Should namespace objects have "length" and "name" own properties?

[foolip]

I can't find anything in https://heycam.github.io/webidl/#namespace-object to suggest that a namespace object should have length or name properties, but looking at namespace objects in implementations the results are mixed:

• CSS.length is 0 and CSS.name is "CSS" in Chrome, Firefox and Safari
• console.length is 0 and console.name is "console" in Firefox, but both are undefined in Chrome and Safari
• WebAssembly.length and WebAssembly.name are undefined in Chrome, Firefox and Safari

I don't think this has caused any problems, but it could be a future nuisance making it harder to introduce a new namespace which has a name member. (Wanting a length member seems less likely.)

With confirmation that these shouldn't be there, I could add negative tests for it.

[domenic]

They should not. This is due to incorrect implementations of them as functions instead of namespace objects. You could also test that typeof x === "object" (not "function").

[TimothyGu]

It appears to be a bit more nuanced. On Chrome 88, CSS.length and CSS.name are both undefined, contrary to @foolip's findings. OTOH typeof CSS is "object" on both Chrome and Firefox here, though it's "function" on WebKit.

[foolip]

I was testing in Chrome 87. In Chrome Canary, I see that CSS.length and CSS.name are indeed both undefined. And typeof CSS has changed from "function" to "object". This was a recent fix, perhaps in https://chromium-review.googlesource.com/c/chromium/src/+/2463092. @yuki3 was that an intentional effect of that CL?

Regardless, sounds like we should have tests for each of the namespace objects to ensure they're objects, not functions. Checking that they don't have length and name is then a bit redundant, but maybe we should.

[foolip]

Turns out idlharness.js tests this. In https://wpt.fyi/results/css/cssom/idlharness.html?label=experimental&label=master&aligned, the 'CSS namespace: typeof is "object"' test fails in Safari. I'll file bugs.

[foolip]

OK, I've filed https://bugs.webkit.org/show_bug.cgi?id=220855.

[yuki3]

Yes, the change was intentional. https://chromium-review.googlesource.com/c/chromium/src/+/2434404 is the patch to make CSS an IDL namespace.

[saschanaz]

I think Gecko does have a bug here, filed https://bugzilla.mozilla.org/show_bug.cgi?id=1688335

[foolip]

Thanks for confirming @yuki3, I didn't think to check if CSS was recently changes to a namespace and just assumed it was a bug with namespace objects generally.

And thanks @saschanaz for filing that Gecko bug, I forgot about console when I closed the issue.