[css-align] "the relevant computed self-alignment property" is unnecessarily vague (it can only be "align-self")

[dholbert]

https://drafts.csswg.org/css-align/#baseline-align-content has some hand-wavy language that tries to be generic over self-alignment properties:

If its baseline alignment preference is “first” (“last”),
the relevant computed self-alignment property is either align-self/stretch or self-start (self-end)

This "relevant computed self-alignment property" text is now unnecessarily vague/generic -- it can only be "align-self", right? (i.e. that phrase should just be replaced with "its computed align-self", I think)

That's the only "relevant ... self-alignment property" that could be cooperating here in the align-content axis. (And this chunk of spec is only relevant in the align-content axis, ever since #1184 was resolved)

CC @fantasai @tabatkins

[dholbert]

Quoting my suggestion:

(i.e. that phrase should just be replaced with "its computed align-self", I think)

Actually, this^ isn't quite right either. The spec doesn't quite want to check whether the computed align-self is precisely the value stretch, because that doesn't account for align-self:auto (which behaves as stretch under default conditions) and align-self:normal (which behaves as stretch for some boxes, e.g. flex items).

Perhaps we should keep the sentence hand-wavy, though (e.g. "computed align-self is...stretch"), and then add an explanatory afterthought to say that for the purposes of this check, "stretch" includes other values that are defined to behave as stretch, including e.g. auto when the parent has align-items:stretch...? We already have an explanatory afterthought paragraph to generalize from the slightly-too-specific usage of self-start / self-end, here -- so I think we want something like that for stretch as well.

[tabatkins]

This "relevant computed self-alignment property" text is now unnecessarily vague/generic -- it can only be "align-self", right? (i.e. that phrase should just be replaced with "its computed align-self", I think)

Nope, which axis is align/justify for the -self properties is determined by the containing block's writing mode, which has no relation to your own writing mode. So it can be either one (tho it will often by align-self, yes).

Actually, this^ isn't quite right either. The spec doesn't quite want to check whether the computed align-self is precisely the value stretch

You're right; we've fixed this to use "behaves as".

[dholbert]

Ah, OK - I think I understand.

So for a grid item whose writing-mode is orthogonal to its parent grid, for example: if the item has align-content: baseline; justify-self:baseline, then align-content would defer to justify-self in determining its sizing/positioning in that axis (the parent's inline axis i.e. the item's block axis) -- correct?

(I initially filed this bug with flexbox in mind, where justify-self has no effect, which is why I was confused.)

[tabatkins]

Yes.