VisionOS 2.0 App Crashes Faster with Each Upload from XCode

We developed an app for VisionOS 2.0 Beta in XCode 16 Beta. The development was done in the beta versions since we needed features for our app which were not available in VisionOS 1.0, which was the most recent stable release at the time we developed the app. The app was fully functional on our AVP running VisionOS 2.0 Beta Version 5, and we never had any errors. We did not publish the app to the app store since we are a research lab using the app for teleoperating a custom robot.

Last week, we upgraded the AVP from VisionOS 2.0 Beta Version 5 to VisionOS 2.0 (stable release). Unfortunately, once we upgraded to 2.0, we began to have an issue with the app. While the app is running, at seemingly random times, without any new functionality being used within the app (no new buttons being pressed, etc), we encounter the following console error:

assertion failure: 'index < m_size' (operator[]:line 1011) Index out of range. index = 18446744073709551615, size = 0

We could re-upload the app to the AVP and successfully operate the app for several minutes until the same error occurred again. We thought to use Apple Configurator to flash VisionOS 2.0 Beta 5 to the AVP since the error wasn't happening on the previous firmware, but we were unable to flash a beta version of VisionOS via Apple Configurator, so we simply performed a factory reset of the device (on VisionOS2.0, by pressing restore in Apple Configurator with the AVP connected via the developer strap) to see if this might fix the issue.

After doing a factory reset, we thought the console error completely went away. We were able to operate the app for ~3 hours on Sunday with no issues. Then, yesterday (Monday) we operated the app for another 2 hours, and at the very end of using the app, it crashed with the same error. We re-uploaded the app with XCode, and the error occurred again after about 20 mins of using the app. This cycle repeated, and every time we re-uploaded the app, the time it took for the error to occur decreased, until we uploaded the app and the error occurred in <20 seconds.

We decided to test our hypothesis by upgrading VisionOS to 2.1 and using XCode 16. Similarly, we were able to run the app on the AVP for 2 hours, then the error occurred. The next time we ran the app, the error occurred within 20 minutes, then after reloading, occurred within 5 mins, then 2 mins, etc.

We are pretty stumped on why the app would work after a factory reset or a firmware upgrade for hours, then fail faster and faster every time we re-upload the app from Xcode. We are not experienced in debugging Swift and ObjC, so we wanted to inquire if this is an issue that you have ran into before to point us in the right direction. We think that it could be a problem with the cached memory that persists on the device across uploads from Xcode, but that's the extent of our understanding.

P.S., we also experienced this error during some of the app failures, but the one above is the most common:

assertion failure: Index out of range (operatorl]:line 858) index = 576460752303423487, max = 1

Answered by DTS Engineer in 812147022

Hello @AVPDeveloper,

Thanks for sending me that crash log. It appears that your code is calling HasTransform.position(relativeTo:) off of the MainActor (thread 3 of your crash log shows this call happening on your own DispatchQueue). HasTransform (and Entity) are isolated to the MainActor.

This is likely the explanation for the "Index out of range" assertion failure you are seeing from time to time. Effectively, I suspect that an array that should be isolated to the MainActor is being manipulated concurrently, resulting in this "seemingly random crash".

Best regards,

Greg

Hello @AVPDeveloper,

Can you provide a symbolicated crash log that shows the issue?

Best regards,

Greg

Hi @DTS Engineer,

I could not attach a .ips file in the developer forums, so I sent it in an email to you.

Thanks!

Hello @AVPDeveloper,

Thanks for sending me that crash log. It appears that your code is calling HasTransform.position(relativeTo:) off of the MainActor (thread 3 of your crash log shows this call happening on your own DispatchQueue). HasTransform (and Entity) are isolated to the MainActor.

This is likely the explanation for the "Index out of range" assertion failure you are seeing from time to time. Effectively, I suspect that an array that should be isolated to the MainActor is being manipulated concurrently, resulting in this "seemingly random crash".

Best regards,

Greg

VisionOS 2.0 App Crashes Faster with Each Upload from XCode
 
 
Q