[css-color-5] Clarification on `none` components in relative color syntax.

[tiaanl]

I have the following examples from wpt:

rgb(from rebeccapurple none none none)
rgb(from rebeccapurple none none none / none)
rgb(from rebeccapurple r g none)
rgb(from rebeccapurple r g none / alpha)
rgb(from rgb(20% 40% 60% / 80%) r g none / alpha)

all using none somewhere in the components. My expectation from the spec is that each component should stay none in the resulting code: rgb(from rebeccapurple none none none) becomes rgb(none none none).

But the current wpt contradicts this assumption.
https://github.com/web-platform-tests/wpt/blob/1f9e1af4e5bbd84e79d8c8cd17a135bcf20f6e2d/css/css-color/parsing/color-computed-relative-color.html#L115

[svgeesus]

OK it looks like these tests need review as they encapsulate incorrect assumptions.

[tiaanl]

Has there been any clarification on the use of none here?

[LeaVerou]

Yeah these WPTs are wrong.

[svgeesus]

@tiaanl if you make a PR on WPT to fix them, ping me for a review

[kbabbitt]

@LeaVerou @svgeesus @tiaanl

I'm working on updating Blink to match the discussion in this thread. Here's the WPT change; does it look correct? https://chromium-review.googlesource.com/c/chromium/src/+/5905968/3/third_party/blink/web_tests/external/wpt/css/css-color/parsing/color-computed-relative-color.html

[LeaVerou]

@kbabbitt Yes, that looks correct! Thank you!

[cdoublev]

Note that there is a resolution to serialize with rgb() rather than color() (as currently specified).

[kbabbitt]

Thanks Lea.

@cdoublev:

Note that there is a resolution to serialize with rgb() rather than color() (as currently specified).

From what I saw, the text in that resolution applies to declared values. My understanding was that computed values would continue to be serialized according to the text in https://drafts.csswg.org/css-color-5/#serial-relative-color starting with "The serialization of the result of a relative color function..." and referring to the table that follows. Several of the examples in #10328 (comment) appear to be consistent with that understanding. Is there something I missed?

[cdoublev]

You are right, sorry. I think I was misled by this in the op:

My expectation from the spec is that each component should stay none in the resulting code

... which should only apply to declared values.