Tests (and spec) for :user-valid and :user-invalid are... light

[mfreed7]

Test List

https://wpt.fyi/results/css/selectors/user-invalid.html
https://wpt.fyi/results/css/selectors/user-valid.html

Rationale

These tests of :user-valid and :user-invalid will 100% pass if:

1. Those two pseudo classes are parsed as valid by the CSS parser.
2. They never match any elements.

Here is the :user-valid test, in its entirety:

  const input = document.querySelector('input');
  assert_true(input.validity.valid);
  // The selector can't match because no interaction has happened.
  assert_false(input.matches(':user-valid'));

The spec is located here:
https://w3c.github.io/csswg-drafts/selectors-4/#user-pseudos

It uses phrasing like "the user has significantly interacted with it" and "interacted again" as a criterion for when to apply these pseudos, but I'm not clear that we've defined those cases clearly enough to be interoperable. There's an open discussion in the interop issue for this set of features, and this comment (#178 (comment)) attempts to define the behavior in terms of links to Gecko code. That's great, and I'm sure the proposed behavior is reasonable. But I think that this Interop2023 issue is kind of pointless unless we have significantly more testing, and that will require significantly more discussion about exactly how to spec the feature. Chromium is supportive of this feature, but only if it can be implemented, spec'd, and tested interoperably.

How should we proceed?

@emilio @annevk

[annevk]

Right, see whatwg/html#8452.

[mfreed7]

Right, see whatwg/html#8452.

Ahh, thanks! I missed that issue. Looks like my comments here are still correct though - it isn't standardized yet.

[nt1m]

Yeah I do think we should work towards precisely defining the behavior and then writing more precise tests. WebKit uses a different model than Gecko, which I think should be easier to spec (since it just maps directly to change events).

[nt1m]

I'm going to take on the work to spec user-invalid/valid

[mfreed7]

I'm going to take on the work to spec user-invalid/valid

Awesome, thanks!

[nt1m]

Here's my first attempt at specifying: whatwg/html#9047

[nt1m]

@mfreed7 Can you let us know if the new tests are sufficient and close the issue if needed? I'll work on the spec & further tests when I have more time.

[nt1m]

These are the changes fwiw: web-platform-tests/wpt@b669c6c

[mfreed7]

@mfreed7 Can you let us know if the new tests are sufficient and close the issue if needed? I'll work on the spec & further tests when I have more time.

Thanks - the new tests look pretty good. I can't think of another case to check, off-hand, but we'll possibly find more as we start to implement it. But for now I think this is great, so I'll close it.

[mfreed7]

See my comment here whatwg/html#9047 (comment)

I'll re-open this tentatively.

[nt1m]

Given Chromium is implementing :user-valid/:user-invalid, I'm going to close this: https://chromium-review.googlesource.com/4800294

I'm hoping the spec PR can be merged soon as well.