Changeset 106191 in webkit

Timestamp:

Jan 28, 2012 11:28:44 AM (12 years ago)

Author:

jchaffraix@webkit.org

Message:

REGRESSION (r94016): Element with visibility:hidden but visible descendant may not be properly repainted
​https://bugs.webkit.org/show_bug.cgi?id=76126

Reviewed by Simon Fraser.

Source/WebCore:

Tests: fast/layers/scroll-no-visible-content-but-visible-descendant-expected.html

fast/layers/scroll-no-visible-content-but-visible-descendant.html

The optimization missed out that if the current layer does not have some visible content
(m_hasVisibleContent is false), a descendant could totally be visible and would need to
have its repaint rectangles recomputed.

• rendering/RenderLayer.cpp:

(WebCore::RenderLayer::updateLayerPositionsAfterScroll):
Bail out only if the layer doesn't have visible content AND no visible descendants.

LayoutTests:

Added testing for a layer with no visible content but one of its descendant has some visible content.
The reverse case is already profusely tested.

• fast/layers/scroll-no-visible-content-but-visible-descendant-expected.html: Added.
• fast/layers/scroll-no-visible-content-but-visible-descendant.html: Added.

Location:

trunk

Files:

• LayoutTests/ChangeLog (modified) (1 diff)
• LayoutTests/fast/layers/scroll-no-visible-content-but-visible-descendant-expected.html (added)
• LayoutTests/fast/layers/scroll-no-visible-content-but-visible-descendant.html (added)
• Source/WebCore/ChangeLog (modified) (1 diff)
• Source/WebCore/rendering/RenderLayer.cpp (modified) (1 diff)

trunk/LayoutTests/ChangeLog

@@ +1 @@
+2012-01-28  Julien Chaffraix  <jchaffraix@webkit.org>
+
+        REGRESSION (r94016): Element with visibility:hidden but visible descendant may not be properly repainted
+        https://bugs.webkit.org/show_bug.cgi?id=76126
+
+        Reviewed by Simon Fraser.
+
+        Added testing for a layer with no visible content but one of its descendant has some visible content.
+        The reverse case is already profusely tested.
+
+        * fast/layers/scroll-no-visible-content-but-visible-descendant-expected.html: Added.
+        * fast/layers/scroll-no-visible-content-but-visible-descendant.html: Added.
+
 2012-01-28 Hajime Morita  <morrita@google.com>
 

trunk/Source/WebCore/ChangeLog

@@ +1 @@
+2012-01-28  Julien Chaffraix  <jchaffraix@webkit.org>
+
+        REGRESSION (r94016): Element with visibility:hidden but visible descendant may not be properly repainted
+        https://bugs.webkit.org/show_bug.cgi?id=76126
+
+        Reviewed by Simon Fraser.
+
+        Tests: fast/layers/scroll-no-visible-content-but-visible-descendant-expected.html
+               fast/layers/scroll-no-visible-content-but-visible-descendant.html
+
+        The optimization missed out that if the current layer does not have some visible content
+        (m_hasVisibleContent is false), a descendant could totally be visible and would need to
+        have its repaint rectangles recomputed.
+
+        * rendering/RenderLayer.cpp:
+        (WebCore::RenderLayer::updateLayerPositionsAfterScroll):
+        Bail out only if the layer doesn't have visible content AND no visible descendants.
+
 2012-01-27  Chris Marrin  <cmarrin@apple.com>
 

trunk/Source/WebCore/rendering/RenderLayer.cpp

@@ -452 +452 @@
 void RenderLayer::updateLayerPositionsAfterScroll(UpdateLayerPositionsAfterScrollFlags flags)
 {
+    ASSERT(!m_visibleDescendantStatusDirty);
     ASSERT(!m_visibleContentStatusDirty);
 
-    // If we have no visible content, there is no point recomputing our rectangles as
-    // they will be empty. If our visibility changes, we are expected to recompute all
-    // our positions anyway.
-    if (!m_hasVisibleContent)
+    // If we have no visible content and no visible descendants, there is no point recomputing
+    // our rectangles as they will be empty. If our visibility changes, we are expected to
+    // recompute all our positions anyway.
+    if (!m_hasVisibleDescendant && !m_hasVisibleContent)
         return;