Problems with Modelica.Math.Matrices.inv()

[Francesco Casella]

The following test model:

model TestInverse
  parameter Integer N = 3;
  Real A[N,N] = identity(N);
  Real B[N,N] = identity(N);
  Real C[N,N];
equation
  C = Modelica.Math.Matrices.inv(A) * B;
end TestInverse;

gives the following error:

"[TestInverse.mo:8:3-8:38:writable] Error: Cannot resolve type of 
expression Modelica.Math.Matrices.inv(A) * B. 
The operands have types Real[:, :], Real[3, 3] 
in component <NO COMPONENT>.

The problem persists even if all matrix sizes are replaced by integer literals, as in

model TestInverse2
  Real A[3,3] = identity(3);
  Real B[3,3] = identity(3);
  Real C[3,3];
equation
  C = Modelica.Math.Matrices.inv(A) * B;
end TestInverse2;

Last, but not least, the following simpler model is translated, but then the simulation executable triggers an unhandled exception and crashes

model TestInverse3
  Real A[3,3] = identity(3);
  Real C[3,3];
equation
  C = Modelica.Math.Matrices.inv(A);
end TestInverse3;

[Martin Sjölund]

Reassigning this to Per since I'll be busy until the 15th.

[Per Östlund]

The typing errors are now fixed in r15135. But the models still segfaults during simulation, in the LAPACK library. Might be some issues with our LAPACK interface. I will probably be busy the next two days though, so I added CC to Adrian in case he has time to look at it before me.

[Per Östlund]

The segmentation faults should hopefully be fixed in r15241, at least it works for me on 64-bit Linux now.