Title: [290254] trunk/Source/WebCore
- Revision
- 290254
- Author
- [email protected]
- Date
- 2022-02-21 11:15:35 -0800 (Mon, 21 Feb 2022)
Log Message
[LBSE] Implement SVG2 rules to establish a stacking context
https://bugs.webkit.org/show_bug.cgi?id=236193
Reviewed by Rob Buis.
Follow the rules specified in https://www.w3.org/TR/SVG2/render.html#RenderingOrder,
when to create a stacking context. This is necessary to enforce layer creation.
Unlike SVG 1.1, SVG2 defines its whole rendering model, as well as painting operations
such as masking/clipping around the 'stacking-context' term that stems from CSS.
The definition is backwards compatible with SVG 1.1, except for un-spec'ed territory
missing from SVG 1.1, that's now specified for SVG2 (e.g. <foreignObject> + stacking context).
LBSE handles clipping / masking / etc. via RenderLayer, which is also responsbile for
for stacking-context managment. Therefore we need to follow the existing rules to enforce
a stacking context / RenderLayer creation in StyleAdjuster, by enforcing a non-auto 'used z-index'.
Covered by existing tests, no change in behaviour.
* style/StyleAdjuster.cpp:
(WebCore::Style::Adjuster::adjustSVGElementStyle):
Modified Paths
Diff
Modified: trunk/Source/WebCore/ChangeLog (290253 => 290254)
--- trunk/Source/WebCore/ChangeLog 2022-02-21 18:57:52 UTC (rev 290253)
+++ trunk/Source/WebCore/ChangeLog 2022-02-21 19:15:35 UTC (rev 290254)
@@ -1,3 +1,27 @@
+2022-02-21 Nikolas Zimmermann <[email protected]>
+
+ [LBSE] Implement SVG2 rules to establish a stacking context
+ https://bugs.webkit.org/show_bug.cgi?id=236193
+
+ Reviewed by Rob Buis.
+
+ Follow the rules specified in https://www.w3.org/TR/SVG2/render.html#RenderingOrder,
+ when to create a stacking context. This is necessary to enforce layer creation.
+
+ Unlike SVG 1.1, SVG2 defines its whole rendering model, as well as painting operations
+ such as masking/clipping around the 'stacking-context' term that stems from CSS.
+ The definition is backwards compatible with SVG 1.1, except for un-spec'ed territory
+ missing from SVG 1.1, that's now specified for SVG2 (e.g. <foreignObject> + stacking context).
+
+ LBSE handles clipping / masking / etc. via RenderLayer, which is also responsbile for
+ for stacking-context managment. Therefore we need to follow the existing rules to enforce
+ a stacking context / RenderLayer creation in StyleAdjuster, by enforcing a non-auto 'used z-index'.
+
+ Covered by existing tests, no change in behaviour.
+
+ * style/StyleAdjuster.cpp:
+ (WebCore::Style::Adjuster::adjustSVGElementStyle):
+
2022-02-21 Myles C. Maxfield <[email protected]>
[WebGPU] Tracer bullet part 2: Fix infinite recursion in GPUDevice::ref()
Modified: trunk/Source/WebCore/style/StyleAdjuster.cpp (290253 => 290254)
--- trunk/Source/WebCore/style/StyleAdjuster.cpp 2022-02-21 18:57:52 UTC (rev 290253)
+++ trunk/Source/WebCore/style/StyleAdjuster.cpp 2022-02-21 19:15:35 UTC (rev 290254)
@@ -9,7 +9,7 @@
* Copyright (c) 2011, Code Aurora Forum. All rights reserved.
* Copyright (C) Research In Motion Limited 2011. All rights reserved.
* Copyright (C) 2012, 2013 Google Inc. All rights reserved.
- * Copyright (C) 2014 Igalia S.L.
+ * Copyright (C) 2014, 2020, 2022 Igalia S.L.
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
@@ -632,7 +632,45 @@
if (!isPositioningAllowed)
style.setPosition(RenderStyle::initialPosition());
- // LegacyRenderSVGRoot handles zooming for the whole SVG subtree, so foreignObject content should
+#if ENABLE(LAYER_BASED_SVG_ENGINE)
+ // SVG2: A new stacking context must be established at an SVG element for its descendants if:
+ // - it is the root element
+ // - the "z-index" property applies to the element and its computed value is an integer
+ // - the element is an outermost svg element, or a "foreignObject", "image", "marker", "mask", "pattern", "symbol" or "use" element
+ // - the element is an inner "svg" element and the computed value of its "overflow" property is a value other than visible
+ // - the element is subject to explicit clipping:
+ // - the "clip" property applies to the element and it has a computed value other than auto
+ // - the "clip-path" property applies to the element and it has a computed value other than none
+ // - the "mask" property applies to the element and it has a computed value other than none
+ // - the "filter" property applies to the element and it has a computed value other than none
+ // - a property defined in another specification is applied and that property is defined to establish a stacking context in SVG
+ //
+ // Some of the rules above were already enforced in StyleResolver::adjustRenderStyle() - for those cases assertions were added.
+ if (svgElement.document().settings().layerBasedSVGEngineEnabled() && style.hasAutoUsedZIndex()) {
+ // adjustRenderStyle() has already assigned a z-index of 0 if clip / filter is present or the element is the root element.
+ ASSERT(!style.hasClip());
+ ASSERT(!style.clipPath());
+ ASSERT(!style.hasFilter());
+ ASSERT(!svgElement.isOutermostSVGSVGElement());
+
+ auto isInnerSVGElement = [] (const SVGElement& svgElement) -> bool {
+ return svgElement.hasTagName(SVGNames::svgTag) && svgElement.parentNode() && is<SVGElement>(svgElement.parentNode());
+ };
+
+ if (svgElement.hasTagName(SVGNames::foreignObjectTag)
+ || svgElement.hasTagName(SVGNames::imageTag)
+ || svgElement.hasTagName(SVGNames::markerTag)
+ || svgElement.hasTagName(SVGNames::maskTag)
+ || svgElement.hasTagName(SVGNames::patternTag)
+ || svgElement.hasTagName(SVGNames::symbolTag)
+ || svgElement.hasTagName(SVGNames::useTag)
+ || (isInnerSVGElement(svgElement) && (style.overflowX() != Overflow::Visible || style.overflowY() != Overflow::Visible))
+ || style.hasPositionedMask())
+ style.setUsedZIndex(0);
+ }
+#endif
+
+ // (Legacy)RenderSVGRoot handles zooming for the whole SVG subtree, so foreignObject content should
// not be scaled again.
if (svgElement.hasTagName(SVGNames::foreignObjectTag))
style.setEffectiveZoom(RenderStyle::initialZoom());
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
