What is the rationale behind Apple's pattern of checking return value rather than error?

Apple's guide on Using and Creating Error Objects gives the following code example:

NSError *theError;
BOOL success = [myDoc writeToURL:[self docURL] ofType:@"html" error:&theError];

if (success == NO) {
    // Maybe try to determine cause of error and recover first.
    NSAlert *theAlert = [NSAlert alertWithError:theError];
   [theAlert runModal]; // Ignore return value.
}

and accompanies it with the statement:

Important: Success or failure is indicated by the return value of the method. Although Cocoa methods that indirectly return error objects in the Cocoa error domain are guaranteed to return such objects if the method indicates failure by directly returning nil or NO, you should always check that the return value is nil or NO before attempting to do anything with the NSError object.

I have always been wondering why is this pattern so important? Why should we ALWAYS check the return value? What's wrong if we check whether the error is nil or not?

Solution

This design is not terribly unusual, compare also errno in standard C.

The design has a number of potential advantages:

The function does not have to write through a pointer on success. This does not only make the implementation of such functions easier and less error prone, it can also be a small performance advantage (e.g. this prevent CPU caches from being invalidated if the function succeeds).
If we always check that the function failed before accessing an error, we can use that same error pointer for multiple functions. Otherwise, we might get a previous failure rather than the failure of the most recent function.
This makes validation code easier to write. E.g. a function could set the error by default. If all validations pass, the function can simply return success instead of having to reset the error variable.
A function can use the same error pointer when calling other functions, but a failure of these helpers doesn't necessarily imply a failure of the top function.

In your specific case, the variable NSError *theError; has not been initialized. Accessing that variable without assigning to it first would invoke undefined behaviour. The documentation only guarantees that the variable will be set in case of an error.