Add the miou implementation

[dinosaure]

#443 rebased with string. /cc @hannesm feel free to remove the execution of the fuzzer which can take a long time.

[hannesm]

looks great apart from minor comments.

[dinosaure]

Not sure that you want to execute this executable every time, it takes a very long time.

[hannesm]

how long is "very long"? And what are the options? marking it as executable? then we can't attach it to a package without a public_name...

[dinosaure]

how long is "very long"?

2mins on my (good) machine.

And what are the options? marking it as executable? then we can't attach it to a package without a public_name...

We can set it as a simple executable if the test bother you too much (because it takes to much time).

[hannesm]

I'm fine leaving it as test. I usually develop in ocaml 4 land where I don't have miou anyways, and I prefer to have tests that are useful enabled and running (otherwise we hit the "what was the last time that feedback.exe worked" again).

[hannesm]

This looks fine to merge, but:

When given the input:


    ["\203\171\240\014\003\191\176Q\174TZ\212\134\127&\171\029\0315}\1720I\207\132\156\172\238:gy\157\214\180\152\003\204T\161kT\016\224$O\224\183@2\140\142\004\146\155\022]\177\185i\163\229\218u;";
     [[[#1 client-to-server;
        ">8\127%\158\206\218\161u\131T\141\\&\217\178\016>q\028\1402)"]];
      ([#3 noop; #0 close]) => [[[#1 client-to-server;
                                  "\229g\229,`yr\172\252\217\209\026\243\006\223\030\207\n!\175>\174R\185{1\172\255O\215\025\194$7\159f\221\166dA\191\253\173<^\135/!\154\169\132H\202f\024\015\173x?U\135\232Fr"]];
                                ([#0 close; #1 shutdown-recv]) => [[[#1 client-to-server;
                                                                    "\244,#gnm\028\244\241(\144\131v\006\157Z\236\191\243\252\020*\027$\168y\235\154\174\128\031=~\251\235\180\192\184\163\243\170\222\173\171\140\166{\170G\152\135vIa\149\227\194\rD\180C\135L\136"];
                                                                    [?; ?]];
                                                                   ([#0 close;
                                                                    #1 shutdown-recv]) => 
                                                                   [[[?; ?]];
                                                                    (
                                                                    [?; ?]) => 
                                                                    [[[?; ?]];
                                                                    ([?; ?]) => ?]]]]]]
    
the test failed:


    [ERROR] Unexpected result: Ok [] & Error
                                     Unix.Unix_error(Unix.EMFILE, "socket", "")

so is there by chance somewhere a socket leak? or can we reduce the number of sockets being opened?

[dinosaure]

so is there by chance somewhere a socket leak? or can we reduce the number of sockets being opened?

I'm currently on it, due to the random path produced by the fuzzer and taken by the state machine, it's not sure that all sockets are correctly closed. I will ping you back and I will fix that.

[hannesm]

Thanks, I rebased on the main branch and added 6328427. This works nicely on my laptop :)