Re: Compiler warnings with the new base code

[Richard Frith-Macdonald]

On 18 Feb 2010, at 11:02, David Chisnall wrote:

> On 18 Feb 2010, at 10:59, Richard Frith-Macdonald wrote:
> 
>> PS. the above test is checking conformance with OSX behavior
>> 
>> This will be a cached NSNumber (one of the small integer values), and 
>> presumably OSX will just return the cached value rather than a copy of it, 
>> because they think that using cached values is more important than using the 
>> correct zone.
> 
> 
> That's true.  OS X will always return the same singleton instance.  If this 
> level of compatibility with OS X is desirable, then I can make the change, 
> but then we'd be copying undocumented (and subject to change) OS X behaviour 
> at the expense of doing the right thing...
> 
> Your call.

I'm afraid that, unless there really is a very strong reason to 'do the right 
thing', OSX compatibility wins.   It's simple pragmatism that, since people 
complain about incompatibilities even when we are doing the right thing 
according to the documentation and even when our behavior seems more sane, it's 
easier to track OSX than to keep wasting time dealing with complaints.

One of the points of having the testsuite is to track undocumented behaviors 
(or behaviors which contravene the documentation).  We can run it with OSX 
Foundation and with gnustep-base and see any incompatibilities (and also see if 
Apple change their behavior).