PK should not be generated with random_bytes

[kpp]

https://github.com/TokTok/c-toxcore/blob/2abc00693cc99f277c7c8fe1a201ec33becb01f1/toxcore/DHT.c#L2743..L2751

Since you have a function named public_key_valid, PK from random_bytes generates an invalid PK each 1/2 attempt.

void random_bytes(uint8_t *data, size_t length)
{
    randombytes(data, length); // from NaCl
}

bool public_key_valid(const uint8_t *public_key)
{
    if (public_key[31] >= 128) { /* Last bit of key is always zero. */
        return 0;
    }

    return 1;
}

[iphydf]

Thanks. The keys generated there aren't real keys, they are random keys to seed the DHT search list.

This issue still makes sense to fix, but it's not breaking anything. We never actually find or even want to find these keys.

[kpp]

We never actually find or even want to find these keys.

Actually DHT sends NodesRequest with PKs of fake friends to fill the Kademlia tree faster. It does not break anything, but it smells a little.

[iphydf]

Correct, we send requests for them, but we don't want to find them. I agree it smells and we should fix that. The correct fix involves stronger typing and ensuring that values of type "Public_Key" are never invalid.