ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Markus Achtelik's profile - activity

2015-07-07 06:51:44 -0500 commented answer "fatal error: 'QMetaType' file not found" when building ROS Indigo on OS X 10.10.3

it seems to be solved now: https://github.com/Homebrew/homebrew-... haven't tested it though, since I don't want to touch my working version now :)

2015-04-15 06:24:27 -0500 received badge  Good Question (source)
2015-02-18 14:11:08 -0500 received badge  Taxonomist
2014-08-28 02:56:53 -0500 commented answer Two versions of libboost?

sorry for missing the "more" button ;) . This is quite weird, I was sure we had it running on 64-bit 12.04 machines. There is one last thing you could try to do: remove the boost libs from the sdk directory, and try to link your system's boost_system the catkin way.

2014-08-26 11:16:18 -0500 commented answer Two versions of libboost?

Do you mean edit 3? It looks like it doesn't even connect to the vicon computer. the segfault usually happened after it connected successfully. Can you ping the vicon computer with "ping vicon" ? you can change the hostname/address in vicon.launch

2014-08-25 04:57:22 -0500 answered a question Two versions of libboost?

I'm not sure if this is fixable the easy way. To me, it looks like the problem is boost thread, which appears to be not header only. I'm not too much into linking, so here is my naive interpretation of the problem: The boost thread header compiled in the vicon-lib is compatible to the thread symbol for 1.5x , while ros asks for 1.4x . Since both symbols have the same name, there is no chance for the linker to decide which lib is the correct one to link... Back then, I played quite a bit with linking, the result was always the same: either the vicon sdk segfaulted or our ros interface, depending on which boost lib the linker decided to link.

The only (but ugly) way out I see for now, is to install a matching boost version and compile ros from source. We did this on a few of our machines successfully, since we ran into other problems with the old boost version.

Even though the linking problem still exists, it didn't crash on our 64 bit machines so far.

2014-06-17 17:07:03 -0500 received badge  Nice Question (source)
2014-01-28 17:30:00 -0500 marked best answer RQT not showing icons in OSX

Hi,

I just upgraded to the latest releases of rqt as announced on the mailing list earlier today, hoping this would help. rqt does not show icons for the plugins, e.g. the "refresh" or "help" icon under osx 10.8. It looks worst for rqt_bag, which is somewhat unusable in this state, as shown in the screenshot. Any ideas what that could be?screenshot

Thanks a lot!

2014-01-28 17:28:57 -0500 marked best answer clearing all markers / view in RVIZ remotely

Hi,

I'm drawing a lot of markers in rviz to visualize the outcome of an algorithm. As I'm often interrupting/killing it, it's infeasible to keep track of what has been drawn in order to delete it later. Is there any possibility to send something like "wipe" or "clear" to rviz to clear all markers? Currently, I'm sending a message with many markers with action=DELETE but that seems a bit brute force. Is there any better option?

Thanks a lot!

2014-01-28 17:28:16 -0500 marked best answer Undefined reference for string constant in service

Hi,

when I define a simple service, e.g.

string my_string
string CONST_STRING=some_value
---

I get an undefined reference to ...::CONST_STRING during linking. Looking at the generated header file, I noticed that CONST_STRING gets declared, but never defined. I tested the same with messages, here it works fine and I can see the definition of CONST_STRING.

The same applies for floatXX-- I guess that problem exists for every constant that can not be encoded into an enum.

Am I doing something wrong,or is this a bug in gensrv? I'm working with Fuerte

Thanks a lot for your help!

Markus

2014-01-28 17:22:42 -0500 marked best answer Best practices for logging messages and parameters

Hi all,

I was wondering about best practices to log datasets with rosbag and save the parameters that were used during logging. As long as parameters do not change, I could just run rosparam dump paramfile and then run rosbag record my_topics. However, when e.g. using dynamic reconfigure during logging, parameters will change. One could now run rosparam dump paramfile every time dynamic reconfigure was used, but that's rather tedious and error prone. Another way would be to create a parameter message and publish the current parameters at a constant rate - less error prone but also not really nice... any better ideas?

Would it be possible to configure the parameter server such that it dumps the parameters on every change, or can I find something like this already somewhere in the logs?

Best, Markus

2014-01-28 17:22:29 -0500 marked best answer strange namespace behavior when setting the ROS_NAMESPACE environment variable

Hi all,

I have a node publishing a topic let's say "my_topic". Simply running this node, rostopic list tells me as expected:

/my_topic
/rosout
/rosout_agg

Now, when I want to push this node to a namespace by: export ROS_NAMESPACE=my_ns

rostopic list says:

/my_ns/my_ns/my_ns/my_topic
/rosout
/rosout_agg

Any idea where three times "my_ns" comes from? When I run the node from a launch file and set the "ns" attribute, everything is alright. I'm not doing anything fancy with namespaces etc. in the node, I put some example code below. It seems like this came with ros 1.4.8. It used to work with 1.4.6 which I had installed before.

Best, Markus

#include <ros/ros.h>
#include <std_msgs/String.h>

int main(int argc, char** argv)
{

  ros::init(argc, argv, "ns_test");
  ros::NodeHandle nh;
  ros::Publisher pub = nh.advertise<std_msgs::String> ("my_topic", 1);

  ros::Rate r(5);
  while (ros::ok())
  {
    std_msgs::String msg;
    msg.data = "my string message";
    pub.publish(msg);
    r.sleep();
  }
  return 0;
}
2013-11-27 23:59:06 -0500 received badge  Famous Question (source)
2013-10-24 23:17:47 -0500 received badge  Notable Question (source)
2013-10-13 04:58:50 -0500 received badge  Famous Question (source)
2013-09-17 21:54:12 -0500 received badge  Nice Question (source)
2013-09-17 21:51:08 -0500 received badge  Notable Question (source)
2013-09-17 21:49:31 -0500 commented answer rosbuild migration - mixing workspaces?

Thanks a lot for the answers and suggestions! I wasn't fully aware of the different folders to check-out to with one rosinstall file. Since indeed multiple people will have to use this setup, that (and probably vcstool) seems to be the best solution for now.

2013-09-17 11:33:41 -0500 received badge  Popular Question (source)
2013-09-16 23:20:11 -0500 asked a question rosbuild migration - mixing workspaces?

Hi,

we're in the process of migration to catkin, but have a lot of packages that we won't migrate all at once. Keeping track of overlays, and which package is migrated in two rosinstall files, one for the rosbuild workspace, and one for the catkin workspace seems a bit tedious... Can we just put all packages (rosbuild + catkin) into the src folder of the catkin workspace, add src to ROS_PACKAGE_PATH and create CATKIN_IGNORE / ROS_NOBUILD files in the packages according to their migration status? Or will that just create more problems?

Thanks! Markus

2013-09-11 06:29:29 -0500 received badge  Popular Question (source)
2013-09-11 06:28:54 -0500 commented question c++11 with ROS on OSX

I thought it would be worth discussing it here as well. c++11 works fine for us under linux since gcc 4.6 if I remember it correct. I just checked, it links to libstdc++. However, it seems quite old on osx. I'll let you know when I find something out,but I'm almost giving up on this for near future.

2013-09-10 01:29:04 -0500 asked a question c++11 with ROS on OSX

Hi all,

we have some packages developed with c++11 features, which was no problem on Ubuntu. We'd like to use it on osx now as well, which seems to be harder than we thought. From what I understood so far, clang (shipped with the latest version of Xcode) does support c++11, when using the -stdlib=libc++ flag. However, this causes issues (especially std::string) when linking against libstdc++, which is the case for all homebrew-built libraries and for ROS by default as well. Linking is fine, but runtime crashes occur whenever std objects are passed between the libraries.

As libstdc++ seems not to be further developed, the cleanest solution would probably be to get homebrew to use libc++. There is a discussion about it here: https://github.com/mxcl/homebrew/issues/9314 . However, it seems like this is not going to move forward soon.

We could now start fixing the homebrew dependencies to use libc++, but that would most likely cause issues with some other libs built against libstdc++ ...

Any idea or thought to work around this? Or am I missing anything? I'd dislike porting back all our c++11 code.

Thanks a lot, Markus

2013-07-30 10:59:06 -0500 received badge  Famous Question (source)
2013-07-26 08:41:51 -0500 received badge  Famous Question (source)
2013-07-14 23:04:19 -0500 commented question Asctec_Proc - Timeout: 0 bytes available 3 bytes requested

is the Atomboard connected properly to the low level processor of the pelican? This might help: http://asctec-users.986163.n3.nabble....