runtime: Use [[Define]] for the Symbol.toStringTag property

[ExE-Boss]

See tc39/proposal-iterator-helpers#115 and Firefox’s Bug 1644581 (comment #6).

[ljharb]

I filed #400 which uses a conditional; defineProperty won't work in IE < 9, which is an important target for this package.

[ExE-Boss]

This will work in IE < 9, since it only uses defineProperty if the symbol property already exists in GeneratorPrototype’s prototype chain.

[benjamn]

That's a fair point @ExE-Boss! Do you see value in overriding the existing string, if it's already defined? PR #400 lets the existing string remain unchanged, whereas defineProperty would presumably change it. I have no strong opinion here.

[ExE-Boss]

Well, PR #400 makes it so that once engines start shipping Iterator.prototype[Symbol.toStringTag], then Object.prototype.toString will return [object Iterator] instead of [object Generator] for regenerator’s generator objects.

[ljharb]

That sounds like the right behavior to me, since that's what the standard will dictate.

[ExE-Boss]

But the %Symbol.toStringTag% property of %GeneratorFunction.prototype.prototype% is "Generator", which PR #400 ends up breaking for polyfilled generators.

[ljharb]

I'm a bit confused.

This change was motivated by a web compat concern, where Gp[Symbol.toStringTag] was being assigned, but this was breaking it. Are you saying that it's because it should be defining an own property instead of using [[Set]] and hitting a non-writable property up the [[Prototype]] chain?

[ExE-Boss]

Are you saying that it's because it should be defining an own property instead of using [[Set]] and hitting a non-writable property up the [[Prototype]] chain?

Yes.

[ljharb]

In that case, i think the helper @benjamn added should be modified to conditionally use Object.defineProperty, when it's available. It will need to be used in a try/catch, since IE 8 has the method, but it throws on things that aren't HTML elements.

[ExE-Boss]

That’s what I’m doing in #402.