strtoul

[P.J. Plauger]

Fred Tydeman asks:

> Assuming that unsigned long int is 32-bits wide, what are the return
> values (in base 16) and errno for the following:

I have annotated his examples along the right margin.
  
                                         returns:
>     strtoul( " 100000000", NULL, 16 ); ULONG_MAX  set errno
>     strtoul( "  FFFFFFFF", NULL, 16 ); FFFFFFFF
>     strtoul( "  FFFFFFFE", NULL, 16 ); FFFFFFFE
>     strtoul( "  80000001", NULL, 16 ); 80000001
>     strtoul( "  80000000", NULL, 16 ); 80000000
>     strtoul( "  7FFFFFFF", NULL, 16 ); 7FFFFFFF
>     strtoul( "         1", NULL, 16 ); 00000001
>     strtoul( "         0", NULL, 16 ); 00000000
>     strtoul( "        -1", NULL, 16 ); FFFFFFFF
>     strtoul( " -7FFFFFFF", NULL, 16 ); 80000001
>     strtoul( " -80000000", NULL, 16 ); 80000000
>     strtoul( " -80000001", NULL, 16 ); 7FFFFFFF
>     strtoul( " -FFFFFFFE", NULL, 16 ); 00000002
>     strtoul( " -FFFFFFFF", NULL, 16 ); 00000001
>     strtoul( "-100000000", NULL, 16 ); ULONG_MAX  set errno

>          I do not see a definition of negative of an unsigned
> long.  It seems to me that the unsigned must be converted to signed,
> then the negation done, then converted back to unsigned.

The negative of an unsigned long follows from its required arithmetic
properties, plus the nature of the representation. For X-Y to
equal X+(-Y), the rules of twos complement arithmetic must obtain,
except that there is no overflow. The correct answer is the low
order N bits of the arithmetic result. No need to convert to
signed.

For further enlightenment, X3J11 has formed a response to
Defect Report #006, submitted by Walter J. Murray,
on the same subject:

The relevant citations are the ones supplied by you from subclause 
7.10.1.6:

o	If the subject sequence begins with a minus sign, the value 
resulting from the conversion is negated.

o	If the correct value is outside the range of representable values, 
ULONG_MAX is returned, and the value of the macro ERANGE 
is stored in errno.

The Committee believes that there is only one sensible interpretation 
of a subject sequence with a minus sign: If the subject sequence (neglecting 
the possible minus sign) is outside the range [0, ULONG_MAX], 
then the range error is reported. Otherwise, the value is negated 
(as an unsigned long int).

> What is the range of representable values for unsigned long?  Is it
> [0..ULONG_MAX], [-ULONG_MAX..ULONG_MAX], everything is representable
> modulo ULONG_MAX+1, or something else?

The range is [0, ULONG_MAX-1]

This is my own unofficial interpretation. I cannot speak for WG14
or for SC22.

P.J. Plauger