Stop Core Not Called Upon Actor Core Error Input

[LVB]

I just recently experienced some unexpected behavior with the Actor Framework:

• Store Core is not called if ActorFramework.lvlib:Actor Core.vi has an error input

I was calling an Init function inside of Actor Core before the Call Parent node.  I wired the Init error output terminal into the Actor Core error input terminal.  The Init VI was launching some nested actors and storing the send queues in the Actor class data.  Some code inside the Init VI was returning an error after the nested actors were launched.  This caused ActorFramework.lvlib:Actor Core.vi not to execute the Stop Core and the nested actors were left running.

Can I suggest that Stop Core be placed outside of the error case structure in ActorFramework.lvlib:Actor Core.vi?  This would result in Stop Core always being called.  Is there any downside to this suggestion?

[KvZ]

I would second that suggestion. I actually made the same assumption on the Stop Core that it will always be called when a Launched Actor Stops.

One other option is to put your initialization code in the Pre Launch Init.vi (if possible). If this initialization fails the actor core is not called, Actor Launch.vi return error 6006 indicating Pre Launch Init failure and the actual error returned by Pre Launch Init is returned with the Lask Ack Message.

[LVB]

I placed the Init.vi code in the Actor Core because:

• I wanted to launch nested actors (this rules out Pre-Launch Init)
• The Init code can take a few seconds to run, and I did not want to block the caller (this rules out Launch MyActor.vi)

There is actually a good discussion on where to place init code at https://decibel.ni.com/content/thread/12162

In addition, there is a lot of static data creted in the Init.vi and never changed during the lifetime of the Actor.  I use this information to populate controls and indicators on the Actor Core override UI.  This code only occurs on "Init", so creating messages for all of the init code and populating the class with controls/indicator refs would be a lot of additional work for only the gain of the Stop Core to be called.

[AristosQueue (NI)]

My thought was that you wouldn't want to call Stop if the loop never got started in the first place.

I guess the scenario you're thinking about is "I'm going to launch two nested actors in my Actor Core. The first one works. The second errors out."

I could trap the error and immediately stop the already-launched nested actor. Or I could just let my own Stop Core handle it, assuming there's nothing in my stop funciton that would keep it from stopping stuff if one of the references wasn't allocated.

Hm... it's not a bad suggestion. Anyone else have thoughts?

(PS: We've released the second beta of LV 2012 now. We are *really* late in the release cycle, so making this change now would require some serious determination on my part, and I like to save my late-in-the-game arguments with managers for the most significant changes. It's not impossible, but it gets harder with each passing day.)

[AristosQueue (NI)]

Well, that was easier than I expected. The change proposal flew through several reviewers and got a nod. I put the change in this morning. You'll see it in August.

And, now, having said that, I'm going back to not talking about the private beta on a public forum before The Powers That Be notice. 🙂

[LVB]

This is very impressive, I hope NI realizes what an effort you have made to create and support this framework.  It is great to work with a company that listens to it's customers, and then reacts so quickly to feedback.

IMO, having the guarantee of a Stop Core call greatly reduces the chance of leaking references.  This will be a great change to the framework.

[AristosQueue (NI)]

LVB wrote:

and then reacts so quickly to feedback.

It helps that I have their home e-mail addresses. 🙂

[maynej1]

Hmmm.... This is interesting.  Just because an error occurred in my actore core init doesn't mean I want to shut down my actor.  I would leave that decision to Handle Error.vi for the specified actor.  In my case I have all the Handle Error.vi's propogate errors to my main actor which then logs the issue to a file sends it off to my UI actor.  If a critical error were seen then the main actor would send a stop to all nested actors. 

I know the way I am handling the error is specific to my application, but that is kinda my point.  Don't run stop core on error.  I still want the actore core's message handler to run....  Just propagate it to Handle Error.vi.  I can then shutdown my actor from there if desired (or do whatever I want).

This post is probably too late...

[AristosQueue (NI)]

The change was not to make Stop Core run when an error occurs. The change was to make Stop Core run when the loop stops *even if* an error occurred. In other words, no matter why the loop stops, whether it be through clean shutdown or through error, run Stop Core.

[maynej1]

I apologize if I misunderstood your post.

I was referring to the problem of the original post where the issue is with error in when calling the parent actor core.vi (shown below in 3.0.7).  An error in will cause the entire core to be skipped and the child actor would simply exit.  If this child actor launched nested actors they remain running. 

I was simply suggesting to allow the error to propogate into the while loop, skip dequeue, skip do.vi, and run into the Handle Error routine.  One would then be able to choose whether to shutdown the actor (via Stop msg or simply setting boolean to True) or propagate the error. 

I have run into this scenario where error in into the parent Actor Core.vi caused problems with my nested actors (same scenario as described by initial post).  For now I am clearing/reporting the error before the parent Actor Core.vi is called.  I wish it would just enter Handler Error.vi like every other error in my actor.