[css-display] Use "computes to" instead of "behaves as" for display: contents on unusual elements.

[emilio]

https://drafts.csswg.org/css-display/#unbox says:

display: contents behaves as display: none.

Which means that the computed value of display should still be contents. This adds unnecessary complexity, and this is not how all engines implementing this section right work.

In particular, it requires a mapping from element to "is really display contents", or something of the sort. Firefox used to have that, but it caused a lot of performance issues for no gain, so I removed it in https://bugzilla.mozilla.org/show_bug.cgi?id=1303605.

(Note that I removed that mapping earlier without realizing the correctness issue it introduced, https://bugzilla.mozilla.org/show_bug.cgi?id=1453702, which I fixed computing to none instead of contents, which also aligns with Blink and WebKit).

[emilio]

cc @lilles since he implemented the Blink style fixup, but I think he wouldn't have any objection about changing this, given it would adapt the spec to how Blink, Firefox and WebKit implement.

[lilles]

Huh. I thought we did store the original display for computed style before adjusting. I think we're fine with either resolution.

[lilles]

I misunderstood our OriginalDisplay() thing. Not used for getComputedStyle(). I support resolving this as proposed.

[tabatkins]

Sounds fine to me. @fantasai, any objection?

[fantasai]

WFM.

[css-meeting-bot]

The Working Group just discussed Use "computes to" instead of "behaves as" for display: contents on unusual elements, and agreed to the following:

• RESOLVED: serialize it out as none
• RESOLVED: compute to none

The full IRC log of that discussion

<dael> Topic: Use "computes to" instead of "behaves as" for display: contents on unusual elements
<dael> github: https://github.com//issues/2755
<dael> fantasai: We had said that display contents behaves as display:none on certain elements like some SVG. There's an appendix of exactly which ones. emilio filed an issue to jsut say computes to display:none since that's easier from impl then handling that at used style time
<dael> fantasai: TabAtkins and I have not much of an opinion. Up to impl.
<dael> emilio: Do we clear everyone that impl computes to none?
<fantasai> s/Do we/To be/
<astearns> s/Do we/To be/
<dael> Rossen_: Any reason to do anything other than that? Sounds like there's interop.
<dael> Rossen_: Objections to serialize it out as none?
<dael> RESOLVED: serialize it out as none
<dael> RESOLVED: compute to none