search

web-platform-tests / wpt

Test suites for Web platform specs — including WHATWG, W3C, and others
https://web-platform-tests.org/
Other
5k stars 3.09k forks source link

UAX 14 not strictly required by css-text #46493

Open frivoal opened 5 months ago

frivoal commented 5 months ago

I suspect the following two tests ought to be marked as "should" tests rather than the default must, as per https://web-platform-tests.org/writing-tests/css-metadata.html#requirement-flags, For the reasons covered in the Note at the end of https://drafts.csswg.org/css-text-4/#line-breaking

https://github.com/web-platform-tests/wpt/blob/master/css/css-text/line-breaking/line-breaking-028.html https://github.com/web-platform-tests/wpt/blob/master/css/css-text/line-breaking/line-breaking-029.html

cc: @makotokato, @jfkthame (as authors/reviewers)

jfkthame commented 5 months ago

I haven't tried to check extensively, but I suspect there are probably quite a few tests where this is the case -- they assume line-breaking behavior that is not strictly required by the spec, but in practice is what everyone expects. Unless there's an implementation that is actually wanting to deviate here, without being called out as non-compliant, I'm not sure there's much reason to soften the tests to use "should". What would the practical results of this be -- how do "should" tests (whether passing or failing) show up on wpt.fyi and other forms of reporting?

frivoal commented 5 months ago

I don't believe wpt currently distinguishes. The old csswg test system did distinguish: when reporting results, it would separate "must" tests from "should" ones, which is useful when trying to progress a spec, as must failures are a sign that either implementations or the spec need fixing, while should failures can be allowed to persist. I hope that wpt will eventually gain this ability