add ./setup_sut.sh wolfssl-5.8.2

[c-southwest]

I don't have enough time, I added the wolfssl-5.8.2 into the setup script, so you can run the experiments locally.

you can install wolfssl-5.8.2 by
./setup_sut.sh wolfssl-5.8.2

and run experiments like this:

Client (CERT OPT)

java -jar target/dtls-fuzzer.jar @args/wolfssl/learn_wolfssl_client_x_cert_nreq -Dwolfssl.version=5.8.2 -eqvThreads 8

Client (CERT REQ)

java -jar target/dtls-fuzzer.jar @args/wolfssl/learn_wolfssl_client_x_cert_req -Dwolfssl.version=5.8.2 -eqvThreads 8

[kostis]

Thanks @c-southwest for taking the lead on this!

I've run the experiment with CERT_OPT. We are learning a model (with 47 states) after quite a lot of time and many tests (15798 when the last hypothesis was generated). Learning the DTLS 1.2 model suffers from non-determinism, which we sort of knew already because @pfg666 has reported three different issues to the wolfSSL developers which are all acknowledged as bugs and two of them have already been fixed.

Given that the numbers we get suffer from bugs in the underlying SUL, I do not think we can use them for comparison with the DTLS 1.3 learning experiments. Too bad for this, but at least we've tried. So, we'll leave the table in the paper incomplete.

I will also leave this PR open for a while so that we can revisit it with a newer version of wolfSSL that contains the fixes to the bugs that @pfg666 reported, so that we see if the learning time and tests have improved.

[c-southwest]

@kostis I forget to append argument -eqvThreads 8, without this argument learning could be slow. (now I edited my post)

I also added ./setup_sut.sh wolfssl-5.8.2 function in pull request #219