[css-overflow] Which descendants are skipped in max-lines?

[Loirooriol]

max-lines is defined with

A region break is forced after its Nth descendant in-flow line box, where N is the specified value of max-lines. The contents of descendants that are scroll containers or that are not block containers are skipped over while counting line boxes.

But I'm not sure if that's enough:

• Lines inside table captions

  Despite @FremyCompany's complaints, the table wrapper box establishes a BFC and thus is technically a block container. That's not really the case on most browsers, though. Table captions are also block containers, so their lines are not skipped. Is this expected? Not sure.

• Lines inside inline-blocks

  As said in [css-overflow] Which kind of boxes accept max-lines? #2408 they are block containers but should be skipped

• Ruby

  I don't know much about ruby but I guess ruby annotations should probably be skipped. Less sure about ruby bases.

• overflow: clip

  This is not a scroll container but visually is identical to overflow: hidden, whose lines are skipped because it's a scroll container.

• Out of flow

  The spec mentions "in-flow line box". But only things with float or position can be out-of-flow. Line boxes live in the fragment tree and don't own float not position properties, so I think they will always be in-flow (but possibly inside an out-of-flow container).

  Either properly define what "in-flow" and "out-of-flow" means for line boxes or skip the contents of out-of-flow descendant boxes.

• Other BFC roots

  I'm not convinced max-lines should work across formatting context boundaries, even if it's only a BFC.

[fantasai]

• We should clarify that the table wrapper box is skipped.
• Neither line boxes nor out-of-flows are introspected. I would have thought this was obvious, but we can add a clarification to this effect.
• overflow: clip is covered in [css-overflow] How does max-lines interact with hidden content? #2421 due to your comment there.
• working across formating context boundaries is covered in [css-overflow] Which kind of boxes accept max-lines? #2408 afaict.

Removing from agenda; first two issues are bugfix/editorial.

[Loirooriol]

working across formating context boundaries is covered in #2408 afaict.

That issue covers FFC and GFC but not BFC boundaries. You can use display: flow-root, contain, etc. to have an inflow block box that is not a scroll container but establishes a BFC. Should line boxes inside it be skipped?

[fantasai]

Sure it does, they're discussed in multiple comments in that issue.

[frivoal]

it was, but the original issue was about something else, and the conclusion did not cover that question, so it is relevant to discuss here.

Personally, I don't have a particularly strong opinion, but here is the rationale: (copying a comment from another thread):

@fantasai and I considered that possibility, but we thought it was better not to skip:

• We did not think it would make a big difference in implementation complexity
• Counting lines in nested BFCs doesn't seem to introduce strange corner cases
• The various reasons block container get BFCfied did not seem to be particularly relevant to the issue of counting lines, do we expect authors will prefer it this way
• Authors can always turn their BFCs into some other kind of formatting contexts if they want to exclude them (e.g. a single item vertical flexbox)

[Loirooriol]

Exactly. And CSS Display says

Exclusions are able to affect content across formatting context boundaries. (At time of writing, they are the only layout feature that can.)

If max-lines starts hiding contents across formatting context boundaries, the quote should be updated. But breaking this invariant doesn't seem nice.

[css-meeting-bot]

The Working Group just discussed Which descendants are skipped in max-lines?, and agreed to the following resolutions:

• RESOLVED: Having max-lines not apply across BFC boundaries

The full IRC log of that discussion

<dael> Topic: Which descendants are skipped in max-lines?
<dael> github: https://github.com//issues/2429
<dael> florian: Within the broad topic of what's skipped, when you're counting lines do we cross BFC bounderies? Yes we shouldn't count inside obviously insane things like orthogonal flows, but what about BFCs?
<dael> dbaron: What sorts of BFC are okay?
<dael> fantasai: We excluded scroll containers. Orthogonal flow are insane.
<dael> florian: Nested BFCs. [gives example]
<dael> dbaron: Other then the display:flow-root that exists to create a BFC, others?
<dael> fantasai: Align-content is probably okay. Tirggers when not normal
<dael> dbaron: Which axis?
<dael> fantasai: Vertical
<dael> Rossen: Is anyone impl?
<dael> fantasai: Don't think so.
<dael> iank_: In your impl we skipped all this for line counting?
<dael> Rossen: Yes.
<dael> Rossen: Also when you start reasonint it becomes insane. What would it mean on a grid?
<dael> TabAtkins: Those are all excluded.
<dael> TabAtkins: There's several in the issue.
<dael> florian: Table caption is another I"m not sure.
<dael> fantasai: WE skip the table.
<dael> iank_: Someone uses display:flex it triggers a formatting context and it will display a part.
<dbaron> What alignment happens when something with align-content is fragmented?
<dael> florian: If you have a BFC that's not excluded and it's not what you wanted you make it a flex.
<dael> TabAtkins: align content makes me think we want it to work. WE want to do it on an align content thing
<dael> dbaron: When something with align content is fragmented what happens?
<dael> fantasai: I don't think we define.
<dael> fantasai: I'm not sure that's what we want. If you have a box supposed to be centered and it's frag I don't think you want each of them centered.
<dael> TabAtkins: Regardless, boxes don't have size.
<dael> fantasai: they do
<dael> TabAtkins: Fragments have size
<dael> fantasai: Box has computed size
<dael> florian: Different topic.
<fantasai> The fragments eat up the computed size, they don't each get 100px if you specify 'width: 100px'
<dael> Rossen: Is there a reason why you wouldn't want to make the easier part work, get a modal that makes sense, and then extend
<dael> florian: I'm not against get rid of BFC
<dael> Rossen: I'm sympathetic to excelude all BFCs, get stable, and then see what we can include. But until we're they're we're hypothetically speculating.
<dael> TabAtkins: Easier to cut off BFCs.
<dael> Rossen: Also that will make it compat with line-clamp.
<dael> florian: Yeah.
<dael> Rossen: Prop: Having max-lines not apply across BFC bounderies
<dael> RESOLVED: Having max-lines not apply across BFC boundaries