Wrong result in Buildings.Fluid.HydronicConfigurations.PassiveNetworks.Examples.SingleMixing

[mwetter]

Description

The model Buildings.Fluid.HydronicConfigurations.PassiveNetworks.Examples.SingleMixing (and others in this package that use similar constructs) have wrong results with OpenModelica.

The result below is from Buildings, commit 9b4be2033c1765ad5f518352ee164b8c7d78e775
The error is in the instance T2SetMod:

The OpenModelica results are as follows:

Note that index changes, but the output y does not.
It looks like the line
https://github.com/lbl-srg/modelica-buildings/blob/9b4be2033c1765ad5f518352ee164b8c7d78e775/Buildings/Controls/OBC/CDL/Routing/RealExtractor.mo#L21
is not computed properly.

Steps to Reproduce

Download Buildings, commit 9b4be2033c1765ad5f518352ee164b8c7d78e775
Simulate Buildings.Fluid.HydronicConfigurations.PassiveNetworks.Examples.SingleMixing with dassl, 1E-6.

Expected Behavior

See above.

Version and OS

• OpenModelica Version: OpenModelica 1.22.0~dev-41-g8a5b18f
• OS: Linux Ubuntu 20.04

[mwetter]

It turns out that this work-around gives the right result:
lbl-srg/modelica-buildings@48ad1f6

[casella]

That's kind of weird, the two formulations should be totally equivalent.

[rfranke]

It works with the Cpp runtime after PR #10642. See also #10603 for another example generating different events for C and C++.

[mahge]

I am not a 100% sure yet but this seems to be an issue with the removeSimpleEquations module. It appears the module constant evaluates the equation when it is in an equation section. It probably should not.

The library version in question is not available in the OpenModelica package manager (?) so I am working with a MWEs that I think are similar. It would be very helpful if you can provide a MWE for it. Otherwise I will check more and let you know.

[casella]

@mahge I tried to come up with an MWE only containing the signal blocks, but if I do so, everything works fine. I attach an MWE package to reproduce the issue (compare SingleMixing with SingleMixingAlgorithm, which uses the tweaked block with the algorithm instead of the equation)
MWE.zip

[mahge]

@mahge I tried to come up with an MWE only containing the signal blocks, but if I do so, everything works fine

Yes it was annoying me as well 🙂. One thing I found that made the two (algorithm vs equation) behave different was to have index as a parameter with no binding in my MWEs. However, I was not 100% sure if that represented the actual problem here.

I will check your MWE and will let you know.

[casella]

I will check your MWE and will let you know.

It's not such a big deal, it is basically just a copy of the original model, just set up so it depends on the installed version.