[css-anchor-position] Recent change in acceptable anchor element seems incomplete

[dshin-moz]

With commit 184472e, as commented here:

I believe the new edit is missing the blurb about the possible anchor element's penultimate containing block, that it should either be statically-positioned, or absolutely-positioned and occurring after the positioned element.

As is, given:

<div id=abs-cb>
  <div id=positioned-el></div>
  <div id=penultimate-abspos>
    <div id=anchor></div>
  </div>
</div>

#anchor isn't absolutely positioned, and its containing block, #penultimate-abspos is a descendant of #abs-cb, which is the abspos containing block of #positioned-el, satisfying this condition. However, it's laid out after #positioned-el.

[tabatkins]

Argh, you're right. This is so tricky to get right!

Okay, Elika and I worked over the text again, and found a way to significantly simplify it. The "penultimate CB" condition is really just asking for the element generating that CB to also be an acceptable anchor element; you're allowed to anchor to anything CB-contained inside of an acceptable anchor. So we can just recurse, to handle that. This then lets us focus exclusively on the case where they have the same CB (or we've recurse up to an ancestor who has the same CB), where we have the meaningful conditions - earlier top layer, earlier in flat tree, etc.

Text is in a4f08e4. Look reasonable to you?

[fantasai]

(And now fixed the marked issue by using the correct term. ^_^)

[dshin-moz]

Reacted +1, but just noticed the label - So to make sure, yeah, the revision makes sense.