[css-fonts-4] Stop claiming that 'ch' uses the first available font, since it doesn't

[dbaron]

css-fonts-4 claims that the 'ch' unit uses the "first available font", but it doesn't. This revises the spec to stop making this claim.

Note that the same claim is made in css-fonts-3. Given its status, I'm not sure whether to suggest the same edit there...

[AmeliaBR]

Your "it doesn't" link leads to the following text:

Equal to the used advance measure of the “0” (ZERO, U+0030) glyph in the font used to render it.

This text could use with some clarification, too. It took me a moment to realize that "it" referred back to the "0", and not the element as a whole. And since there might not be a "0" in the element's text content, maybe a better wording would be (borrowing from the definition of "first available font"):

Equal to the used advance measure of the “0” (ZERO, U+0030) glyph in the first available font that would match that character, given font families in the ‘font-family’ list (or a user agent's default font if none are available).

(and similar wording for the ic unit).

[litherum]

It’s pretty unfortunate that ch units can cause fonts to download.

WebKit doesn’t do this; it just uses the primary font, and if the primary font doesn’t support “0” then it uses that font’s .notdef glyph. No downloads necessary.

https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/platform/graphics/Font.cpp#L123

https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/css/CSSPrimitiveValue.cpp#L661

[emilio]

I'm not the most familiar with Gecko's font metrics code, but it seems that we use the first valid font as well:

https://searchfox.org/mozilla-central/rev/0b8ed772d24605d7cb44c1af6d59e4ca023bd5f5/layout/style/ServoBindings.cpp#2557

And fall back to aveCharWidth if there's no 0 glyph, which is computed in a somewhat complicated way which I'm not confident enough to describe.

@jfkthame can probably correct if anything above is wrong.

[jfkthame]

It’s pretty unfortunate that ch units can cause fonts to download.

Can they?

It seems to me the definition of "first available font" is slightly unclear:

The first available font, used for example in the definition of font-relative lengths such as 'ex' and 'ch' or in the definition of the 'line-height' property is defined to be the first available font that would match the U+0020 (space) character given font families in the 'font-family' list (or a user agent’s default font if none are available).

But this doesn't say what it means for a font to be "available". If the font-family list includes @font-face families for which resources have not (yet) been downloaded, are those fonts "available" for the purpose of computing font-relative lengths or not?

AFAICT, the spec usually uses "available" to refer to a font resource that has been downloaded and is ready for use, so I've assumed that was the intent here as well; but in §4.1 it seems to regard any font defined by a @font-face rule as being "available":

A given set of @font-face rules define a set of fonts available for use within the documents that contain these rules.

without regard for whether the relevant resources have actually been fetched.

[1] https://drafts.csswg.org/css-fonts-4/#first-available-font

[svgeesus]

Note that the same claim is made in css-fonts-3. Given its status, I'm not sure whether to suggest the same edit there...

Once we have agreement, the same change can be listed as an erratum on Fonts 3.

[svgeesus]

Hmm that would also call into question this test currently passed by Gecko, Edge, Blink and even Trident.

[dbaron]

So this is a PR to make a referencing spec agree with the spec that normatively defines the relevant behavior. If folks disagree with the behavior, I think it would be good to open an issue against the latter spec (css-values).

[dbaron]

Also, I don't think https://test.csswg.org/harness/test/css-fonts-4_dev/first-available-font-002/ really distinguishes the case, since it's a font that has neither a space nor a zero (it has only a lowercase a). It would be more relevant to the question of how to define ch units to try variants of that test that have one but not the other of space and zero (probably both ways).

[svgeesus]

I agree that V&U defines ch to use the font that renders it, not the first available font. Thus, I agree with this PR.

[svgeesus]

I agree, brings fonts-4 into alignment with V&U.

[litherum]

It’s pretty unfortunate that ch units can cause fonts to download.

Can they?

Right now, Values and Units says:

Equal to the used advance measure of the “0” (ZERO, U+0030) glyph in the font used to render it.

Seems fairly unambiguous to me.