Starling breaks on release build not on debug.

jefflemon

When I debug from flash builder, starling launches fine. When I do a release build starling won't launch. My iOS app is just stuck before starling initialises.

This is strange because sometimes a release build works, up until yesterday, release builds were working fine, then they broke, so I started a new project and copied everything across and release builds worked again but now they have broken again.

Does anyone have a fix?

Daniel

Sorry, I haven't had such issues before!
That is, 2-3 times, I couldn't start the app, but after deleting it and copying it again to the device, it worked. So I guess some process was simply stuck (or something like that). Try to stop and delete the app from the device, reboot the device, and copy it again -- perhaps that helps!

wideeye

Hmm, I'm getting this - or very similar- as well. Target is Android device. Tend to do builds directly onto device using USB cable. Works absolutely fine when launching from FDT over USB connection, but when launching the app directly on the phone it shows the "bootstrap" loader I've got (displays movieclip on stage then passes control to starling) but it doesn't initialise the Starling portion of the app. Odd!

blemmo

Are you calling Starling.start() when you initialize?

wideeye

me or jefflemon? if me, yes, calling start() method on starling instance... same happens for debug or release builds. Been absolutely fine for a couple of weeks of development, now won't start "from cold" (e.g. restart phone, unplugged from anything, launch app... hangs on Starling start)

blemmo

I had a problem with starting my app when I didn't call Starling.start() in the Starling init process. Turns out that since AIR 3.2, Flash won't render anything if you have a stage3d context but you don't refresh it. So you cannot create Starling without starting it too, because this will freeze your app completely. Unfortunately, this isn't the case when you use the (IPhone) simulator - maybe it's also not freezing if you use the direct deploy to Android, but does so when you build a proper release?

If you're sure to call Starling.start() it should be another problem tho...

wideeye

Erm, oops, my fault entirely, it was an unhandled exception in my code early on. Apologies all, ignore all my comments

/hangs head in shame

Kramer

Hi guys, so, wideeye, how did you find the unhandled exception ? The reason I'm asking is that I am experiencing that my release build breaks on Starling initialization.

When I install and run it on my HTC Desire via USB cable, I see the HiRes stats box for a little while, hear 1/2 second of music, then my app just shows me a full screen of the starling background color.

If I install my app on my desktop PC and run it, it will just open up, play the half second of music and terminate.

Surely there must be something, I just dont know how and where to look... Any ideas ?
Thanks guys, cheers, Claus

EDIT: Forgot to mention that my debug versions all run well on my desktop as well as on my phone.

eRistocracy

Hi. I am having the same, but with inlining used on iOS. I have noticed some NullPointerException java error, which I now don't remember exactly ;(
I have reduced the number of [Inline] functions and it fixed my problem. Don't use more than 50 expressions.

On Android 2.2 (Samsung Galaxy Tab) and iPhone4 my game runs much faster thanks to ASC 2.0.

When we will see Starling and Feathers versions with inline improvements?

joshtynjala

When we will see Starling and Feathers versions with inline improvements?

Most functions in Feathers are protected so that they can be overridden. With that in mind, Feathers and inline won't mix very well.

eRistocracy

Why is that? Could you explain? Plaease

joshtynjala

As I said, most functions in Feathers are protected so that they can be overridden. Functions need to be final to allow inlining. That makes it so you can't override them.

Kramer

Hi again,
so today I have been commenting out code until I reached the line that kills my release builds. My app consists of 5 screens that scroll left and right. All screens are initialized when my app starts up and they all extend the Starling.display.Sprite class. So, upon initialization, each screen instantiates its own number of Starling.display.Button objects, and this, apparently, crashes the release build.
Instead of five screens, for testing purposes, I now only have one screen with one button. Instantiating this Button with a Texture from a TextureAtlas or with an empty Texture, makes absolutely no difference, my app just shows me the HiRes stats box until the Starling engine is up, then the box disappears and I am left with the background color on my screen.
To me, it makes no sense so far, any ideas ?
Cheers, Claus

EDIT: Replaced the Button object with an Image object with the same Texture from my TextureAtlas. The app is running, but with no Buttons comes less interactivity...

iddqd

I just had the same problem and finally figured it out, after commenting out pretty much everything 😉

In my case it was not related to starling, this line that detects if i'm running a release or debug build seems to break the whole app:

public static var isDebug:Boolean = (new Error().getStackTrace().search(/:[0-9]+]$/m) > -1);

I'm pretty sure it was working fine before ASC 2.0. A clean of the project is required after removing the line, otherwise it still won't work...

phew, finally.

Kramer

Hi Guys,
I finally solved it too, at least I hope so, the app now runs on my Android devices as well as on my desktop pc.
If you read my earlier posts, you will see that my problem only presented itself in release builds, the debug versions all worked with no signs of faulty code.
Since a release build will give you no clue as to what is going wrong (no traces, no output windows), here's what I did, to get some information from my app:

Exported a release build for installation on desktop and installed the app.
Read this post: http://backroom.bostonproductions.com/?p=902 - this page tells you how to get debug information logged in a textfile. I never got that to work, however, followed step 1 and 2, which resulted in a debug dialog window when running my app, which showed me the stack trace of my faulty code.

The debug dialog window told me, that a stack underflow error had occured in my AddedToStageEventHandler... This had gotten me a bit closer, but nowhere near the solution. I googled a bit and stumbled on this post (http://stackoverflow.com/questions/6270837/how-to-debug-a-runtime-stack-underflow-error) talking about how to debug this error. The later entries of this post mention that trace statements may get compiled out in release mode and that it might lead to the stack underflow error (VerifyError #1024).
Adding the following statement into the additional compiler arguments in flash builder got rid of the trace statements in my release builds and thus relieved me of my problem:
-omit-trace-statements=false

Hope to have shed some light on this so that the next to experience this nasty bug will find a quick solution.
Cheers, Claus

yono

Kramer - you are a legend! I just spent a day trying to track down this error that happened on a HTC Desire with release build but NOT on a Samsung Galaxy II and it had me totally stumped. Added the -omit-trace-statements=false and totally fixed.
Why Flash Builder would compile a trace statement into a release build I still don't really understand but for now its fixed so I thought best to let people know to try this!

Kramer

@yono, thanks, great to hear that it solved your problem relatively quickly. I spent days as my posts reveal. I just wanted to add that for me it happened using flash builder 4.6. I have no idea if this has been fixed in 4.7 nor as to why some trace statements spilled over into the release build.
I am still working on the same project and I have had NO trouble since adding the -omit-trace-statement=false - I am hoping neither you nor anyone else will see this reemerging !!!
Cheers, Claus

yono

Kramer - just an update on FB4.7 - I just spent a few hours scratching my head about a Release build app not working in Flashbuilder 4.7 (debug worked fine) and then remembered this problem from my last project. Low and behold I added the -omit-trace-statements=false and it all started working again.
Come on Adobe! Surely when we hit compile stripping out trace statements should be pretty straightforward!

asking

Wow, so, I might as well post this here, because my issue is somewhat similar:

With a debug build, I can run my whole app, no issues, no bugs, can deploy to Android no problem. Fine...all is good...

However, packaged for release on Android OR even when I run (NON DEBUG version) from IntelliJ, deploying straight to the device, I get strange uncaught errors, stating that event listeners are not functions. The errors are:

TypeError: Error #1006: value is not a function.
	at Main/initCtrl()
	at Main/onLoaderInfoComplete()
[Starling] Initialization complete.
[Starling] Display Driver: OpenGL (Baseline Constrained)
TypeError: Error #1006: value is not a function.
	at Main/onRootCreated()
	at starling.events::EventDispatcher/invokeEvent()
	at starling.events::EventDispatcher/dispatchEvent()
	at starling.events::EventDispatcher/dispatchEventWith()
	at starling.core::Starling/initializeRoot()
	at starling.core::Starling/initialize()
	at starling.core::Starling/onContextCreated()

What duh??? Mother?

asking

Hmm... my problem seems to be that that [Inject] tags, and perhaps other metadata tags, are being stripped out in release build... I thought passing the compiler option:

-keep-as3-metadata

...would keep all metadata, no?