Acitviti Modeler issue on Error Boundary Event

[skay]

An Error Boundary Event with no(null) Error Code attribute is not supported when deploying from the Explorer Workspace / after having using the Activiti Modeler.

This post is reformulating my previous post I answered myself
http://forums.activiti.org/content/error-boundary-event-and-modeler-bpmn-xml-failure

CREATE OF BPMN XML FAILED

Errors while parsing:
[Validation set: 'activiti-executable-process' | Problem: 'activiti-boundary-event-no-event-definition'] : Event definition is missing from boundary event - [Extra info : processDefinitionId = MainProcess | processDefinitionName = MainProcess | | id = boundaryerror1 | | activityName = Error | ] ( line: 9, column: 82)

An Error Boundary Event with no error code must be allowed and trap any BPMN Exception (like a try/catch Exception)

Thanks to # ACT-953, a Boundary Event with no Error code is supported when directly import the bpmn into Exporer process deployment / but this is not suppported when editing with Activiti Modeler.

Thank you for raising this issue.

[jbarrez]

So you mean the Modeler removes it?

[skay]

No : the Modeler is not supporting an error boundary with errorCode=null (exception when deploying).
This MUST be supported at it is explained in  #ACT-953 and means : get ANY BPMN Error thown from that node / no special code trigged.

The use case I described in my previous was

1) Using Activiti Designer (Eclipse), design a .bpmn with an error boundary / error code=NULL.
2) Deploy the .bpmn it in Activiti : deployment is OK and the process is running (OK).
3) Convet the process into editable model (workspace) / step into Activiti Modeler.
4) Do nothing (!) just save the model from Activiti Modeler.
5) Try to deploy it
=> EXCEPTION / validation error.

So the validation logic is different if you upload a .bpmn and deploy it into Activiti or if you edit it within Activiti Modeler.
According the exception, it sounds like the *parser* is different / coming from uploading .bpmn or Activiti Modeler.

[jbarrez]

I don't think the validation logic is wrong. I think the Modeler throws away the errorCode altogether, which is a bug