Mask/background-repeat: round should round the number of tiles of an image to the nearest natural n…

Source/core/rendering/RenderBoxModelObject.cpp

 /*
  * Copyright (C) 1999 Lars Knoll (knoll@kde.org)
  *           (C) 1999 Antti Koivisto (koivisto@kde.org)
  *           (C) 2005 Allan Sandfeld Jensen (kde@carewolf.com)
  *           (C) 2005, 2006 Samuel Weinig (sam.weinig@gmail.com)
  * Copyright (C) 2005, 2006, 2007, 2008, 2009 Apple Inc. All rights 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
@@ skipping 1044 matching lines @@
     geometry.setTileSize(fillTileSize);
 
     EFillRepeat backgroundRepeatX = fillLayer->repeatX();
     EFillRepeat backgroundRepeatY = fillLayer->repeatY();
     RenderView* renderView = view();
     int availableWidth = positioningAreaSize.width() - geometry.tileSize().width();
     int availableHeight = positioningAreaSize.height() - geometry.tileSize().height();
 
     LayoutUnit computedXPosition = minimumValueForLength(fillLayer->xPosition(), availableWidth, renderView, true);
     if (backgroundRepeatX == RoundFill && positioningAreaSize.width() > 0 && fillTileSize.width() > 0) {
-        int nrTiles = ceil((double)positioningAreaSize.width() /
-            fillTileSize.width());
+        long nrTiles = lroundf((float)positioningAreaSize.width() / fillTileSize.width());

[eseidel, 2013/09/17 18:07:00]

So this is intentionally causing whitespace when the author doesn't match the
background tile size precisely to the size of the div?  I'm confused.  Why
wouldn't we want to ceil here?

[eseidel, 2013/09/17 18:08:32]

I see.  The author has *specifically* requested that we round instead of ceil
and that's the problem.  They want this whitespace for whatever reason. :)

[Julien - ping for review, 2013/09/17 18:37:48]

This mandated by the specification.

Thanks for making the change to long even if it turns out I got confused and
it's not needed as we are taking the result of the division of 2 integers. The
new code is more future proof though.

 
         if (fillLayer->size().size.height().isAuto() && backgroundRepeatY != RoundFill) {
             fillTileSize.setHeight(fillTileSize.height() * positioningAreaSize.width() / (nrTiles * fillTileSize.width()));
         }
 
         fillTileSize.setWidth(positioningAreaSize.width() / nrTiles);
         geometry.setTileSize(fillTileSize);
         geometry.setPhaseX(geometry.tileSize().width() ? geometry.tileSize().width() - roundToInt(computedXPosition + left) % geometry.tileSize().width() : 0);
     }
 
     LayoutUnit computedYPosition = minimumValueForLength(fillLayer->yPosition(), availableHeight, renderView, true);
     if (backgroundRepeatY == RoundFill && positioningAreaSize.height() > 0 && fillTileSize.height() > 0) {
-        int nrTiles = ceil((double)positioningAreaSize.height() /
-            fillTileSize.height());
+        long nrTiles = lroundf((float)positioningAreaSize.height() / fillTileSize.height());
 
         if (fillLayer->size().size.width().isAuto() && backgroundRepeatX != RoundFill) {
             fillTileSize.setWidth(fillTileSize.width() * positioningAreaSize.height() / (nrTiles * fillTileSize.height()));
         }
 
         fillTileSize.setHeight(positioningAreaSize.height() / nrTiles);
         geometry.setTileSize(fillTileSize);
         geometry.setPhaseY(geometry.tileSize().height() ? geometry.tileSize().height() - roundToInt(computedYPosition + top) % geometry.tileSize().height() : 0);
     }
 
@@ skipping 1695 matching lines @@
     ASSERT(!beforeChild || toBoxModelObject == beforeChild->parent());
     for (RenderObject* child = startChild; child && child != endChild; ) {
         // Save our next sibling as moveChildTo will clear it.
         RenderObject* nextSibling = child->nextSibling();
         moveChildTo(toBoxModelObject, child, beforeChild, fullRemoveInsert);
         child = nextSibling;
     }
 }
 
 } // namespace WebCore