Android apps should catch & handle JNI crashes as exception

[mhutti1]

From @kelson42 on September 26, 2016 10:57

To avoid any wild crashes of the apps. Here is a how to:
http://blog.httrack.com/blog/2013/08/23/catching-posix-signals-on-android/

Copied from original issue: kiwix/kiwix-xulrunner#330

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be now be reviewed manually. Thank you for your contributions.

[macgills]

This requires a kiwix-lib update, the native code must re-throw exceptions as java ones for the android app to begin handling them.

[kelson42]

@macgills I think:

• Newer version of the libkiwix tends to generate exceptions (general approach)
• The JNI tends to catch them already.

@mhutti1 @macgills Could you confirm? If the libkiwix does not behaves properly from your perspective, please open a ticket and linked it here. But, I believe this ticket can be closed.

[macgills]

This ticket cannot be closed. When the native code causes an error we get fatal signals, such errors are present in 2.5.1

If Java Exceptions were getting thrown I would get a stacktrace for such an error but what actually happens is I get this

Once it is actually throwing exceptions then on the android side we must catch them.

[kelson42]

I think this is a crash of the libzim, so a bug in it. You can not handle that neither in c++ nor in Java IMO. Do you have a reprocase for that?

[macgills]

I have absolutely no idea how that crash happened but the library @mhutti1 linked to claims it can catch SIGSEGV signals and if that was the case I would have a stacktrace and at least some idea of what code execution caused it.

[kelson42]

OK, thx for the explanation. Then, yes, if CoffeCatch works, then it should be setup.

[macgills]

@mgautierfr I believe this issue should be moved to kiwix-lib, can you confirm this?

[macgills]

I am moving this ticket to 3.1, @mgautierfr if you can review the information on this ticket and give your opinion on it's viability we can progress with it

[mgautierfr]

Yes it is feasible. At least we will be able to have some trace and solve issues.

Next question will be : "What to do when native code crash ?" If we crash because of a sigsegv, the library is in an undefined state so unusable.

[macgills]

Is there much internal state of the library? Can we supply a method to reset the state of the library.
On android we can relaunch the process which is fairly nasty but gives you a guarantee there is no lingering state

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be now be reviewed manually. Thank you for your contributions.

[kelson42]

@macgills Your last comment would probably benefit of a dedicated ticket on the kiwix-lib repo.

[kelson42]

@macgills We can not fix this ticket now. Postponing..

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be now be reviewed manually. Thank you for your contributions.

[kelson42]

@mgautierfr @macgills I believe it would be good to implement this if this is the path to follow to be able to investigate properly crash reports. Does that sounds doable now?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be now be reviewed manually. Thank you for your contributions.

[kelson42]

@mgautierfr @MohitMaliFtechiz This has to be fixed in new JNI binding. What is the status?