Summary: | REGRESSION(r113588): 15-30% perf. regression on CSS/CSSPropertySetterGetter | ||||||||
---|---|---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Ryosuke Niwa <rniwa> | ||||||
Component: | CSS | Assignee: | Andreas Kling <kling> | ||||||
Status: | RESOLVED FIXED | ||||||||
Severity: | Normal | CC: | abarth, haraken, japhet, kling, koivisto, macpherson, menard, webkit.review.bot | ||||||
Priority: | P1 | ||||||||
Version: | 528+ (Nightly build) | ||||||||
Hardware: | Unspecified | ||||||||
OS: | Unspecified | ||||||||
Bug Depends on: | 83426 | ||||||||
Bug Blocks: | |||||||||
Attachments: |
|
Description
Ryosuke Niwa
2012-04-09 19:39:41 PDT
Taking for tomorrow. We're triggering CSSOM clone instantiation from named CSS property getters in the bindings layer. Created attachment 136502 [details]
Proposed patch
Comment on attachment 136502 [details]
Proposed patch
r=me
Comment on attachment 136502 [details] Proposed patch View in context: https://bugs.webkit.org/attachment.cgi?id=136502&action=review > Source/WebCore/css/CSSComputedStyleDeclaration.cpp:2689 > +PassRefPtr<CSSValue> CSSComputedStyleDeclaration::getUnsafePropertyCSSValueInternal(CSSPropertyID propertyID) Can we reflect the idea that this CSSValue can't be exposed to the Web? e.g. getPropertyCSSValueInternalWithoutCSSOMWrapper? We should eliminate getPropertyCSSValueInternal entirely. The only clients are in the editing code and that can just access the internal values. Comment on attachment 136502 [details] Proposed patch Attachment 136502 [details] did not pass win-ews (win): Output: http://queues.webkit.org/results/12385221 Created attachment 136543 [details]
Simpler patch
Change the behavior of getPropertyCSSValueInternal() instead, yielding a much simpler patch.
Committed r113774: <http://trac.webkit.org/changeset/113774> |