[css-values] Specify how to handle negative percentage basis in mathematical expressions #6298
Description
Some properties like object-position can have a negative percentage basis.
This causes an issue where some browsers calculates/simplifies clamp(100%, 50%, 0%) to 100% ignoring the negative percentage basis, but solves clamp(100%, 50% + 0px, 0%) to 50%. With a percentage basis of -100px the original equation is equal to clamp(-100px, -50px, 0px) which would solve to -50px. Solving the equation without accounting for the negative basis creates unexpected behaviour.
(See https://bugzilla.mozilla.org/show_bug.cgi?id=1709018).
It would be helpful if the spec specified how to handle negative percentage basis in mathematical expressions https://drafts.csswg.org/css-values-4/#calc-func
Activity
Loirooriol commentedon May 19, 2021
See #4227 (comment):
So it's a bug if the implementation assumes that
0% <= 50% <= 100%and doesn't wait until the percentages can be resolved into a length. This is already implied in the spec, https://drafts.csswg.org/css-values/#calc-simplificationemilio commentedon May 19, 2021
I think reading the spec to the letter, we should never simplify percentages. However that's a bit of a hard one to fix with properties that only parse percentages because that requires preserving the calculation a lot longer than browsers are now.
emilio commentedon May 19, 2021
Hah, I mid-aired with oriol :)
Loirooriol commentedon May 19, 2021
But properties that don't resolve percentages into another (non-number) type, like
opacity, compare percentages as percentages. So you can resolve immediately.tabatkins commentedon May 19, 2021
No, @emilio is right - the spec (mistakenly) never simplifies %s in min()/max(). It assumes they'll turn into another type, at which point they can simplify.
I should fix that to specify that %s which would resolve to another type in the context shouldn't simplify away in this case, but %s which stand on their own can.
tabatkins commentedon May 19, 2021
That said, the OP is indeed invalid - the spec definitely does specify that you don't resolve a min() or max() in that case until simplification can change the %s into lengths, at which point you get the correct values.
[css-values-4] Fix min()/max() partial simplification to not imply th…
tabatkins commentedon May 20, 2021
I did just now add a small note making the behavior of percentages clearer, tho.
johannesodland commentedon May 20, 2021
@tabatkins
This was indeed specified in the spec, and I just missed it. The spec makes sense as it is, but the added note is really helpful.
Thank you for clearing this up :)