This copy also had a bug in it, it fails to incorporate the length
into the hash by using it as the loop variable so it'll always be -1
by the time it is XORed in.
As such: I'm doing this only in Python 3.3 and not backporting as it
would change the existing hash behavior of datetime objects.
return clone;
}
-/*
- Borrowed from stringobject.c, originally it was string_hash()
-*/
static Py_hash_t
generic_hash(unsigned char *data, int len)
{
- register unsigned char *p;
- register Py_uhash_t x;
-
- p = (unsigned char *) data;
- x = (Py_uhash_t)*p << 7;
- while (--len >= 0)
- x = (1000003U*x) ^ (Py_uhash_t)*p++;
- x ^= (Py_uhash_t)len;
- if (x == -1)
- x = -2;
-
- return x;
+ return _Py_HashBytes(data, len);
}