Bug 215941 – Add ::marker { font-variant-numeric: tabular-nums; } to UA stylesheet per spec

• Status: RESOLVED
• Resolution: WORKSFORME
• Component: CSS

[Tyler Wilcock]

Quoting https://www.w3.org/TR/css-pseudo-4/#marker-pseudo:

> UAs must add the following rule to their default style sheet:
>     ::marker { font-variant-numeric: tabular-nums; }

[Tyler Wilcock]

Created attachment 407493 [details]
Patch

[Darin Adler]

Comment on attachment 407493 [details]
Patch

Code change looks great. Can we add a test, please? Surprised there is no test of this in Web Platform Tests.

[Tyler Wilcock]

Sure, I'll add one and give it another r?.

[Tyler Wilcock]

Actually, I think there is a WPT for the addition of this property, but it started passing here with the addition of `::marker { text-transform }` for some reason:

https://bugs.webkit.org/show_bug.cgi?id=214359

[Darin Adler]

That is very strange. Maybe we don’t need this change, then?

[Darin Adler]

If the test is passing that makes it sound like we might not need this code.

[Darin Adler]

It the test already passes without this fix, then there are a few different possibilities:

1) The test is not a very good test, and we need to make a better test that can detect that this is missing.

2) The test is being run incorrectly or skipped, and we need to correct how it’s being run by WebKit’s test machinery.

3) This code change isn’t actually needed for some reason.

We should not land the change without figuring this out.

[Darin Adler]

Another possibility is:

4) There is another bug in the implementation of markers that causes the test to pass. A "false negative". Once we fix that other bug then the test will start failing.

[Radar WebKit Bug Importer]

<rdar://problem/68358696>

[Oriol Brufau]

(In reply to Darin Adler from comment #7)
> 3) This code change isn’t actually needed for some reason.
Basically this one. font-variant-numeric is already applied in RenderListItem::computeMarkerStyle
https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/rendering/RenderListItem.cpp#L59

However, before bug 214359, I think this wouldn't be reflected in getComputedStyle for unstyled ::marker.

BTW, not sure if it's the same in WebKit, but directly adding ::marker in the UA sheet would have disabled some optimizations in Chromium. I had to do it more carefully. See this comment from the link above:

> Although the CSS Pseudo-Elements Module Level 4 spec. saids to add ::marker to the UA sheet we apply it here as an optimization because it only applies to markers. That is, it does not apply to all elements.

[Darin Adler]

So I think there is no bug to fix here. Resolving for now.

If we can find something wrong, then it’s fine to reopen this bug.