[wip] unprefix filter

third_party/WebKit/Source/core/layout/svg/LayoutSVGResourceContainer.cpp

 /*
  * Copyright (C) Research In Motion Limited 2010. 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.
  *
  * This library is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
@@ skipping 229 matching lines @@
         if (!layoutObject)
             continue;
 
         const ComputedStyle& style = layoutObject->styleRef();
 
         // If the client has a layer (is a non-SVGElement) we need to signal
         // invalidation in the same way as is done in markAllResourceClientsForInvalidation above.
         if (layoutObject->hasLayer() && resourceType() == FilterResourceType) {
             if (style.hasFilter())
                 toLayoutBoxModelObject(layoutObject)->layer()->filterNeedsPaintInvalidation();
-            // If this is the SVG root, we could have both 'filter' and
-            // '-webkit-filter' applied, so we need to do the invalidation
-            // below as well, unless we can optimistically determine that
-            // 'filter' does not apply to the element in question.
-            if (!layoutObject->isSVGRoot() || !style.svgStyle().hasFilter())
+            // If this is the SVG root, we could have 'filter' applied and

[fs, 2016/05/23 12:29:53]

This comment does make a lot of sense anymore. If !style.hasFilter() this
resource does not apply at all, so 'continue' should be enough in that case. I
guess for LayoutSVGRoot we might need to call
markForLayoutAndParentResourceInvalidation (like clientStyleChanged does now) to
notify any dependents. filterNeedsPaintInvalidation() will mark for paint
invalidation and don't think marking for layout should be needed anymore (used
to be needed because some bounds needed to be recomputed I think.)

So maybe something like:

if (!style.hasFilter())
    continue;
toLayoutBoxModelObject(layoutObject)->layer()->filterNeedsPaintInvalidation();
if (layoutObject->isSVGRoot())
    markForLayoutAndParentResourceInvalidation(layoutObject, false);
continue;

[Noel Gordon, 2016/05/29 04:15:28]

Thanks for the code.  I tried it and the two tests in question failed

  svg/custom/cross-referenced-resources.html [ Failure ]
  svg/filters/filter-on-svg-root-w-layer.html [ Failure ]

sorry to say.  If I let the svgRoot fall into the full invalidation code that
the original comment mentioned, then these tests pass.  So I did that for now
(and adjusted the comment).

[fs, 2016/05/30 11:58:52]

Yes, this was under the assumption that FilterPainter was applying the filter,
and no state was kept in SVGResources. If the PaintLayer isn't painting the
filter, then the filterNeedsPaintInvalidation() ought to be redundant instead.

+            // so need to do the invalidation below as well, unless we can
+            // optimistically determine that filter does not apply to the
+            // element in question.
+            if (!layoutObject->isSVGRoot() || !style.hasFilter())
                 continue;
         }
 
         StyleDifference diff;
         diff.setNeedsFullLayout();
         SVGResourcesCache::clientStyleChanged(layoutObject, diff, style);
         layoutObject->setNeedsLayoutAndFullPaintInvalidation(LayoutInvalidationReason::SvgResourceInvalidated);
     }
 }
 
@@ skipping 56 matching lines @@
             // This will process the rest of the ancestors.
             toLayoutSVGResourceContainer(current)->removeAllClientsFromCache();
             break;
         }
 
         current = current->parent();
     }
 }
 
 } // namespace blink