comedi and rtai 3.1

Roberto Bucher roberto.bucher at supsi.ch
Mon Oct 4 21:40:08 CEST 2004


I'm working for 1 year with 3.1 CVS-> 3.1 and comedi under kernel 2.4 and 2.6 
using the RTAI-Lab tool and the different blocks under Matlab/Simulink/RTW 
and under Scilab/Scicos/RTAICodeGen without problems.

First of all you need the comedi CVS with support Adeos!
As second step check which modules you loaded:

You need:
1) comedi drivers

2) rtai_hal, rtai_lxrt, rtai_fifos, rtai_sem, rtai_mbx, rtai_msg, 
(rtai_netrpc),

AND:

3) rtai_shm, rtai_comedi_lxrt, kcomedilib

Roberto


On Monday 04 October 2004 16:53, you wrote:
> Wolfgang Pichl wrote:
> > On Fri, 1 Oct 2004, Peter Dettori wrote:
> > Hi,
> >
> >>Hi,
> >>
> >>I am trying to build rtai-3.1-test5 with comedi support.  I have comedi
> >>comedi-0.7.68 and comedilib-0.7.21 installed.
> >
> > I have tried the same thing last week and did not get it to work. The
> > kernel-space RTAI support in Comedi works very well, but the user-space
> > LXRT interface seems to be buggy.
> >
> >>I then noticed in a september post
> >>(https://mail.rtai.org/pipermail/rtai/2004-September/008734.html)
> >>that it should work with comedi cvs.  Is that right?
> >
> > Unfortunately not. My personal experience:
> > If I use the latest stable comedi and comedilib, RTAI does not even
> > compile if the Comedi-LXRT is enabled.
> > Using the comedi cvs, it does compile, but the comedi-rtai module cannot
> > be loaded due to unresolved symbols.
> > Please note: in the compilation of RTAI, the header comedilib.h from
> > the *comedi* package, not that from *comedilib* must be used. This
> > information was somewhere in an earlier posting in this mailing list.
> > However, even if this is done correctly, it does not work.
>
> I have it working well enough for my purposes now.  I can submit a
> tests.c for the showroom if it will be added.  I have only tested the
> analog output though.
>
> I encountered 2 problems, the first is that comedi_close function in the
> comedi distribution does not check if the device is NULL.  So one gets
> an oops if the comedi_config is not done before running tests, as it is
> coded in 0.7.68.
>
> calling comedi_data_write seemed to cause a crash a couple of times when
> i was trying to gets things running.  I guess this was a result of
> something else I had done wrong since it seems stable now.  but my
> feeling is that there is not enough error checking in the kernel modules
> (presumably comedi since rtai just passes things on.) to safeguard
> against "beginners" mistakes.
>
> > One last question: if you only want analog input, do you really need
> > LXRT? Using the "command" facility of comedi and the kernel-based RTAI
> > support, you get a data stream in real-time to the comedi device (e.g.
> > /dev/comedi0), which your application can read continuously in
> > non-realtime. At least for me, this turned out to be completely
> > sufficient, so I did not bother any more about LXRT.
>
> I want analog output not input.
>
> ,
> pete
>
> > best regards
> > Wolfgang Pichl
> >
> >
> > ******************************************************************
> > a.o.Prof.Dr.Wolfgang Pichl
> > Institut f. Materialphysik         University of Vienna
> > Boltzmanngasse 5
> > A-1090 Wien     AUSTRIA
> >
> > Tel. (+43 1) 4277 51382
> > Fax  (+43 1) 4277 9513
> > email pichl at ap.univie.ac.at
> > *****************************************************************
>
> _______________________________________________
> RTAI mailing list
> RTAI at rtai.org
> https://mail.rtai.org/cgi-bin/mailman/listinfo/rtai

-- 
University of Applied Sciences
of Southern Switzerland
Dept CS & EE
http://www.dti.supsi.ch/~bucher





More information about the Rtai mailing list