[chromium] Clipping/Transforms applied in wrong order in opaque paint tracking

• Component: New Bugs
• Status: RESOLVED FIXED

[Dana Jansens]

[chromium] Clipping/Transforms applied in wrong order in opaque paint tracking

[Dana Jansens]

The clip is applied in device coordinates, before transforming the painted rect into device coordinates. This makes any translations get doubly represented, and gives incorrect paint tracking results.

[Dana Jansens]

Created attachment 127301 [details]
Patch

[WebKit Review Bot]

Comment on attachment 127301 [details]
Patch

Attachment 127301 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/11534680

New failing tests:
compositing/scrollbar-painting.html
compositing/overflow/clip-content-under-overflow-controls.html

[Dana Jansens]

(I edited the scrollbar-painting.html test to remove the body margin/padding for easier numbers.)

===

Before this change, we see the draw in the bottom left corner as:

diddraw 85.000000 85.000000 15.000000x15.000000

The device clip is 0,0 15x15. So the draw is ignored for opaque tracking as it assumes it is not drawn at all.

===

After this change, we see it as:

diddraw 85.000000 85.000000 15.000000x15.000000
canvasTransform 0.000000 0.000000 15.000000x15.000000 1
clipped 0.000000 0.000000 15.000000x15.000000 1
targetTransform 0.000000 0.000000 15.000000x15.000000 1
opaque 1

ie. The transform puts us at 0,0 where the clip is applied, and the opaque paint can be used now for opaque tracking (and turning off blending). Making this our lovely off-by-1 friend again.

===

The paint in question here is

#0  WebCore::OpaqueRegionSkia::markRectAsOpaque (this=0x7fffde4523a0, rect=...) at ../../third_party/WebKit/Source/WebCore/platform/graphics/skia/OpaqueRegionSkia.cpp:293
#1  0x00007fffedcea9a2 in WebCore::OpaqueRegionSkia::didDraw (this=0x7fffde4523a0, context=0x7fffde452370, transform=..., rect=..., paint=..., drawsOpaque=true, fillsBounds=true) at ../../third_party/WebKit/Source/WebCore/platform/graphics/skia/OpaqueRegionSkia.cpp:271
#2  0x00007fffedcea1d8 in WebCore::OpaqueRegionSkia::didDrawRect (this=0x7fffde4523a0, context=0x7fffde452370, transform=..., fillRect=..., paint=..., bitmap=0x0) at ../../third_party/WebKit/Source/WebCore/platform/graphics/skia/OpaqueRegionSkia.cpp:169
#3  0x00007fffedcee2f6 in WebCore::PlatformContextSkia::didDrawRect (this=0x7fffde452370, rect=..., paint=..., bitmap=0x0) at ../../third_party/WebKit/Source/WebCore/platform/graphics/skia/PlatformContextSkia.cpp:611
#4  0x00007fffedce1f37 in WebCore::GraphicsContext::fillRect (this=0x7fffdd8cdaa0, rect=..., color=..., colorSpace=WebCore::ColorSpaceDeviceRGB) at ../../third_party/WebKit/Source/WebCore/platform/graphics/skia/GraphicsContextSkia.cpp:865
#5  0x00007fffed5f0f48 in WebCore::RenderLayer::paintScrollCorner (this=0x7fffe03a2438, context=0x7fffdd8cdaa0, paintOffset=..., damageRect=...) at ../../third_party/WebKit/Source/WebCore/rendering/RenderLayer.cpp:2496

Which is:
        context->fillRect(absRect, Color::white, box->style()->colorSpace());

With absRect = {85, 85, 15x15}. Clearly an opaque paint.

===

I will include new linux baselines in this CL.

[Dana Jansens]

Created attachment 127407 [details]
Patch

Includes new baselines for the off-by-1 fails. I suspect the mac platform baselines can go away after this.

[Stephen White]

Comment on attachment 127407 [details]
Patch

OK.  r=me

[WebKit Review Bot]

Comment on attachment 127407 [details]
Patch

Clearing flags on attachment: 127407

Committed r107988: <http://trac.webkit.org/changeset/107988>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.