[patch] [cla] doh._objPropEq() gives wrong result when one, and only one, of its arguments is 'null'.

[Joseph Scheuhammer]

The problem shows up in both doh.assertEqual() and doh.asserNotEqual() just before calling this._objPropEq(). They first determine if the typeof their arguments is "object" before making the call.

The problem is that (typeof null) evaluates to "object". Thus, for example, if expected is null, and actual is a non-null object, then doh._objPropEq() is called like this:

	...
	this._objPropEq(null, nonNullObject);
	...

Inside doh._objPropEq() are two for(x in ...) loops for examining the equality of expected's vs, actual's properties. But, trying to access any property of a null throws a type error. FF/Firebug reports, "TypeError: expected is null".

However, doh.assertNotEqual(null, nonNullObject), for example, should succeed since actual is, in fact, not null.

There needs to be a check to insure that both arguments to doh._objPropEq() are non-null before looking at their properties.

[Joseph Scheuhammer]

The attached file, "nullExample.html" shows the error using doh.assertNotEqual().

If running in FF3 with Firebug installed, the error shows up in Firebug's console.

If running in IE7, an alert is posted saying, "'null' is null or not an object".

[Joseph Scheuhammer]

The following patch, "8097.patch", is a way fix the problem.

The rationale is to check for null in doh._objPropEq() itself since, already, two public methods of doh call it. The alternative is to check for null before calling _objPropEq(). The likelihood, if anything, is that the popularity of _objPropEq() will increase; better to put the check in one place.

The logic of the patch is:

• if both expected and actual are null, return 'true' (their "properties", or lack thereof, do match, OR, they are equal objects).
• if only one of expected or actual is null, return 'false'.
• otherwise, do the check as before.

[Becky Gibson]

needed for dialog keyboard doh test

[bill]

Fixed in [15864].

[Adam Peller]

see also [15865]

[Adam Peller]

(In [15865]) missing an '='. Refs #8097 !strict