Issue with analyzeInitialSystem in HanserModelica

[Francesco Casella]

There are three models in the HanserModelica library that fail for unspecified reasons during analyzeInitialSystem, see:

• ​HanserModelica.Machines.TestDoubleLayer7over9
• ​HanserModelica.Machines.TestDoubleLayer10over12
• ​HanserModelica.Machines.TestSingleLayer12over12

@Karim, would you mind having a look at your earliest convenience? I think we should solve these for the next release, so we can claim 100% support over this nice library for teaching.

[Christian Kral <dr.christian.kral@…>]

The new frontend drops the following error message:

[/var/lib/jenkins2/ws/LINUX_BUILDS/tmp.build/openmodelica-1.16.0~dev-548-gd3091d6/OMCompiler/Compiler/NFFrontEnd/NFEvalFunction.mo: 732:9-735:52]: 
Internal error NFEvalFunction.assignArrayElement: unimplemented case for fill(0, if false then div(24, 2) else 24)[1] = 1

Unfortunately I do not understand what exactly this error message means. Might this be a Modelica implementation issue of HanserModelica?

[Karim Adbdelhak]

Looks like the fill() operator is indexed ... does this appear in the model like that? I don't think that this would be modelica language compliant.

If that is the case just replace it with an auxiliary variable which is created with fill and index that variable instead.

[Francesco Casella]

First of all, I didn't notice that these three models work fine with the old frontend. I guess this can give some clue so as to what is going wrong.

Then, it is a bit weird that I am not getting the same error message that @christiankral is getting. If I try out TestDoubleLayer7over9 with the new frontend using v1.16.0-dev-545-ga1be2e6252 under Windows, the backend fails with

[1] 18:40:26 Translation Error
No system for the symbolic initialization was generated

which is the same that happens with the ​Jenkins testing under Linux, currently with v1.16.0-dev.548+gd3091d6954. There is no error in the frontend, apparently, even though the way the new frontend handles this model eventually breaks the backend.

@christiankral are you installing OMC using APT or are you compiling it yourself? Maybe there is something wrong with that?

[Francesco Casella]

I checked the flattened model with the old and new frontend. The main difference I see is that the OF can split the binding equation for N into its scalar counterparts

  parameter Real N[1].re = 108.2294825561955 "Real part of complex number";
  parameter Real N[1].im = 0.0 "Imaginary part of complex number";
  parameter Real N[2].re = -54.11474127809771 "Real part of complex number";
  parameter Real N[2].im = 93.72948133211004 "Imaginary part of complex number";
  parameter Real N[3].re = -54.11474127809775 "Real part of complex number";
  parameter Real N[3].im = -93.72948133211004 "Imaginary part of complex number";

while the NF keeps it together and apparently defers their evaluation to the backend (which fails for some reason)

  parameter Real N[1].re(fixed = false) "Real part of complex number";
  parameter Real N[1].im(fixed = false) "Imaginary part of complex number";
  parameter Real N[2].re(fixed = false) "Real part of complex number";
  parameter Real N[2].im(fixed = false) "Imaginary part of complex number";
  parameter Real N[3].re(fixed = false) "Real part of complex number";
  parameter Real N[3].im(fixed = false) "Imaginary part of complex number";
...
initial equation
  N = HanserModelica.Machines.Functions.complexTurns(winding);

I'm not sure why the backend cannot handle that. @Karim, maybe you want to give a look?

[Christian Kral <dr.christian.kral@…>]

Replying to casella:

@christiankral are you installing OMC using APT or are you compiling it yourself? Maybe there is something wrong with that?

I am using APT to install and update OMC.

[Francesco Casella]

Weird.

I am using the ​latest released version of HanserModelica, are you?

[Christian Kral <dr.christian.kral@…>]

I am using the exact same version v1.1.0 (2019-03-24).

[Christian Kral <dr.christian.kral@…>]

Screen shot of error message

[Christian Kral <dr.christian.kral@…>]

One more note to not miss any important detail: I suppose that the path /var/lib/jenkins2/... shown in the error message is something like a virtual path as /var/lib/jenkins2 does physically not exist on my machine.

[Francesco Casella]

Ping...

[Francesco Casella]

Retargeted to 1.18.0 after 1.17.0-dev.beta2 release

[Francesco Casella]

Ticket retargeted after milestone closed

[Francesco Casella]

1.18.0 blocker tickets moved to 1.19.0