Regression: Shadow offsets seem to have changed in 10.5.3 in canvas

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[Francisco Tolmasky]

Having a shadow offset of 1.0,1.0 doesn't seem to produce a shadow anymore (see below).  Adding 0.01 to the shadow offsets appears to provide consistent results however.

<canvas width="100" height="100" id="c" style="background: lime"></canvas>
<script>
window.onload = function () {
    var ctx = document.getElementById('c').getContext('2d');

//    ctx.globalCompositeOperation = 'xor';
    ctx.shadowColor = 'black';
    ctx.shadowBlur = 1.0;
    ctx.shadowOffsetX = 1.0;
    ctx.shadowOffsetY = 1.0;
    ctx.fillStyle = '#f00';
    ctx.fillRect(10, 10, 20, 20);
}
</script>

CanvasRenderingContext2D::applyShadow() needs to have the same extraShadowOffset logic GraphicsContext::setPlatformShadow() has in GraphicsContextCG.cpp to guarantee consistent results.

I have a patch but I cannot generate test results because DRT crashes on every test.

[Mark Rowe (bdash)]

<rdar://problem/5977300>

Created attachment 21452 [details]
Account for changes in  CoreGraphics shadow behavior; fix recent Y-axis-direction regression

The test covers both fixes.

[Mark Rowe (bdash)]

Does SVG have some form of shadows that may require a similar fix?

[Darin Adler]

Comment on attachment 21452 [details]
Account for changes in  CoreGraphics shadow behavior; fix recent Y-axis-direction regression

r=me

Fixed in <http://trac.webkit.org/changeset/34287>.

(In reply to comment #5)
> Does SVG have some form of shadows that may require a similar fix?

No, but looking for calls to CGContextSetShadow* I just realized that canvas has many more which may need to be patched as well.

Additional related fixes landed in <http://trac.webkit.org/changeset/34317>.