Disable canvas line dashing when dash sequence is all zeros.

third_party/WebKit/Source/modules/canvas2d/CanvasRenderingContext2DState.cpp

Index: third_party/WebKit/Source/modules/canvas2d/CanvasRenderingContext2DState.cpp
diff --git a/third_party/WebKit/Source/modules/canvas2d/CanvasRenderingContext2DState.cpp b/third_party/WebKit/Source/modules/canvas2d/CanvasRenderingContext2DState.cpp
index 94b6d969cc35ec778d0894bd298375ebb4331097..6e75481849f09163c196e8f1f4d46dd3bcb383c3 100644
--- a/third_party/WebKit/Source/modules/canvas2d/CanvasRenderingContext2DState.cpp
+++ b/third_party/WebKit/Source/modules/canvas2d/CanvasRenderingContext2DState.cpp
@@ -199,12 +199,21 @@ void CanvasRenderingContext2DState::setLineDash(const Vector<float>& dash)
     m_lineDashDirty = true;
 }
 
+static bool hasANonZeroElement(const Vector<float>& lineDash)
+{
+    for (size_t i = 0; i < lineDash.size(); i++) {
+        if (lineDash[i] != 0.0f)
+            return true;
+    }
+    return false;
+}
+
 void CanvasRenderingContext2DState::updateLineDash() const
 {
     if (!m_lineDashDirty)
         return;
 
-    if (m_lineDash.size() == 0) {
+    if (!hasANonZeroElement(m_lineDash)) {

[Stephen White, 2015/10/01 15:46:02]

Please create a Skia bug for this (diff between hw & sw rendering in this case),
and mention that this workaround could be removed when the Skia bug is fixed.

[Justin Novosad, 2015/10/01 16:06:42]

I'll create a skia bug, but I don't think this workaround should be removed
since there is no point in create a no-op path effect.

[Stephen White, 2015/10/01 16:54:44]

Acknowledged.

         m_strokePaint.setPathEffect(0);
     } else {
         RefPtr<SkPathEffect> dashPathEffect = adoptRef(SkDashPathEffect::Create(m_lineDash.data(), m_lineDash.size(), m_lineDashOffset));