[Libnifalcon-devel] cube test

[Stephen S. <rad...@gm...>]

Hi,

Okay I finally got around to spending a day playing with libnifalcon.  (hurrah!)

I had some build issues first...

But finally I managed to write an additional "sub program" for
falcon_test_cli which presents a touchable cube surface.
This adds a nice basic test for the device motors and position reading
within a feedback loop.

First of all, the build problems.  I've attached several patches to
this email which resolved them for me, but I can't be sure these are
the "right" solutions but they worked for me.   Anyways, I'll just
list them:

1) I got errors including "gmtl/gtml.h".  I had to add /falcon to the
include path.

2) Got some segfaults during initialization with libftdi.  I found a
couple bugs in the device list iteration.

3) If I forgot to specify --libftdi the test_cli program crashed on
me, so I added an error condition for when no communication method is
specified. I guess this should be done with a proper default instead,
but it would depend on what was compiled it.  I'm sure it could be
done.

4) When I saw my bad timing with --loop_time_test (10!), I decided to
download libusb-1.0 and compiled it without problems.  I usually use
~/.local for installing locally compiled things, so I had to figure
out how to instruct cmake to find it.  But even when CMake found it, I
still got errors related to "libusb-1.0/libusb.h".  The solution I
found was to make sure that LIBUSB_1_INCLUDE_DIR is added to the list
of include_directories, and to change "libusb-1.0/libusb.h" to just
"libusb.h", since the LIBUSB_1_INCLUDE_DIR points to the libusb-1.0
dir.

The fifth patch just contains the cube subprogram, which works for me.

Hope this is useful.  I should mention this was all on Ubuntu 8.10.
My --loop_time_test reports 3 now, which still isn't ideal, but at
least I can present a stable surface with a stiffness of 5.

Steve