Losing class modifiers during flattening

[Adrian Pop]

Flattening: ModelicaTest.Fluid.TestComponents.Sensors.TestTemperature1
should give orifice1.port_b.h_outflow having a start value from:

package StandardWater = WaterIF97_ph;

package WaterIF97_ph  
        extends WaterIF97_base(ThermoStates = Modelica.Media.Interfaces.Choices.IndependentVariables.ph, final ph_explicit = true, final dT_explicit = false, final pT_explicit = false, smoothModel = false, onePhase = false);
end WaterIF97_ph;

partial package WaterIF97_base
        extends Interfaces.PartialTwoPhaseMedium(mediumName = "WaterIF97", substanceNames = {"water"}, singleState = false, SpecificEnthalpy(start = 1.0e5, nominal = 5.0e5), Density(start = 150, nominal = 500), AbsolutePressure(start = 50e5, nominal = 10e5, min = 611.657, max = 100e6), Temperature(start = 500, nominal = 500, min = 273.15, max = 2273.15), smoothModel = false, onePhase = false, fluidConstants = waterConstants);
...
end WaterIF97_base;

Seems we're loosing class modifier:

SpecificEnthalpy(start = 1.0e5, nominal = 5.0e5)

when instantiating StandardWater.

[Willi Braun]

Is that also connected to #3051?

[Adrian Pop]

Yes. Seems similar to #3051.

I think it only happens if you have a short class definition that you extends in the mix something like:

type X = Y(mods).
type Z
  extends X(more mods);
end Z;

[Francesco Casella]

One more example of this behaviour is shown by ThermoPower.PowerPlants.SteamTurbineGroup.Tests.TestST3LRh_bypass, after the changes in revision 377 of the library have been applied.

Self-contained test package (TestST3LRh_bypassTotal-update1.zip) and test script (test-update1.mos) can be found attached to #3104.

This is a very serious bug for thermo-hydraulic models, which depend critically on start attributes. Great care has been taken at providing reasonable defaults in the Medium models, if they get lost and the default of zero is out of range, all kind of crazy runtime errors can take place.

[Adrian Pop]

Fixed in r24990.