prng_bleagh.c

/* MacOS libc PRNG is garbage!
 * Test this on MacOS 10.12.3 (may be the same with other versions)
 * You'll see three times the same output.
 * Cheers by @antirez. */

#include <stdio.h>
#include <stdlib.h>

int main(void) {
    /* Note, 1000000, 1000001, 1000002 are not special.
     * You can use any other "near" seeds and MOD 7 the sequence will be the same. */
    srand(1000000);
    printf("%d\n", rand() % 7);
    srand(1000001);
    printf("%d\n", rand() % 7);
    srand(1000002);
    printf("%d\n", rand() % 7);
    return 0;
}

^ makes sense.

using the first seed gives

4
1
1
3
0

or

21585
18586
29373
4301
3304

without the modulus

every time.

Less a bug, more an intended, but sucky, feature.

There is some confusion here about crypto-PRNG, my example above, which is not limited to the seeds I selected, and so forth. So:

1. The problem is not that the rand() sequence is deterministic. Here I'm not even checking that its distribution is random-looking.
2. The special seeds I wrote above are just examples, the % 7 same results for different X, X+1, X+2 seeds will happen for almost all the seeds.
3. It is trivial to have a fast PRNG that provides a decent distribution, and that by changing the seed produces a very different sequence.
4. My claim is documented in the implementation itself... so the code acknowledges it.
5. srand(time(NULL)) is a huge programming pattern to just say: give me a different sequence every different second I run it.

@colmmacc read my last post. You misunderstood the point completely. Happens for every possible N seeds that are just similar numbers.

TLDR: it happens even with this code.

#include <stdio.h>
#include <stdlib.h>
#include <time.h>

int main(void) {
    int seed = time(NULL);
    srand(seed+1);
    printf("%d\n", rand() % 7);
    srand(seed+2);
    printf("%d\n", rand() % 7);
    srand(seed+10);
    printf("%d\n", rand() % 7);
    return 0;
}