Clarify execution order of statements in an algorithm

[mauss]

The following model contains an algorithm with three
statements a,b,c. These seem to be executed in the
order a,c,b, which I find not completely obvious when
reading the specification.

model A
  Real x;
  Real y;
algorithm
  when time>0.5 then
    x := 1;
  end when;
  y := x;
  when time>0.5 then
    x := 2;
  end when;
end A;

When running this, y jumps from 0 to 2 at time=0.5,
not from 0 to 1 as one might expect.
Dynasim confirmed that this is the intended behaviour.

Proposal: In section 11.2.7 of the Modelica Spec 3.0,
clarify that all activated when statements of an algorithm
section are executed in their order of occurrence in the
algorithm section, but BEFORE all other statements
of the algorithm section.

[HansOlsson]

The description is incorrect.

The actual execution order is not a,c,b but a,b,c,b or in more detail:

first event iteration: a,b,c,
following event iterations: (skip a), b, (skip c)
(this can be repeated several times and the result is the same - as defined for event iterations)

This can be confirmed by changing this to:

when time>0.5 then

x := 1;

end when;
y := 1/(1-x);
when time>0.5 then

x := 2;

end when;

The proposed change would cause a change in semantics for algorithms - with unknown consequences.

However, a tool may often perform it as an optimization according to the general rule allowing optimizations.
---
Basically when the algorithm is executed once again during the event iteration y := 2 according to the algorithm. Thus y has several values at t=0.5, but you do not see the value y = 1.