search

karlpauwels / simtrack

A simulation-based framework for tracking
BSD 3-Clause "New" or "Revised" License
103 stars 40 forks source link

Stopped working with Kinect #7

Closed balloch closed 8 years ago

balloch commented 8 years ago

Hey there,

Suddenly my previously working system, using Kinect (not Kinect2) has stopped working. Was wondering if you could help look into this. <roslaunch simtrack_nodes main_kinect.launch> still appears to be running without an issue, and topics are being opened, but nothing is being published, even with the test objects. Where can I look for debug info?

Cheers,

Jon

balloch commented 8 years ago

to make this more interesting, I tried to rerun catkin make, it bombs on

d_multiple_rigid_pose_sparse.cpp

simtrack-catkin_make-errors.txt

UPDATE: I have tried deleting the source, re-downloading with wstool update and rebuilding. still errors in the same place.

karlpauwels commented 8 years ago

Hi Jon,

I would first check if the kinect (and/or its driver) is still working. You can try running freenect, e.g. through camera_kinect.launch and see if the kinect topics are still published. They're easy to see using for example rqt.

For your compilation problem, try a fresh clone of the repository and recompile it. Maybe something has changed in your code?

Hope this helps.

balloch commented 8 years ago

Hey Karl,

I have tried a fresh clone to no avail. I am a little lost as to what this could be.

one thought: the vector calls in the file I listed in the text file output, what is the namespace/class they are suppose to belong to? I tried changing the offending terms to std::vector (and later, when getting similar errors on a few string calls without an explicit namespace I changed them to std::string), which allowed it to build successfully, but then the topics wouldn't quite initiate properly when I run roslaunch simtrack_nodes main.launch. That is to say, I could still get an output from the camera topics, but all of the simtrack topics showed nothing. More indicative of a problem, I tried to prepare a new model, for tracking using the rosrun interface cmd_line_generate_sift_model code, and this simply bombed with a "Segmentation fault (core dumped)" exit. no additional information.

Looking forward to your input. I will continue to update letting you know what is going on

karlpauwels commented 8 years ago

Thanks Jon,

There is definitely some namespace pollution going on. I have been able to replicate your problem but the error seems to be due to Ogre. It's a bit tricky to fix so I'm going to need some more time. Will get back to you on this asap.

Karl

balloch commented 8 years ago

Thanks Karl. Functionality for me is of the upmost importance, so do you have any recommendations as to what version still works, and with what version of ROS and Ubuntu? -Jon

balloch commented 8 years ago

Also, is there a way to debug this in Eclipse? Or did you build the whole project in a text editor?

karlpauwels commented 8 years ago

Sorry about the delay. The current version works with Ubuntu 14.04 and ROS Indigo. Both are LTS versions and thus recommended in general. As IDE I use QtCreator which is very convenient for debugging. You may want to stick to version 3.2.2 though since higher versions have issues with finding all headers for text completion. Also have a look here for hints: http://wiki.ros.org/IDEs#QtCreator

karlpauwels commented 8 years ago

I cleaned up the namespace references in the last commit: 2f7cc0777dd517cf62294dc7fe00c39ed5412c1b Are you still experiencing problems now?

balloch commented 8 years ago

Hey karl, thanks, namespace issues gone, but the main problem persists - getting a core dump on the rosrun interface cmd_line_get_sift_model for the samples. I'll start a new issue so you can close this one, thanks