[ros-users] bug: SOEM stack: Errors in master component re-configuration

[Ruben Smits]

Hi Santiago,

On Thu, Feb 23, 2012 at 1:15 AM, Santiago Focke <focke.85 at gmail.com> wrote:

> Hi
>
> I've been trying to make some reconfigurations with the Orocos SOEM master
> component and I saw that after doing it the component stopped working
> correctly. This was caused because every time the component (re)configures
> itself, it creates again all the slaves found in the network, and adds the
> into the m_drivers vector. So, if I have one slave and configure 3 times my
> master, there will be 3 different slave drivers in the vector for the same
> slave. Also, when reconfiguring, it cannot add the service of the
> last-created driver, and the first-created one will be pointing to some
> region of memory that no longer corresponds to the slave image (in case
> SOEM remaps in a different place). I'm thinking on using a class variable
> (set by an operation of the master service) that will define if the master
> makes the whole configuration (searching for slaves, mapping, etc..) or
> just the slaves' parameter configuration, but I'd like to listen to better
> options.
>

This problem has been troubling me too, I haven't had the time to implement
a nice solution though. I think we should split up the current
configuration into to different steps:

* A configuration of the interface so we can talk to the EtherCAT device
* Slave specific configuration

On top of this I also would like to add some functions to the SoemDriver to
change the operational mode per-slave instead of doing it for the entire
stack at once during configuration/starting.



>
>
> Regards,
>
> --
> Santiago
>

Ruben


>
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20120223/4ca902d3/attachment-0004.html>