Flattening a model seems to change the model

[nickerso]

If I have things right, flattening a model should not in any way change the model. If I have this wrong, stop reading here :)

It seems that using the bug.py in the attached zip file I'm seeing the original model being changed following the call to flatten it. Found this when I was accidentally validating the wrong model object and first thought this was a bug in the flattening.

bug.zip

This is using a windows build of the current main HEAD code (if I have things right..)

[agarny]

To flatten a model should clearly not change the model.

I have just tried your bug.py script using the main branch on macOS and I got the following output:

The method "parse_model" found 1 issues:
    - (Message) Given model is a CellML 1.1 model, the parser will try to represent this model in CellML 2.0.
The method "validate_model" found 0 issues:
The method "resolve_imports" found 1 issues:
    - (Message) Given model is a CellML 1.1 model, the parser will try to represent this model in CellML 2.0.
The method "validate_resolved_model" found 0 issues:
The method "validate_flattened_model" found 0 issues:
The method "validate_original_model" found 1 issues:
    - (Error) Imported component 'child_main' is not valid because:
  -> Component 'child_main' importing 'main' from 'child.cellml' has an error:
   - Variable 'TAveabsE' in component 'main' has a units reference 'perS' which is neither standard nor defined in the parent model.
swig/python detected a memory leak of type 'std::shared_ptr< libcellml::Parser > *', no destructor found.
swig/python detected a memory leak of type 'std::shared_ptr< libcellml::Model > *', no destructor found.
swig/python detected a memory leak of type 'std::shared_ptr< libcellml::Validator > *', no destructor found.
swig/python detected a memory leak of type 'std::shared_ptr< libcellml::Importer > *', no destructor found.
swig/python detected a memory leak of type 'std::shared_ptr< libcellml::Model > *', no destructor found.

Is that what you are getting? If so, then yeah, the original model gets changed and this is bad. :/

[hsorby]

Technically it is not the original model that is getting changed, it is the imported child model. The child's units are getting removed resulting in the validation error. But otherwise I agree, it is bad.