[Paraview] Strange Multigroup crashing/behaviour

John Biddiscombe biddisco at cscs.ch
Wed Jan 18 18:03:26 EST 2006


When generating data from netCDF, I can tell the reader to either 
generate UnstructuredGrid datasets (if the variables have bounds array 
which define the cells), or to generate RectilinearGrid datasets (if the 
coordinate variables have one dimensional arrays only) (optionally, 
uniform grid if they are evenly spaced).
When I generate data on the server with a single process(or) suplplying 
the data, the output of the filter is a multigroup dataset with just one 
object inside. Everything works fine. Animating over timesteps produces 
the correct results.
When the datatype ins Unstructured, running on N processors also works 
and all is well,
But when outputting Rectilinear data, the server processes segfault on 
the second timestep - always -
There are two interesting odd things -
1) If I display timestep N initially, it displays ok, but then if I 
display N+1 or N-1 or any other, it crashes - but if I display N+1 or 
other initially it is ok, then crashes when I display N. It's always the 
second timestep (irrespective or order)
2) If I display timestep N, then chage some scaling parameter so that 
the entier reader is updated again and all geometry and point/cell data 
is regenerated, then it does not crash - so the second timestep is ok if 
it hasn't changed.

But it all works with unstructed - only rectilinear it doesn't like

I'm stumped. I welcome any idea, no matter how far fetched as to why 
this behaviour occurs. I've checked N times, the extents, the allocation 
everything I can check, but debugging the server is like chewing your 
own elbows.

thanks

JB




More information about the ParaView mailing list