Passing component references from outside of the function in the external function call

[henrikt-ma]

I don't think it is clear what to do in this case:

constant Integer c = 5;

pure function extFun
  output Integer y;
external "C"
  y = extFun(c);
end extFun;

What makes me unsure is that I expected that all component references in the external function call would be possible to classify as either input, output, or protected.

Note that making c an array shows that this can not be understood as a way to pass a constant argument, since only scalar constant arguments are allowed:

constant Integer[1] c = {5};

pure function extFun
  output Integer y;
external "C"
  y = extFun(c);
end extFun;

[HansOlsson]

After some thought I think this should be illegal, as it isn't clear how the argument should passed.

[HansOlsson]

Based on #3809 I realize that such constants from enclosing scopes should also be passed similarly as protected.

[henrikt-ma]

Based on #3809 I realize that such constants from enclosing scopes should also be passed similarly as protected.

Please clarify. In #3809 (comment) you suggested this (my emphasis):

Constants and size calls are passed as inputs, whereas protected non-constant variables (i.e., temporaries) are passed in the same way as outputs.

That is, saying that a constant variable should be treated similarly to a protected variable makes no sense.

My take from #3809 is that these constants from enclosing scopes should only – and clearly – be supported as valid subexpressions of a constant scalar argument expression.

[HansOlsson]

Based on #3809 I realize that such constants from enclosing scopes should also be passed similarly as protected.

Please clarify. In #3809 (comment) you suggested this (my emphasis):

Constants and size calls are passed as inputs, whereas protected non-constant variables (i.e., temporaries) are passed in the same way as outputs.

That is, saying that a constant variable should be treated similarly to a protected variable makes no sense.

My take from #3809 is that these constants from enclosing scopes should only – and clearly – be supported as valid subexpressions of a constant scalar argument expression.

Agreed. That was my intent as well, but I must have messed it up.

[henrikt-ma]

Agreed. That was my intent as well, but I must have messed it up.

Good. Then the fix for #3809 should be marked as a fix for the present issue as well.