|20 Feb 2024
|Noam Rosenthal: Hm? What security implication?
| Noam Rosenthal: the test is failing in chrome because it's parsing
:grammar-error even though it's not valid CSS syntax
|Something about not using this to find out if there is a particular highlight?
|Noam Rosenthal: the security implication is that you have to return the style as if there was a highlight, but the test reflects that
|Noam Rosenthal: But it seems that the issue is that chrome is just not rejecting the single-colon syntax as per spec (https://drafts.csswg.org/cssom/#dom-window-getcomputedstyle step 3)
|Or am I missing something?
|Noam Rosenthal: annevk: I think Firefox is right unless I'm missing something obvious
In reply to @emilio:mozilla.orgI’m working on the test and locally only those two fail after fixing the other ones
|They fail because we don’t create a UA shadow DOM which means there is no placeholder pseudo
| Noam Rosenthal: I think that's wrong. You still return e.g. a
::marker style even if there's no marker, a
::before style even if there's no
|Though you're right the spec isn't super clear about "when pseudo-elements are supposed to exist"
|What Gecko does is resolve the style anyways, the same way you do for every other pseudo
In reply to @emilio:mozilla.orgIt’s ambiguous in the spec. We discussed it in the csswg in the context of ::view-transition pseudos
|Noam Rosenthal: still the firefox behavior seems more consistent on how we deal with e.g. ::marker
In reply to @emilio:mozilla.orgSomeone said in the WG that marker was like this so that we don’t make the existence of the marker web observable
| Noam Rosenthal: you could argue that if you don't return a style for ::file-selector-button if you're not an
<input type=file>, you shouldn't return a style for ::marker if you're not a list item
|hm? The existence of the marker is observable by querying used-value properties like width
In reply to @emilio:mozilla.orgThat’s fair but needs to be in the spec IMO
| e.g., a marker that exists would resolve
width: auto to a pixel value
|I'd file a spec issue and discuss it in the WG first before changing the test, I don't see why ::file-selector-button or ::placeholder would be special but...
|emilio: the initial context was this CSSWG resolution: https://github.com/w3c/csswg-drafts/issues/9880#issuecomment-1939408219 which I now think was wrong.
| emilio: the question there was should we match the pseudo-element when there is no view-transition, or no view-transition with the given name. We resolved that it should match an actual VT object (which sorta aligns with the spec) while for consistency it should perhaps just match a hypothetical style (which aligns with
::marker). I'm impartial to what we end up with as long as we make the CSSOM spec clear about it
|Noam Rosenthal: yeah so... my understanding of that resolution is that the only thing that VT pseudos matched were names etc, just so that you could have an unambiguous style and return something sensible from gCS and co
|But yeah the written down resolution is a bit different than that
In reply to @emilio:mozilla.orgAlso about returning an empty style when there is no active transition. It’s in the minutes
|I'm getting an odd lint error in https://github.com/web-platform-tests/wpt/pull/44644 ./wpt lint --all --github-checks-text-file=/home/test/artifacts/checkrun.md ERROR:lint:css/css-highlight-api/historical.window.js: Testcase file must have a link to a spec (MISSING-LINK) But a) there are two rel=help spec links in the test and b) what on earth is the relevance of css highlight api to the (png spec) test? How do I resolve that?
|Seems like zcorpan commented on the issue (I'm not actually sure you need to rebase; anything that causes the jobs to run again would be enough, but rebasing is probably the best way to do that)
|21 Feb 2024
|Sam Sneddon [:gsnedders]
|this has all turned into https://github.com/web-platform-tests/wpt/issues/44701, erghhhhhh 😐️