[burgreen]

I am trying to time-integrate only a single source term defined by:

fvScalarMatrix eqn( fvm::ddt(x) == -fvm::Sp(coeff, x) );
eqn.relax(); 
eqn.solve();

It is failing with this error:

[4] --> FOAM FATAL ERROR:
[4] lowerPtr_ or upperPtr_ unallocated
[4]
[4]     From function const scalarField& Foam::lduMatrix::lower() const
[4]     in file matrices/lduMatrix/lduMatrix/lduMatrix.C at line 271.
[4]
FOAM parallel run aborting

The solver is defined as:

"(x)"
    {
        solver          PCG;
        preconditioner  diagonal;
        tolerance       1e-8;
        relTol          0.001;
        maxIter         400;
    }

If I replace the fvm source term with the fvc operator:

fvScalarMatrix eqn( fvm::ddt(x) == -fvc::Sp(coeff, x) );

I get same error.
If I comment out the eqn.relax() and just solve, I get a different error:

Foam::error:printStack(Foam::Ostream&) at ??:?
[34] #1  Foam::sigFpe::sigHandler(int) at ??:?

Is this type of equation supported?

[Geb1313]

Hi burgreen,

Did you solve it? I got a similar problem.

[ruloz]

Originally Posted by Geb1313

Hi burgreen,

Did you solve it? I got a similar problem.

Hi everyone, this has been discused on this post


lowerPtr_ or upperPtr_ unallocated


and hjasak says it is related with a bug in the matrix classes.


I'm working on a custom solver in OpenFOAM 10 and it works with an explicit source,

fvScalarMatrix Eqn(fvm::ddt(phi) == -source);
Eqn.relax();
Eqn.solve();

but when I addapted the solver for OpenFOAM 2212 it didn't worked, I've manage to run it as

solve (fvm::ddt(phi) == -source);

So it seems that in the Foundation version they solved the bug, and maybe a bug report for the OpenCFD Ltd version will help.