csswg-drafts [css-sizing] Does indefinite `stretch` behave as automatic size or as initial value?

First, consider

<!DOCTYPE html>
<div style="min-height: max-content; height: 0; width: 200px; border: solid">lorem ipsum</div>

Spec, Blink	Gecko, WebKit

@bfgeek told me that Blink had to deal with some breakage when they adopted this behavior, but that it's the correct thing to do. But what about stretch?

<!DOCTYPE html>
<div style="min-height: -moz-available; min-height: -webkit-fill-available; min-height: stretch; 
            height: 0; width: 200px; border: solid">lorem ipsum</div>

Spec	Gecko, Blink	WebKit

(WebKit seems to stretch elements with an indefinite height that contain some -webkit-fill-available, so it's always definite)

From https://drafts.csswg.org/css-sizing-4/#stretch-fit-sizing

Formally, its behavior is the same as specifying an automatic size together with a self-alignment property value of stretch

But here Blink is not implementing this, not sure if due to an oversight.

And then the situation is that browsers are implementing stretch, but nobody follows the spec here. It may be too late for *-content, but I would like to have an agreement for stretch before it ships. So Agenda+.

Oct 05 '24 07:10 Loirooriol

For Servo following the spec seems a bit simpler than aligning with Blink, so I guess I will go with that for now.

Oct 05 '24 11:10 Loirooriol

Formally, its behavior is the same as specifying an automatic size together with a self-alignment property value of stretch

So I think this was intended just to be applicable to main sizes, (e.g. not min/max-sizes).

IMO this should be consistent with min-height: 100%, e.g. if a percentage is indefinite it behaves as zero for min-height, and Infinity for max-height.

(e.g. The Blink/Gecko behaviour).

For example it'd be pretty weird if:

<!DOCTYPE html>
<div style="max-height: stretch; height: 200px; width: 200px; border: solid">lorem ipsum</div>

got clamped by the intrinsic size (when stretch is an extrinsic constraint).

cc/ @tabatkins

Oct 08 '24 19:10 bfgeek

I'm not convinced we should treat it like a percentage. For example, with a min/max-content of 100px, fit-content is basically clamp(100px, stretch, 100px), and in Blink max-height: fit-content behaves as 100px but max-height: clamp(100px, 100%, 100px) behaves as none. And doing this like Blink seems more annoying to implement in Servo.

So I think we should either:

Keep the spec as-is, or
Treat both fit-content and stretch as containing a cyclic percentage. More specifically, treat fit-content as fit-content(100%) and stretch as 100%, so:
- For preferred and max properties they both behave as the initial value (with some special behavior for replaced elements)
- For min properties see #10969

Oct 09 '24 17:10 Loirooriol

I'm not convinced we should treat it like a percentage. For example, with a min/max-content of 100px, fit-content is basically clamp(100px, stretch, 100px)

This isn't correct - we recently did a change to follow Firefox in this regard. See: https://github.com/w3c/csswg-drafts/issues/10721 When the stretch is indefinite - min-size: fit-content resolves to min-size: min-content, and max-size: fit-content to max-size: max-content respectively.

And doing this like Blink seems more annoying to implement in Servo.

I'm curious why this is difficult in Servo? This might result in difficulty when attempting to implement calc-size() for example.

I'm pretty convinced that we should just be consistent with 100% here, (again see the max-height:stretch example above, it'd be super weird if we coerce an explicitly extrinsic constraint to be intrinsic).

Oct 09 '24 21:10 bfgeek

we recently did a change to follow Firefox in this regard. See: https://github.com/w3c/csswg-drafts/issues/10721

Not sure if that is the best behavior, I have commented there.

I'm curious why this is difficult in Servo?

I misunderstood what you were doing for fit-content, I thought it was trickier. Resolving keywords in different ways depending on the property and having a potentially indefinite stretch size during layout still adds some complexity. But not a big deal I guess.

it'd be super weird if we coerce an explicitly extrinsic constraint to be intrinsic

Maybe a bit weird indeed.

Oct 17 '24 09:10 Loirooriol

It's worth noting that there is a difference among major browsers:

Gecko treats an indefinite min-height: stretch as the initial auto
Blink and WebKit treat it as 0px

<!DOCTYPE html>
<div style="display: flex; flex-direction: column; border: solid cyan; min-height: 100px; width: 200px;">
  <div style="min-height: -moz-available; min-height: -webkit-fill-available; min-height: stretch;
              flex-basis: 0; border: solid">lorem ipsum</div>
</div>

Gecko	Blink, WebKit

Interestingly, they all treat min-height: 0% as 0px.

Oct 17 '24 20:10 Loirooriol

I wouldn't read too much into the block direction behaviour with FF (support was lacking last time I checked).

Oct 17 '24 21:10 bfgeek

Indeed, the FF behavior is just because I haven't implemented stretch behavior in the block axis yet (keep getting distracted by other tasks, but made a little local progress this week, hope to be done soonish).

Oct 17 '24 21:10 dholbert

Sure, but if the spec changes, we need to decide whether min-height: stretch should behave as auto or as 0px or whatever 100% would do (the spec says that 100% should behave as 0px, but browsers don't follow the spec in general).

Oct 18 '24 10:10 Loirooriol

Formally, its behavior is the same as specifying an automatic size together with a self-alignment property value of stretch

So I think this was intended just to be applicable to main sizes, (e.g. not min/max-sizes).

IMO this should be consistent with min-height: 100%, e.g. if a percentage is indefinite it behaves as zero for min-height, and Infinity for max-height.

(e.g. The Blink/Gecko behaviour).

For example it'd be pretty weird if:
<!DOCTYPE html>
<div style="max-height: stretch; height: 200px; width: 200px; border: solid">lorem ipsum</div>
got clamped by the intrinsic size (when stretch is an extrinsic constraint).

cc/ @tabatkins

I talked to @tabatkins about this a few weeks ago. He confirmed that the spec editors intended the Blink/Gecko behavior (min-size:stretch resolves to 0 and max-size:stretch resolves to infinity when available size is indefinite) and that it was indeed on the editors to better define the spec around min/max sizes behaving as auto, which should hopefully clear up the confusion here.

Dec 09 '24 20:12 davidsgrogan

The CSS Working Group just discussed [css-sizing] Does indefinite `stretch` behave as automatic size or as initial value?, and agreed to the following:

RESOLVED: min-size: stretch behaves as zero, max-size; stretch as infinity, when in an indefinite container

The full IRC log of that discussion

<fantasai> oriol: Spec says stretch size is same behavior as automatic size + stretch alignment
<fantasai> oriol: but it doesn't restrict this defintion to preferred size property
<fantasai> oriol: my understanding is you can set this in min or max property
<fantasai> oriol: I've been implementing in servo, but doesn't match other browsers
<fantasai> oriol: they don't support keywords in the block axis
<fantasai> oriol: it ignores it in min-size, and in max-size it treats it as none
<fantasai> oriol: from what davidsgrogan said, this was the intention of the editors of the spec ?
<TabAtkins> fantasai: that seems a bit weird, if we're not supporting it we just shoudln't parse it
<TabAtkins> oriol: the status is inconsistent because browsers have been implementing these things...
<fantasai> oriol: state is inconsistent because actively changing implementations
<fantasai> oriol: in Firefox these keywords work in inline axis, but not in block one
<fantasai> oriol: and then you need to parse the value because you don't know the writing mode
<fantasai> oriol: Blink obeys these keywords in both axes
<fantasai> fantasai: is there a reason we can't make it mean what it says in the min/max properties?
<TabAtkins> fantasai: from an authors' perspective, stretch means "i want to fill the container"
<TabAtkins> fantasai: so if you said "i would like this box to have a width of 100px but a max-width of stretch", then it should be 100px if the container is larger than 100px but filling it exactly if it's smaller
<TabAtkins> fantasai: that's what i'd expect
<TabAtkins> oriol: this is just when the stretch size is indefinite, so we don't know the container size
<TabAtkins> oriol: so it happens in the block axis in block layout
<fantasai> fantasai: Ah, in that case that make ssense
<fantasai> oriol: Ian says 'stretch' is an extrinsic value, would be weird to treat it as intrinsic.
<fantasai> oriol: what I did in servo is slightly simpler to implement, but aligning with Blink is ok
<fantasai> astearns: resolving stretch to zero, infinity when available size is indefinite
<fantasai> astearns: can we resolve on it?
<fantasai> astearns: any objections
<fantasai> RESOLVED: min-size: stretch behaves as zero, max-size; stretch as infinity, when in an indefinite container

Feb 19 '25 17:02 css-meeting-bot

csswg-drafts csswg-drafts copied to clipboard

[css-sizing] Does indefinite `stretch` behave as automatic size or as initial value?

csswg-drafts
csswg-drafts copied to clipboard