dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

There seems to be more than one problem here...

1. You need to allocate the array as

    real* xx = new real[3*mesh.numVertices()];

2. Yes indeed, the size is never computed and it needs to be computed 
correctly or otherwise you'll most likely get a segmentation fault.

For this to work, a user-defined function must know the dimensions of 
the value type and it currently has no idea... I suggest we make the
two functions dim() and rank() in Function virtual so that it is 
possible to compute this. Any further thoughts?

/Anders


Jake Ostien wrote:
> Hi,
>
> I may have found a bug in the interpolation method for the 
> UserFunction.  In the private data there is a uint size that gets 
> initialized to 1 in the constructor, but, as near as I can tell, does 
> not get set anywhere else.  When I try and do something like
>    
>    
>     <f is a member of a BodyForce class that returns a 3D force vector 
> and is derived from Function, thus its a UserFunction>
>    
>     real* xx = new real[mesh.numVertices];
>     f.interpolate(xx);
>
>
> I get a memory error from glibc.  To make a long (painful) story 
> short(er), I realized that in the interpolation function, local_values 
> was of size 1 and needed to be of size 3.  So I hard coded in a 3 in 
> place of size, and things magically got better.  Unfortunately, it is 
> not clear to me how to fix this in the general case.  If anyone has a 
> suggestion, I'd be happy to try it out.
>
> -Jake
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev