[css-position][css-align] How do other alignments affect the inset-modified containing block?

[Loirooriol]

https://drafts.csswg.org/css-position/#abspos-insets says

If both inset properties in a given axis are auto, then, depending on the box’s self-alignment property in the relevant axis:

• for self-start alignment or its equivalent
  Set its start-edge inset property to the static position, and its end-edge inset property to zero.
• for self-end alignment or its equivalent
  Set its end-edge inset property to the static position, and its start-edge inset property to zero.
• for center alignment
  ...

What should happen for other alignments? Like, the initial justify-self is auto, which is typically normal. https://drafts.csswg.org/css-align-3/#justify-abspos

For all other absolutely-positioned boxes, normal behaves as stretch.

So what's the inset-modified containing block for stretch? Are we instead supposed to use https://drafts.csswg.org/css-align-3/#abspos-sizing ?

Treat normal as start

That seems to match implementations. But what if I specify justify-self: stretch explicitly?

This is all very confusing because both css-position and css-align monkeypatch both css2 and each other.

[fantasai]

@Loirooriol Fixed by inserting “(treating normal as start and any distributed, baseline, or stretch alignment value as its fallback alignment)”. Let us know if this addresses the issue.

[Loirooriol]

I don't think self-alignment (align-self/justify-self) accepts distributed alignment values, though? Instead, it accepts other positional alignments which are not handled?

[fantasai]

@Loirooriol Which other alignments are not handled?

[Loirooriol]

Well, it's not clear what to do if <overflow-position> is specified. Is it just ignored and we only consider the <self-position>? safe can make the alignment behave as start, which I guess is handled anyways as an equivalent of either self-start or self-end. But here we are determining the inset-modified containing block into which it will be aligned. So I guess we don't know yet if it will overflow or not.

And normal is not explicitly handled, but it behaves as some other value, so maybe it's handled as that value? But not all handled cases say "or its equivalent".

[fantasai]

@Loirooriol Fixed. Please close out if you're satisfied!