members of Object.prototype shouldn't be treated as undeclared variables in no-undef

[michaelficarra]

See discussion in #164. Related: #6.

[nzakas]

JSHint warns on these. Is there a reason we should act differently?

[michaelficarra]

The rule is described as "locate potential ReferenceErrors resulting from mispellings [sic] of variable and parameter names or accidental implicit globals". The members of Object.prototype are in scope, so usage will not cause a ReferenceError and assignments will not cause an own-property of the global object to be created.

[anba]

Strictly speaking the current behaviour is compliant to ECMAScript, because the [[Prototype]] of the global object is implementation specific. That means a compliant ECMAScript implementation can set the global object's [[Prototype]] to null or any other object.

[nzakas]

That's a very good point. Given that, I think we should stick with the current behavior.

[michaelficarra]

@anba: That's a really good point. I had always taken it for granted that the global object's [[Prototype]] was Object.prototype. But you're right that it's not specified that way. I am going to ask es-discuss whether this was an omission or intentional, in which case a spec NOTE making it explicit would be handy.

[michaelficarra]

ECMA-262 bug filed here: https://bugs.ecmascript.org/show_bug.cgi?id=1714. The spec will soon enforce that Object.prototype is in the prototype chain of the global object.