On 2013-02-20 10:49, Stefan Farfeleder <stef...@freebsd.org> wrote:
>On Wed, Feb 20, 2013 at 09:32:43AM +0000, David Chisnall wrote:
>>On 20 Feb 2013, at 08:25, m...@freebsd.org wrote:
>>> These should be declared const int *. And the cast shouldn't be
>>> needed in C, since void * can be assigned to any other pointer type.
>>
>> In fact, the entire function body can be replaced with:
>>
>> return (*(int*)p1 - *(int*)p2);
>>
>> qsort doesn't require that you return -1, 0, or 1, it requires you return
>> <0, 0, or >0.
>
> The subtraction might overflow and give wrong results. It won't for
> these specific elements, but it would be a bad example, IMHO.
That's a good point. The Linux version of the manpage uses a string
comparison function as an example, *and* a subtraction, which then
requires a lengthy comment to explain what's happening and why all the
casts:
static int
cmpstringp(const void *p1, const void *p2)
{
/* The actual arguments to this function are "pointers to
pointers to char", but strcmp(3) arguments are "pointers
to char", hence the following cast plus dereference */
return strcmp(* (char * const *) p1, * (char * const *) p2);
}
Now I prefer sticking with the rather explicit and rather simple to
understand version:
/*
* Custom comparison function that can compare 'int' values through
pointers
* passed by qsort(3).
*/
static int
int_compare(const void *p1, const void *p2)
{
const int *left = p1;
const int *right = p2;
if (*left < *right)
return (-1);
else if (*left > *right)
return (1);
else
return (0);
}
Even the comment is not stricly needed. The code is simpler than the
version with the casts, especially if the casts have to be repeated to
avoid subtraction induced underflows.
_______________________________________________
svn-src-head@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-head
To unsubscribe, send any mail to "svn-src-head-unsubscr...@freebsd.org"
