[Paraview] Mesa/Offscreen Rendering

[Rick Angelini(CISD/CIV)]

Curiously (or maybe not so curiously) if I "xhost  +" on the client 
side, the errors on the server side go away!    8-)      So, this 
problem did not exist in 3.2.1, but was somehow introduced the 
development tree?      It seems to work quite well with the production 
release ... no X errors, etc.



Kent Eschenberg wrote:
> Rick,
>
> A few months ago Utkarsh worked with me to get pvbatch to run without 
> an X server. He didn't have the time to figure out what was creating 
> those warnings. I haven't updated from CVS for a while but this sounds 
> like the same thing, only from pvserver instead of pvbatch.
>
> We ran pvbatch with 256 nodes on an XT3 where there was not only no X 
> server but no way for it to connect back to a display. Many warnings 
> but it worked.
>
> Kent
> Pittsburgh Supercomputing Center
>
> Kent
>
> Rick Angelini(CISD/CIV) wrote:
>> Is there an issue with offscreen rendering and mesa in the CVS?   I'm 
>> building Paraview from a CVS download from Monday.     I'm 
>> configuring this to run on one of our clusters that does not have GL 
>> cards and I believe that I'm building against mesa.
>>
>> When I fire up pvserver and connect back to the client, and load up 
>> some geometry (max'd out sphere, for instance), I get a number of 
>> X-windows errors on the server side, like it's trying to connect back 
>> to my display.
>>
>> Process id: 3 >> ERROR: In 
>> /build/src/ParaView-08.25.08/VTK/Rendering/vtkXOpenGLRenderWindow.cxx, 
>> line 1478
>> vtkXOpenGLRenderWindow (0xb62160): bad X server connection. 
>> DISPLAY=hpc-cluster-l7:10.0
>>
>> The HPC node does not have GL cards and is not running an X server.   
>> Other than spewing X-errors, I am getting updated GL on the client - 
>> it's just not clear to me that it's actually doing distributed SW 
>> rendering on the server side. I have a build from the last production 
>> release (3.2.1) that behaves properly ... but I didn't do that 
>> build!   8-)