Boundary Events and cancel activity

[jcosano]

I see that now cancelActivity is working fine but…

If i put cancelActivity to false, now, activity still in same state, but boundary event only is launched 1 time,

My target is:
I'll have an user Task that every day launch a mail if activity is not complete.

Now, I have a cycle with cancelActivity true, user task- boundary event- mail- user task, but this cause that task id changes  (causes by cycle, and I prefer I can keep same id)

If i put cancelActivity to true, only 1 mail is sended…

Which is the best way to model my target???

thanks!!

[frederikherema1]

AFAIK, non-interupting boundry events aren't supported yet in activiti (cancelActivity=false).

Is there a specific reason why the task-id (in case of your workaround, using flow that goos back in userTask after sending mail?

[jcosano]

In mail, I'll put a url, that opens this task…

If every day, task id changes (really is the same activity), I'm sure that someone will try to go to task with a old url (from a previous mail), and this url reference a task that not exists.

[frederikherema1]

I see,

Is the link to the explore UI or an own UI? Maybe, as workaround, you can set unique key in process-variable and copy it into a task-variable when task is started (listener), and use this as ID in your link. The query for the task should be something like taskService.createTaskQuery().taskVariableValueEquals("firstTaskIdentifier", id).

I should check with Joram and Tom, but don't think there is any concrete planning for non-interuptive boundary event on the roadmap.

[jcosano]

I'm using Activiti Explorer…

http://localhost:8080/activiti-explorer/start?taskId=119, where 119 is the ID of the task with an associated form.

[jcosano]

Maybe we need a reminder option, configure as a listener maybe?

[frederikherema1]

As a matter of fact, we're looking into (task) notifications this release. More details on that will folow.

[jcosano]

Thanks…

[jbarrez]

non-interrupting events are indeed not yet supported. There is some code that could work, but it's not yet documented nor is it tested enough.