Crash on checkThread()

[Macarse]

After some investigation of a crash we are seeing in our app, we would like to share some information.

The stacktrace is the following:

de.greenrobot.dao.DaoException: Method may be called only in owner thread, use forCurrentThread to get an instance for this thread
at de.greenrobot.dao.query.AbstractQuery.checkThread(SourceFile:80)
at de.greenrobot.dao.query.AbstractQuery.setParameter(SourceFile:70)
at de.greenrobot.dao.query.AbstractQueryWithLimit.setParameter(SourceFile:48)
at de.greenrobot.dao.query.Query.setParameter(SourceFile:30)
setParameter
setParameter

This is happening mostly on RIM devices. AbstractQueryData's forCurrentThread() is using Process.myTid() and this method might return 0. There was an attempt to fix this problem in #171 and the ticket was closed saying it was fixed with 013af9c

While that fix might have worked for Robolectric, the issue mentioned by @dcowley is real and it is still there.

Can this be revised?

[Macarse]

ping?

[greenrobot]

This is happening mostly on RIM devices.

Mostly? What are other devices?

[Macarse]

@greenrobot based on the crash reporting tool we use, most of the crashes come from:

• RIM PASSPORT
• RIM Z30
• RIM CLASSIC
• RIM Z10
• RIM Q10
• RIM LEAP

We are also seeing some instances of Samsung/motorola/nexus devices. Apparently, Process.myTid() might return 0 in most devices but it's more frequent in RIM devices.

[greenrobot]

Current code looks like this:

        int threadId = Process.myTid();
        if (threadId == 0) {
            // Workaround for Robolectric, always returns 0
            long id = Thread.currentThread().getId();
            if (id < 0 || id > Integer.MAX_VALUE) {
                throw new RuntimeException("Cannot handle thread ID: " + id);
            }
            threadId = (int) id;
        }

How should that not work? What greenDAO version are you using?

[lexi-sr]

We are also getting this crash reported with the same stack trace for these devices:

Samsung: 75%
SM-N910V 39%
SM-G930V 15%
SAMSUNG-SM-J321AZ 8%
Other… 38%

LGE: 15%
LG-V930 25%
Nexus 5 25%
VS986 25%
Other… 25%

motorola: 8%
XT1080 75%
XT907 25%

other…: 2%

I think that there could be a case where Process.myTid(); returns a non-zero value that is not the owner thread's ID. If what dcowley said is correct (on #171),

From my understanding, Process.myTid() is OS-dependant, whereas Thread.currentThread().getId() (Java) is platform independent.

Using Thread.currentThread().getId() solves the issue of the application crashing on BB, and works for both BB and Android devices.

Then I would guess that the fix would be changing int threadId = Process.myTid(); to int threadId = Thread.currentThread().getId()? It looks like someone made a pull request for this in the past: https://github.com/greenrobot/greenDAO/pull/174/files But I think they closed it because their specific case was handled (a zero ID) but the general case is not handled (which is, an incorrect ID)? This is just my guess from what I read of these two issues.

[Macarse]

@lexi-sr: Could you find a way to reproduce? We couldn't.
@greenrobot I am using version 2.1.0 but this section of the code didn't change.

[lexi-sr]

@Macarse No, couldn't reproduce. For all the crash reports but one, it only occurred once per user. For one user, it occurred twice. So chances are it only happens in rare circumstances, but normally works fine.

[greenrobot]

In theory, Thread.currentThread().getId() may return a long while Process.myTid() is int. It is used with SparseArray here, which supports ints only.

[greenrobot]

Please keep an eye on this with the 3.1 release. Let me know if it will resolve the issue.

[Macarse]

@greenrobot any chance you can patch the 2.x release? It's hard for us to update to 3.x and it will take us some time.

[Macarse]

@greenrobot plz? :)