Remove redundant seed

[FloopCZ]

I believe those seed commands are unnecessary after the constructor with random device.
Moreover, this only causes that seeding in the same second will produce the same number.

[beniz]

Hi, default seed in C++ for the Mersenne twister engine is is 589u, see http://www.cplusplus.com/reference/random/mersenne_twister_engine/mersenne_twister_engine/
We really want a random seed here.

If you have use case where the seed needs a sub-second init, let's review it here.

[FloopCZ]

Hi @beniz, this is a misunderstanding. Naturally, I would not bother you with a PR initializing mt19937 with a fixed value, I will try to explain the problem. 🙂

Take a look at the affected lines:

    std::random_device rd;
    _uhgen = std::mt19937(rd());
    _uhgen.seed(static_cast<uint64_t>(time(nullptr)));  // This line is removed by this PR.

You create std::random_device (i.e., a link to /dev/urandom or a similar source of truly random numbers) and pass a random number from this device to the Mersenne twister's constructor as a seed. So there is no need to seed it again using the current time. On the contrary, if you seed it again like this:

    _uhgen.seed(static_cast<uint64_t>(time(nullptr)));

, then the seed value changes only every second, so constructing multiple Mersenne twisters at the same second actually means that they will generate the same sequence of random numbers.

[FloopCZ]

Hi @beniz, it has been a while. 🙂
Do you think it would be possible to take another look at this PR and my last comment? I hope it explains the current bug and why it is fixed by the proposed PR.

[beniz]

Hi @FloopCZ this may have slipped below my radar at the time. You are correct the twister is wrongly seeded twice, with the second seeding being worse than the first one.

Your last commit is breaking this PR though, what you want is to do a rebase to master, or resubmit a PR, and I'll merge it asap.

Thanks for your patience and persistence!

[FloopCZ]

Your last commit is breaking this PR though, what you want is to do a rebase to master, or resubmit a PR, and I'll merge it asap.

Rebased and force pushed!