elm is a flex item in #innerFlex, and its main size is 100%. Its
containing block is #innerFlex, so why not just set it to 100% of the width of
#innerFlex? Why ask the ancestry for permission? Fear of cyclic dependencies?

+        return true;
     if (box->hasOverrideContainingBlockLogicalWidth())
         return box->overrideContainingBlockContentLogicalWidth() != -1;
     if (box->style()->logicalWidth().hasPercent())