Missing properties are not enumerable when checking enumerability

[whitlockjc]

When checking property enumerability (Object.prototype.propertyIsEnumerable) on a String, missing properties should return false instead of throwing a Property {0} not found. exception.

Fixes #415 (regression in 1.7.7.2)

[gbrail]

Thank you for looking at this and fixing it -- it's a real regression.

I'd like to take a little time and figure out if there's a way to fix this without catching the exception and checking the error message on it. Is it just the String class that's throwing an exception from "has"? Perhaps we should fix that, or even fix the implementation of "has" in ScriptableObject so it can't leak out an exception.

[whitlockjc]

Fair enough. I just did it this way since the method throwing is used in other places. This approach makes the special handling (false instead of Exception) isolated the changes to the lowest place without affecting other callers. If you'd rather just not try/catch, I can find all consumers and put the Exception throwing logic in there.

[gbrail]

Given how many things out there extend ScriptableObject and may have the same undesirable behavior for "has," I think that this is a fine fix.

However I think we have more to fix, as I see that V8 and Rhino differ on the enumerability of string properties and other stuff. So I may open an issue to fix the underlying problem, which requires that classes that override "has" should also override "getAttributes" to return the proper attributes for these properties, as NativeArray already does.

[whitlockjc]

Agreed.