Unknown and unreproducible crash causes App Store rejection

Question

After submitting our application several times, we continue to receive the following response:

Thank you for submitting My App to the App Store. We've reviewed your application and determined that we cannot post this version of your iPad application to the App Store because My App is crashing on iPad running iPhone OS 3.2 and Mac OS X 10.6.2. My App crashes upon launch. Unfortunately, crash logs have not been generated.

However, resigning the same build with the AdHoc entitlements and loading the build onto the device yields no such crash. After a number of attempts, the application simply does not crash as reported by the reviewer. Furthermore, the reviewer does not provide any useful logging that may have been generated by SpringBoard such as an exit status or event if it had worked properly for any other device. There are no calls to explicitly exit or quit the application in the code line and yet the application terminates on startup.

What might cause an application to terminate in such a manner? Under what conditions is an application tested that might not be found under a development environment? Could it be a result of a signing issue that the submission validation system is simply unable to catch?

Thanks in advance.

Answer 1

Ouch! That's a tough one. I've had this kind of thing happen before, but it seemed it was reproduce-able when I switched to a different device (iPhone 3G vs 3Gs, etc.). In your case though, it sounds like you're iPad only, so that might not help. There's a difference between 3G and Wifi, I suppose, but I would be really surprised if that was causing a crash on one and not the other.

One thing you could try if you think Apple's not doing their part is to change the binary name and re-submit it under another app name/record. See if you get the same response. If you get the same reviewer, you might, but I think there's a good chance you would get a different reviewer with a new app. Just set your app release date to some time in the distant future and it will never actually show up in the app store if they approve it. If the new app comes back rejected for the same reason, you've got work to do, but if they do in fact approve it, then I would get on the phone with them and point out the discrepancy.

Answer 2

I think you have two options: try harder to reproduce the defect using the tools other people are mentioning or to catch those crashes in the field.

PLCrashReporter will trap on an uncaught exception and store all relevant information. Next time your app is run, it can send the crash report that you can then symbolificate and view a stack trace of.

Answer 3

A few suggestions:

1. Try to use the leaks tool of the static analyzer to see if there are any memory leaks or problems your not seeing.

2. Does your app use a web service? This happened to me one time because the day Apple was reviewing the app the web service went down. That cause a crash. If thats the case you would want to add something to catch that.

3. Finally, In the logs Apple sent you did they send you the dsym file? If they did you can run atos from the command line and it will convert the address to symbols. That will show you what thread and symbol it's crashing on.

Answer 4

My approach would be to make sure that the compilation is absolutely clean both in the static analyzer (this is NOT optional, it will save you time!) and for the build itself. Then set everything as described on CocoaDev' NSZombie page. Then make sure you have an iPod running 3.2 (actual hardware, not simulator) and test, test, test. Do it with network available and not available, with the device nearly full and freshly restored, every other variation you can think of. Run it with the debugger active and as a release build disconnected. Use Leaks, Instruments and all the other tools to peek in there and get a better feel for what is happening. If you exhaust every possibility you are going to have to beg for more detail from Apple but I bet you find a problem - one of the most important things in debugging an issue like this is to try to forget everything you think you know for sure and start at the beginning.

Answer 5

Be sure you test with a variety of settings: with Wifi disabled/enabled, with 3G disabled, Airplane mode on/off, location services disabled, etc.

As a last resort, assume there is a problem in your code that executes at startup. Remove half your startup code, set your release date in the far future (just in case it gets approved), then resubmit and see if they have the same problem. If not the problem is in the half you removed... it's a binary search.

Answer 6

After fighting with this for weeks, the application has finally been approved. The key: signing corruption. Hence why the application would start (or at least appear to start by showing the splash screen) and then suddenly vanish without a crash log. It failed Springboard's preflight sighing check.

A good tool to keep in mind to check for signing issues. It tells if the application has a valid signing:

codesign -vvvv MyApp.app

When I built the application, I would cp it to a network storage device where our product manager would pull it down and submit it to Apple. If decompressed on the NAS, the code signing was valid. But if you coped the compressed application back off the NAS and validate it, it would fail.

The lesson: make use of the new XCode utility to submit applications.