CSS attribute selectors cause unnecessary style recalc when setting attribute to same value.

• Component: CSS
• Status: RESOLVED FIXED

[Andreas Kling]

Setting an attribute to the same value it already has should never trigger a style recalc due to attribute selectors.
The code is currently too dumb to figure this out, and ends up restyling the same stuff over and over.

[Radar WebKit Bug Importer]

<rdar://problem/13727709>

[Andreas Kling]

Created attachment 199498 [details]
Proposed patch

[Simon Fraser (smfr)]

Comment on attachment 199498 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=199498&action=review

> Source/WebCore/ChangeLog:13
> +        This reduces unnecessary style recalculation in MAS content.

"MAS" is pretty meaningless in this context.

[Antti Koivisto]

Comment on attachment 199498 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=199498&action=review

Better fixes involve some significant refactoring, this is ok for now.

> Source/WebCore/dom/Element.cpp:-878
>      else if (name == HTMLNames::pseudoAttr)
>          shouldInvalidateStyle |= testShouldInvalidateStyle && isInShadowTree();
>  
> -    shouldInvalidateStyle |= testShouldInvalidateStyle && styleResolver->hasSelectorForAttribute(name.localName());

It seem wrong that attributeChanged() is called when attribute does not change. It would be better to fix that.

> Source/WebCore/dom/Element.cpp:2726
> +    if (oldValue != newValue) {
> +        if (attached() && document()->styleResolver() && document()->styleResolver()->hasSelectorForAttribute(name.localName()))
> +            setNeedsStyleRecalc();
> +    }

Bit sad that the invalidation code gets spread around like this.

[Andreas Kling]

Committed r149047: <http://trac.webkit.org/changeset/149047>