Re: [Xenomai] Failed testsuite/latency

[Philippe Gerum <rpm@xenomai.org>]

On 02/25/2015 09:33 AM, Bart Vissers wrote:
> Hi Philippe,
> 
> Before making any changes, I ran xeno-config --info:
> Xenomai version: Xenomai/cobalt v3.0-rc3 --
> Linux rtpc 3.16.7-xenomai-3.0-rc3 #1 SMP Mon Feb 23 12:30:46 CET 2015
> i686 GNU/Linux
> Kernel parameters: BOOT_IMAGE=/boot/vmlinuz-3.16.7-xenomai-3.0-rc3
> root=UUID=0925426e-e058-4093-963c-8b7de072e229 ro quiet
> I-pipe release #2 detected
> Cobalt core 3.0-rc3 detected
> Build args: --prefix=/usr --includedir=/usr/include/xenomai
> --mandir=/usr/share/man --with-testdir=/usr/lib/xenomai/testsuite
> --build i486-linux-gnu build_alias=i486-linux-gnu
> 
> namelist did not work, as latency does not contain any symbols. How
> can I include these when building debian packages?

I don't have much knowledge of the Debian pkg system, but the way I used
to do this was:

export DEB_BUILD_OPTIONS="debug nostrip"

before building the package, maybe adding "noopt" if you also want to
disable the optimizer, although this is not required for keeping the
namelist. Gilles certainly has more on this.

This said, at this stage you may want to build the user-space code
manually, leaving aside the debian pkg builder. It would be much easier
to control the build configuration process, at least for me:

http://xenomai.org/installing-xenomai-3-x/#Building_the_x86_libraries_3264bit

> Unfortunately, I got compilation errors after applying the patch to
> xenomai-3.0-rc3: error: ‘XENO_WARNING’ undeclared (first use in this
> function)

Yes, I'm always referring to the current development tip in this discussion.

> So I decided to repeat my steps with the git next branch, 52fb91e...
> So, same linux kernel, i-pipe patch and kernel config. Also included
> the printk patch. Output:
> 
> == Sampling period: 100 us
> == Test mode: periodic user-mode task
> == All results in microseconds
> warming up...
> latency: failed to set periodic, code 9
> ^C---|-----------|-----------|-----------|--------|------|-------------------------
> RTS|  10000.000|      0.000| -10000.000|       0|     0|    00:03:26/00:03:26
> 
> Notice the ^C. I killed it because it was running indefinitely. This

That is expected, the main thread just hangs paused in this case. If you
don't hit ^C, the real-time core should report two cobalt threads as
follows:

~ # cat /proc/xenomai/sched/threads
CPU  PID    CLASS  TYPE      PRI   TIMEOUT       STAT       NAME
  0  0      idle   core       -1   -             R          [ROOT]
  0  1206   rt     cobalt      0   -             X          latency
  0  1208   rt     cobalt      0   -             W          display-1206

Do you confirm this on your end? If so, then can you look a bit further,
listing the file descriptors currently open for the main thread, e.g. in
my case:

~ # ls -l /proc/1206/fd
total 0
lrwx------    1 root     root            64 Feb 25 09:21 0 -> /dev/pts/0
lrwx------    1 root     root            64 Feb 25 09:21 1 -> /dev/pts/0
lrwx------    1 root     root            64 Feb 25 09:21 2 -> /dev/pts/0
lrwx------    1 root     root            64 Feb 25 09:21 3 ->
anon_inode:[cobalt-timerfd]

I'm interested to know whether you can see the last one.

> was also the case in my previous attempts.
> 

So you received no message visible from dmesg?

-- 
Philippe.