Some WPT are failing because of colors being in the wrong format

[Tim Nguyen (:ntim)]

Example: 
https://webkit-search.igalia.com/webkit/source/LayoutTests/imported/w3c/web-platform-tests/css/css-backgrounds/parsing/background-image-computed.sub-expected.txt

[Sam Weinig]

Created attachment 424555 [details]
Minimal test case

[Sam Weinig]

Looks like something (likely CSSGradientValue) is not resolving color identifiers down to Colors before serializing.

[Sam Weinig]

I haven't dropped into the debugger yet, but I am pretty sure this happening in ComputedStyleExtractor::valueForPropertyInStyle():

        case CSSPropertyBackgroundImage:
        case CSSPropertyWebkitMaskImage: {
            auto& layers = propertyID == CSSPropertyWebkitMaskImage ? style.maskLayers() : style.backgroundLayers();
            if (!layers.next()) {
                if (layers.image())
                    return layers.image()->cssValue();
                return cssValuePool.createIdentifierValue(CSSValueNone);
            }
            auto list = CSSValueList::createCommaSeparated();
            for (auto* currLayer = &layers; currLayer; currLayer = currLayer->next()) {
                if (currLayer->image())
                    list->append(currLayer->image()->cssValue());
                else
                    list->append(cssValuePool.createIdentifierValue(CSSValueNone));
            }
            return list;
        }


When the `layers.image()->cssValue()` call happens, the `image()`, which is of type CSSGradientValue, has not had its color styles resolved. You can see in the lines above it what a call with color resolving looks like:

        case CSSPropertyBackgroundColor:
            return m_allowVisitedStyle ? cssValuePool.createColorValue(style.visitedDependentColor(CSSPropertyBackgroundColor)) : currentColorOrValidColor(&style, style.backgroundColor());


So, somehow, we need to resolve those colors.

[Sam Weinig]

Antti, any ideas on a good pattern for this?

[Radar WebKit Bug Importer]

<rdar://problem/76226756>