Intermittent NetworkError caused by TypeError in verification Lambda

[athewsey]

Hi folks,

I've been seeing an intermittent issue which manifests as NetworkError in the front end, but the underlying cause seems to be an unhandled exception in the verification Lambda (because CORS headers don't get set on the error response through API Gateway, so the issue appears as a CORS/network issue on browser side).

Stack trace below but I haven't had chance to dig in to it yet - I wonder if there's something about the number of frames uploaded for the check that can affect the dimensionality of the variable in question?

[ERROR] TypeError: only size-1 arrays can be converted to Python scalars
Traceback (most recent call last):
  File "/var/task/app.py", line 52, in lambda_handler
    status_code, response = execute_if_token_is_valid(body['token'], challenge_id, verify_challenge, challenge_id)
  File "/var/task/app.py", line 68, in execute_if_token_is_valid
    return func(*args)
  File "/var/task/app.py", line 143, in verify_challenge
    state_manager.process(frame)
  File "/var/task/states/manager.py", line 19, in process
    success = self.current_state.process(frame)
  File "/var/task/states/nose.py", line 66, in process
    verified = self.verify_challenge(rek_landmarks, rek_pose, self.challenge_in_the_right)
  File "/var/task/states/nose.py", line 92, in verify_challenge
    trajectory_error = math.sqrt(residuals/len(self.nose_trajectory))

[rafaelwmartins]

Hi. Yes, you're right regarding why errors manifest as NetworkError. That needs to be fixed although is not as important as the underlying error, which is totally unexpected and should never happen. We'll take a look on that exception soon. Thanks for the details.

[helbertDev3c]

I am facing the same problem, how can it be solved?

[eddturtle]

I appreciate this might not be that useful, but I think the issue lies in nose.py inside function verify_challenge().

The math sqrt() function only takes in 2 variables, residuals and nose_trajectory - so it has to be in one of those 2. And I think nose_trajectory is ok, so I'm wondering if residuals could be an empty array of some kind?

[eddturtle]

Based on the info here: https://stackoverflow.com/a/26438018/402585

It does seem to work if you change the code in nose.py verify_challenge() to:

_, residuals, _, _, _ = np.polyfit(nose_trajectory_x, nose_trajectory_y, 2, full=True)

# if empty array, set to zero
if not residuals:
    residuals = 0

trajectory_error = math.sqrt(residuals/len(self.nose_trajectory))

[rznadri]

I ran to this issue and the change suggested by @eddturtle fixed the issue.