IssueTracker

Please attach a sample project (maybe from the new project wizard) that demonstrates this problem, and specify the method name.

Unable to attach due to 10mb file limit, so I used a sample project as suggested.

Bug can be reproduced by creating new 'login activity' sample project in Android Studio 1.51. Place a breakpoint on the second last line of the onCreate method. Proceed to follow steps in description above.

Ok, you mention: "Actual result: the app crashes with no stacktrace (tried several apps, including sample Google IO demo apps)". Could you attach the stack trace from logcat? Thanks.

There is no stack trace from logcat, the log is empty.

Indeed, I can reproduce this issue with the login template. On logcat, I only see:

02-12 06:55:48.011  1197  1197 I Zygote  : Process 3381 exited due to signal (11)

+shertz: The app just crashes on the device. The only thing I see on logcat is:

02-12 06:58:33.277  1197  1197 I Zygote  : Process 3887 exited due to signal (11)
02-12 06:58:33.282  1526  3333 I ActivityManager: Process com.foo.debugcrash (pid 3887) has died

To repro:

- Use http://tools.android.com/download/studio/canary/latest & open attached project (or if you already have some version of Studio, just follow steps in comment #2).
- Set a breakpoint 2 lines before end of onCreate()
- Click Step Over thrice until you go past the end of the method. After the last step over, the app dies on the device.

Yes I can reproduce too (on device and emulator running both M). I'm a bit surprised we do not see any stacktrace that would indicate what is the cause of the crash.

Since it happens when exiting the method, I guess this is related to deoptimization of the caller. Though I don't know what happens exactly.

+shertz: Do you think this could be a recent change somewhere in the frontend (Android Studio, or the build tools) that is triggering this more often now?

The 2.0 builds are also instant run enabled, and I wonder if that is leading to the crash. I'll get our QA to verify if this issue existing with older Studio, or with the older build tools.

QA tested this and it looks like this behavior is new with Studio 2.0. I will check to see what has changed. This is tracked in Issue 36949180.

I'm going to leave this bug open to track the core ART issue in case you guys are interested. Feel free to close otherwise.

This isn’t true - if you read the issue steps to replicate you will see that this behaviour happens on Android Studio 1.5 as well as 2.0 and is specific to testing with Marshmallow devices - any other Android version works fine.

I reproduced with a sample project on Nexus 5 (MMB29Q). This is the stack trace:

09-08 20:14:49.036 201-201/? A/DEBUG: *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
09-08 20:14:49.036 201-201/? A/DEBUG: Build fingerprint: 'google/hammerhead/hammerhead:6.0.1/MMB29Q/2480792:userdebug/dev-keys'
09-08 20:14:49.036 201-201/? A/DEBUG: Revision: '0'
09-08 20:14:49.036 201-201/? A/DEBUG: ABI: 'arm'
09-08 20:14:49.037 201-201/? A/DEBUG: pid: 4900, tid: 4900, name: eactionprovider  >>> com.example.android.actionbarcompat.shareactionprovider <<<
09-08 20:14:49.037 201-201/? A/DEBUG: signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 0xc
09-08 20:14:49.047 201-201/? A/DEBUG:     r0 bee6c4a4  r1 b4db6500  r2 00000000  r3 00000000
09-08 20:14:49.049 201-201/? A/DEBUG:     r4 bee6c4a4  r5 b6d51ec0  r6 b4db658c  r7 bee6c498
09-08 20:14:49.050 201-201/? A/DEBUG:     r8 ab1f08e0  r9 bee6c4ac  sl bee6c50c  fp 00000000
09-08 20:14:49.050 201-201/? A/DEBUG:     ip b6d5a2fc  sp bee6c488  lr b4943a33  pc b4b9fa62  cpsr 800e0030
09-08 20:14:49.069 201-201/? A/DEBUG:     #00 pc 00348a62  /system/lib/libart.so (art::Thread::PopDeoptimizationReturnValue()+9)
09-08 20:14:49.070 201-201/? A/DEBUG:     #01 pc 000eca2f  /system/lib/libart.so (art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*)+450)
09-08 20:14:49.070 201-201/? A/DEBUG:     #02 pc 0032ec6f  /system/lib/libart.so (art::InvokeMethod(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jobject*, _jobject*, unsigned int)+674)
09-08 20:14:49.070 201-201/? A/DEBUG:     #03 pc 002e40cd  /system/lib/libart.so (art::Method_invoke(_JNIEnv*, _jobject*, _jobject*, _jobject*)+32)
09-08 20:14:49.070 201-201/? A/DEBUG:     #04 pc 71e89e11  /data/dalvik-cache/arm/system@framework@boot.oat (offset 0x1ec8000)
09-08 20:14:49.389 201-201/? E/DEBUG: AM write failed: Broken pipe
09-08 20:14:49.391 820-852/system_process I/BootReceiver: Copying /data/tombstones/tombstone_00 to DropBox (SYSTEM_TOMBSTONE)

It looks very similar to https://code.google.com/p/android/issues/detail?id=198983

#12 - Thanks for reminding me of that. This is definitely a bug in ART, but there is very little we can do to fix API 23. Follow Issue 36949180 for the ART fix.

I only wanted to make sure that if Android Studio was doing something different that might trigger this more often, then we should look at that. In Issue 36949180, I'll try and see if I can repro on 1.5 as well. If so, there is probably little we can do from Studio. If its new w/ 2.0, then maybe we can change something.

I'm facing exactly the same issue - it crashes on stepping out from any method when debugging in Android Studio 1.5 and up on Android M. However, note that stepping out works fine in Eclipse ADT (tried it on the same source code).

I'm experiencing this with A.S. 2.2.2. It was happening also with the prior version and now it's worse. If I single step too fast, the debugger appears to crash, that is, my app dies and there is no further logcat info.

Also experiencing this - to the point where I am forced to use breakpoints
everywhere and constantly hit 'continue' to progress rather than 'step
into' or 'step over' - inconvenient to say the least.

Please fix that it makes debugging a nightmare.

Any fix???

Discussion on Stack Overflow: http://stackoverflow.com/questions/37495131/debugging-android-studio-causes-app-to-crash-when-stepping-over-to-exit-a-meth

The issue is still occurring on the latest build using android 6 on asus ze500kl.

I am also experiencing this issue.

(sending to the right assignee in the internal b/37077971 issue)

Using Studio 3.0 with Android 6 (Nexus) while steeping over still exits the app without any indication !

Having this problem on also, Android is version 6.0, Android Studio 3.0.1

Atonio, see suggestion on bug #37077971:

> I think moving to N is the best suggestion.

@26 Unless Google forces all Android 6.0 devices to update to newer versions, this means we won't be able to debug those problematic versions.
Question is, how could it be that Google released a version that can't be debugged.

Real men test on prod

With Mingyao leaving, we have to find new owners for CHA and LSE flavored issues. I am assigning this bug to me for now to get proper ownership for an initial investigation (but I may dispatch this later).

Debugging related, Alex would you mind having a look?

Does this mean the issue is gone?