Native Extensions

**AGulev** · May 10, 2017, 10:59am

Do you have any approximate estimates about implementation supports of *.aar? (just want to understand should I work with jar, or have sense to wait)

**Mathias_Westerdahl** · May 10, 2017, 12:02pm

Sorry, no good estimate other than that it’ll probably happen after all platforms are supported. We rarely change the manifest though so you should be able to start using jars for now.

**AGulev** · May 15, 2017, 8:56am

One more question:
How to use imports in java from my jar file?
For now I receive message that package not exist, but my jar file in “lib/android” folder.
(unfortunately I can’t find information about ext.manifest parameters for java sources compilation)

@Mathias_Westerdahl can you help me with it? This is my blocker for now (

**sven** · May 15, 2017, 9:42am

I would think you should be able to just reference the package name + class in your Java file. Say you have a Jar which has the following class and package; com.agulev.MyClass, then you should be able to import it in your Java as;

import com.agulev.MyClass;
...

If this does not work, let us know and we will try to debug it here.

**AGulev** · May 15, 2017, 10:01am

I try to work with alredy builded classes.jar by unity https://github.com/Unity-Technologies/unity-ads-android/releases/download/2.1.0/unity-ads.zip
It works without any problem with JNI, but for callbacks I need to create my own java classes with custom code using the same classes.jar.

When I try to
import com.unity3d.ads.UnityAds;

I recieve an error that package com.unity3d.ads does not exist.

**Mathias_Westerdahl** · May 16, 2017, 10:52am

Here’s a small update from the native extension support.
We’ve added support for HTML5 and we anticipate it being shipped in the 1.2.105 release.

As you might be aware, we are using Emscripten to generate our Html5 builds from C++. So, you’ll be able to use C++, JavaScript and JavaScript libraries which is Emscriptens glue between C++ and JavaScript.

You’ll be able to make function calls in both directions (C++ <-> JavaScript) and here is a screen shot of the simple example app, demonstrating the features:

**sven** · May 16, 2017, 12:48pm

We have now deployed a new build server with fixes related to Java import errors.

**ross.grams** · May 16, 2017, 12:50pm

“Small update”? This is awesome! When is the .105 release scheduled for? We’re still on .103, right?

**Mathias_Westerdahl** · May 16, 2017, 1:04pm

Hopefully, the 1.2.104 should be released tomorrow, and the 105 release is scheduled for May 29’th.
(I still have some code cleanup and code reviews to go through )

**AGulev** · May 16, 2017, 2:23pm

Cool! All works good! Thank you!

**Oleg_The_Evangelist** · May 17, 2017, 8:03am

This is osom! I wonder who;d be the first with a Defold instant game on the Facebook =]

**Christian_Ostman** · May 17, 2017, 10:39am

I would love to - but i think instant games is still only for selected companies?

**Pkeod** · May 17, 2017, 10:41am

I think I tried to sign up for that before but never heard back from Facebook. Just tried the form to sign up again and it’s giving an error. ¯\_(ツ)_/¯

**Oleg_The_Evangelist** · May 17, 2017, 11:37am

there’s a King power. Please reach out to me at oleg.pridiuk@king.com and I shall make the intro into FB so you get hooked up with FB instant games.

**Mathias_Westerdahl** · May 23, 2017, 2:10pm

Another small update…

I’ve cleaned up our AdMob example and it’s now published online:

It works on iOS and Android, for Banners, Interstitials, Rewarded Videos and Native Express ads.

Note that there is one (quite big) known issue for Banners on iOS: I got a crash when deleting the banners. I’ve reported it as an issue to the Firebase (AdMob) team, and they are helping me debug it.

It is designed to work as a Defold library, so you can add it as a dependency.

Again, it’s designed as a starting example, but it should still work for most people.

We hope you’ll like it!
/The Defold team

**juette.sven** · May 23, 2017, 2:14pm

Thank you right in the time when i need it

**Mathias_Westerdahl** · May 23, 2017, 2:15pm

Nice! Take it for a spin and let us know how it works out for you!

**Mathias_Westerdahl** · May 26, 2017, 2:42pm

Some small news, we’ve finally started on the Windows support for Native Extensions! \o/

Have a nice weekend!

**AGulev** · May 26, 2017, 2:54pm

Hi @sven , @Mathias_Westerdahl I have an issue and I wasted a whole day trying to understand whats wrong… and… nothing…
I really need your help.
Did you try to use callbacks with android native extensions?
I have a problem with my DefUnityAds extension on Android (On iOS it works without any problems)

It crashes in situation when I try to recieve few callbacks.
It is easy to reproduce with standart DefUnityAds example:

just run an example on android
then tap to Initialize button
then fast tap 2-3 times to ShowAd button.

On iOS it receives few errors by callback - it is internal error of ads skd like “not initialized” and so on… it’s ok.
But on Android it crashes.

The crash is in lua part, something like:

Build fingerprint: 'Xiaomi/nikel/nikel:6.0/MRA58K/V8.2.1.0.MBFMIDL:user/release-keys'
Revision: '0'
ABI: 'arm'
pid: 14358, tid: 14373, name: Thread-13844  >>> com.agulev.test <<<
signal 6 (SIGABRT), code -6 (SI_TKILL), fault addr --------
Abort message: '../src/gameobject/comp_script.cpp:212: dmGameObject::UpdateResult dmGameObject::CompScriptUpdate(const dmGameObject::ComponentsUpdateParams&): assertion "top == lua_gettop(L)" failed'
    r0 00000000  r1 00003825  r2 00000006  r3 f342c978
    r4 f342c980  r5 f342c930  r6 00000000  r7 0000010c
    r8 ee91e1d0  r9 f342c56c  sl 00000000  fp eea001b0
    ip 00000006  sp f342c4d8  lr f7184ebd  pc f71872ac  cpsr 400e0010

backtrace:
    #00 pc 000432ac  /system/lib/libc.so (tgkill+12)
    #01 pc 00040eb9  /system/lib/libc.so (pthread_kill+32)
    #02 pc 0001c81b  /system/lib/libc.so (raise+10)
    #03 pc 000199cd  /system/lib/libc.so (__libc_android_abort+34)
    #04 pc 00017580  /system/lib/libc.so (abort+4)
    #05 pc 0001b42f  /system/lib/libc.so (__libc_fatal+16)
    #06 pc 00019a55  /system/lib/libc.so (__assert2+20)
    #07 pc 000c6b2c  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_ZN12dmGameObject16CompScriptUpdateERKNS_22ComponentsUpdateParamsE+212)
    #08 pc 000b5d90  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_ZN12dmGameObject6UpdateEPNS_10CollectionEPKNS_13UpdateContextE+764)
    #09 pc 0009ecfc  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_ZN8dmEngine4StepEPNS_6EngineE+1268)
    #10 pc 00101e5c  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_ZN10dmGraphics18RunApplicationLoopEPvPFvS0_EPFiS0_E+20)
    #11 pc 000a152c  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_ZN8dmEngineL7InitRunEPN15dmEngineService13EngineServiceEiPPcPFvPNS_6EngineEPvES9_S7_+168)
    #12 pc 000a15cc  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_ZN8dmEngine6LaunchEiPPcPFvPNS_6EngineEPvES6_S4_+72)
    #13 pc 0009d964  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_Z11engine_mainiPPc+112)
    #14 pc 0009bcb4  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (main+24)
    #15 pc 0023cb9c  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (_glfwPreMain+236)
    #16 pc 0009bcdc  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (android_main+20)
    #17 pc 0023d3bc  /data/app/com.agulev.test-1/lib/arm/libDefUnityAds.so (android_app_entry+176)
    #18 pc 000407bb  /system/lib/libc.so (_ZL15__pthread_startPv+30)
    #19 pc 0001a051  /system/lib/libc.so (__start_thread+6)

It is almost all time assertion problem but in different places.

I thougth that the problem is in my code, but:

It works perfect on iOS;
Second strange thing, I made few logs:

    dmLogInfo("top %i",top);
    lua_rawgeti(L, LUA_REGISTRYINDEX, defUtoLua->listener.m_Callback);
   	lua_rawgeti(L, LUA_REGISTRYINDEX, defUtoLua->listener.m_Self);
   	lua_pushvalue(L, -1);
    dmLogInfo("top1 %i",lua_gettop(L));
    dmScript::SetInstance(L);
    dmLogInfo("top2 %i",lua_gettop(L));
    lua_pushnumber(L, type_of_msg);
    dmLogInfo("top3 %i",lua_gettop(L));
    int count_table_elements = 1;
    if (key_2 != NULL){
        count_table_elements = 2;
    }
    lua_createtable(L, 0, count_table_elements);
    dmLogInfo("top4 %i",lua_gettop(L));
    luaL_push_pair_str_str(L, key_1, value_1);
    if (key_2 != NULL){
        luaL_push_pair_str_num(L, key_2, value_2);
    }
    dmLogInfo("top5 %i",lua_gettop(L));
    int ret = lua_pcall(L, 3, LUA_MULTRET, 0);
    dmLogInfo("top6 %i",lua_gettop(L));
    if (ret != 0) {
        dmLogError("Error running defUtoLua callback: %s", lua_tostring(L, -1));
        lua_pop(L, 1);
    }
    dmLogInfo("top7 %i",lua_gettop(L));

and receive next result:

INFO:DEFAULT: top 0
INFO:DEFAULT: top1 3
INFO:DEFAULT: top2 5 --between top1 and top2 only dmScript::SetInstance(L); - WHY STACK IS SO BIG? O_o
INFO:DEFAULT: top3 1
INFO:DEFAULT: top4 2
INFO:DEFAULT: top5 2

the same trace is on ios almost all time

INFO:DEFAULT: top1 3
INFO:DEFAULT: top2 2
INFO:DEFAULT: top3 3
INFO:DEFAULT: top4 4
INFO:DEFAULT: top5 4

On iOS I have stable trace of stack size all the time.
On Android first and second time - it’s ok, but after then it randomly changes size of stack (it is not all time place between top1 and top2).
It looks like something changing stack in other thread or something like this. I can’t explain and can’t solve this problem (

Sorry for this long post, I hope you understand what I mean.
I reread all c++/lua docs, recheck all methods and ask @britzl about lua_gettop … Nothing helps…

Hope you can help me, thank you.

**Mathias_Westerdahl** · May 26, 2017, 3:07pm

Well, to answer your first question, yes, we use callbacks in the AdMob example: googlemobileads.cpp

As for the discrepancy in the stack, I’m not sure, I’m going to read through your code when I get home tonight.
But, a tip, is to make sure you push/pop the number of expected values in each function.
See the documentation for DM_LUA_STACK_CHECK() and examples here and here

Also, sometimes when forgetting to return the correct value (number of values pushed onto the stack) from the function, it can also misbehave (e.g. crash)