test_models: add tx fuzzy test

[bongbui321]

resolve #32425

Uses FuzzyGenerator to generate fuzzy CarControl messages

hypothesis example generation is extremly slow.

[github-actions[bot]]

Thanks for contributing to openpilot! In order for us to review your PR as quickly as possible, check the following:

• Convert your PR to a draft unless it's ready to review
• Read the contributing docs
• Before marking as "ready for review", ensure:
  • the goal is clearly stated in the description
  • all the tests are passing
  • the change is something we merge
  • include a route or your device' dongle ID if relevant

[sshane]

Let me know when you want a review!

[adeebshihadeh]

Nice work - it's a great sign that it's finding bugs!

[adeebshihadeh]

Bounty locked!

hypothesis example generation is extremly slow.

Might be worth looking into https://github.com/commaai/openpilot/issues/32693

[bongbui321]

yeah, ideally, I want to create a list of CarControl messages rather than only one such as now, but it takes nearly 18min on my machine to finish half the tx_fuzzy test.

Might be worth looking into https://github.com/commaai/openpilot/issues/32693

Our current Hypothesis is already the faster version, the latest release of hypothesis is 2x slower :).

[bongbui321]

Might merge tx fuzzy with carstate fuzzy.

Occasionally catches https://github.com/commaai/panda/pull/1948. Not entirely sure why hypothesis doesn't consistenly repro that failed data generation

[bongbui321]

@sshane ready for review

[bongbui321]

I don't think we merged all your bugfixes yet, so how is this passing? Are we not getting good coverage on a single run?

I don't think we merged all your bugfixes yet, so how is this passing? Are we not getting good coverage on a single run?

Nissan max steering angle [issue](https://github.com/commaai/openpilot/issues/32978) is opened and included in this PR

Tesla [issue](https://github.com/commaai/openpilot/issues/32999): This might happen with a few runs, and I suspect that if run locally, hypothesis will try to recreate that failing examples pretty consistently. That requires a more thoughtful fix, which I'm hesitant to open a fix PR since I'm not familiar with the expected behaviour of the platform

The Nissan speed issue is addressed here:

    # Nissan calculate vEgoRaw with avg speed of 4 wheels
    # while panda get vehicle_speed from rear wheels only,
    # so there might be a big enough difference between the two
    if self.CP.carName == "nissan":
      panda_vehicle_speed_last = self.safety.get_vehicle_speed_last() / VEHICLE_SPEED_FACTOR
      if abs(self.CI.CS.out.vEgoRaw - panda_vehicle_speed_last) > 0.2:
        cs_msg = self.CI.CS.out.to_dict()
        cs_msg["vEgoRaw"] = panda_vehicle_speed_last
        self.CI.CS.out = car.CarState.new_message(**cs_msg).as_reader()

This realistically shouldn't happen front wheels and rear wheel speed should be close

[adeebshihadeh]

Can you rebase?

[maxime-desroches]

This PR was closed because of a git history rewrite. Please read https://github.com/commaai/openpilot/issues/33399 for what to do with your fork and your PRs.