Re: NSNumber Bug I can't fix
Hi Richard, > Thanks, I added a quick hack to try to fix this and put a few very simple > tests in > svn://svn.gna.org/svn/gnustep/libs/base/trunk/Tests/base/NSNumber/test01.m thanks for the quick response & hotfix. I can confirm that it works here (FreeBSD 8.2rc3 amd64). Cheers, Marcus -- Marcus Mueller . . . crack-admin/coder ;-) Mulle kybernetiK . http://www.mulle-kybernetik.com Current projects: http://www.mulle-kybernetik.com/znek/ smime.p7s Description: S/MIME cryptographic signature ___ Gnustep-dev mailing list Gnustep-dev@gnu.org http://lists.gnu.org/mailman/listinfo/gnustep-dev
Re: NSNumber Bug I can't fix
On 17 Feb 2011, at 16:01, Marcus Müller wrote:
>
>> As you can see above, asking an NSIntNumber instance for its doubleValue
>> works properly, however, it fails at NSNumber.m:165. I don't know the
>> reason, but I also don't understand the implications of the macros used. Can
>> somebody elaborate? Please note that I'm on a 64-bit (amd64) system.
>
> Thinking about it, I guess my hint was pretty bogus. I believe the real
> problem here is the COMPARE macro which compares a double with a NaN, which
> simply won't work:
>
> #define COMPARE(value, other) \
> if (value < other)\
> {\
> return NSOrderedAscending;\
> }\
> if (value > other)\
> {\
> return NSOrderedDescending;\
> }\
> return NSOrderedSame;
>
> If either is a nan, this will return NSOrderedSame which is, well, not
> incorrect. However, in the context of NSNumber.m:165 it returns a wrong
> result.
>
> This raises the question, what should be done to - (NSComparisonResult)
> compare: (NSNumber*)aNumber?
> Comparing a NaN with another number is undefined and might lead to
> NSOrderedSame, but it should never lead to [2 isEqualToNumber:NaN] == YES.
> I'm not sure how to fix this properly, but not returning NSOrderedSame in
> this case would certainly help.
Thanks, I added a quick hack to try to fix this and put a few very simple tests
in svn://svn.gna.org/svn/gnustep/libs/base/trunk/Tests/base/NSNumber/test01.m
___
Gnustep-dev mailing list
Gnustep-dev@gnu.org
http://lists.gnu.org/mailman/listinfo/gnustep-dev
Re: NSNumber Bug I can't fix
For reference, under Mac OS X 10.6.6: 2 compare: nan = NSOrderedDescending nan compare: 2 = NSOrderedAscending nan compare: nan = NSOrderedSame 2 isEqual: nan = NO nan isEqual: n = NO nan isEqual: nan = YES Just for fun, I tried it with [NSNumber numberWithDouble:INFINITY] instead of 2: inf compare: nan = NSOrderedDescending nan compare: inf = NSOrderedAscending -inf compare: nan = NSOrderedAscending nan compare: -inf = NSOrderedDescending -- Jens Ayton ___ Gnustep-dev mailing list Gnustep-dev@gnu.org http://lists.gnu.org/mailman/listinfo/gnustep-dev
Re: NSNumber Bug I can't fix
> As you can see above, asking an NSIntNumber instance for its doubleValue
> works properly, however, it fails at NSNumber.m:165. I don't know the reason,
> but I also don't understand the implications of the macros used. Can somebody
> elaborate? Please note that I'm on a 64-bit (amd64) system.
Thinking about it, I guess my hint was pretty bogus. I believe the real problem
here is the COMPARE macro which compares a double with a NaN, which simply
won't work:
#define COMPARE(value, other) \
if (value < other)\
{\
return NSOrderedAscending;\
}\
if (value > other)\
{\
return NSOrderedDescending;\
}\
return NSOrderedSame;
If either is a nan, this will return NSOrderedSame which is, well, not
incorrect. However, in the context of NSNumber.m:165 it returns a wrong result.
This raises the question, what should be done to - (NSComparisonResult)
compare: (NSNumber*)aNumber?
Comparing a NaN with another number is undefined and might lead to
NSOrderedSame, but it should never lead to [2 isEqualToNumber:NaN] == YES. I'm
not sure how to fix this properly, but not returning NSOrderedSame in this case
would certainly help.
Cheers,
Marcus
--
Marcus Mueller . . . crack-admin/coder ;-)
Mulle kybernetiK . http://www.mulle-kybernetik.com
Current projects: http://www.mulle-kybernetik.com/znek/
smime.p7s
Description: S/MIME cryptographic signature
___
Gnustep-dev mailing list
Gnustep-dev@gnu.org
http://lists.gnu.org/mailman/listinfo/gnustep-dev
NSNumber Bug I can't fix
Hi,
I just stumbled across a bug in NSNumber that I can't fix as I don't understand
the NSNumber code. The problem is easy to see, though:
#import
int main(int argc, char *argv[]) {
NSAutoreleasePool *pool = [[NSAutoreleasePool alloc] init];
NSNumber *n = [NSNumber numberWithInt:2];
NSNumber *nan = [NSDecimalNumber notANumber];
NSLog(@"%@ (class:%@ double:%g) == %@ -> %s",
n, NSStringFromClass([n class]), [n doubleValue],
nan, [n isEqualToNumber:nan] ? "YES" : "NO");
[pool release];
return 0;
}
On FreeBSD-8.2rc3 (amd64) this produces the following output:
2011-02-17 16:02:15.316 NumberTest[59574] 2 (class:NSIntNumber double:2) == NaN
-> YES
Which is clearly wrong.
I came across this issue via NSNumberFormatter, which instead of formatting any
integer numbers always printed NaN. In my project I tried to debug into the
issue and found this:
769 if ([anObject isEqual: [NSDecimalNumber notANumber]])
(gdb) s
-[NSNumber isEqual:] (self=0x803e8d1d0, _cmd=0x801410980, anObject=0x805e625c0)
at NSNumber.m:376
376 if ([anObject isKindOfClass: NSNumberClass])
(gdb) n
378 return [self isEqualToNumber: anObject];
(gdb) po anObject
NaN
(gdb) po [anObject class]
NSDecimalNumber
(gdb) s
-[NSNumber isEqualToNumber:] (self=0x803e8d1d0, _cmd=0x80140d300,
aNumber=0x805e625c0) at NSNumber.m:371
371 return ([self compare: aNumber] == NSOrderedSame) ? YES : NO;
(gdb) s
-[NSSignedIntegerNumber compare:] (self=0x803e8d1d0, _cmd=0x80140d2e0,
aNumber=0x805e625c0) at NSNumber.m:107
107 if (aNumber == self)
[…]
(gdb) n
165 double value = [self doubleValue];
(gdb) p other
$3 = nan(0x8)
(gdb) p value
$4 = 6.9533556002420585e-310
As you can see above, asking an NSIntNumber instance for its doubleValue works
properly, however, it fails at NSNumber.m:165. I don't know the reason, but I
also don't understand the implications of the macros used. Can somebody
elaborate? Please note that I'm on a 64-bit (amd64) system.
Thanks,
Marcus
--
Marcus Mueller . . . crack-admin/coder ;-)
Mulle kybernetiK . http://www.mulle-kybernetik.com
Current projects: http://www.mulle-kybernetik.com/znek/
smime.p7s
Description: S/MIME cryptographic signature
___
Gnustep-dev mailing list
Gnustep-dev@gnu.org
http://lists.gnu.org/mailman/listinfo/gnustep-dev
