Title: [236632] trunk
- Revision
- 236632
- Author
- [email protected]
- Date
- 2018-09-28 18:53:34 -0700 (Fri, 28 Sep 2018)
Log Message
RenderLayer::removeOnlyThisLayer() should not call updateLayerPositions()
https://bugs.webkit.org/show_bug.cgi?id=190093
Reviewed by Dean Jackson and Zalan Bujtas.
Source/WebCore:
It's wrong for RenderLayer::removeOnlyThisLayer() to call updateLayerPositions(),
because this is called at style update time, and layout will be stale.
It was added (see webkit.org/b/25252) so that opacity changes, which can destroy layers, correctly update
descendants. However, RenderStyle::changeRequiresLayout() checks for opacity <=> no opacity
changes and triggers layout accordingly, which will result in a full post-layout
updateLayerPositions().
This also revealed that changes to the "isolate" property fail to trigger any kind of style recalc or layout;
we need it to trigger layout (for now) because it affects z-order.
Covered by existing tests.
* rendering/RenderLayer.cpp:
(WebCore::RenderLayer::removeOnlyThisLayer):
* rendering/style/RenderStyle.cpp:
(WebCore::RenderStyle::changeRequiresLayout const):
LayoutTests:
The ordering of the repaints changes.
blend-mode-turn-off-isolation-no-effect.html now issues a repaint, which is expected now.
Isolation is so rare that this is not a problem.
* css3/blending/repaint/blend-mode-isolate-stacking-context-expected.txt:
* css3/blending/repaint/blend-mode-turn-off-isolation-expected.txt:
* css3/blending/repaint/blend-mode-turn-off-isolation-no-effect-expected.txt:
* fast/repaint/absolute-position-change-containing-block-expected.txt:
Modified Paths
Diff
Modified: trunk/LayoutTests/ChangeLog (236631 => 236632)
--- trunk/LayoutTests/ChangeLog 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/LayoutTests/ChangeLog 2018-09-29 01:53:34 UTC (rev 236632)
@@ -1,3 +1,20 @@
+2018-09-28 Simon Fraser <[email protected]>
+
+ RenderLayer::removeOnlyThisLayer() should not call updateLayerPositions()
+ https://bugs.webkit.org/show_bug.cgi?id=190093
+
+ Reviewed by Dean Jackson and Zalan Bujtas.
+
+ The ordering of the repaints changes.
+
+ blend-mode-turn-off-isolation-no-effect.html now issues a repaint, which is expected now.
+ Isolation is so rare that this is not a problem.
+
+ * css3/blending/repaint/blend-mode-isolate-stacking-context-expected.txt:
+ * css3/blending/repaint/blend-mode-turn-off-isolation-expected.txt:
+ * css3/blending/repaint/blend-mode-turn-off-isolation-no-effect-expected.txt:
+ * fast/repaint/absolute-position-change-containing-block-expected.txt:
+
2018-09-28 Devin Rousso <[email protected]>
Web Inspector: crash in InspectorNetworkAgent::didReceiveResponse when loading denied x-frame resources
Modified: trunk/LayoutTests/css3/blending/repaint/blend-mode-isolate-stacking-context-expected.txt (236631 => 236632)
--- trunk/LayoutTests/css3/blending/repaint/blend-mode-isolate-stacking-context-expected.txt 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/LayoutTests/css3/blending/repaint/blend-mode-isolate-stacking-context-expected.txt 2018-09-29 01:53:34 UTC (rev 236632)
@@ -20,7 +20,6 @@
(rect 28 290 60 60)
(rect 48 290 60 60)
(rect 28 290 60 60)
- (rect 48 290 60 60)
(rect 28 526 60 60)
(rect 48 526 60 60)
(rect 48 408 60 60)
@@ -28,10 +27,11 @@
(rect 48 644 60 60)
(rect 68 644 60 60)
(rect 48 644 60 60)
- (rect 68 644 60 60)
(rect 28 290 60 60)
(rect 48 644 60 60)
+ (rect 48 290 60 60)
(rect 28 526 60 60)
(rect 88 644 20 60)
+ (rect 68 644 60 60)
)
Modified: trunk/LayoutTests/css3/blending/repaint/blend-mode-turn-off-isolation-expected.txt (236631 => 236632)
--- trunk/LayoutTests/css3/blending/repaint/blend-mode-turn-off-isolation-expected.txt 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/LayoutTests/css3/blending/repaint/blend-mode-turn-off-isolation-expected.txt 2018-09-29 01:53:34 UTC (rev 236632)
@@ -3,7 +3,8 @@
(repaint rects
(rect 8 68 100 100)
(rect 58 68 100 100)
+ (rect 8 68 100 100)
+ (rect 8 68 100 100)
(rect 58 68 100 100)
- (rect 8 68 100 100)
)
Modified: trunk/LayoutTests/css3/blending/repaint/blend-mode-turn-off-isolation-no-effect-expected.txt (236631 => 236632)
--- trunk/LayoutTests/css3/blending/repaint/blend-mode-turn-off-isolation-no-effect-expected.txt 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/LayoutTests/css3/blending/repaint/blend-mode-turn-off-isolation-no-effect-expected.txt 2018-09-29 01:53:34 UTC (rev 236632)
@@ -1,3 +1,6 @@
This test checks that removing isolation from an element being stacking context for other reasons will not trigger any repaint.
+(repaint rects
+ (rect 8 68 100 100)
+)
Modified: trunk/LayoutTests/fast/repaint/absolute-position-change-containing-block-expected.txt (236631 => 236632)
--- trunk/LayoutTests/fast/repaint/absolute-position-change-containing-block-expected.txt 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/LayoutTests/fast/repaint/absolute-position-change-containing-block-expected.txt 2018-09-29 01:53:34 UTC (rev 236632)
@@ -2,11 +2,11 @@
(rect 8 5000 100 100)
(rect 108 5100 100 100)
(rect 8 5000 100 100)
- (rect 108 5100 100 100)
- (rect 100 100 100 100)
(rect 8 8 784 2000)
(rect 16 5008 100 100)
(rect 16 5008 100 100)
(rect 8 8 100 100)
+ (rect 108 5100 100 100)
+ (rect 100 100 100 100)
)
Modified: trunk/Source/WebCore/ChangeLog (236631 => 236632)
--- trunk/Source/WebCore/ChangeLog 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/Source/WebCore/ChangeLog 2018-09-29 01:53:34 UTC (rev 236632)
@@ -1,3 +1,28 @@
+2018-09-28 Simon Fraser <[email protected]>
+
+ RenderLayer::removeOnlyThisLayer() should not call updateLayerPositions()
+ https://bugs.webkit.org/show_bug.cgi?id=190093
+
+ Reviewed by Dean Jackson and Zalan Bujtas.
+
+ It's wrong for RenderLayer::removeOnlyThisLayer() to call updateLayerPositions(),
+ because this is called at style update time, and layout will be stale.
+
+ It was added (see webkit.org/b/25252) so that opacity changes, which can destroy layers, correctly update
+ descendants. However, RenderStyle::changeRequiresLayout() checks for opacity <=> no opacity
+ changes and triggers layout accordingly, which will result in a full post-layout
+ updateLayerPositions().
+
+ This also revealed that changes to the "isolate" property fail to trigger any kind of style recalc or layout;
+ we need it to trigger layout (for now) because it affects z-order.
+
+ Covered by existing tests.
+
+ * rendering/RenderLayer.cpp:
+ (WebCore::RenderLayer::removeOnlyThisLayer):
+ * rendering/style/RenderStyle.cpp:
+ (WebCore::RenderStyle::changeRequiresLayout const):
+
2018-09-28 Jiewen Tan <[email protected]>
[WebAuthN] Polish WebAuthN auto-test environment
Modified: trunk/Source/WebCore/rendering/RenderLayer.cpp (236631 => 236632)
--- trunk/Source/WebCore/rendering/RenderLayer.cpp 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/Source/WebCore/rendering/RenderLayer.cpp 2018-09-29 01:53:34 UTC (rev 236632)
@@ -1751,9 +1751,6 @@
removeChild(current);
m_parent->addChild(current, nextSib);
current->setRepaintStatus(NeedsFullRepaint);
- // updateLayerPositions depends on hasLayer() already being false for proper layout.
- ASSERT(!renderer().hasLayer());
- current->updateLayerPositions(); // FIXME: use geometry map.
current = next;
}
Modified: trunk/Source/WebCore/rendering/style/RenderStyle.cpp (236631 => 236632)
--- trunk/Source/WebCore/rendering/style/RenderStyle.cpp 2018-09-29 01:38:26 UTC (rev 236631)
+++ trunk/Source/WebCore/rendering/style/RenderStyle.cpp 2018-09-29 01:53:34 UTC (rev 236632)
@@ -784,6 +784,13 @@
return true;
}
+#if ENABLE(CSS_COMPOSITING)
+ if (m_rareNonInheritedData->isolation != other.m_rareNonInheritedData->isolation) {
+ // Ideally this would trigger a cheaper layout that just updates layer z-order trees (webit.org/b/190088).
+ return true;
+ }
+#endif
+
if (m_rareNonInheritedData->hasFilters() != other.m_rareNonInheritedData->hasFilters())
return true;
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
