[torqueusers] Submitting a graphical job to Torque in batch mode

[David Golden]

> Hi David,
> 
> Xvfb solution worked perfectly. Thanks. I have a question, though.
> According to the users, this same scilab thing was working fine before. 
> Then I performed a complete OS upgrade from SuSE Pro 9.0 to 9.3 in 
> October, 2005. I modified /etc/inittab to set the default init level to 
> 3. Would any of these affect how scilab works in batch mode. It did not 
> require a DISPLAY environment before, but started complaining about it 
> recently.

Not really a SuSE or a scilab user, I'm afraid.  Changing the 
init level to 3 probably on worker nodes probably turned off a 
local X server, of course,  but it seems pretty unlikely it 
would have been usable by batch processes unless the X server 
had also been configured to allow it (out of box on modern linux 
distros, the [kxg]dm-managed X11 server typically wouldn't allow 
unauthorised connections from clients, to the best of my knowledge), 
and even then the batch job would have to happen to have a 
correct DISPLAY env var, assuming scilab needs a valid display.

Was scilab itself upgraded? (I have no idea if it's "in SuSE" -
all I know is that it's non-free): could just be a bug in scilab, 
failingto do whatever it's supposed to do when given -nwni, assuming
what it's supposed to do when given -nwni includes not even
trying to use any X display.