Invalidate border/outline on color change only if they use current color

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

Index: third_party/WebKit/Source/core/layout/LayoutObject.cpp
diff --git a/third_party/WebKit/Source/core/layout/LayoutObject.cpp b/third_party/WebKit/Source/core/layout/LayoutObject.cpp
index 57050627de17fafbb5179957e1627db37958916b..6f64ffa686e2f3d531afab7d770550bb5ad0564c 100644
--- a/third_party/WebKit/Source/core/layout/LayoutObject.cpp
+++ b/third_party/WebKit/Source/core/layout/LayoutObject.cpp
@@ -1543,7 +1543,8 @@ StyleDifference LayoutObject::AdjustStyleDifference(
   // needed if we have style or text affected by these properties.
   if (diff.TextDecorationOrColorChanged() &&
       !diff.NeedsFullPaintInvalidation()) {
-    if (Style()->HasBorder() || Style()->HasOutline() ||
+    if (Style()->HasBorderColorReferencingCurrentColor() ||

[wkorman, 2017/04/24 22:05:41]

Could add unit tests for this, BorderData, and ComputedStyle (in separate
*Test.cpp each). Would be trivial tests but I think nice to start building them
up. BorderDataTest does not exist but the others do.

Layout test obviously covers end behavior.

[Xianzhu, 2017/04/24 22:36:48]

Added ComputedStyleTest.HasOutlineWithCurrentColor and
ComputedStyleTest.HasBorderColorReferencingCurrentColor. Not BorderDataTest
because ComputedStyle controls how BorderData mutates. Using friend test would
just exposes BorderData as a struct and the tested code is too different from
actual executed code.

+        Style()->IsOutlineColorCurrentColor() ||
         Style()->HasBackgroundRelatedColorReferencingCurrentColor() ||
         // Skip any text nodes that do not contain text boxes. Whitespace cannot
         // be skipped or we will miss invalidating decorations (e.g.,