PreOptModules messes up model

[Martin Sjölund]

The following model passes the frontend, but some preOptModule messes up the structure very badly:

model M
  record R
    String name "Unique name used as identifier";
    Real c "Coefficient";
  end R;
  function name
    input R r;
    output String n=r.name;
  end name;
  R[3] dictionary = {R(name="A", c=1), R(name="B", c=2), R(name="C", c=3)} "Data dictionary";
  Boolean[:] b = {name(r) == "B" for r in dictionary};
  Integer i = Modelica.Math.BooleanVectors.firstTrueIndex(b);
  Real c = dictionary[i].c "Coefficient";
  annotation(uses(Modelica(version="3.2.1")));
end M;

Error: Model is structurally singular, error found sorting equations 
 3: 3.0 = 3.0;
4: "A" = false;
5: "B" = false;
6: "C" = false;
 for variables 
 dictionary[1].c(8), b[3](3), b[2](4), b[1](5), dictionary[3].c(6)

[Lennart Ochel]

This is an issue of the module removeSimpleEquations. I will have a look.

[Lennart Ochel]

It seems that this model covers also another issue. instantiateModel generates following output:

class M
  String dictionary[1].name = \"A\" \"Unique name used as identifier\";
  Real dictionary[1].c = 1.0 \"Coefficient\";
  String dictionary[2].name = \"B\" \"Unique name used as identifier\";
  Real dictionary[2].c = 2.0 \"Coefficient\";
  String dictionary[3].name = \"C\" \"Unique name used as identifier\";
  Real dictionary[3].c = 3.0 \"Coefficient\";
  Boolean b[1];
  Boolean b[2];
  Boolean b[3];
  Integer i = Modelica.Math.BooleanVectors.firstTrueIndex({b[1], b[2], b[3]});
  Real c = dictionary[i].c \"Coefficient\";
equation
  b = {false, false, false};
end M;

However, b should be {false, true, false} .

[Per Östlund]

Using the old frontend the model now passes the backend, so I assume that issue was fixed. It still evaluate b incorrectly, but that's fixed in the new frontend.