Bug 213171 – [WPT] infrastructure/assumptions/html-elements.html fails due to changes in style when all: initial is used

• Status: RESOLVED
• Resolution: FIXED
• Component: CSS

[Sam Weinig]

Created attachment 401845 [details]
Testcase

WPT: infrastructure/assumptions/html-elements.html fails due to changes in style when all: initial is used.

Test: https://wpt.fyi/results/infrastructure/assumptions/html-elements.html?run_id=586650004&run_id=579020002&run_id=594530003&run_id=583030005

Issue: For the simple tree:

<style>
span.b {
  all: initial;
}
</style>
<span class="a"><span>
<span class="b"><span>

The computed values of the two spans differ for two properties: "stroke-color" and "-webkit-mask-box-image-slice"

Attached reduction / reproduction which tests all the computed properties for both spans. (Unselect the "Pass" checkbox to just see the two failing cases).

[Sam Weinig]

The two failing cases are:

- Different value for stroke-color expected "rgb(0, 0, 0)" but got "rgba(0, 0, 0, 0)".
- Different value for -webkit-mask-box-image-slice expected "0" but got "0 fill".

[Sam Weinig]

For stroke-color, the first thing that pops out to me is that in CSSProperties.json, "stroke-color" is set to: 

    "initial": "currentColor"

but, RenderStyle::initialStrokeColor() is set to:

    static Color initialStrokeColor() { return Color::transparent; }

[Antti Koivisto]

Yeah, seems like a bug. The spec indicates 'transparent' is the correct value so I suppose CSSProperties.json is wrong here.

[Sam Weinig]

(In reply to Antti Koivisto from comment #3)
> Yeah, seems like a bug. The spec indicates 'transparent' is the correct
> value so I suppose CSSProperties.json is wrong here.

Ok. Cool, that's easy to fix. For -webkit-mask-box-image-slice, it was a bit harder to track down due to macros, but it looks like it gets the fill = false from ApplyPropertyBorderImageModifier::applyInitialValue() while NinePieceImage when initialized with Type::Mask (as it is in StyleRareNonInheritedData) gets initialized with fill = true.

My guess is to match the one in NinePieceImage.

[Sam Weinig]

Created attachment 401939 [details]
WIP

[Sam Weinig]

Created attachment 401966 [details]
Patch

[Antti Koivisto]

Comment on attachment 401966 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=401966&action=review

> Source/WebCore/css/CSSProperties.json:3734
> +                "initial": "initialStrokeColor",

I think calling RenderStyle::initialFoo is the default so just not specifying anything here should work too.

> Source/WebCore/style/StyleBuilderCustom.h:543
> -            image.setFill(false);
> +            image.setFill(type == BorderImage ? false : true);

type != BorderImage

[Sam Weinig]

Created attachment 402110 [details]
Patch

[EWS]

Committed r263156: <https://trac.webkit.org/changeset/263156>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 402110 [details].

[Radar WebKit Bug Importer]

<rdar://problem/64451750>

[Sam Weinig]

*** Bug 187052 has been marked as a duplicate of this bug. ***