[Paraview] EGL build of ParaView

[Alexandre Ancel]

Indeed, when I use an ssh -Y for example with the EGL version of ParaView,
I end up having this error in the pvserver:
"
Waiting for client...
Connection URL: cs://irma-atlas:11111
Accepting connection(s): irma-atlas:11111
Client connected.
ERROR: In
/data/software/src/ParaView/ParaView-v5.0.0-source/VTK/Rendering/OpenGL2/vtkEGLRenderWindow.cxx,
line 315
vtkEGLRenderWindow (0x2efe490): Only EGL 1.4 and greater allows OpenGL as
client API. See eglBindAPI for more information.

ERROR: In
/data/software/src/ParaView/ParaView-v5.0.0-source/VTK/Rendering/OpenGL2/vtkOpenGLRenderWindow.cxx,
line 575
vtkEGLRenderWindow (0x2efe490): GLEW could not be initialized.
"
and ParaVIew crashes.
I noticed that at the beginning so now I make sure to have no DISPLAY
variable set.


Regarding to what is shown in:
Help, About, Connection Information, OpenGL Renderer

In the OpenGL Renderer section, I have the reference of the graphic card I
use, either the K80 or GTX 470. So it should be OK.


Additionaly, when I built ParaView, I didn't initially put the
-DPARAVIEW_BUILD_QT_GUI=OFF option, and I ended having errors with
XCreateWindow during the compilation, which seems normal as I used
-DVTK_USE_X=OFF.
Maybe the use of PARAVIEW_BUILD_QT_GUI=ON should be prevented when using
VTK_USE_X=OFF ? Or if it is already the case, I might have had a mixup in
my configuration.


In any case and if it might help someone else on this list, I wrote some
notes about how to make everything work together:
https://github.com/aancel/admin/wiki/Compile-ParaView-with-EGL-support-on-Ubuntu-14.04


Hope I could have been of some help,
Regards,

Alexandre Ancel

On Thu, Feb 11, 2016 at 5:33 PM, Dan Lipsa <dan.lipsa at kitware.com> wrote:

> Alexandre,
>
>
> On Thu, Feb 11, 2016 at 11:13 AM, Alexandre Ancel <
> alexandre.ancel at cemosis.fr> wrote:
>
>> Hello Dan,
>>
>> Regarding to the values of the variables, I have:
>>
>> //No help, variable specified on the command line.
>> EGL_INCLUDE_DIR:UNINITIALIZED=/data/software/install/EGL/include
>>
>> //Path to a library.
>> EGL_LIBRARY:FILEPATH=/usr/lib/x86_64-linux-gnu/libEGL.so
>>
>> //No help, variable specified on the command line.
>>
>> EGL_gldispatch_LIBRARY:UNINITIALIZED=/usr/lib/x86_64-linux-gnu/libGLdispatch.so.0
>>
>> //Path to a library.
>> EGL_opengl_LIBRARY:FILEPATH=/usr/lib/x86_64-linux-gnu/libOpenGL.so
>>
>> The libraries are OK (They come from the manual installation of the
>> Nvidia driver which should be at least 355 if I remember well). I have a
>> 358.16.
>>
>> I didn't previously saw that EGL_INCLUDE_DIR was marked as UNINITIALIZED,
>> however ParaView was built successfully with the previously mentioned
>> options.
>>
>> I am just not sure about the libGLdispatch.so.0 library, as it is not a
>> symlink to a libGLdispatch.so.358.16.
>> When I do a "strings libGLdispatch.so.0", I have gl*NV functions, so I
>> guess it should be ok.
>>
>
> Al this looks fine.
>
>
>>
>> As I said before, remote rendering is working but only when I add
>> "--disable-xdisplay-test" to the pvserver.
>>
>
> I see. It seems your build works fine but we should have authomatically
> pass that option and now show the option in ParaView at all.
> When you run through EGL there is no X involved. Opening an X display as
> this test does probably screws up things.
>
> You can also try (with this option passed as you have done) to see what is
> shown in:
> Help, About, Connection Information, OpenGL Renderer.
>
> It is interesting that I have not seen this problem. I tested on a machine
> that was already running X. Maybe this made a difference.
> I will file a bug report as it makes sense to disable the xdisplay test.
>
> Dan
>
>


-- 
Alexandre Ancel
Docteur, Ingénieur de recherche / Phd, Research Engineer
Cemosis <http://www.cemosis.fr> - alexandre.ancel at cemosis.fr
Tel: +33 (0)3 68 8*5 02 06*
IRMA - 7, rue René Descartes
67 000 Strasbourg, France
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20160211/da09cb01/attachment.html>