Ensure chttpd pids have a proper random seed

[chewbranca]

Overview

Currently chttpd pids are not properly seeded for random number generation. This ensures that they are.

Testing recommendations

GitHub issue number

Related Pull Requests

Checklist

• Code is written and works correctly;
• Changes are covered by tests;
• Documentation reflects the changes;

[nickva]

+1

[davisp]

Don't you need an /integer specification on those?

[davisp]

Also might make sense to pull it out into a function.

[nickva]

Integer is the default, I think

[davisp]

Huh. I would've expected binary as the default but I guess I just add specifiers out of habit.

[chewbranca]

Yeah I'm thinking we would be better served by a function like:

maybe_seed() ->
    case get(random_seed) of
        undefined ->
            <<A:32, B:32, C:32>> = crypto:rand_bytes(12),
            Seed = {A, B, C},
            random:seed(Seed),
            Seed;
        Seed ->
            Seed
end.

[chewbranca]

It's a shame we can't use the rand module, as rand takes care of this automatically, but it's not in all the versions of Erlang we support. The random module has a terrible default seed: https://github.com/erlang/otp/blob/master/lib/stdlib/src/random.erl#L44-L45

[nickva]

@chewbranca why terrible? It's a perfectly fine randomly chosen number :P

https://www.xkcd.com/221/

[nickva]

@davisp agree. Binary would seem to be a default type for a ... binary. Not sure why integers were picked.

[chewbranca]

@nickva well if {3172, 9814, 20125}. was chosen by way of random dice roll then we should be good. I feel bad for whoever had to roll dice that many times, although perhaps they had help from Emanuel Czuber.

[chewbranca]

Closing this out.