Re: [NSNumber numberWithFloat:] issue

[Luboš Doležel]

Hi,

that's strange - I believe that CFRetain() already worked with small 
NSStrings, because it calls object_getClass() and that returns a special 
class type for these, hence CFRetain() should just forward the call to 
[obj retain], which is a NOP.

Could you please send a backtrace with the crash?

Luboš

On Wed, 18 Feb 2015 20:39:27 -0600, Stefan Bidigaray wrote:
> That's because CoreBase does not understand "small objects".  These
> are objects that have their contents stored as part of the pointer,
> instead of being an actual object.  Because they aren't actually
> objects, they don't have any reference counting.  Passing a small
> object around is like passing a int or double, and do not have to 
> call
> malloc or free, making them very fast.
>
> I can implement a quick fix, but it might not until the weekend.  It
> would also be very helpful if you created a test case that I could
> include in the test suite.  Nothing fancy.  Or, if you're not
> familiar with the test suite, a simple program with the expected
> output.
>
> Thanks
>
> On Wed, Feb 18, 2015 at 6:16 PM, Amr Aboelela  wrote:
>
> > Hello, 
> >
> > I found that [NSNumber numberWithFloat:] is not toll free bridged
> > with CoreFoundation
> > if you do CFRetain on an object result from [NSNumber
> > numberWithFloat:] it will crash, and if you do a retainCount on it,
> > it will always give you -1
> >
> > I feel suspicious about this line:
> >
> > return boxDouble(aValue, SMALL_FLOAT_MASK);
> > _______________________________________________
> > Discuss-gnustep mailing list
> > Discuss-gnustep@gnu.org [1]
> > https://lists.gnu.org/mailman/listinfo/discuss-gnustep [2]