Bug 130779 – FEGaussianBlur: unify and const-ify calculateKernelSize

• Component: SVG
• Status: RESOLVED FIXED

[Adenilson Cavalcanti Silva]

Some methods can benefit of using const refs as also it makes sense to unify the interface (i.e. parameters) in calculateKernelSize and calculateUnscaledKernelSize.

[Adenilson Cavalcanti Silva]

Created attachment 227857 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 227857 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=227857&action=review

> Source/WebCore/platform/graphics/filters/FEGaussianBlur.cpp:408
> +IntSize FEGaussianBlur::calculateKernelSize(Filter* filter, const FloatPoint& std)

const Filter* ?
Please rename "std" to something better (is it stdDeviation?)

[Adenilson Cavalcanti Silva]

Created attachment 227860 [details]
Patch

[Adenilson Cavalcanti Silva]

FilterEffect() provides the filter and it uses a pointer as constructor parameter (which can be null).

But we can have a const pointer, since applyVerticalScale/HorizontalScale() are const methods.

[Adenilson Cavalcanti Silva]

Created attachment 227864 [details]
Patch

[Adenilson Cavalcanti Silva]

Last patch uses const Filter*.

[Simon Fraser (smfr)]

Comment on attachment 227864 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=227864&action=review

> Source/WebCore/platform/graphics/filters/FEGaussianBlur.cpp:410
> +    FloatPoint stdFilterScaled(filter->applyHorizontalScale(stdDeviation.x()), filter->applyVerticalScale(stdDeviation.y()));

Since you dereference filter without a null check, you should make it a const Filter&

[Adenilson Cavalcanti Silva]

Created attachment 227879 [details]
Patch

[Adenilson Cavalcanti Silva]

Agreed, FilterEffect will now return a ref.

[Simon Fraser (smfr)]

Comment on attachment 227879 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=227879&action=review

> Source/WebCore/platform/graphics/filters/Filter.h:50
> +    virtual float applyHorizontalScale(const float value) const { return value * m_filterResolution.width(); }
> +    virtual float applyVerticalScale(const float value) const { return value * m_filterResolution.height(); }

no need for const float

[Adenilson Cavalcanti Silva]

Created attachment 227902 [details]
Patch

[Adenilson Cavalcanti Silva]

Removed unnecessary const.

[Simon Fraser (smfr)]

Comment on attachment 227902 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=227902&action=review

> Source/WebCore/platform/graphics/filters/Filter.h:49
> +    // TODO: verify descendants that overload this method and turn them to const.

Not sure why you need the comment. AFAICT all subclasses use "override" so you would get a build failure with mismatched signatures.

[WebKit Commit Bot]

Comment on attachment 227902 [details]
Patch

Clearing flags on attachment: 227902

Committed r166341: <http://trac.webkit.org/changeset/166341>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.