Snap inline box boundaries to layout units during line breaking.

third_party/WebKit/Source/core/layout/line/LineWidth.h

Index: third_party/WebKit/Source/core/layout/line/LineWidth.h
diff --git a/third_party/WebKit/Source/core/layout/line/LineWidth.h b/third_party/WebKit/Source/core/layout/line/LineWidth.h
index 2db3736d6cb6e073958ae4eafd213a5df11e4daf..55a9c0bea0b8c08e2eeca1e91b20805aca1497c1 100644
--- a/third_party/WebKit/Source/core/layout/line/LineWidth.h
+++ b/third_party/WebKit/Source/core/layout/line/LineWidth.h
@@ -68,6 +68,7 @@ public:
     void applyOverhang(LineLayoutRubyRun, LineLayoutItem startLayoutItem, LineLayoutItem endLayoutItem);
     void fitBelowFloats(bool isFirstLine = false);
     void setTrailingWhitespaceWidth(float width) { m_trailingWhitespaceWidth = width; }
+    void snapUncommittedWidth() { m_uncommittedWidth = LayoutUnit(m_uncommittedWidth).toFloat(); }

[eae, 2015/09/29 14:50:00]

Is the currentWidth already snapped? What about snapping currentWidth?

[szager1, 2015/09/29 16:48:32]

Based on conversation with eae: currentWidth is used while laying out words in a
single LayoutText, and at that time, it's expected that offsets are not snapped.
 Snapping currentWidth() isn't obviously wrong, but it's not obviously right
either.  Since the layout tests are passing without that change, I'm going to
leave it out of this patch.

I added a comment to the code about this.

 
     bool shouldIndentText() const { return m_shouldIndentText == IndentText; }