1836563 - WPT css/selectors/invalidation/nth-child-of-has.html and nth-last-child-of-has.html fail in Firefox

Status: RESOLVED DUPLICATE of bug 1792501

(Core :: CSS Parsing and Computation, defect)

Description

[Daniel Holbert [:dholbert]]

Right now, we fail this WPT (part of the interop-2023 test set):
https://wpt.fyi/results/css/selectors/invalidation/nth-last-child-of-has.html
Direct link to test:
http://wpt.live/css/selectors/invalidation/nth-last-child-of-has.html

Using a fresh Firefox profile, I can trigger 3 distinct renderings:
(1) by default, we render all of the text black.
(2) if I toggle layout.css.has-selector.enabled to true and reload, we render the 3rd and 5th line as green.
(3) If I then open devtools and hover anything, then we instead render the 2nd and 4th line as green. And that seems to be what the test expects.

(My regular browsing profile often goes directly to the "good" rendering -- possibly due to a race condition?

So there seems to be a bug here where we get the initial state wrong, but we correct it eventually if you force some sort of restyle flush.

Possibly some work on bug 418039 or another preexisting associated bug will fix this, but I figured I'd spin off this dedicated bug, just in case this is a special case that needs additional work.

(Note, this test was added by Zach in bug 1818135 ; adding dependency on that bug.)

Comment 2

[Tess Hoffman [:zrhoffman]]

Note that css/selectors/invalidation/nth-child-of-has.html also fails (with layout.css.has-selector.enabled enabled).

Comment 3

[Tess Hoffman [:zrhoffman]]

wpt.fyi result: https://wpt.fyi/results/css/selectors/invalidation/nth-child-of-has.html
Direct link to the test http://wpt.live/css/selectors/invalidation/nth-child-of-has.html

Comment 4

[Daniel Holbert [:dholbert]]

Thanks! Indeed, that one fails in roughly the same way roughly as described in comment 0 (with 3 distinct renderings), though now with the 4th and 6th lines initially green, changing to 5th and 7th after I interact with devtools. And as noted there, it automatically swaps to the good rendering in my regular browsing profile, due to a race condition or something.

(Sorry for overlooking your bug 1818155 that was already tracking these; and thanks duping that over to clean up after me. :))

Comment 5

[Tess Hoffman [:zrhoffman]]

Fixed in bug 1792501!