Christoph's profile - activity

Basically it seems to be related to a misconfiguration of ROS. rosbridge requires the BSON module that is delivered with MongoDB (or compatible). If doing a "pip install bson" a non compatible version is installed. rosbridge (for now) fails silently so only one way direction is possible.

Potential fixes include pip install mongodb or using the rosdep system.

As I was using a slim robot system (Intel Edison based) with only very limited functionality, I fell into this trap. I resolved it manually by copying the bson directory from the python-mongodb sources to /opt/ros/<distro>/libs/python2.7/site-packages.

After that all issues seen to be resolved :-)

Btw. this topic is also discussed on GIT HUB, but I wanted to explicitely share my findings here, so people don't have to unnecessarily waste any time.

There is no error. The moment I link libroscpp with an executable, the executable will not start. :-(

Hi there,

I tried to install ROS on a x86 based embedded system (Linux installed using the meta-ros layer for yocto).

In genereal (with python libraries only) it works fine.

Unfortunately I observed a strange behavior with libroscpp.so . My initial issue has been that after starting roscore, everything was up and running besides rosout. I tried to restart it manually, but the shell just blocks after calling rosout and with top I could observe that the process already stopped.

I did some tests afterwards, and wrote a small program, that does not use ROS at all. If I link it with libroscpp I get the same strange behavior as with rosout. When I leave libroscpp out, no issue at all. I can even link librostime... It is only libroscpp that creates the troubles.

I already checked with ldd all libraries that are additionally linked to my executable, and in general they are all on the harddisk...

Any ideas what I could try as next steps? As it is already the linking that causes the trouble, I am running a bit out of ideas, what to try...

Best regards Christoph

Since beginning of this year there is a project that provides a build-layer for Yocto:

http://wiki.ros.org/hydro/Installatio...

Stumbled over it by pure luck, but initial installation worked ok, so I can recommend it as a good starting Point :-)

Best regards Christoph

Meanwhile I switched to Intel Edison (as Galileo turned out to be quite on the lowest possible edge of what is still doable for ROS). I have good experiences with meta-ros.

Good luck everyone! :-)

Hi community,

I'd like to create a Linux for my brand new x86 Intel Galileo board. The installed Linux is based on a Yocto-build, which is source-code available.

Has anyone got experience on how to package ROS with Yocto? I found only a year old thread on this topic, where it has been discussed for ARM devices. State back then was more or less, no further info available (yet)...

Best regards Christoph

There seems to be a pointcloud_to_laserscan stack. Haven't used it myself, but might be worth a try.

Best regards Christoph

The Node-architecture is imho not very suitable for a global variable. To keep track of a FIFO-load level you would typically go for a handshake protocol, like the proposal with the ack.

If you really want to have a shared memory my proposal would be to create an additional Node, let's call it "RAM", that offers a suitable set of services and topics to read/write to it. This way you have global storage for all your Nodes.

To have an even more convenient way to handle this, you could create a small custom launch file for usb_cam that sets this parameter.

I finally looked a bit more closely at the mjpegcanvasjs code. The reload issue could be resolved for me by adding a image.src = image.src line. This raised another issue, because the animation routine was not working as expected. For my patch I now solely rely on the setInterval() mechanism, skipping the browser specific calls.

I am not sure, if this "hack" will make it into main branch, but I uploaded it to github. So if you experience a similar issue, feel free to have a look ;-)

Regards Christoph

Hi community,

I tried to connect the camera of my self-built robot with mjpegcanvasjs to my Firefox webbrowser as described in the tutorial (unfortunately I am not allowed to post the link because of my bad karma).

Loading the initial picture works quite well. In the debugging web console I checked the javascript, and draw is also called repeatedly. Unfortunately the picture does not update.

My guess is that the browser is always redelivering the cached version of the image, instead of re-requesting an updated version. When I access the link to the image-source directly (by entering h__p://my.robot:8080/stream?topic=/usb_cam/image_raw into the address line) the image reloads automatically.

I tried a "cache-breaking" method by appending "#some_random_value" to the request in the JS. This method seems to work in the address line(everything after # is just stripped as it was explained in the cache breaking tutorial I read), but in the JS it just does not load, when adding the # value.

Anybody got experience with this and has some hints for me, how to resolve it?

Regards Christoph