Bug 3537 – object.constructor returns incorrect object

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[David Wheeler]

A growing trend (IMO) in JavaScript is to use objects for namespaces. However, when an object 
constructor is an attribute of an object, the Function.toString() method thinks its an internal function. 
For example, this code:

  var Foo = {};
  Foo.Bar = function () {
    this.bar = 1;
  }; 
  var f = new Foo.Bar();
  document.write(Function.toString.call(f.constructor));

Outputs "(Internal Function)" in Safari and "function () { this.bar = 1; }" in Firefox.  Can Safari be updated 
to output the same string as Firefox?

[David Wheeler]

Actually, looking more closly at it, I realize that the problem is not the stringification of the constructor 
of an object created in a class that uses objects for namespaces. The problem is that 
Function.constructor() returns the wrong function object!

  var Foo = {};
  Foo.Bar = function () {
    this.bar = 1;
  };
  var f = new Foo.Bar();
  document.write(Function.toString.call(f.constructor));

Again, this outputs "(Internal Function)" because the return value of f.constructor is not the actual 
constructor function. I have no idea what function it's returning.

Thanks!

[David Wheeler]

> Function.constructor() returns the wrong function object!

Uh, make that the constructor attribute of the object has the wrong object.

[David Wheeler]

Here's a simpler test case:

  var Foo = { Bar: function () {}};
  var f = new Foo.Bar();
  document.write(f.constructor == Foo.Bar);

This code outputs true in Firefox, but false in Safari. It should output true.

[Joost de Valk (AlthA)]

I'll add the testcase in the comment below as a real testcase, please do so next time you post a bug :). And 
i will confirm it, since the behavior IS different from firefox and some will have to look in to it.

[Joost de Valk (AlthA)]

Created attachment 2347 [details]
testcase

[Eric Seidel (no email)]

Created attachment 4069 [details]
Test case showing broken nameless function support.

[Eric Seidel (no email)]

*** Bug 4042 has been marked as a duplicate of this bug. ***

[Eric Seidel (no email)]

Created attachment 4071 [details]
A small patch to fix this.

[Maciej Stachowiak]

I suggested to Eric that he also test the case of the Function constructor.

[Maciej Stachowiak]

Comment on attachment 4071 [details]
A small patch to fix this.

r=me

Make sure to make the test case also cover the new Function() case.

[David Wheeler]

Shouldn't the first example in the test case also evaluate to true?