[css-anchor-position] Anchor Element's "Containing Block"

[dshin-moz]

Acceptable anchor element uses containing blocks of positioned and anchor elements.

If the anchor element is statically positioned, then its containing block is the nearest block container. The positioned element may use absolute positioning, then the containing block is the nearest 'position: absolute|relative|fixed' container, irrespective of it being a block or inline container.

According to above, given:

<div id=block>
  <span id=inline style="position: relative;">
    <span id=anchor style="anchor-name: --a;">Anchor</span>
    <span id=positioned style="position: absolute; position-anchor: --a; left: anchor(right); top: anchor(bottom);">Positioned</span>
  </span>
</div>

#positioned can't anchor to #anchor, because #anchor's containing block is #block, but #postioned's containing block is #inline, and hence #positioned's containing block is not an ancestor of #anchor?

[tabatkins]

This is better specified in the new text for acceptable anchor element, where we properly reference "absolute positioning containing block" where relevant.

[fantasai]

Yeah, but, I think the spec is probably wrong and we should clarify that this is an acceptable anchor. If the anchor is an in-flow within the abspos's containing block, it should be acceptable.

[fantasai]

Considering the new text, I think we need an additional clause. Something like (patching this very strictly):

• possible anchor is an [=inline-level box=] or [=internal ruby box=], and positioned el's [=absolute positioning containing block=] is formed by an ancestor of possible anchor whose [=containing block=] is also possible anchor's [=containing block=]

Or perhaps

• possible anchor is an [=inline-level box=] or [=internal ruby box=], and positioned el's [=absolute positioning containing block=] is formed by an ancestor of possible anchor.

Or possibly even

• positioned el's [=absolute positioning containing block=] is formed by an ancestor of possible anchor (in the same [=top layer=] as possible anchor)?.

if that doesn't actually break anything here.

[fantasai]

Aside from top-layer shenanigans, I think the containing block chain is always a subset of the box-tree ancestor chain.