test: Add a protocol dump test to generate initial fuzzer input.

[iphydf]

This change is 

[codecov]

Codecov Report

Merging #2291 (50094b7) into master (3a5da35) will decrease coverage by 0.10%.
The diff coverage is 82.35%.

@@            Coverage Diff             @@
##           master    #2291      +/-   ##
==========================================
- Coverage   79.02%   78.92%   -0.11%     
==========================================
  Files         127      127              
  Lines       24021    24063      +42     
==========================================
+ Hits        18983    18992       +9     
- Misses       5038     5071      +33     

Impacted Files	Coverage Δ
toxcore/crypto_core.c	90.28% <ø> (ø)
auto_tests/auto_test_support.c	79.70% <33.33%> (-3.47%)	⬇️
toxcore/network.c	87.39% <97.43%> (+0.64%)	⬆️
toxcore/TCP_server.c	78.58% <0.00%> (-1.87%)	⬇️
toxcore/ping.c	84.56% <0.00%> (-1.24%)	⬇️
toxcore/Messenger.c	82.76% <0.00%> (-1.07%)	⬇️
toxcore/TCP_connection.c	69.45% <0.00%> (-0.87%)	⬇️
toxcore/friend_connection.c	83.25% <0.00%> (-0.70%)	⬇️
toxav/toxav.c	69.08% <0.00%> (-0.15%)	⬇️
toxcore/tox.c	64.65% <0.00%> (-0.08%)	⬇️
... and 5 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3a5da35...50094b7. Read the comment docs.

[sudden6]

Reviewed 8 of 11 files at r1, 6 of 7 files at r2, all commit messages.
Reviewable status: 1 change requests, 0 of 1 approvals obtained (waiting on @iphydf)

---

testing/fuzzing/bootstrap_harness.cc, line 170 at r1 (raw file):

    uint8_t pub_key[TOX_PUBLIC_KEY_SIZE] = {0};

    const bool udp_success = tox_bootstrap(tox, "192.168.0.127", 33446, pub_key, nullptr);

Any reason to use IP addresses which might actually be used on the host machine? also how about reading them from fuzzer input? (just a thought and it will invalidate the corpus)

---

testing/fuzzing/e2e_fuzz_test.cc, line 131 at r1 (raw file):

}

static char tox_log_level_name(Tox_Log_Level level)

This function appears multiple times, make it constexpr and put it in a header?

---

testing/fuzzing/e2e_fuzz_test.cc, line 196 at r2 (raw file):

        // Move the clock forward a decent amount so all the time-based checks
        // trigger more quickly.
        sys.clock += 200;

I think this needs to be synced between all the fuzzers to be effective, right? This should be documented somewhere, and maybe a named constant. Also I know I was against this earlier, but I think now it really makes sense to read the time step from fuzzer input data so that all the behavior is contained in the corpus files.

---

testing/fuzzing/protodump.cc, line 188 at r2 (raw file):

{
    std::printf("%zu bytes: %s\n", recording.size(), filename);
    std::ofstream out_init(filename);

Might want to set the binary flag here

[iphydf]

Reviewable status: 1 change requests, 0 of 1 approvals obtained (waiting on @sudden6)

---

testing/fuzzing/bootstrap_harness.cc, line 170 at r1 (raw file):

Previously, sudden6 wrote…

Any reason to use IP addresses which might actually be used on the host machine? also how about reading them from fuzzer input? (just a thought and it will invalidate the corpus)

Reading them from the fuzzer input seems either risky or useless. In case something in toxcore structurally depends on the IP address, the fuzzer will have only a 1 in 4 billion chance of getting the right number. I don't think anything does depend on it, but even if it does, I don't think we'll find anything interesting from twiddling this particular number. If nothing indeed does depend on it, the input here only tests the "IP to string" and then "IP from string" functions, which are better tested outside the bootstrap fuzz test in a more specific test scenario.

Anyway, I've changed them back to (almost) the number you had here.

---

testing/fuzzing/e2e_fuzz_test.cc, line 131 at r1 (raw file):

Previously, sudden6 wrote…

This function appears multiple times, make it constexpr and put it in a header?

Done.

---

testing/fuzzing/e2e_fuzz_test.cc, line 196 at r2 (raw file):

Previously, sudden6 wrote…

I think this needs to be synced between all the fuzzers to be effective, right? This should be documented somewhere, and maybe a named constant. Also I know I was against this earlier, but I think now it really makes sense to read the time step from fuzzer input data so that all the behavior is contained in the corpus files.

Done.

---

testing/fuzzing/protodump.cc, line 188 at r2 (raw file):

Previously, sudden6 wrote…

Might want to set the binary flag here

Done.

[sudden6]

Reviewed 6 of 6 files at r3, all commit messages.
Reviewable status: complete! 1 of 1 approvals obtained