My USRP is in a “lab” hooked to a pc, but I like to ssh -Y into it from
more pleasant parts of the house using my netbook. It worked great for
gnuradio and most other applications until I installed NVIDIA drivers on
the lab PC in order to enable cuda GPU processing. (Cuda made my
processing applications much faster and I hope to integrate it with USRP
output soon.) Most applications still work fine remotely, including
fairly graphic-intensive ones such as Google-earth and several QT-based
About that time I re-built gnuradio and now if I try to run any of the
graphic sinks remotely I get the error:
Xlib: extension “NV-GLX” missing on display “localhost:10.0”.
If I run the same application on another PC that uses the Nouveau
drivers instead of the Nvidia drivers I can ssh into that PC from my
netbook and everything runs fine. My guess is that the compile
processes used the best available graphics on the computer it was built
on even though the graphics requirements are fairly modest.
I tried VNC but it’s much too sluggish to use, especially for wiggly
scope plots. When it works, X over ssh is almost like being directly on
the machine in the messy room with the antenna drop and the noisy beige
box, except I’m comfy on the couch.
Is there some sort of build configure option that will not cause the
executable to use the Nvidia proprietary graphics calls, but more
portable ones instead?