[webkit-reviews] review granted: [Bug 61388] [chromium] edge aliasing on 3D transformed elements : [Attachment 98258] Layer anti-aliasing

[bugzilla-daemon at webkit.org]

Stephen White <senorblanco at chromium.org> has granted David Reveman
<reveman at google.com>'s request for review:
Bug 61388: [chromium] edge aliasing on 3D transformed elements
https://bugs.webkit.org/show_bug.cgi?id=61388

Attachment 98258: Layer anti-aliasing
https://bugs.webkit.org/attachment.cgi?id=98258&action=review

------- Additional Comments from Stephen White <senorblanco at chromium.org>
View in context: https://bugs.webkit.org/attachment.cgi?id=98258&action=review

Looks good.  r=me

> Source/WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:406
> +			      matrix.m44()).inverse();

Is this matrix always guaranteed to be invertible?  There is an "isInvertible"
function on TransformationMatrix if you need to check.

> Source/WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:494
> +		   p1 = matrix.mapPoint(p1);
> +		   p2 = matrix.mapPoint(p2);
> +
> +		   // Compute vertical edge.
> +		   edgeX = computeEdge(p1, p2, sign);

Just FYI, if there's a chance that p1 == p2, computeEdge will return something
meaningless.  Don't know if that's an issue here.

> Source/WebCore/platform/graphics/gpu/TilingData.cpp:52
> +    int numTiles = max(1, 1 + (totalSizeWithBorder - 1 - 2 * borderTexels) /
(maxTextureSize - 2 * borderTexels));

Note that this code is shared with the legacy accelerated canvas
implementation, so changes here will affect that too.  Probably academic at
this point, since it's no longer being used or tested.

> LayoutTests/platform/chromium/test_expectations.txt:3988
> +BUGWK61388 GPU : compositing/color-matching/image-color-matching.html =
IMAGE
> +BUGWK61388 GPU : compositing/direct-image-compositing.html = IMAGE
> +BUGWK61388 GPU : compositing/geometry/fixed-in-composited.html = IMAGE
> +BUGWK61388 GPU : compositing/geometry/fixed-position.html = IMAGE
> +BUGWK61388 GPU : compositing/geometry/horizontal-scroll-composited.html =
IMAGE

(etc) Please try to rebaseline these as soon as possible, so we don't lose test
coverage for very long.  Let me know if you need help with that.