[css-transforms-2] Should a no-op <div> flatten?

[smfr]

How does a <div> with no specific transform-style or transform behave w.r.t. flattening?

Current spec

[mattwoodrow]

As mentioned in the meeting, gecko is currently using the DOM parent to determine if we're participating in a 3d rendering context, so no-op <div>s flatten. I'm not aware of any major compat issues with doing this.

Note that this makes things simpler since we can no longer have a <div style="overflow:hidden"> in the middle (which implies flattening, but not containing block or stacking context), and we don't have to resolve the complexities with that.

The children of the overflow:hidden div decide that they are the root of their own rendering context (since their parent doesn't have preserve-3d), flatten themselves, and then we render into the overflow:hidden as 2d.

[AmeliaBR]

I think I'm missing some context for this issue.

The Gecko behavior essential removes transform-style: auto as a value, so that the default becomes flat unless preserve-3d is explicitly specified.

I assume you still use the grouping properties to override transform-style: preserve-3d, though. Don't you still have to deal with complexities then? preserve-3d is only supposed to preserve the 3D matrix in the face of parent transforms or perspective settings; it shouldn't cancel out other factors like hidden overflow.

PS, Demo showing what we're talking about, for anyone else who was confused. In Firefox, you don't get the 3D perspective effect because of the extra semantic elements in between the images and the body with the perspective set:
https://codepen.io/AmeliaBR/pen/BmLgxj/

[AmeliaBR]

I assume...

I shouldn't have assumed. It looks like Firefox still applies the perspective from the body when there is hidden overflow plus transform-style: preserve-3d on the intermediary element, but then applies a cropping effect that seems to also be affected by the perspective? I've added various tests in the form of commented-out CSS rules in the CodePen.

[mattwoodrow]

Which version of firefox do you see that with?

We definitely need preserve-3d on the intermediate elements, but I see the perspective being disabled by overflow:hidden.

[AmeliaBR]

@mattwoodrow That was in 58. Whatever's going on, it's not the same as transform-style: flat; overflow: hidden;, which is what the specs currently say it should be.

[mattwoodrow]

Very weird, I see identical behaviour for the initial example, with the second rule un-commented, and with the second rule un-commented and modified to be explicitly transform-style: flat.
Only un-commenting the first rule results in perspective being applied for me.

Anyway, the important bit is what the desired behaviour is, and I'm still of the opinion that we should flatten for no-op divs, unless we have more evidence that it's unacceptable from a webcompat standpoint.