Roku Developer Program

[btpoole]

Have an app that worked great on the box with version  7.5.0.  Once 7.6 hit app stopped. The app is done in scene graph. Any reported bugs for such?

[belltown]

That's extremely vague, and not even posted in the right forum. Good luck with that.

[btpoole]

Well as for forum, I posted in general discussion because other post concerning 7.6 updates were posted here.  As for vague, was looking for any general info that may have been reported under such conditions, not looking for a debug of the app.

[RokuShawnS]

Moved to the appropriate forum.

[NB_]

A (well-constructed) channel that worked on 7.5 should work as well with 7.6.
We go over great pains to ensure compatibility with the existing apps but a fact of life is that with each new version some apps would be unintentionally affected.


No, i am not aware of why "an RSG app" would not work in 7.6.
Side-load it and check the console for diagnostics, hopefully would be easy to tell.

[btpoole]

Well, I haven't received the 7.6 update on the Roku2 the pack was developed on so I can't really check it. As for well constructed, the app was running on approx 2,000 boxes in the past month. Since the 7.6 hit the Roku 2 boxes some of the users have, the app no longer runs.  It still runs on the express and Roku 3.

[btpoole]

Edit: If the task is indefinite, is it possible that the next step in the process, the parse begins before the task is complete? If so this would explain why it might be crashing in the function where the parse occurs with the execution timeout.  I am guessing using an observerfield to wait on the "DONE" state of the task would be too simple of a solution.


According to the doc on line

As of firmware version 7.5, thread rendezvous no longer timeout and will wait indefinitely

I have updated the box and the app crashes on 7.6 and I receive a Execution timeout (runtime error &h23) error. I have a task that reads in data from a server and puts in an array that is passed to the content node. Once there, a function parses the information in the node. In this particular run, the array has 5000 items and is process or cycled thru 67 times. In V 7.5 it never timed out. In V7.6 the crash occurs around 3700. From what I read in the docs state that 

If a timeout occurs while getting a field, the response will be invalid, which may crash the application if the script is not prepared for the invalid response[/color]

So, I guess my question is did the update to 7.6 change from a wait indefinitely to a limited wait?

I do have a catch in the code to stop invalid or empty returns, but that doesn't seem to stop the crash.

[JethroTSquirrel]

A production channel is allowed to run for a maximum of 10 seconds on the Scene Graph render thread. This means that if a channel needs to perform significant processing, that work must be moved to a Task node, to avoid occupying the render thread. To enforce this requirement, a channel that executes for longer than 10 seconds on the render thread will be stopped with an EXECUTION_TIMEOUT. Note that for a side-loaded channel, the timeout is reduced to 3 seconds, so that problems will be found before a channel is published.

The channel's main thread and Task threads do not have this restriction.

Unfortunately, the enforcement of the EXECUTION_TIMEOUT on the render thread was spurious in firmware version 7.5. The enforcement should now be consistent in 7.6.

[btpoole]

Thank you JethroTSquirrel you have validated the conclusion I had come to in my mind. I had already started the process of rewriting some of the code to move the process to a Task to see if this stops the crash.