search

ensenso / ros_driver

Official ROS driver for Ensenso stereo cameras.
http://wiki.ros.org/ensenso_driver
BSD 3-Clause "New" or "Revised" License
29 stars 25 forks source link

Fixed sporadic bug in hand-eye calibration where feedback was attempted before progress was available #40

Closed jornb closed 4 years ago

jornb commented 4 years ago

I noticed a sporadic crash because the Progress node did not exist during hand-eye calibration. This was with a virtual camera.

Cause

During hand-eye calibration, the following code is executed

NxLibCommand calibrateHandEye(cmdCalibrateHandEye, serial);
calibrateHandEye.execute(false);
while (!calibrateHandEye.finished()) {
    ensenso_camera_msgs::CalibrateHandEyeFeedback feedback;
    feedback.number_of_iterations = calibrateHandEye.result()[itmProgress][itmIterations].asInt();
    feedback.residual = getCalibrationResidual(calibrateHandEye.result()[itmProgress]);
    calibrateHandEyeServer->publishFeedback(feedback);
}

I concluded that this must be because of a race condition, in which the main thread is reading the Progress node created by a worker thread. Printing the contents of calibrateHandEye.result() reveals that the Progress node is not created until some time has passed:

[ INFO] [1579082632.274600800]: progress: {
        "Time": 2.2157,
        "TimeExecute": 0.432599999999999651,
        "TimeFinalize": 0.0316999999999998394,
        "TimePrepare": 1.24019999999999997
}
[ INFO] [1579082632.775791000]: progress: {
        "Progress": {
                "Iterations": 11,
                "Residual": 0.0903454391874089907
        },
        "Time": 404.227700000000027,
        "TimeExecute": 0,
        "TimeFinalize": 0.129799999999928806,
        "TimePrepare": 0.317699999999999982
}

Fix

I stopped sending feedback until the Progress node is created. The action is still preemtable as before.

Setup