signal 11 (SIGSEGV), code 1 (SEGV_MAPERR) art_quick_imt_conflict_trampoline

Solar2D General Questions/Discussion
#1

Hi Corona

Since my last build (using 3200) I am seeing a lot of the following…

Today, 16:39 on app version 1110 Lenovo YT3-X50F/Yoga3 Tablet (YT3-X50F), 1024MB RAM, Android 6.0 Report 18 \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* \*\*\* pid: 0, tid: 0 \>\>\> com.spheregamestudios.spacecity \<\<\< backtrace: #00 pc 00000000000ea678 /system/lib/libart.so (art\_quick\_imt\_conflict\_trampoline+7) #01 pc 0000000001f2d6f3 /system/framework/arm/boot.oat

… note this is only happening on android 6.  I appreciate the trace is tiny but that is all I can see.

Any ideas?

#2

So lets examine this a bit… so I moved from vungle SDK 4 to SDK 5 and have noticed a massive increase in crashes.  I will represent this with a graphic.

The only thing I added with this release was Vungle SDK 5 to monetize Amazon players (this was unsupported in SDK 4).

Now bear in mind some 15k DAU then 50 crashes might seem acceptable (low mem, etc) but a 4x increase certainly is not.  Do note this is entirely limited to Android 6 so therefore is a framework/plugin issue.

Now I am sure Corona will say “oh Vungle maintain their own plugin” and whilst I get that, you are surfacing that as a Corona plugin so surely there is some ownership there?  After all Amazon take ownership of all their sellers, etc.

Please also note this recent post which is related https://forums.coronalabs.com/topic/71859-crash-report-signal-11-sigsegv-code-1-segv-maperr-zn3art11interpreter6docallilb0elb1eeebpns-9artmethodepns-6threaderns-11shadowframeepkns-11i/

So perhaps there can be some investigation? 

#3

Hi, I wanted to add that in addition to seeing the crash report above, I’ve seen other issues with the latest Vungle plugin. One that I reported directly to Vungle a couple weeks ago.

Issue 1:

One of the ads will keep playing a music loop after it is dismissed. What’s worse is it keeps playing even when leaving the app unless you completely close the app. I’ve only seen this on iOS so far but it may happen on Android as well.

Issue 2:

On Amazon when clicking on some ads it will take you to a browser and then bring up a completely different app than what is being advertised. I realize Vungle is new on Amazon and inventory is low, but this needs to be reported. If you need more detail, I will ask a tester to send it to me or try to re-produce the problem myself.

#4

I’m not clear about your release schedule. Is the change to 3200 timed with the increase in crashes, or have you been using 3200 for previous releases and the change to Vungle 5 is reflected by this change?

I’ll be happy to email our contact at Vungle and have them look into this. As you mentioned, we have no access to their source code. They are pretty good about watching their specific forum for issues, but with this being in a general content forum, I’ll make sure they are aware of this thread.

Rob

#5

Move to 3200 coincided with moving from vungle sdk 4 to 5 - so I’m not sure which is at fault (or both).  I was previously 3184 and sdk 4 plugin.

All I can say is this crash is now some 80+% of all crashes in Space City so it is kinda major for me.  I am holding back doing this for Designer City until this can get resolved.

#6

I bet my life that this has nothing to do with Vungle or other plugins.

Corona has major bugs related to 1) “openal” and 2) the listeners involved when app changes state (pause/resume/exit). I guess there is a race condition or a destroy trap there.

They released a bug fix last month but it didn’t fixed all the bugs. Corona team keeps asking for projects that can reproduce the bug. This is not possible! This happens @0.5% os the sessions so we developers cannot reproduce it easily, but it is a huge percentage that marks an app for a penalty in google play rankings. Worst of all, a game needs to have over 5K daily active users to start facing a stable number of the bug occurrences. Since most devs are not that successful not everybody screams about it. However, the bugs are there for good! I also face the same problem for a lot of months (probably since the beggining years ago - now google vitals just undercover the situation).

I really don’t expect something good from the Corona team anymore. It seems that this project is just (softly) maintaned and will be soon abandoned. They don’t even fix a major bug for 6+ months now. Search the forums for “crashes” and “openal” and you will see what is happening right now. But nobody cares. The devs are focusing in the html5 feature promised 5 years ago. Just abandon things you have not the resources to do and focus on bugs…

A dissapointed long time (Carlos times) user…

#7

I think it was more of a “shot in the dark” – my impression was that Vlad did find “something” irregular, addressed it, and “hoped” that it might be the root cause.  I was pretty excited to maybe finally have a fix.  Unfortunately, whatever it may have fixed, it didn’t fix ANY of the ANR’s – my ANR rate (for that same error) remains where it was 3 months after submitting update.

SGS’s report is a different one, though I suspect you are correct that it is still somehow related to state-change handling (as that’s where “most” of the current serious errors seem to track toward).

#8

We are replacing the underlying OpenAL code with a new system. Hopefully this will address the OpenAL issues.  I’ll make sure Engineering see’s this thread.

Rob

#9

@Rob I’ve just raised a ticket on this with back traces and more info.

#12934639

#10

@Rob, Some more insight…  Had a few customers emailing in saying their game wouldn’t start and I’ve tracked it to this problem.

Basically what is happening is the crash occurs some 3- 5 seconds into the load procedure but before any heavy lifting is done.  

So I don’t think it is a state change as such and is more likely plugin initialisation as that is about when they are instantiated.

Hope this helps?

#11

I’ve added a note to your bug report.

Rob

#12

Is there a way to follow the bug report/ticket or will you update us here when there is a fix?

#13

In general if the person filing the bug report includes a link to the forum post, we can come back and update the forum post. If not, you have to watch the change log. The bug reporter will get an email, with follow up questions as necessary and updates as we process the bug.

I don’t know if @sgs included the forum link in the submission form or not.

Rob

#14

UPDATE: Vungle are going to be patching the SDK. 

They have said “there will be an iOS release and corona plugin will be released with the fix for this android issue as well on 12th of March.”

Hopefully, this will address the crashes and the duplication of events being raised by the SDK.  If not then it points to a framework issue as I doubt Google have done any updates to Android 6 recently.

@Rob, if they do patch the plugin March 12th, how long til it is available for sim builds?

Thanks

#15

Once they upload the plugin to our servers, it will be live in about an hour.  

Rob

#16

Hey Rob, can you see if Vungle have patched today?

#17

It does not appear that Vungle has updated today, but it’s still early.

Rob

#18

Seems the Vungle patch fixed the multiple events bug but NOT the crashes on Android 6.

The stack trace for the crashes is 

java.lang.NullPointerException: Attempt to invoke virtual method 'void android.app.Activity.runOnUiThread(java.lang.Runnable)' on a null object reference Java Stack Trace: plugin.vungle.LuaLoader.init(LuaLoader.java:176) plugin.vungle.LuaLoader$InitWrapper.invoke(LuaLoader.java:132) com.ansca.corona.JavaToNativeShim.nativeRender(Native Method) com.ansca.corona.JavaToNativeShim.render(JavaToNativeShim.java:182) com.ansca.corona.Controller.updateRuntimeState(Controller.java:347) com.ansca.corona.graphics.opengl.CoronaGLSurfaceView$CoronaRenderer.onDrawFrame(CoronaGLSurfaceView.java:425) com.ansca.corona.graphics.opengl.GLSurfaceView$GLThread.guardedRun(GLSurfaceView.java:1623) com.ansca.corona.graphics.opengl.GLSurfaceView$GLThread.run(GLSurfaceView.java:1378) stack traceback: [C]: in function 'init' ?: in function 'vungle\_Init' ?: in function 'method' /Users/jenkins/slaveroot/workspace/Templates/label/android/platform/resources/init.lua:221: in function 'dispatchEvent' ?: in function 'gotoScene' ?: in function 'listener' /Users/jenkins/slaveroot/workspace/Templates/label/android/subrepos/timer/timer.lua:210: in function 'method' /Users/jenkins/slaveroot/workspace/Templates/label/android/platform/resources/init.lua:221: in function \</Users/jenkins/slaveroot/workspace/Templates/label/android/platform/resources/init.lua:190\>
#19

We have asked Vungle to look into this.

Rob

#20

Hi, any updates from Vungle or other developments on this?