Superstate-enter event not fired

[jeanbobby]

Hi,

This is my first usage of super states, as phases in a process.

here's an extact:

<start-state name="start">
 <transition to="buildPhase"/>
</start-state>

<super-state name="buildPhase">

 <event type="superstate-enter">
 <action class="com..." />
 </event>

...

 <state name="waitForValidation">
 <transition to="../deliveryPhase"/>
 </state>

 <event type="superstate-leave">
 <action class="com..." />
 </event>

The super state leave event is fired, but enter is not.

From what I read I cant say the doc is precise about this specific behaviour.
Looking at the code this seems 'normal' :

- Transition.take:

if ( destination.getSuperState()!=null ) {
 ... fire the event ...

- SuperState.execute :

Node startNode = (Node) nodes.get(0);
 startNode.enter(executionContext);

Going from a node to a super-state taking an explicit transition wont cause the event to be fired as dest.getSuperState == null.

Then the super state will be executed and the first node will be entered without any kind of 'transient' transition being taken.

What do you guys think about that ?

AFAIAC I would expect the super-enter event being fired whatever the enter context (thru an innder node, or referencing the super-state itself).

Thanks for your input.

Best regards

Olivier

[jeanbobby]

from the doc : (http://docs.jboss.com/jbpm/v3/userguide/processmodelling.html#superstateevents)

9.6.2. Superstate events
...
These events will be fired no matter over which transitions the node is entered or left respectively

I cant find anything more precise about the expected behaviour

[estaub]

jb,

It looks like a bug to me - I'd JIRA it and see what response you get.

Maybe the intent was to not allow transitions to the superstate. One way or another, though, there's probably something that needs doing.

-Ed Staub

[jeanbobby]

Ok, thanks Ed.
Copy/paste from the post is detailed enough ?

[estaub]

jb,

Copy/paste from the post is detailed enough ?

It would be for me. I'd include the doc reference too.

-Ed Staub

[jeanbobby]

FYI : http://jira.jboss.com/jira/browse/JBPM-980

[tom.baeyens]

before the test "if (destination.getSuperState()!=null)", look at how the destination is set to the first node inside the superstate:

Node destination = to;
 while (destination instanceof SuperState) {
 destination = (Node) ((SuperState) destination).getNodes().get(0);
 }

 ...

 // performance optimisation: check if at least there is a candidate superstate to be entered.
 if ( destination.getSuperState()!=null ) {

the following test succeeds:

public void testSuperStateEnterViaTransitionToSuperState() {
 processDefinition = ProcessDefinition.parseXmlString(
 "<process-definition>" +
 " <start-state name='start'>" +
 " <transition to='superstate'/>" +
 " </start-state>" +
 " <super-state name='superstate'>" +
 " <event type='superstate-enter'>" +
 " <action class='org.jbpm.graph.exe.SuperStateActionExecutionTest$Recorder' />" +
 " </event>" +
 " <super-state name='nestedsuperstate'>" +
 " <event type='superstate-enter'>" +
 " <action class='org.jbpm.graph.exe.SuperStateActionExecutionTest$Recorder' />" +
 " </event>" +
 " <state name='insidenestedsuperstate' />" +
 " </super-state>" +
 " </super-state>" +
 "</process-definition>"
 );
 // create the process instance
 processInstance = new ProcessInstance(processDefinition);
 processInstance.signal();
 assertEquals(3, executedActions.size());

 // the first action called is the superstate-enter on the 'superstate'
 ExecutedAction executedAction = (ExecutedAction) executedActions.get(0);
 assertEquals("superstate-enter", executedAction.event.getEventType());
 assertSame(processDefinition.getNode("superstate"), executedAction.event.getGraphElement());
 assertSame(processDefinition.getNode("superstate"), executedAction.eventSource);
 assertSame(processInstance.getRootToken(), executedAction.token);
 assertNull(executedAction.node);

 // the second action called is the superstate-enter on the 'nestedsuperstate'
 executedAction = (ExecutedAction) executedActions.get(1);
 assertEquals("superstate-enter", executedAction.event.getEventType());
 assertSame(processDefinition.findNode("superstate/nestedsuperstate"), executedAction.event.getGraphElement());
 assertSame(processDefinition.findNode("superstate/nestedsuperstate"), executedAction.eventSource);
 assertSame(processInstance.getRootToken(), executedAction.token);
 assertNull(executedAction.node);

 // the third action called is the *propagated* event of the 'nestedsuperstate' to the 'superstate'
 executedAction = (ExecutedAction) executedActions.get(2);
 assertEquals("superstate-enter", executedAction.event.getEventType());
 assertSame(processDefinition.findNode("superstate"), executedAction.event.getGraphElement());
 assertSame(processDefinition.findNode("superstate/nestedsuperstate"), executedAction.eventSource);
 assertSame(processInstance.getRootToken(), executedAction.token);
 assertNull(executedAction.node);
 }

do you see any difference with your scenario ? i can't seem to reproduce like this. i'll give it another try and see if persistenting the execution makes any difference.

[tom.baeyens]

persistence seems to make a difference... i'm investigating.

[kukeltje]

Tom, persistence seems to make a difference in a lot more cases. See e.g. the combobox test with selectItems.... Sorry for threadjacking ;-)