Mandar's profile - activity

Sorry, I guess I had some left-over statements that I didn't snip out properly while copying here. But I removed the `close(sd)` and `return status` commands, made a new package & ran it. I get the same leaks. Can't/Shouldn't really change to ubuntu12.10 right now. Guess it's valgrind 3.7

Well I'm on the latest one for Ubuntu 12.04 - running valgrind-3.7.0

Wow, that was simple to fix. I wonder which method would be better to use - this answer or using classes as suggested by @Wolf.

I modified the example code you provided to follow the tutorial http://wiki.ros.org/roscpp_tutorials/Tutorials/UsingClassMethodsAsCallbacks more closely. This way, only the function write_enocean and the topic_name were in the class, while the subscriber was defined outside, rather than class method

I'm not sure I understand what you mean. Do I use it something like boost::function(bind, ...) or? I've not really used boost before, hence the confusion.

I'll try this out and report back in case of errors. Thanks!

Hey, Kind of newbie in using boost, and this is driving me crazy. I wanted to pass on the topicname to the callback function of subscriber, so I followed the various questions already asked and wrote the following: My callback function:

void write_enocean(const palgate_madynes_msg::EnOceanFrame& msg,std::string topic)

Subscriber:

sub[i] = nh.subscribe(Topiclist[i].name,1000,boost::bind(&write_enocean,_1,Topicname[i].name));

If I do this, I get the following error:

error: no matching function for call to ‘ros::NodeHandle::subscribe(std::string&, int, boost::_bi::bind_t<void, void (*)(const palgate_madynes_msg::EnOceanFrame_<std::allocator<void> >&, std::basic_string<char>), boost::_bi::list2<boost::arg<1>, boost::_bi::value<std::basic_string<char> > > >*)’

followed by the various candidates, and then:

node_handle.h:785:14: note:   candidate expects 1 argument, 3 provided

On the other hand, if write the subscriber as such:

 sub[i] = nh.subscribe<palgate_madynes_msg::EnOceanFrame>(Topiclist[i].name,1000,boost::bind(&write_enocean,_1,test));

I get the error:

/usr/include/boost/bind/bind.hpp:313:9: error: invalid initialization of reference of type ‘const palgate_madynes_msg::EnOceanFrame_<std::allocator<void> >&’ from expression of type ‘const boost::shared_ptr<const palgate_madynes_msg::EnOceanFrame_<std::allocator<void> > >’

What am I doing wrong with boost? My CMakeLists.text already includes the rosbuild_add_boost_directories() but gives me an error if write rosbuild_link_boost(topic_listener bind) saying it could not locate bind.

I managed to solve this problem (Note to self: Read the python docs). Turns out that a new process thread creates a copy of all the data from the parent, and I had not taken care of this properly.

The way I solved it was thus: I called the Popen (listen_stat in this case) from the device method get_status:

 def get_status(self):
    varnum = 0  # for testing only
    while not rospy.is_shutdown():
       self.response_line = self.listen_stat.stdout.readline().split()
       self.new_detect.value = int(self.response_line[varnum])

The important detail was declaring the variable new_detect as a multiprocessing.Value() and then calling this function instead for the multiprocess.

stat_procs = []
for device in device_props:
    sp = multiprocessing.Process(target=device.get_status)
    stat_procs.append(sp)
    sp.start()

The actual publisher then was just called in a normal loop:

try:
    while not rospy.is_shutdown():
        for device in device_props:
            device.device_talker()
except (KeyboardInterrupt,rospy.ROSInterruptException,IOError) as e:
    for p in stat_procs:
       p.join()

Hope this helps the others.

I realize that I may be using the multiprocessing module wrong to share variables (the publisher for instance). It would be helpful if someone could suggest a way to fix that.

I currently have a setup where I have about 10 shell scripts that listen to 10 different KNX addresses (using the shell command grouplisten) and publish their state to a ROS topic. These scripts can run in the background, so it's quite convenient to 'listen' to the state of these devices.

I tried to replace these 10 scripts with a single python program that basically does this:

1. Declare a subprocess.Popen for each device, to listen to its knx device.

   self.listen_stat = subprocess.Popen(self.cmds,stdout=subprocess.PIPE,preexec_fn=os.setsid)
   

2. Declare the node (in __main__) , and topic for each device ( in device.__init__()).

   class DeviceProp:
   def __init__(self):
     ...
     self.pub = rospy.Publisher(self.name,Int32,latch=True)
   

3. Call a talker function like so:

try:
  for device in device_props:
       p = multiprocessing.Process(target=device.device_talker)
       procs.append(p)
       p.start()
except (KeyboardInterrupt,rospy.ROSInterruptException,IOError) as e:
    for p in procs:
       p.join()

(The device_talker basically just publishes new data to device.pub)

What happens is that the logs suggest data is being published, but I don't see anything on when I subscribe to those topics or if I do rostopic echo. If I run rostopic pub from a new shell, I get the data on the topic as expected. I've also added a rospy.sleep(1) at multiple places (in loops too) so that even if I lose the first few messages, I _should_ get the later ones. But I don't.

Any suggestions on what I should do? (In case you are wondering why - I want to replace 10 scripts with 1 for easier code management. Also, Since I can store all address and device props in a yaml file, it will be trivial to add, or edit the configuration file).