The VGL Transport is more or less a legacy feature and isn't used much anymore except to supplement existing remote X environments (Linux-to-Linux, primarily) on fast networks, so I would not recommend its use.
It sounds like you need to use the -xstartup argument to vncserver rather than TVNC_WM. That lets you completely bypass ~/.vnc/xstartup.turbovnc and load your own custom VNC startup script that does whatever you want. On 8/3/18 4:21 PM, qwofford wrote: > I was digging through the vncserver script and found a handy variable, > TVNC_3DWM, which I am using to simplify the user experience for my > TurboVNC/VirtualGL setup. It's nice that users don't need to type > vglrun before every command. This capability led me down a rabbit hole... > > I use the TVNC_WM variable to control how interactive the remote > desktop experience is. I typically set this to mate-session or xterm, > depending on how interactive we need to be. > > I have been abusing TVNC_WM a bit for testing. For instance, > TVNC_WM=glxgears is a quick way to test VGL. > > Some people I support find it cumbersome to run an entire desktop > environment when a single application is needed. A subset of those > users don't even want to use an xterm window. I can use/abuse TVNC_WM > to simplify their workflow, as long as the application they want to > run is a binary, and as long as the application doesn't require any > module loading (two very big if's in my HPC environment). > > I made a wrapper script for calling vncserver, and part of this script > takes an arbitrarily long list of arguments and sets them as the > TVNC_WM environment variable. In practice, the script might be used > like this: > > | > vnc_go xterm > | > > or > | > vnc_go glxspheres64 > | > > The user can connect their local client to localhost:1 and view the > terminal or opengl demo with no problem. If I write a traditional bash > script, this paradigm fails: > > | > -bash-4.2$ cat ~/my_startup.sh > #!/bin/bash > moduleload anaconda/Anaconda3; > jupyter-notebook > | > > | > vnc_go ~/my_script.sh # doesn't work > | > > It's pretty clear why this would fail based on the vncserver script > (specifically, line 159). If I mangle the traditional bash scripting > format, this script works like a dream... > > | > -bash-4.2$ cat ~/my_startup.silly_sh > module load anaconda/Anaconda3;jupyter-notebook > | > > > | > vnc_go ~/my_script.silly_sh # does work > | > > Of course, I don't want to train people to script this way. I do > wonder if you might have thought about this use case before, or if you > might otherwise have a suggestion on how to achieve this goal without > invoking TVNC_WM. > > PS- I realize VGLTransport is functionally similar to what I'm trying > to achieve, but I'm still in the process of /motivating/ the use of > TurboVNC/VirtualGL over other closed source products. I want to be > sure that performance is maximized for these demos; so not looking for > the VGLTransport solution at this time. I want to do this with TurboVNC. > -- You received this message because you are subscribed to the Google Groups "TurboVNC User Discussion/Support" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/turbovnc-users/f2c015ea-163d-5512-8b08-aad9ad280d28%40virtualgl.org. For more options, visit https://groups.google.com/d/optout.
