(t'other thread's getting a bit mixed up - thought I'd start a new one for this.)
Again, this might or might not be an issue, due to the weirdity of how I did this.
I've got this communication diagram that investigates one flavour of the dreaded login use case. This is because the system is unathenticated access, new user's can add themeselves etc etc - basically it's a sort of
"web shopping trolley with an additional twist of user selects the specific shop" type of pattern. This is coupled with two additional features in that the user can a) swap shops during a session and b) even change identity and swap shops during a session.
I have always had trouble with C2SEQ for this model, in terms of the layout and message sequencing. But version 6.766 seems to have added some new twists, some amusing, some downright weird.
Dealing with the amusing first -
a) the lifelines don't appear to be resizable in any way
b) new objects added to the generated diagram are of a completely different (unresizable) size.
c) no horizontal scroll bars appear until you close and re-open the generated diagram.
Now the weird...
a) auto-garbage collection seems to have gone totally bonkers. The auto death-X's sometimes even appear before an object is finished its' activity and/or returned a result. Let A instantiate a B, B's constructor then does a couple of things (like retrieve a list of known user's from persistent storage) before telling A that it exists. A wants to databind the resultant user list to a form control. However, according to the diagram B is GORN!. There appears to be no way to extend B's lifeline - note I'm talking about the lifeline not the activation. So I turned auto garbage collection off.
b) activations have gone even screwier. Extending the above example, A is actually instantiated by an object even further to the left. When I add, delete or modify messages between A and B, A's activation behaviour changes to include variants of it starting even before A is instantiated to not having an activation at all.
Now I don't know at this stage whether it's the complexity of this diagram, the fact that it was generated by C2SEQ or 6.766 is at present not handling seq diagrams too well.
So, has any else seen weirdness in the seq diagrams?
bruce
