[css-overflow-4] Draft spec for continue: collapse (#7708)

[andreubotella]

No description provided.

[bfgeek]

@emilio - Does this look reasonable to you?

[andreubotella]

Before now, this PR allowed clamping between any two blocks, to make it closer to the continue: discard behavior, but in hallway conversations in TPAC we agreed that having it clamp to the last line before clamp, even if there are lineless boxes after it that will still fit, was a better behavior. I just updated the spec text to reflect this.

[tabatkins]

but in hallway conversations in TPAC we agreed that having it clamp to the last line before clamp, even if there are lineless boxes after it that will still fit, was a better behavior

In particular, the argument was that auto clamping should be equivalent to the largest <integer> clamp that avoids overflow; it's weird if auto actually lets you include an amount of content between two <integer> values, and thus can't be reproduced directly.

[frivoal]

This seems mostly good. I've made a few small suggestion for normative changes, and more suggestions for editorial changes (most of which are about source formatting).

Most of the remaining issues are orthogonal to this PR, so I think we can merge first and keep working on those later, as soon as the comments on this PR are addressed.

[frivoal]

Editorial:

Suggested change

	must be pushed to that <a>fragmentation container</a>. If it is placed before a [=clamp point=],
	must be pushed to that <a>fragmentation container</a>.
	If it is placed before a [=clamp point=],

Reason: this spec uses semantic line-breaks, and I'd prefer to stay consistent with that. See https://rhodesmill.org/brandon/2012/one-sentence-per-line/

[frivoal]

I'm ok with that, but I think this has been controversial, with some advocating that the displaced content be dropped / invisible instead, in order to avoid that inserting the ellipsis would cause an extra layout pass. Should we open a separate issue to discuss that aspect?
Or maybe this is unavoidable due to bidi processing of the ellipsis, which means we have to relayout anyway?

[frivoal]

Suggested change

	Sets 'continue' to ''collapse''
	Issue(7708): Whether this should imply ''discard'' or ''collapse'' is an unresolved question.
	Sets 'continue' to ''collapse''

[frivoal]

Suggested change

	a <a>region break</a> or
	a <a>region break</a>

[frivoal]

Editorial:

Suggested change

	If the value of 'max-lines' is not <dfn for=max-lines dfn-type=value>none</dfn>, then, where
	<var>N</var> is the computed value of 'max-lines':
	If the value of 'max-lines' is not <dfn for=max-lines dfn-type=value>none</dfn>,
	then, given a [=computed value=] of <var>N</var>:

[frivoal]

Editorial

Suggested change

	- A point immediately after a [=line box=] in the line-clamp container's [=block formatting
	context=].
	- A point immediately after a [=line box=]
	in the line-clamp container's [=block formatting context=].

See https://rhodesmill.org/brandon/2012/one-sentence-per-line/

[frivoal]

Editorial

Suggested change

	- The end of the line-clamp container, if there are any boxes after the last [=inline formatting
	context=] in the line-clamp container's [=block formatting context=].
	- The end of the line-clamp container,
	if there are any boxes after the last [=inline formatting context=]
	in the line-clamp container's [=block formatting context=].

See https://rhodesmill.org/brandon/2012/one-sentence-per-line/

[frivoal]

Editorial

Suggested change

	Any boxes in a [=line-clamp container=] that follow its [=clamp point=] in the box tree, as well
	as any [=line boxes=] that follow it inside an [=inline formatting context=], will be invisible
	(behaving like ''visibility: hidden''). This includes [=independent formatting contexts=] as
	well as [=out-of-flow=] boxes, and all of their descendants. Any overflow such boxes and line
	boxes might have is always counted as [=ink overflow=] rather than [=scrollable overflow=].
	Any boxes in a [=line-clamp container=] that follow its [=clamp point=] in the box tree,
	as well as any [=line boxes=] that follow it inside an [=inline formatting context=],
	will be invisible (behaving like ''visibility: hidden'').
	This includes [=independent formatting contexts=]
	as well as [=out-of-flow=] boxes,
	and all of their descendants.
	Any overflow such boxes and line boxes might have
	is always counted as [=ink overflow=] rather than [=scrollable overflow=].

See https://rhodesmill.org/brandon/2012/one-sentence-per-line/

[frivoal]

Editorial:

Suggested change

	If the [=line-clamp container=]'s [=block formatting context root=] has a [=computed value=] of
	'max-lines' other than ''max-lines/none'', then that property will determine the [=clamp point=].
	Otherwise, the clamp point will be set to the last possible clamp point such that, for it and
	all previous possible clamp points, the line-clamp container's [=automatic block size=] (as
	determined below) is not greater than the [=block size=] the box would have if its automatic
	block size were infinite; or if that is not the case for any clamp points, to the first clamp
	point in the block formatting context.
	If the [=line-clamp container=]'s [=block formatting context root=]
	has a [=computed value=] of 'max-lines' other than ''max-lines/none'',
	then that property will determine the [=clamp point=].
	Otherwise, the clamp point will be set to the last possible clamp point
	such that, for it and all previous possible clamp points,
	the line-clamp container's [=automatic block size=]
	(as determined below)
	is not greater than the [=block size=] the box would have
	if its automatic block size were infinite;
	or if that is not the case for any possible clamp points,
	to the first possible clamp point in the block formatting context.

See https://rhodesmill.org/brandon/2012/one-sentence-per-line/

I also added the word "possible" before "clamp point" in the last two lines.

[frivoal]

Editorial

Suggested change

	This property only affects a line box if it immediately precedes
	eiher a [=region break=] or a [=clamp point=] in the [=block formatting context=].
	If the [=clamp point=] is placed at the end of the [=line-clamp container=], then the line will
	not be affected.
	This property only affects a line box
	if it immediately precedes eiher a [=region break=]
	or a [=clamp point=] in the [=block formatting context=].
	If the [=clamp point=] is placed at the end of the [=line-clamp container=],
	then the line will not be affected.

Reason: this spec uses semantic line-breaks, and I'd prefer to stay consistent with that. See https://rhodesmill.org/brandon/2012/one-sentence-per-line/

[frivoal]

@tabatkins , for your comment in #10816 (comment), I'd suggest following up in #10868, because that aspect exists before this PR. The text of the spec may need to change depending on where we land on that issue (and others), but this PR isn't changing it, so I think we should take one problem at a time.