Add default case for non-member types in assert_type_is

[stephenmcgruer]

Fixes #23346

[stephenmcgruer]

Stability check failure was Firefox timing out. Results from wpt.fyi suggest no regressions, which is reasonable minimum bar.

Josh, Domenic - wdyt of this as a fix for #23346 ?

[domenic]

The logic here LGTM. And if this causes no regressions then I'm happy to approve.

I worry a bit that without this, people will often forget to input important dependent IDL into their testharness tests. But probably the benefit in simple cases outweighs that.

[inexorabletash]

Hmmm... Is the set of intrinsic types from WebIDL enumerated anywhere?

For example, I see "Function" here but it's not listed anywhere in webidl2.js's source. It's a defined type in WebIDL that "coincidentally" has the same name in JS.

What I'm worried about: say someone writes an IDL with RegExp - a JS type but not a WebIDL type. Would that get caught by idlharness before (giving an error), but slip through now?

Approving though since this assert function is not where such things should get caught.

[inexorabletash]

Following up to my previous comment: looks like idlharness indeed doesn't validate that types are known. Having a function return any of (Function, VoidFunction, NoSuchThing, RegExp) don't cause errors. Plenty of open issues against idlharness already, though.

Still LGTM, thanks!

[stephenmcgruer]

Ok. I'm a bit hesitant here because the experts (y'all) are unsure (and I know so very little). But then again, experts also approved this so I guess its good :D.

Admin-merging because Firefox stability check is timing out due to number of tests affected.