Bug 234145 – When transform-style: preserve-3d is used with a grouping property it should still create a containing block

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

[Martin Robinson]

https://drafts.csswg.org/css-transforms-2/#interpolation-of-transform-functions says:

> The use of this property with any value other than none establishes a stacking context. It also establishes a containing block for all descendants, just like the transform property does.

Currently we are handling this via the StyleAdjuster, which is why preserve-3d-flat-grouping-properties-containing-block.html is failing the WPT test suite.

[Radar WebKit Bug Importer]

<rdar://problem/86630387>

[Martin Robinson]

Created attachment 448397 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 448397 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=448397&action=review

> Source/WebCore/rendering/style/RenderStyle.cpp:1188
> +        if (usedTransformStyle3D() != other.usedTransformStyle3D()

I wonder if this will be a perf regression, since usedTransformStyle3D() does more work now. This code is pretty hot. Might be better to keep the /StyleAdjuster code but have it set a usedTransformStyle field.

[Martin Robinson]

Created attachment 448501 [details]
Patch

[Martin Robinson]

(In reply to Simon Fraser (smfr) from comment #3)
> Comment on attachment 448397 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=448397&action=review
> 
> > Source/WebCore/rendering/style/RenderStyle.cpp:1188
> > +        if (usedTransformStyle3D() != other.usedTransformStyle3D()
> 
> I wonder if this will be a perf regression, since usedTransformStyle3D()
> does more work now. This code is pretty hot. Might be better to keep the
> /StyleAdjuster code but have it set a usedTransformStyle field.

Thanks for the review!

The new patch takes the approach you suggest. I ended up adding a boolean field to StyleRareNonInheritedData because that makes managing style changes a little more straight-forward (and saves a bit). I can probably change this to be an enum field if that's a better path. In addition, I've rewritten the existing tests that checked this behavior. Now they test both the used and computed value.

[Martin Robinson]

Created attachment 448503 [details]
Patch

[EWS]

Committed r287742 (245812@main): <https://commits.webkit.org/245812@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 448503 [details].