Need to repaint its ::first-line background when a block moves.

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

 /*
  * Copyright (C) 2000 Lars Knoll (knoll@kde.org)
  * Copyright (C) 2003, 2004, 2006, 2007, 2008, 2009, 2010, 2011 Apple Inc. All right reserved.
  * Copyright (C) 2010 Google Inc. All rights reserved.
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Library General Public
  * License as published by the Free Software Foundation; either
  * version 2 of the License, or (at your option) any later version.
  *
@@ skipping 2061 matching lines @@
     return logicalLeft;
 }
 
 void LayoutBlockFlow::invalidateDisplayItemClientsOfFirstLine()
 {
     ASSERT(childrenInline());
     if (RootInlineBox* firstRootBox = this->firstRootBox())
         firstRootBox->invalidateDisplayItemClientsRecursively();
 }
 
+PaintInvalidationReason LayoutBlockFlow::invalidatePaintIfNeeded(PaintInvalidationState& paintInvalidationState, const LayoutBoxModelObject& paintInvalidationContainer)

[Xianzhu, 2015/12/18 17:24:54]

This will miss the LayoutButton case, but I think it's fine to leave this corner
case as-is which will be automatically fixed when we fix crbug.com/226252.

+{
+    PaintInvalidationReason reason = LayoutBlock::invalidatePaintIfNeeded(paintInvalidationState, paintInvalidationContainer);
+    if (reason == PaintInvalidationNone)
+        return reason;
+    RootInlineBox* line = firstRootBox();
+    if (!line || !line->isFirstLineStyle())
+        return reason;
+    // It's the RootInlineBox that paints the ::first-line background.
+    invalidateDisplayItemClient(*line);

[Xianzhu, 2015/12/18 07:01:20]

Will this always invalidate the first line, even if there is no first-line
style?

[mstensho (USE GERRIT), 2015/12/18 10:33:04]

Yeah, looks like it. That wasn't my intention.

[mstensho (USE GERRIT), 2015/12/18 13:11:11]

Fixed it. Not sure if it's worth it, though. Could it generally be more
expensive to figure out that you have no ::first-line that applies, than to just
invalidate it (which would also involve rebaselining a bunch of tests)? What do
you think?

[Xianzhu, 2015/12/18 17:24:54]

Hmm enclosingFirstLineStyleBlock() seems not cheap (especially there is no
first-line style) and it would be tested for all LayoutBlockFlows. Normally a
RootInlineBox's painting is just one drawRect which is cheap. This invalidation
also won't cause more rasterization invalidation. I think we can use this
always-invalidation version.

Patch Set 1 lgtm.

[mstensho (USE GERRIT), 2015/12/21 10:54:10]

Done. Added a comment too.

+    return reason;
 }
+
+}