Segmentation Fault when analysing model with an error.

Question

Segmentation Fault when analysing model with an error.

Closed this issue 4 days ago · 13 comments

I have a model with an error. When I try to analyse this model it throws a segmentation fault instead of telling me the error

Here is the model

lung_ROM.zip

The code I use to analyse is:

model = cellml.parse_model(os.path.join(self.project_dir + 'lung_ROM.cellml'), False)
# resolve imports, in non-strict mode
importer = cellml.resolve_imports(model, self.project_dir False)
# need a flattened model for analysing
flat_model = cellml.flatten_model(model, importer)
model_string = cellml.print_model(flat_model)

# this if we want to create the flat model, for debugging
with open(os.path.join(self.output_dir, self.filename_prefix + '_flat.cellml'), 'w') as f:
    f.write(model_string)

# analyse the model
a = Analyser()

a.analyseModel(flat_model)
analysed_model = a.model() # here it segfaults.

The error when I open the model in OpenCOR is: "Error: a problem occurred during the generation of the model code (CellML Variable u_out in component pp_T_RICRI_type has in interfaces, but is not connected (directly or indirectly) to any variable with no in interfaces;)."

I'm not worried about the error in the CellML model, I am fixing that. The issue is to do with libCellML segfaulting when analysing a model with an error instead of specifying the error.

This is in the new libcellml0.6 with Python3.12

Answer 1 · 2024-11-01T02:15:29.000Z

The seg fault still occurs after I have fixed the CellML model.

Fixed model:
file:///home/farg967/Documents/physiology_models/pulmonary_CVS_Alfred/patient_4/pre/generated_models/lung_ROM/lung_ROM_fixed.zip

Answer 2 · 2024-11-01T02:53:26.000Z

Will look into it next week. In the meantime, @FinbarArgus, could you upload your fixed version?

Answer 3 · 2024-11-01T03:00:33.000Z

lung_ROM_fixed.zip

Answer 4 · 2024-11-01T03:10:47.000Z

Has anyone tried this with the previous version of libCellML? Does that version segfault with this model?

Answer 5 · 2024-11-01T03:11:39.000Z

The previous version worked fine.

Answer 6 · 2024-11-01T07:31:56.000Z

On second thought, the previous version worked fine with a very similar model, but not exactly the same. I will check with the same model on Monday.

Answer 7 · 2024-11-04T04:35:40.000Z

I confirm that with the previous version and the same CellML model there is no seg fault.

Answer 8 · 2024-11-04T05:56:18.000Z

Thanks @FinbarArgus. Will look into it once I am done with my current task (testing of generated code when tracking/untracking some variables).

Answer 9 · 2024-11-05T03:40:48.000Z

Ok, using @hsorby, I can confirm @FinbarArgus's issue. That's indeed some code that got introduced in 0.6.x... and our tests clearly didn't catch this particular (?) case. Am on it.

Answer 10 · 2024-11-05T03:41:37.000Z

Should be fixed either today or tomorrow and from there, @hsorby should be able to release 0.6.3.

Answer 11 · 2024-11-05T06:08:34.000Z

I have fixed the issue and tested the fix against @FinbarArgus's CellML file and it's all good. It was an oversight on my part. My bad!

Will provide a PR tomorrow morning.

Answer 12 · 2024-11-05T23:07:23.000Z

FWIW, the fix is very simple. I am just trying to come up with an easy to reproduce use case and this is not proving easy...

Answer 13 · 2024-11-06T05:31:35.000Z

Ok, got some atomic tests for this issue (and some related ones!).