Changeset 220699 in webkit

Timestamp:

Aug 14, 2017 9:37:01 AM (7 years ago)

Author:

zandobersek@gmail.com

Message:

REGRESSION(r220517-r220521) [GTK] Various compositing tests fail
​https://bugs.webkit.org/show_bug.cgi?id=175478

Reviewed by Sergio Villar Senin.

Source/WebCore:

Fix a regression introduced in r220519. Move a framebuffer binding in TextureMapperGL
away from GraphicsContext3D and onto the OpenGL API to work around the framebuffer
binding state that's kept in the GraphicsContext3D class. This mismatch in how the
framebuffer was bound was causing a couple of test failures in compositing tests.

No new tests -- regressed tests have their failure expectations removed.

• platform/graphics/texmap/TextureMapperGL.cpp:

(WebCore::TextureMapperGL::bindDefaultSurface): Don't use GraphicsContext3D for the
glBindFramebuffer() call. This would be done in subsequent patches that would move
TextureMapperGL off the GraphicsContext3D usage, but this specific use case clashes
with the framebuffer binding that's done in BitmapTextureGL and was causing test
failures.

LayoutTests:

• platform/gtk/TestExpectations: Remove failure and flakiness expectations.

Location:

trunk

Files:

• LayoutTests/ChangeLog (modified) (1 diff)
• LayoutTests/platform/gtk/TestExpectations (modified) (2 diffs)
• Source/WebCore/ChangeLog (modified) (1 diff)
• Source/WebCore/platform/graphics/texmap/TextureMapperGL.cpp (modified) (1 diff)

trunk/LayoutTests/ChangeLog

@@ +1 @@
+2017-08-14  Zan Dobersek  <zdobersek@igalia.com>
+
+        REGRESSION(r220517-r220521) [GTK] Various compositing tests fail
+        https://bugs.webkit.org/show_bug.cgi?id=175478
+
+        Reviewed by Sergio Villar Senin.
+
+        * platform/gtk/TestExpectations: Remove failure and flakiness expectations.
+
 2017-08-14  Ms2ger  <Ms2ger@igalia.com>
 

trunk/LayoutTests/platform/gtk/TestExpectations

@@ -1897 +1897 @@
 webkit.org/b/174354 imported/w3c/IndexedDB-private-browsing/idbfactory_open.html [ Crash Pass ]
 
-webkit.org/b/175478 compositing/images/positioned-image-content-rect.html [ Pass ImageOnlyFailure ]
-webkit.org/b/175478 compositing/images/truncated-direct-png-image.html [ Pass ImageOnlyFailure ]
-
 #////////////////////////////////////////////////////////////////////////////////////////
 # End of Flaky tests
@@ -3415 +3412 @@
 webkit.org/b/175422 imported/w3c/web-platform-tests/beacon/headers/header-content-type.html [ Failure ]
 
-webkit.org/b/175478 compositing/contents-opaque/opaque-with-clip-path.html [ ImageOnlyFailure ]
-webkit.org/b/175478 compositing/masks/compositing-clip-path-and-mask.html [ ImageOnlyFailure ]
-webkit.org/b/175478 compositing/masks/compositing-clip-path.html [ ImageOnlyFailure ]
-webkit.org/b/175478 compositing/masks/compositing-clip-path-origin.html [ ImageOnlyFailure ]
-webkit.org/b/175478 compositing/masks/reference-clip-path-on-composited.html [ ImageOnlyFailure ]
-webkit.org/b/175478 css3/filters/animation-from-initial-values-with-color-matrix.html [ ImageOnlyFailure ]
-
 #////////////////////////////////////////////////////////////////////////////////////////
 # End of non-crashing, non-flaky tests failing

trunk/Source/WebCore/ChangeLog

@@ +1 @@
+2017-08-14  Zan Dobersek  <zdobersek@igalia.com>
+
+        REGRESSION(r220517-r220521) [GTK] Various compositing tests fail
+        https://bugs.webkit.org/show_bug.cgi?id=175478
+
+        Reviewed by Sergio Villar Senin.
+
+        Fix a regression introduced in r220519. Move a framebuffer binding in TextureMapperGL
+        away from GraphicsContext3D and onto the OpenGL API to work around the framebuffer
+        binding state that's kept in the GraphicsContext3D class. This mismatch in how the
+        framebuffer was bound was causing a couple of test failures in compositing tests.
+
+        No new tests -- regressed tests have their failure expectations removed.
+
+        * platform/graphics/texmap/TextureMapperGL.cpp:
+        (WebCore::TextureMapperGL::bindDefaultSurface): Don't use GraphicsContext3D for the
+        glBindFramebuffer() call. This would be done in subsequent patches that would move
+        TextureMapperGL off the GraphicsContext3D usage, but this specific use case clashes
+        with the framebuffer binding that's done in BitmapTextureGL and was causing test
+        failures.
+
 2017-08-14  Chris Dumez  <cdumez@apple.com>
 

trunk/Source/WebCore/platform/graphics/texmap/TextureMapperGL.cpp

@@ -650 +650 @@
 void TextureMapperGL::bindDefaultSurface()
 {
-    m_context3D->bindFramebuffer(GraphicsContext3D::FRAMEBUFFER, data().targetFrameBuffer);
+    glBindFramebuffer(GL_FRAMEBUFFER, data().targetFrameBuffer);
     auto& viewport = data().viewport;
     data().projectionMatrix = createProjectionMatrix(IntSize(viewport[2], viewport[3]), data().PaintFlags & PaintingMirrored);