From: Mathieu Malaterre on
On Tue, Dec 1, 2009 at 6:30 PM, Javier Barroso <javibarroso(a)gmail.com> wrote:
> On Tue, Dec 1, 2009 at 6:03 PM, Mathieu Malaterre
> <mathieu.malaterre(a)gmail.com> wrote:
>> On Tue, Dec 1, 2009 at 5:49 PM, Javier Barroso <javibarroso(a)gmail.com> wrote:
>>> Hi,
>>>
>>> On Tue, Dec 1, 2009 at 5:35 PM, Mathieu Malaterre
>>> <mathieu.malaterre(a)gmail.com> wrote:
>>>> Hi there,
>>>>
>>>>  I am getting confused. I cannot get xclock to run on a remote host
>>>> from my ssh session:
>>>>
>>>> local $ ssh -X mpi0
>>>> remote $ xclock
>>>> Error: Can't open display: :0.0
>>>>
>>>> But I can still do:
>>>>
>>>> local $ ssh -Xf mpi0 xclock
>>>>
>>>> Could someone please remind me what is the difference, and what am I
>>>> missing in the first scenario ?
>>> Do you set DISPLAY variable in your login ?
>>>
>>> If you do you have to unset it (don't assign any value in any session
>>> config file).
>>
>> Ah ! That was the trick.
>> I can now move on to the next issue, how to get mpirun to do ssh -X
>> ... (since ForwardX11 yes is not honored...)
> Maybe the response is in mpi faq (search xterm):
> http://www.open-mpi.de/faq/?category=debugging

Excellent FAQ indeed !

> I think you will have to disable -nolisten option in X see this bug
> for reference:
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=539186

Hum, I did not understand what was going on in this bug report (very
cryptic). Instead I used:

$ mpirun -np 4 --hostfile ~/mpicluster -x DISPLAY=localhost:10 -x
LIBGL_DEBUG=verbose /home/mpiuser/Kitware/VTK-gcc/bin/ParallelBFS

Note that for some reason LIBGL_DEBUG=verbose avoids a crash I was getting:


ERROR: In /home/mpiuser/Kitware/VTK/Rendering/vtkXOpenGLRenderWindow.cxx,
line 344
vtkXOpenGLRenderWindow (0x186b3b0): bad X server connection.
DISPLAY=localhost:10


ERROR: In /home/mpiuser/Kitware/VTK/Rendering/vtkXOpenGLRenderWindow.cxx,
line 344
vtkXOpenGLRenderWindow (0x186b3b0): bad X server connection.
DISPLAY=localhost:10


ERROR: In /home/mpiuser/Kitware/VTK/Rendering/vtkXOpenGLRenderWindow.cxx,
line 344
vtkXOpenGLRenderWindow (0x186b3b0): bad X server connection.
DISPLAY=localhost:10


[dhcp-67-180:28173] *** Process received signal ***
[dhcp-67-180:28173] Signal: Segmentation fault (11)
[dhcp-67-180:28173] Signal code: Address not mapped (1)
[dhcp-67-180:28173] Failing at address: 0xe0
[dhcp-67-180:28173] [ 0] /lib/libpthread.so.0 [0x7ffe19270a80]
[dhcp-67-180:28173] [ 1] /usr/lib/libXt.so.6(_XtAppCreateShell+0x48)
[0x7ffe1cc59198]
[dhcp-67-180:28173] [ 2] /usr/lib/libXt.so.6(XtVaAppCreateShell+0x156)
[0x7ffe1cc87466]
[dhcp-67-180:28173] [ 3]
/home/mpiuser/Kitware/VTK-gcc/bin/libvtkRendering.so.5.5(_ZN26vtkXRenderWindowInteractor10InitializeEv+0x597)
[0x7ffe1e7cd837]
[dhcp-67-180:28173] [ 4]
/home/mpiuser/Kitware/VTK-gcc/bin/ParallelBFS(main+0x320) [0x409760]
[dhcp-67-180:28173] [ 5] /lib/libc.so.6(__libc_start_main+0xe6) [0x7ffe187dc1a6]
[dhcp-67-180:28173] [ 6]
/home/mpiuser/Kitware/VTK-gcc/bin/ParallelBFS(__gxx_personality_v0+0x369)
[0x409349]
[dhcp-67-180:28173] *** End of error message ***
--------------------------------------------------------------------------
mpirun noticed that process rank 0 with PID 28173 on node mpi0 exited
on signal 11 (Segmentation fault).
--------------------------------------------------------------------------


Since I am using nvidia driver, I do not think I can report that.


--
Mathieu


--
To UNSUBSCRIBE, email to debian-user-REQUEST(a)lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster(a)lists.debian.org
From: Bob McGowan on
Mathieu Malaterre wrote:
> On Tue, Dec 1, 2009 at 6:30 PM, Javier Barroso <javibarroso(a)gmail.com> wrote:
>> On Tue, Dec 1, 2009 at 6:03 PM, Mathieu Malaterre

--deleted text--

> Excellent FAQ indeed !
>
>> I think you will have to disable -nolisten option in X see this bug
>> for reference:
>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=539186
>
> Hum, I did not understand what was going on in this bug report (very

I looked at it too, and it seemed to me to be stating the issue was a
user error.

Message 40 indicates the user included "server arguments to startx (-dpi
100)", which apparently prevents reading of the X server startup files.

In Message 45, the OP states "I didn't realise that server arguments
would bypass xserverrc."

So it would seem to be not relevant to your problem.

> cryptic). Instead I used:
>
> $ mpirun -np 4 --hostfile ~/mpicluster -x DISPLAY=localhost:10 -x
> LIBGL_DEBUG=verbose /home/mpiuser/Kitware/VTK-gcc/bin/ParallelBFS
>
> Note that for some reason LIBGL_DEBUG=verbose avoids a crash I was getting:
>
>

--deleted error text--

In my experience, this usually means a bad pointer, which in one case
accesses memory in a location for which it has no permission or some
such, and in the other case is not "important".

In any case, there's a problem you should report to someone, I think.

Perhaps Nvidia?

>
> Since I am using nvidia driver, I do not think I can report that.
>
>

--
Bob McGowan
Symantec
US Internationalization


--
To UNSUBSCRIBE, email to debian-user-REQUEST(a)lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster(a)lists.debian.org