I am rather familiar with the assert() function, but I have just noticed that it behaves differently when called within a function. Here’s an MRE:

model toto

  Real y;
  Real z;

equation 

  y = sin(time);

  assert(
    noEvent(y > 0),
    "y smaller than 0",
    AssertionLevel.warning);

  z = function_titi(y);

  annotation (experiment(
      StopTime=10,
      Interval=0.01,
      __Dymola_Algorithm="Dassl"));

end toto;

and the function:

function function_titi

  input Real y;
  output Real z=1 "Dummy output";

algorithm 

  assert(
    noEvent(y < 0),
    "y greater than 0",
    AssertionLevel.warning);

end function_titi;

In this case, the assert() in the equation section will only be triggered once, when y becomes negative at t≡pi mod 2*pi. However, the assert() in the function_titi() is triggered at every time step while y is positive. Though I understand the logic behind it (i.e. the function has no “memory”), I am looking for a way to get the same behaviour to avoid untimely warnings. Perhaps using flags? Any idea?