Got COMPILER BUG when compiling Modelica_Requirements (minimal example attached)

[anatoly.trosinenko@…]

When compiling examples from the Modelica_Requirements library, I got the COMPILER BUG message. The example was then reduced to quite strangely looking model with setting pre() of a variable of enum type inside the initial equation. Not sure whether this is a correct Modelica code, but compiler asked to send a bug report. :)

How to reproduce

Take the following TestEnum.mo:

model TestEnum
  type Property = enumeration(Undecided);
  Property localProperty;
initial equation
  pre(localProperty) = Property.Undecided;
end TestEnum;

And try to process it:

$ omc --version
OpenModelica 1.13.0~dev-1409-gd2ab68c
$ omc TestEnum.mo 
Error processing file: TestEnum.mo
[/home/trosinenko/TestEnum.mo:5:3-5:42:writable] Error: Got type mismatch error, but matching types enumeration(Undecided).
This is a ***COMPILER BUG***, please report it to https://trac.openmodelica.org/OpenModelica.
Error: Error occurred while flattening model TestEnum

# Error encountered! Exiting...
# Please check the error message and the flags.

Execution failed!

What is expected

The model is either compiled or rejected with an explanation, without COMPILER BUG message.

[Per Östlund]

Fixed in baede2f.

The actual issue was that the old frontend had a special case for component_reference = enum_expression in the part of the instantiation that handles equations, but no other cases involving enums. So e.g.:

model M
  type E = enumeration(a, b, c);
  E e;
equation
  e = E.a;   // This would work.
  E.a = a;   // This wouldn't work.
  E.a = E.a; // This wouldn't work either.
end M;

I don't know why it was implemented like that, but I added the missing cases and it seems to work just fine.

[anatoly.trosinenko@…]

I have tested it with some Modelica_Requirements examples -- most of them can be compiled and simulated now, some don't, but with meaningful error messages -- maybe errors in library. Thank you for blazingly fast fix!