[ESPResSo-devel] Segmentation fault

[Markus Gusenbauer]

Hi all,

We defined new bonding forces for an elastic object (interacting 
particles in a triangular mesh representing the surface of the object, 
like constant volume force, constant area force). For this we have to 
calculate current volume and area of the object. This is done before the 
actual force calculation in Espresso because we need information about 
every particle in the object. We made exactly the same loop through the 
cells and particles as used in calculate_verlet_ia(). The method works 
fine with 7 objects (every
object has around 400 particles forming a triangle mesh). Suddenly with 
8 objects we get a segmentation fault. Looks like that a bond is broken 
(not on the same node). But this error is also when using just 1 node 
(Is this possible?). After analyzing we found out that this bond (that 
holds 1 particle with 2 neighbours forming a triangle) has a wrong 
neighbour id. So exactly the 2nd neighbour of the broken bond points not 
to a particle id, but to a different bond id. Could it be that we came 
into an out-of-array error and reading information from somewhere else? 
We have more than 25000 different bonds because every triangle in the 
mesh need to be defined and also other forces (like stretching between 
only 2 particles). On the tcl-level (in the tcl-start-script) the bonds 
are all correct when printing particle information. We are using version 
3.1.0!

We would really appreciate any ideas that could solve our problem!

Thanks,
Markus