Invalidate first line display item clients when first line style changes

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 2a8d96148b2551660a4739d5e892cbc991cb3fc0..d5425ccf47fe32858fd7c0366dfdb2b3ee5abf9a 100644
--- a/third_party/WebKit/Source/core/layout/LayoutObject.cpp
+++ b/third_party/WebKit/Source/core/layout/LayoutObject.cpp
@@ -1844,13 +1844,15 @@ void LayoutObject::setPseudoStyle(PassRefPtr<ComputedStyle> pseudoStyle)
 
 void LayoutObject::firstLineStyleDidChange(const ComputedStyle& oldStyle, const ComputedStyle& newStyle)
 {
-    StyleDifference diff = oldStyle.visualInvalidationDiff(newStyle);
-    if (diff.hasDifference()) {
-        // TODO(rune@opera.com): We should use the diff to determine whether a repaint vs. layout
-        // is needed, but for now just assume a layout will be required. The diff code
-        // in LayoutObject::setStyle would need to be factored out so that it could be reused.
-        setNeedsLayoutAndPrefWidthsRecalcAndFullPaintInvalidation(LayoutInvalidationReason::StyleChange);
+    StyleDifference diff = adjustStyleDifference(oldStyle.visualInvalidationDiff(newStyle));
+
+    if (diff.needsPaintInvalidation()) {
+        setShouldDoFullPaintInvalidation();
+        if (isLayoutBlockFlow())
+            toLayoutBlockFlow(this)->invalidateDisplayItemClientsOfFirstLine();

[rune, 2015/11/27 09:27:11]

How about FIRST_LINE_INHERITED. firstLineStyleDidChange() will be called on
inline elements on the first line, and necessarily so, even if the ::first-line
style doesn't change. I reported 562418, which I suspect is also a slimming
paint regression. I suspect this CL won't fix that. Could you check?

[Xianzhu, 2015/11/30 18:15:53]

For FIRST_LINE_INHERITED, we'll use the normal paint invalidation on style
change flow: mark the LayoutInline shouldDoFullPaintInvalidation during setStyle
and invalidate all line boxes during paint invalidation
(LayoutInline::invalidateDisplayItemClients()).

crbug.com/562418 seems not a slimming paint regression because it also
reproduces on Safari. It seems a style priority issue or a painting issue. Will
investigate it as a separate issue.

     }
+    if (diff.needsLayout())
+        setNeedsLayoutAndPrefWidthsRecalc(LayoutInvalidationReason::StyleChange);
 }
 
 void LayoutObject::markContainingBlocksForOverflowRecalc()