curvilinear vectors

[Elizabeth Dobbins <dobbins@xxxxxxxxxxxxx>]

Hi;

I'm having a really weird problem with the curvilinear version of the
vector command.  I'm plotting velocities from our model output files, and
often, ferret crashes with a core dump.  The problem doesn't seem
consistent to me, so I'm having a hard time working around it.

It doesn't seem related to the number of vectors plotted.
For instance, this works OK:
set reg/i=1:100/j=1:100/l=1
vector/xskip=1/yskip=1/len=.1 ubar,vbar,lon,lat
But this (smaller!) region core dumps:
set reg/i=1:100/j=1:10/l=1
vector/xskip=1/yskip=1/len=.1 ubar,vbar,lon,lat

There also might be a positional element to what causes a dump.  This
works:
set reg/i=1:31/j=1:31/l=1
vector/xskip=1/yskip=1/len=.1 ubar,vbar,lon,lat
But this doesn't:
set reg/i=201:231/j=1:31/l=1
vector/xskip=1/yskip=1/len=.1 ubar,vbar,lon,lat

It doesn't help to avoid the set regions with
vector/xskip=1/yskip=1/len=.1 \
  ubar[i=201:231,j=1:31,l=1],vbar[i=201:231,j=1:31,l=1],lon[i=201:231,j=1:31],lat[i=201:231,j=1:31]

Usually, there is no helpful error message - just the dump.  Very rarely,
I get this message:
XgksDuplicatePrimi()  300 Storage overflow has occurred in GKS
XgksDuplicatePrimi()  300 Storage overflow has occurred in GKS
Segmentation fault (core dumped)

The datasets are large - 302 x 514 x 30 x 120.  I'm running FERRET v5.40
on a Compaq Tru64 (OSF v5.1) machine.  I'm told that someone who uses the
same machine via an Exceed session on a PC never gets core dumps, but I
can't verify this.  I get the same result in batch or realtime mode.

Anyone else ever seen this?  Any work arounds to suggest?  What else can I
do to diagnose it?

Thanks,
Liz

*********************
Elizabeth L. Dobbins
Research Scientist
JISAO/PMEL
phone: (206) 526-4581

"Various forms of jiggery-pokery were used to suit 
horses to courses."  -- Michael Burleigh