Menu

#6185 ANDROID: Crash when (un)plug keyboard on Asus Transformer

open
fuzzie
5
2012-11-14
2012-11-12
No

SCUMMVM versions tested:
- 1.5.0 from app store
- 1.6.0git2052-g46348e7 from daily builds

Android 4.1.1 (ASUS build: JRO03C.WW_epad-10.4.4.20-20121026), on TF700T.

How to reproduce:
- Start scummvm until game list is shown. Optionally start a game or bring up a dialog.
- Plug or unplug keyboard: after ~2s you should be back to home screen.
Restoring scummvm restarts the app (black screen for ~10s, then main menu is displayed)

I believe scummvm crashes rather than just being sent to background, because on the recent app list the preview is blank. When sending scumvm to the background manually, a screenshot is visible and restoring often works (though not always - but that is another bug which I didn't try to reproduce on 1.6 yet).

Discussion

  • digitall

    digitall - 2012-11-12
    • summary: ASUS transformer: scummvm crashes when (un)plugging keyboard --> ANDROID: Crash when (un)plug keyboard on Asus Transformer
     
  • digitall

    digitall - 2012-11-12

    vpelletier: It would be useful if you could manage to get the relevant logs at this point to help determine what is happening:
    https://developer.android.com/tools/help/adb.html#logcat
    and attach them to this bug as a zipped text file.

     
  • Vincent Pelletier

    logcat

     
  • Vincent Pelletier

    Attached.

     
  • digitall

    digitall - 2012-11-14
    • assigned_to: nobody --> fuzzie
     
  • digitall

    digitall - 2012-11-14

    vpelletier2: Thanks.

    Looking at the log, the relevant section appears to be:
    D/ScummVM (29492): surfaceCreated
    D/ScummVM (29492): surfaceChanged: 1920x1128 (4)
    I/ScummVM (29492): ScummVM 1.6.0git2052-g46348e7 (Nov 10 2012 04:11:12)
    F/ScummVM (29492): Assertion failure: '!_system' in ../../src-master/src/backends/platform/android/jni.cpp:426 (static void JNI::create(JNIEnv*, _jobject*, _jobject*, _jobject*, _jobject*, _jobject*, jint, jint))
    F/libc (29492): Fatal signal 11 (SIGSEGV) at 0xdeadbaad (code=1), thread 29535 (ScummVM)

     
  • digitall

    digitall - 2012-11-14

    fuzzie's analysis:

    "The log clearly shows that we get an onDestroy on the activity, so the bug is that we don't actually shut down completely there. The documentation implies we can check isFinishing() to see if we actually got permanently destroyed, so maybe we should leave ScummVM running in that case."

    Looking at replicating and will see if we can fix...