Coverage area: document.all

[littledan]

V8's test262 test harness doesn't yet expose anything for document.all-like objects. The only test262 tests I see failing due to that omission are related to iteration. I would've expected there to be some additional tests failing, such as ensuring that the document.all-like object will be === to neither null nor undefined (which would fail for us, as it is just undefined). There's a bunch of places in the spec where SameValue(undefined) comes up; a few more of these could be tested.

[bakkot]

See previously at #1184, at which I believe the consensus was that those tests belong in the web platform tests. (Actually, I see you commented in that thread agreeing with that, if I'm reading it correctly, so now I'm confused.)

[littledan]

I'm confused too--why is are there tests like this checked in? https://github.com/tc39/test262/blob/master/test/annexB/language/expressions/yield/star-iterable-return-emulates-undefined-throws-when-called.js#L13

[littledan]

About my previous comment: I'd be OK either way on the question, really. Now that I see it wouldn't be much work to expose in V8, and also that the document.all semantics question comes up in a big way for things like the optional chaining operator, I could see the case for having good test262 tests.

[bakkot]

I'm confused too--why is are there tests like this checked in?

See discussion in #1287, but briefly, if I recall correctly, it's because they're mostly testing ES algorithms which don't directly reference document.all, rather than testing document.all per se.

[rwaldron]

@jswalden ping

[jswalden]

I'm not entirely certain what's being asked here. If the question is why that test's checked in, #1372 (comment) is right. Submitting such test in w-p-t would result in someone there having to learn a bunch of very silly ECMA goo well beyond their usual ambit, and squarely in TC39's wheelhouse and competencies.

I do think we could have more testing of places that do Type(x) is Undefined or SameValue(x, undefined). (The latter is less common, I think.) We should be sure to have tests that throw [[IsHTMLDDA]] objects at those places, that would suss out mis-implementation by loose-equality comparison to undefined. I wouldn't make it a super-high priority, but it looks like this issue just flags it as a concern and nothing more.

[gsnedders]

https://github.com/web-platform-tests/wpt/blob/master/html/obsolete/requirements-for-implementations/other-elements-attributes-and-apis/document-all.html already has tests in WPT for this (and has since 2015), predating TC39 specifying [[IsHTMLDDA]].