[omniORB] A question about sequences

[DSSol@aol.com]

In a message dated 8/4/99 7:33:39 AM US Eastern Standard Time, 
b.keeping@ic.ac.uk writes:

<< To Tony:
 
 You should just be passing seq itself here. A _var can be used like a
 pointer, except that it cleans up the thing it points to when it goes
 out of scope. So think of the call as passing a pointer to a sequence (by
 reference, so that it can be changed).
 
 Your client code could look like this:
 
 doSomething( seq );
 for (int i=0;i<seq->length();i++)
    cout<<seq[i];
 
 incidentally the server code could look like this (assuming unbounded 
sequence)
 
 void doSomething( MySequence *&parm )
 {
   parm=new MySequence(2);
   parm->length(2);
   (*parm)[0]=val1; // where val1 and val2 are some values of whatever
   (*parm)[1]=val2; // type it's a sequence of!
 }
 
 To All:
 
 This question demonstrates a need for an example on sequences in the
 distribution. They took me a while to work out and I didn't find a
 book that seemed to consider them very important!
 
 Also, any comments on the way I'm doing things here welcome.
 
 Ben Keeping
 Imperial College
  >>

Ben,

Thanks for the reply.  I have a book that does have an example of how to pass 
sequences as in, inout, and out parameters.  From the client, they declared 
the sequence with the _var and passed it like doSomething( seq.out() ) as I 
described in my original message.  Their server code is implemented much like 
the code you described above.  The book I am using based their code on the 
Visigenic ORB.

In your example above, how should the sequence be declared on the client side 
(MySequence seq or MySequence_var seq)?  If I declared it with the _var and 
pass it without the .out(), the servant faults, even though it does compile 
in this case.  Is it possible that omniidl2 is not generating the proper code 
for the seq.out() case that is causing it not to compile?

Thanks again.
Tony Thompson