Bug 244025 – `:visited` link color does not propagate to SVG through `currentColor`

• Component: CSS
• Status: RESOLVED FIXED

[Masataka Yakura]

Created attachment 461688 [details]
test case - `:visited` + `fill="currentColor"`

Steps to reproduce:

1. open attachment or go to https://myakura.github.io/test--visited-currentcolor-svg/
2. look at the test case


Expected result:

There is a green square.


Actual result:

There is a red square, in Safari (15.6) and Safari TP (151). It is green in Firefox and Chrome.


Additional comments:

People build customizable icon system by utilizing `fill="currentColor"` set on inline SVGs.

* https://css-tricks.com/cascading-svg-fill-color/
* https://mayashavin.com/articles/svg-icons-currentcolor
* https://gomakethings.com/currentcolor-and-svgs/

However, WebKit doesn't propagate colors set on `:visited` elements. So if such icons are placed inside links, they won't look as expected.

https://stackoverflow.com/questions/68116594/safari-not-showing-visited-color-for-svg-icon-that-is-inside-an-anchor-tag

[Masataka Yakura]

It may be a part of https://bugs.webkit.org/show_bug.cgi?id=210005 .

[Radar WebKit Bug Importer]

<rdar://problem/98776770>

[Karl Dubost]

Created attachment 476445 [details]
rendering in safari, firefox, chrome

Safari Technology Preview  26.0          20622.1.20.1 (125)
Firefox Nightly            143.0a1       14325.8.14
Google Chrome Canary       141.0.7364.1  7364.1

[Karl Dubost]

Both Firefox and Safari shows a red square.

[Ahmad Saleem]

It is quite easy to fix:

https://searchfox.org/wubkat/rev/7da001264416843beaa28b96f6f7f859950ce82d/Source/WebCore/rendering/svg/legacy/LegacyRenderSVGResource.cpp#80

```
// For `currentcolor`, 'color' already contains the 'visitedColor'.
        if (auto visitedPaintColor = visitedPaint.tryColor(); visitedPaintColor && !visitedPaintColor->isCurrentColor()) {
            if (auto visitedColor = style.colorResolvingCurrentColor(*visitedPaintColor); visitedColor.isValid())
                color = visitedColor.colorWithAlpha(color.alphaAsFloat());
        }
```

Change to:

```


```

        if (auto visitedPaintColor = visitedPaint.tryColor()) {
            if (visitedPaintColor->isCurrentColor()) {
                color = style.visitedLinkColor();
            } else {
                if (auto visitedColor = style.colorResolvingCurrentColor(*visitedPaintColor); visitedColor.isValid())
                    color = visitedColor.colorWithAlpha(color.alphaAsFloat());
            }
        }

```

It progresses attached test case.

[Ahmad Saleem]

Pull request: https://github.com/WebKit/WebKit/pull/54500

[Karl Dubost]

Created attachment 477542 [details]
testcase with impact on getComputedStyle color

This is a testcase to make sure the colors are not leaked through getComputedStyle.

[Karl Dubost]

Created attachment 477543 [details]
rendering in safari, firefox, chrome for visited color test with getCS

if my test is right the colors are currently not being leaked inside Chrome.
Firefox doesn't apply the :visited color to the rect
Safari too hence this bug.

[Ahmad Saleem]

@Karl - I think getComputedStyle might not work due to https://developer.mozilla.org/en-US/docs/Web/CSS/Guides/Selectors/Privacy_and_:visited

```
To preserve users' privacy, browsers lie to web applications under certain circumstances:

The window.getComputedStyle method and similar functions, such as element.querySelector, always return values indicating that a user has never visited any of the links on a page.
```

[Ahmad Saleem]

One second thought, I found this - color-mix-currentcolor-visited-getcomputedstyle.html WPT test:

WPT Test case - https://wpt.fyi/results/css/css-color/color-mix-currentcolor-visited-getcomputedstyle.html?label=experimental&label=master&aligned

I can easily convert this for SVG case and add it as well with my PR - if my PR regresses this then we will not merge it else it would be good to get another test coverage plus fix this bug. Will update my PR.

[EWS]

Committed 306387@main (8be100449e65): <https://commits.webkit.org/306387@main>

Reviewed commits have been landed. Closing PR #54500 and removing active labels.