Bug 258229 – Don't cap the FEGaussian max kernel size

• Status: NEW
• Component: SVG

[Ahmad Saleem]

Hi Team,

While going through WebKit's commit, I came across another test case where all browsers are different and by merging Blink commit, we start doing bit better (IMO?).

Blink Commit: https://src.chromium.org/viewvc/blink?view=revision&revision=186886

WebKit SOurce: https://searchfox.org/wubkat/source/Source/WebCore/platform/graphics/filters/FEGaussianBlur.cpp#87

^ Remove Line 87 and change Line 92 as below:

return clampTo<int>(size);

_______

I think getting some input from Simon, Said and Nikolas would be good.

Thanks!

[Ahmad Saleem]

Created attachment 466720 [details]
All browsers result in test case

[Radar WebKit Bug Importer]

<rdar://problem/111241971>

[Ahmad Saleem]

src source is gone so other link - https://chromium.googlesource.com/chromium/src.git/+/76892fe354a50a6a33f17759a2ebc7269eb971f6

Firefox also matches Chromium now, so I think it is worth trying to match other browsers.

```
IntSize FEGaussianBlur::calculateUnscaledKernelSize(FloatSize stdDeviation)
{
    ASSERT(stdDeviation.width() >= 0 && stdDeviation.height() >= 0);
    IntSize kernelSize;

    if (stdDeviation.width()) {
        auto size = std::max<unsigned>(2, static_cast<unsigned>(floorf(stdDeviation.width() * gaussianKernelFactor() + 0.5f)));
        kernelSize.setWidth(size);
    }

    if (stdDeviation.height()) {
        auto size = std::max<unsigned>(2, static_cast<unsigned>(floorf(stdDeviation.height() * gaussianKernelFactor() + 0.5f)));
        kernelSize.setHeight(size);
    }

    return kernelSize;
}
```

and delete 'clampedToKernelSize' completely.

[Simon Fraser (smfr)]

That will kill performance in some cases.