Re: Hmi question

On Monday 24 May 2004 12:27, Andrea Consonni wrote:
> >> We are interested in your HMI support components and we want to use
> >> them in our application.
> >>
> >> We have tested them and we have noticed that in some cases, when the
> >> application is running, all stalls with the exception of the Report
> >> which works. When we "quit", the application destroys the
> >> NonRealTimeThread, but when it tries to terminate the ZeroTimeThread,
> >> the message "Finalizing thread after 1000 tries" appears and the
> >> application doesn't go forward. We have to terminate it pressing
> >> "CTRL+c".
> >
> > This looks like a deadlock in the realtime threads. Moreover, (as you
> > maybe know) there is a bug 'somewhere' (rtai/orocos) so that lxrt
> > semaphores are not destroyed (see /proc/rtai/RTAI Names) after the
> > application finishes. Is this the case in your app ? It could be that
> > it is rtai 3.0 specific...
>
> We think so because using 'dmesg' command we had error about lxrt
> semaphores. Now we are using RTAI 24.1.12.

Ok, I'll see how fast I can fix these warnings / errors. I'll put it on the 
list (and make a new release) if it is fixed.

>
> >> Have you ever had this problem? We have read about HMIConsoleOutput
> >> that "It should not be used in hard realtime programs", have you
> >> written this because of that problem?
> >
> > The HMI's run together with the execution extension of the control
> > kernel you are using. If the HMIOutput prints something, this is done
> >
> > right after the control loop. Therefore, the std::cout is not 'good'
> > in the Zero[Time|Latency]Thread under LXRT It will switch you back to
> > soft realtime (wich normally does not cause a crash). If you replace
> > the std::cout with rtos_printf() and include <os/fosi.h> it is again
> > allowed, But we know that RTAI crashes if to much rt_printk's are done
> > after each other :-(
>
> We think that we have resolved our problem deleting every "std::cout"s or
> changing them into "rt_std::cout"s in each our component and in your
> HMIConsoleOutput.hpp.

Ah, very nice to see that rt_std still works :-) It is much more limited than 
std::cout however... (only int, double,...)

>
> > I have developed the 'index' operator for the next version meaning you
> > can write now :
> >
> > var double6d d6 = double6d(0.0)
> > set d6[0] = 3.14
> > var double x = d6[0]
> >
> > cool, heh ? This opens the way to an 'array' type, which would
> > translate to a vector<double> in the code...
>
> Very good job! We think that this could be useful in several applications.

Thanks :-)  It will be in the next release

Peter

-- 
------------------------------------------------------------------------
Peter Soetens, Research Assistent                  http://www.orocos.org
Katholieke Universiteit Leuven
Division Production Engineering,                      tel. +32 16 322773
Machine Design and Automation                         fax. +32 16 322987
Celestijnenlaan 300B                   [email protected]
B-3001 Leuven Belgium                 http://www.mech.kuleuven.ac.be/pma
------------------------------------------------------------------------