The NF causes vector assignment of discrete variables to functions to look like implicit systems of equations

[Francesco Casella]

Many models in the MSL (e.g. ​Modelica.Blocks.Examples.NoiseExamples.Distributions) fail when compiled with the NF because of this kind of error:

[OpenModelica/OMCompiler/Compiler/BackEnd/BackendDAETransform.mo:307:7-307:48:writable]
Error: Internal error BackendDAETransform.analyseStrongComponentBlock failed 
(Sorry - Support for Discrete Equation Systems is not yet implemented)
$PRE.uniformNoise.state[2]
$PRE.uniformNoise.state[1]
$PRE.uniformNoise.state[4]
$PRE.uniformNoise.state[3]
$PRE.uniformNoise.state = Modelica.Math.Random.Utilities.initialStateWithXorshift64star(uniformNoise.localSeed, uniformNoise.actualGlobalSeed, 4)

The flattened model with the old FE contains this equation

  {pre(uniformNoise.state[1]), pre(uniformNoise.state[2]), pre(uniformNoise.state[3]), pre(uniformNoise.state[4])} = Modelica.Math.Random.Generators.Xorshift128plus.initialState(uniformNoise.localSeed, uniformNoise.actualGlobalSeed);

while the new FE reports this

pre(uniformNoise.state) = Modelica.Math.Random.Generators.generator.initialState(uniformNoise.localSeed, uniformNoise.actualGlobalSeed);

Apparently, there is some issue with the DAEs returned by the NF that makes an explicit vector assignment erroneously look like an implicit system of equations, which the backend cannot (and should not, in this case) handle.

[Francesco Casella]

This issue seems to be fixed as of 07-02-2018, though there are still other issues further on with the reported model