simData.success not working properly?

[mestinso]

I'm running a simulation using the following command: simData = fmiSimulate(....

This particular FMU is terminating during the simulation and I see the following error printed out visually in the Julia REPL:
[Error][IllegalFunctionCall][TestFMIJulia]: fmi2DoStep: model is terminated

This is all fine and good and is expected in this particular case.

However, my surprise is that when I query simData.success it is returning true. Is this correct or should it be returning false?

If this behavior is expected, is there a better/alternative way to detect if the simulation wasn't successful?

Edit: I did a few more tests and inserted some divide by zero calculations into the simulation. Again, I see errors printed on the screen, but the simData.success is still true. Please advise if this is expected.

[ThummeTo]

You are correct. We can totally discuss this.
The success comes from the numerical integrator, this is the behavior for most ODE-libraries I know. success=true means there where no issues from the point of view of the integrator during solving (and this is true, because the FMU often gives "dummy" results or just skips a call if it is not applicable, however the ODE is still solved). If I get this correct, you are using CS now. In this case the numerical integration is done by the FMU and I think I just set the value to true. This is something that should be changed. And I will. I am doing a big update and will provide a solution for this too, so just let this open :-)
Thanks!

[mestinso]

To confirm, yes I am using CS.

Also, not to rush you, but any rough ETA for the update you mentioned?

As an FYI, what I've been doing as a workaround is to add a simtime = time equation inside of my modelica code. Then, I query the simtime output at the end of the simulation, and if it is lless than the stoptime I treat that as a failure.

[ThummeTo]

For CS, I think we should just track the return codes of fmi2XDoStep, if there is a != fmiXStatusOK then we will put success=false I think.

Yep, the library tree is under heavy refactoring for a v1.0.0 release, the current plan is to release one after another (one per week), so I think the final FMI.jl patch will be ready in 4-6 weeks (together with a lot more things solved than just this issue of course).