How do I start a blank x-server?

[Micha]

On 28/02/12 09:51, Nadav Har'El wrote:
> On Mon, Feb 27, 2012, Micha wrote about "How do I start a blank x-server?":
>>     <p>For a project I'm working on at the moment, I need to be able to
>>       log in remotely to a machine (via ssh) and start a blank x-server.
>> ...
>>       I seem to recall that just running X as a user used to do it, up
>>       to the no cursor part, leaving an empty (hetched) screen and
>>       running the content of .xsession or something like that.<br>
>>       Things on modern systems seems to have changed enough with all the
>>       xsession / gdm / gnome etc. that it doesn't seem to happen
>>       properly.</p>
>>     <p>Any idea how I can achieve that on a modern system (red hat
>>       enterprise desktop 6 in this case).</p>
> 
> 
> I'm not sure about RHEL6, but Fedora, which is probably similar,
> unfortunately lost the ability to run the user's ~/.xsession instead of
> those fancy (and bloated) GNOME/KDE. It happened when Fedora 9 was
> released, in 2008.
> 
> Luckily, there's a way to bring the good-old-behavior of following a
> ~/.xsession (if it exists). You can enable it by doing:
> 
> 	yum install xorg-x11-xinit-session
> 
> Then, in the graphical login screen, you get a third option besides KDE
> and Gnome, which is "user config" (or something like that) - and if you
> choose that, your ~/.xsession is used.
> If you only switch to this "user config" once in the login screen, it
> will be the default next time, so from there on you will just log in
> normally, and always get your intended configuration.
> 
> But although this answers your question, looking at the rest of your
> discussion, I'm not sure this is what you actually wanted to do. Perhaps
> what you wanted is simply to run "vncserver" and get a blank X server which
> you can view with "vncviewer".

Ok, I'll try to explain again what I need.

I have a machine with one card connected to two outputs. These are
processed by hardware to create some output which are not human
distinguishable, but as far as the computer is concerned they are
regular displays. I connect to that machine via ssh, and I need to open
an OpenGL window on these two outputs to "display" the required data. To
open these windows I need to create an X server (as I need OpenGL with
acceleration, so a frame buffer device will not do). For testing I'm
connecting screen to see that the right output is produced, but they
won't be there after debugging.

Running X as root takes me most of the way there, but it works for now
on my personal laptop and I haven't managed to get it to run on the said
machine (there are complaints about no access to /dev/nvidia1)

>