OMSimulator CI tests of Buildings are reported to end successfully while they probably don't

[casella]

After #1271 was fixed, the master-fmi report for Buildings 11.0.0, which uses OMSimulator, reports 1462 model simulating successfully, whereas the master-fmi-fmpy, that uses FMPy, only reports 1355. So, it seems OMSimulator has a higher simulation success rate.

However, I compared some results and I found them incorrectly reported. For example, the Buildings.ThermalZones.Detailed.Examples.FFD.NaturalConvectionWithControl simulation is reported to fail during simulation (orange cell) in the master-fmi-fmpy report, more specifically during initialization, with the following error log:

python3 -m fmpy simulate  --output-file Buildings_11_Buildings.ThermalZones.Detailed.Examples.FFD.NaturalConvectionWithControl_res.csv --start-time 0 --stop-time 3600 --timeout 400 --relative-tolerance 1e-06 --interface-type ModelExchange --solver CVode --output-interval 1.44 Buildings_11_Buildings_ThermalZones_Detailed_Examples_FFD_NaturalConvectionWithControl.fmu
src->dim_size[0] != dst->dim_size[0], 0 != -1
LOG_ASSERT        | debug   | Failed to copy array. Dimension sizes are not equal and destination array is not flexible.
LOG_ASSERT        | error   | Failed to solve the initialization problem with global homotopy with equidistant step size.
LOG_ASSERT        | debug   | Unable to solve initialization problem.
Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/home/hudson/saved_omc/libraries/.local/lib/python3.10/site-packages/fmpy/__main__.py", line 4, in <module>
    main()
  File "/home/hudson/saved_omc/libraries/.local/lib/python3.10/site-packages/fmpy/command_line.py", line 169, in main
    result = simulate_fmu(args.fmu_filename,
  File "/home/hudson/saved_omc/libraries/.local/lib/python3.10/site-packages/fmpy/simulation.py", line 759, in simulate_fmu
    result = simulateME(model_description, fmu, start_time, stop_time, solver, step_size, relative_tolerance, start_values, apply_default_start_values, input, output, output_interval, record_events, timeout, step_finished, validate, set_stop_time)
  File "/home/hudson/saved_omc/libraries/.local/lib/python3.10/site-packages/fmpy/simulation.py", line 933, in simulateME
    fmu.exitInitializationMode()
  File "/home/hudson/saved_omc/libraries/.local/lib/python3.10/site-packages/fmpy/fmi2.py", line 280, in exitInitializationMode
    return self.fmi2ExitInitializationMode(self.component)
  File "/home/hudson/saved_omc/libraries/.local/lib/python3.10/site-packages/fmpy/fmi2.py", line 215, in w
    raise FMICallException(function=fname, status=res)
fmpy.fmi1.FMICallException: fmi2ExitInitializationMode failed with status 3 (error).

The same test in the master-fmi log reports a success (green), even though there is no result file to verify and the simulation log also mentions an initialization error.

/home/hudson/saved_omc/OMSimulator/install/bin/OMSimulator -r=Buildings_11_Buildings.ThermalZones.Detailed.Examples.FFD.NaturalConvectionWithControl_res.mat --tempDir=temp_Buildings_11_Buildings_ThermalZones_Detailed_Examples_FFD_NaturalConvectionWithControl_fmu --startTime=0 --stopTime=3600 --stepSize=1.44 --timeout=400 --tolerance=1e-06 Buildings_11_Buildings_ThermalZones_Detailed_Examples_FFD_NaturalConvectionWithControl.fmu
src->dim_size[0] != dst->dim_size[0], 0 != -1
LOG_ASSERT        | debug   | Failed to copy array. Dimension sizes are not equal and destination array is not flexible.
LOG_ASSERT        | error   | Failed to solve the initialization problem with global homotopy with equidistant step size.
LOG_ASSERT        | debug   | Unable to solve initialization problem.
error:   [fmi2logger] /var/lib/jenkins1/ws/OpenModelicaLibraryTestingWork_2/OpenModelicaLibraryTesting/Buildings_11_Buildings.ThermalZones.Detailed.Examples.FFD.NaturalConvectionWithControl/temp_Buildings_11_Buildings_ThermalZones_Detailed_Examples_FFD_NaturalConvectionWithControl_fmu/model-1nct0frg/temp/0001_Buildings_11_Buildings_ThermalZones_Detailed_Examples_FFD_NaturalConvectionWithControl (logFmi2Call): fmi2EnterInitializationMode: terminated by an assertion.
error:   [initialize] fmi2_import_exit_initialization_mode failed for FMU "model.root.Buildings_11_Buildings_ThermalZones_Detailed_Examples_FFD_NaturalConvectionWithControl"
error:   [initialize] Initialization of system "model.root" failed
error:   [simulate] Model "model" is in wrong model state
info:    0 warnings
info:    4 errors

I guess the exit code of OMSimulator is wrong - in this case, OMSimulator should return a non-zero exit code, since the simulation was not successful.

@lochel, please check with @adrpo how OMSimulator should terminate in these cases, in order to get correct CI test reports.

[lochel]

The model appears to get stuck in fmi2_freeInstance, which suggests there might be an issue with the FMU itself. It would be helpful if someone could investigate the FMI export to identify potential problems with the generated FMU.

[casella]

@arun3688 could you please check?

@lochel, what is strange is that the same FMU works fine with FMPy. Is OMSimulator calling the FMU differently?

[lochel]

The code gets stuck inside fmi2_freeInstance, specifically when invoking callExternalObjectDestructors:

  /* call external objects destructors */
  comp->fmuData->callback->callExternalObjectDestructors(comp->fmuData, comp->threadData);

The callExternalObjectDestructors implementation appears as follows:

void NaturalConvectionWithControl_callExternalObjectDestructors(DATA *data, threadData_t *threadData)
{
  if(data->simulationInfo->extObjs)
  {
    omc_Modelica_Blocks_Types_ExternalCombiTable1D_destructor(threadData,(data->simulationInfo->extObjs[1]));
    omc_Modelica_Blocks_Types_ExternalCombiTable1D_destructor(threadData,(data->simulationInfo->extObjs[2]));
    omc_Buildings_ThermalZones_Detailed_BaseClasses_CFDThread_destructor(threadData,(data->simulationInfo->extObjs[0]));
    free(data->simulationInfo->extObjs);
    data->simulationInfo->extObjs = 0;
  }
}

It seems the issue arises during the execution of one of the destructors. This needs further investigation by someone familiar with the runtime system to determine:

• Where exactly the hang occurs (e.g., in a specific destructor or during memory deallocation).
• Whether proper synchronization mechanisms are in place, especially if multiple threads are involved.
• If there are any preconditions violated when calling these destructors (e.g., invalid extObjs array state).

[lochel]

The issue description mentioned a log message that contained an incorrect function name. This has been corrected in OpenModelica/OpenModelica#13312.

[casella]

@arun3688 can you please look into that when you have time?

[arun3688]

@casella sure i will look into this