"Force close" shortly after snoozed alarm triggers

[GoogleCodeExporter]

I'm running v1.6 on CyanogenMod 6.1.1 (Android 2.2.1) on an HTC Desire GSM.

I should preface this by saying that AlarmKlock has been working flawlessly for 
me for months.  I frequently change the times of alarms but don't recall 
adding/removing any nor changing any other settings in the last week 
(AlarmKlock ones or even other Android ones).

What I do and see:

I've only seen the problem with one particular alarm (one which plays "Dog Days 
Are Over" by Florence+The Machine; settings screenshot attached), although I 
usually use a combination of two to wake me up.  DogDays sounds at the alarm 
time just fine, and if left to continue sounding for 15s or so it continues 
just fine.  I hit snooze.  6m later DogDays sounds again.  A few seconds into 
the track, AlarmKlock crashes with the standard "has stopped unexpectedly" 
force close dialog (attached).  I can reproduce this every time at the moment 
and it's been happening for at least the last three days or so.

I've posted the original ADB logcat (http://alarmklock.pastebin.com/XVk71dEX) 
plus one from a later experiment (http://alarmklock.pastebin.com/Lyfa8RjL) with 
alarm soundings highlighted.  The ones at 07:35, 07:41 and 07:47 are DogDays, 
and obviously the 07:47 one is the crasher.  (The ones at 07:40 and 07:44 are a 
"stop snoozing" alarm, but I can set DogDays alone and still reproduce the bug 
so it's probably unrelated.)  The "Uncaught exception thrown by finalizer" in 
the experiment log shows up at other times too -- not just directly before a 
crash.

Turning on AlarmKlock's debug only seems to change the UI, not what's logged at 
crash time.

Original issue reported on code.google.com by zakwil...@gmail.com on 19 Feb 2011 at 11:39

Attachments:

• forceclose.png
• dogdays.png

[GoogleCodeExporter]

Thanks for the detailed bug report.  The stack traces is the log output are 
really helpful.

A couple other people have reported this problem as well over the past couple 
weeks.  I haven't been able to reproduce it yet, but I finally managed to get a 
stack trace of the problem from someone earlier this week.  It seems to match 
up with the one you linked to.

I think there's a race condition in the notification code that triggers when 
the garbage collector is triggered at the same time as the alarm.  The app 
hasn't changed since last summer, so I'm still not sure why this seems to me a 
more common problem now.

Regardless, I put together an apk with a change that I think should fix it, but 
I haven't heard back from anyone yet.  Feel free to give it a try.
http://kraigsandroid.googlecode.com/files/Alarm%20Klock.apk

Thanks again,
Craig

Original comment by kraigs.a...@gmail.com on 19 Feb 2011 at 5:26

• Changed state: Accepted

[GoogleCodeExporter]

Thanks for the pointer/fix (and AlarmKlock generally).  I can no longer 
reproduce the problem having installed that new APK over the top of 1.6, so 
perhaps no news is good news.  I'll carry on using it as my daily alarm and 
report here if it happens again.

Has the repository moved?  I don't see the changeset listed at 
http://code.google.com/p/kraigsandroid/source/list .

Curious that it should only manifest itself now.  The most complicated and 
sensational explanation is normally the correct one, so I'm going to assume 
it's an elaborate Chinese trojan specifically targeting lethargic students like 
me in a long-term cyberwar effort to stifle innovation by the West.

Original comment by zakwil...@gmail.com on 19 Feb 2011 at 7:16