Changeset 110741 in webkit

Timestamp:

Mar 14, 2012 1:52:42 PM (12 years ago)

Author:

jamesr@google.com

Message:

[Chromium] Layout Test compositing/repaint/opacity-between-absolute.html is flaky
​https://bugs.webkit.org/show_bug.cgi?id=79823

Reviewed by Adrienne Walker.

Source/WebCore:

LayerRendererChromium was storing a weak pointer to the current render surface in m_currentRenderSurface and
using this in useRenderSurface() to avoid rebinding if we called useRenderSurface() multiple times in a row on
the same surface. This pointer was never cleared, so if any subsequent surface landed at the same address this
caching would misbehave and we'd fail to correctly initialize the new render surface. The caching wasn't
actually buying us anything anyway since we only call useRenderSurface() once per surface per frame and we
always set the viewport for the default render surface.

• platform/graphics/chromium/LayerRendererChromium.cpp:

(WebCore::LayerRendererChromium::beginDrawingFrame):
(WebCore::LayerRendererChromium::useRenderSurface):

LayoutTests:

• platform/chromium/test_expectations.txt:

Location:

trunk

Files:

• LayoutTests/ChangeLog (modified) (2 diffs)
• LayoutTests/platform/chromium/test_expectations.txt (modified) (6 diffs)
• Source/WebCore/ChangeLog (modified) (1 diff)
• Source/WebCore/platform/graphics/chromium/LayerRendererChromium.cpp (modified) (2 diffs)

trunk/LayoutTests/ChangeLog

@@ -1 @@
-2012-03-14  Jessie Berlin  <jberlin@apple.com>
-
-        WKTR doesn't have an implemenation of hasSpellingMarker yet.
-
-        Add more tests that rely on it to the WK2 Skipped list.
-
-        * platform/wk2/Skipped:
+2012-03-14  James Robinson  <jamesr@chromium.org>
+
+        [Chromium] Layout Test compositing/repaint/opacity-between-absolute.html is flaky
+        https://bugs.webkit.org/show_bug.cgi?id=79823
+
+        Reviewed by Adrienne Walker.
+
+        * platform/chromium/test_expectations.txt:
 
 2012-03-14  Zalan Bujtas  <zbujtas@gmail.com>
@@ -107 +108 @@
 
         * platform/chromium/test_expectations.txt:
+
+2012-03-14  Jessie Berlin  <jberlin@apple.com>
+
+        WKTR doesn't have an implemenation of hasSpellingMarker yet.
+
+        Add more tests that rely on it to the WK2 Skipped list.
+
+        * platform/wk2/Skipped:
 
 2012-03-14  Jessie Berlin  <jberlin@apple.com>

trunk/LayoutTests/platform/chromium/test_expectations.txt

@@ -2615 +2615 @@
 BUGWK47949 MAC : compositing/reflections/nested-reflection-anchor-point.html = IMAGE
 BUGWK47949 LINUX WIN : compositing/reflections/reflection-opacity.html = IMAGE+TEXT
+
 // The following are flaky on Mac 10.6; occasionally the layers fail to show up.
 BUGWK47949 MAC : compositing/reflections/nested-reflection.html = IMAGE PASS
-BUGWK47949 : compositing/reflections/nested-reflection-transformed.html = IMAGE PASS
 BUGWK47949 MAC : compositing/reflections/reflection-opacity.html = IMAGE PASS
 BUGWK47949 : compositing/reflections/reflection-positioning.html = IMAGE PASS
-BUGWK47949 : compositing/reflections/nested-reflection-transformed2.html = IMAGE PASS
 BUGWK47949 : compositing/reflections/nested-reflection-transition.html = IMAGE PASS
 BUGWK47949 MAC : compositing/reflections/nested-reflection-size-change.html = IMAGE PASS
 BUGWK79647 WIN MAC : compositing/reflections/reflection-ordering.html = IMAGE PASS
 BUGWK79647 WIN LINUX : compositing/reflections/remove-add-reflection.html = IMAGE PASS
+
+// Need rebaselines
+BUGWK47949 : compositing/reflections/nested-reflection-transformed.html = IMAGE
+BUGWK47949 : compositing/reflections/nested-reflection-transformed2.html = IMAGE
 
 // Flaky because these are a poorly written tests. Need dino's new animation API
@@ -3539 +3542 @@
 BUGWK74634 WIN : compositing/absolute-position-changed-with-composited-parent-layer.html = IMAGE
 
-BUGWK74731 : compositing/reflections/reflection-positioning2.html = PASS IMAGE
-
 BUGWK74746 : fast/filesystem/workers/file-from-file-entry-sync.html = PASS CRASH
 BUGWK74746 : fast/filesystem/workers/file-entry-to-uri-sync.html = PASS CRASH
@@ -3672 +3673 @@
 
 // Will need rebaselines for mock scrollbars
-
-// Flaky, needs investigation
-BUGJAMESR : compositing/masks/masked-ancestor.html = IMAGE PASS
 
 BUGWK77534 LINUX : plugins/createScriptableObject-before-start.html = PASS MISSING
@@ -3774 +3772 @@
 BUGWK79540 LEOPARD DEBUG : fast/files/workers/worker-read-file-async.html = PASS CRASH
 
-BUGWK79572 : css3/filters/effect-drop-shadow-hw.html = PASS IMAGE
-BUGWK80007 : css3/filters/filtered-compositing-descendant.html = PASS IMAGE
-
 BUGWK79642 : accessibility/aria-describedby-on-input.html = TIMEOUT
 BUGWK79642 : fast/loader/subresource-willSendRequest-null.html = TIMEOUT
@@ -3782 +3777 @@
 BUGWK79702 : fast/regex/lastIndex.html = TEXT
 BUGWK79703 : fast/regex/overflow.html = TEXT
-
-BUGWK79823 : compositing/repaint/opacity-between-absolute.html = PASS IMAGE
-BUGWK79823 : compositing/reflections/masked-reflection-on-composited.html = PASS IMAGE
-BUGWK79823 : compositing/reflections/reflection-on-composited.html = PASS IMAGE
-BUGWK79823 : compositing/geometry/fixed-position.html = PASS IMAGE
-BUGWK79823 : compositing/reflections/compositing-change-inside-reflection.html = PASS IMAGE
-BUGWK79823 : compositing/webgl/webgl-reflection.html = PASS IMAGE
-BUGWK79823 : compositing/masks/direct-image-mask.html = PASS IMAGE
-BUGWK79823 : platform/chromium/compositing/render-surface-alpha-blending.html = PASS IMAGE
-BUGWK79823 : platform/chromium/compositing/child-layer-3d-sorting.html = PASS IMAGE
-BUGWK79823 : platform/chromium/compositing/plugins/webplugin-reflection.html = PASS IMAGE
-
-// Already associated with bug 47949 on Mac.
-BUGWK79823 WIN LINUX : compositing/reflections/nested-reflection.html = PASS IMAGE
 
 BUGWK79862 : fast/js/primitive-property-access-edge-cases.html = TEXT
@@ -3888 +3869 @@
 BUGWK80743 WIN MAC : compositing/direct-image-compositing.html = PASS IMAGE
 BUGWK80743 WIN MAC : platform/chromium/compositing/huge-layer-rotated.html = PASS IMAGE
-
-BUGWK80563 : css3/filters/effect-combined-hw.html = PASS IMAGE
-BUGWK80563 : css3/filters/effect-hue-rotate-hw.html = PASS IMAGE
-BUGWK80563 : css3/filters/effect-opacity-hw.html = PASS IMAGE
-BUGWK80563 : compositing/reflections/nested-reflection-opacity.html = PASS IMAGE
 
 BUGWK80665 : fast/js/string-match.html = TEXT

trunk/Source/WebCore/ChangeLog

@@ +1 @@
+2012-03-14  James Robinson  <jamesr@chromium.org>
+
+        [Chromium] Layout Test compositing/repaint/opacity-between-absolute.html is flaky
+        https://bugs.webkit.org/show_bug.cgi?id=79823
+
+        Reviewed by Adrienne Walker.
+
+        LayerRendererChromium was storing a weak pointer to the current render surface in m_currentRenderSurface and
+        using this in useRenderSurface() to avoid rebinding if we called useRenderSurface() multiple times in a row on
+        the same surface. This pointer was never cleared, so if any subsequent surface landed at the same address this
+        caching would misbehave and we'd fail to correctly initialize the new render surface. The caching wasn't
+        actually buying us anything anyway since we only call useRenderSurface() once per surface per frame and we
+        always set the viewport for the default render surface.
+
+        * platform/graphics/chromium/LayerRendererChromium.cpp:
+        (WebCore::LayerRendererChromium::beginDrawingFrame):
+        (WebCore::LayerRendererChromium::useRenderSurface):
+
 2012-03-14  Zalan Bujtas  <zbujtas@gmail.com>
 

trunk/Source/WebCore/platform/graphics/chromium/LayerRendererChromium.cpp

@@ -390 +390 @@
 
     makeContextCurrent();
-    // The GL viewport covers the entire visible area, including the scrollbars.
-    GLC(m_context.get(), m_context->viewport(0, 0, viewportWidth(), viewportHeight()));
-    m_windowMatrix = screenMatrix(0, 0, viewportWidth(), viewportHeight());
-
     // Bind the common vertex attributes used for drawing all the layers.
     m_sharedGeometry->prepareForDraw();
@@ -1144 +1140 @@
 bool LayerRendererChromium::useRenderSurface(CCRenderSurface* renderSurface)
 {
-    if (m_currentRenderSurface == renderSurface)
-        return true;
-
     m_currentRenderSurface = renderSurface;