[PATCH 1/3] lib: implement strnstr()

[Jerome Forissier]

On 1/6/25 11:14, Rasmus Villemoes wrote:
> On Mon, Jan 06 2025, Jerome Forissier <jerome.forissier at linaro.org> wrote:
> 
>>> + *
>>> + * Return:	pointer to the first occurrence or NULL
>>>   */
>>> -char * strstr(const char * s1,const char * s2)
>>> +char *strnstr(const char *s1, const char *s2, size_t len)
>>>  {
>>> -	int l1, l2;
>>> +	size_t l1, l2;
>>>  
>>> +	l1 = strnlen(s1, len);
>>>  	l2 = strlen(s2);
>>> -	if (!l2)
>>> -		return (char *) s1;
>>> -	l1 = strlen(s1);
>>> -	while (l1 >= l2) {
>>> -		l1--;
>>> -		if (!memcmp(s1,s2,l2))
>>> +
>>> +	for (; l1 >= l2; --l1, ++s1) {
>>> +		if (!memcmp(s1, s2, l2))
>>>  			return (char *) s1;
>>> -		s1++;
>>>  	}
>>> +
>>>  	return NULL;
>>>  }
>>>  #endif
>>
>> This won't return s1 when s2 == NULL, will it?
> 
> Why should it? That's a broken caller. You can't call str* functions
> with NULL pointers. It's ok for s2 to point at an empty string, in which
> case l2 is 0, so the very first memcmp() is guaranteed to succeed and
> yes, this will return s1 in that case.

Sorry I read the strncmp() man page too quickly and mistook empty string
for NULL :-/

> 
> As passing "" as s2 is likely quite rare, the current short-circuiting
> of that case is pointless.

Agreed.
 
>> Why not use a known good implementation such as [1]?
>>
>> [1] https://github.com/freebsd/freebsd-src/blob/main/lib/libc/string/strnstr.c
> 
> Licensing?

Isn't BSD compatible with GPL?


> Also, that's quite unreadable, so please don't.

OK.

> Rasmus

-- 
Jerome