Issue 4188 - Update libadblockplus-android to use the latest libadblockplus

libadblockplus-android/jni/JniFilter.cpp

Index: libadblockplus-android/jni/JniFilter.cpp
===================================================================
--- a/libadblockplus-android/jni/JniFilter.cpp
+++ b/libadblockplus-android/jni/JniFilter.cpp
@@ -23,17 +23,19 @@
 {
   return JniLongToTypePtr<AdblockPlus::FilterPtr>(ptr)->get();
 }
 
 static jlong JNICALL JniCtor(JNIEnv* env, jclass clazz, jlong jsValue)
 {
   try
   {
-    return JniPtrToLong(new AdblockPlus::FilterPtr(new AdblockPlus::Filter(JniGetJsValuePtr(jsValue))));
+    AdblockPlus::JsValue *jsValuePtr = JniGetJsValue(jsValue);
+    return JniPtrToLong(new AdblockPlus::FilterPtr(
+      new AdblockPlus::Filter(std::move(*jsValuePtr))));

[sergei, 2016/08/30 09:45:57]

What about using of
return JniPtrToLong(new
AdblockPlus::FilterPtr(dynamic_cast<AdblockPlus::FilterPtr&>(JniGetJsValuePtr(jsValue)));

and similar for Subscription?

We construct C++ Filter and Subscriptions in C++ FilterEngine and can do
whatever we need there, it's better to keep it that way and avoid constructing
of them manually.
- Subscription can have some members and the things get complicated.
- moving of *jsValuePtr can steal the actual value from FilterEngine, although
it's not true now.

However, in that case the internal state of Java Filter or Subscription is also
shared with others but on practice it's not shared now.

PS:
If there are some issues with address arithmetic in dynamic_cast then we should
rather address
https://hg.adblockplus.org/libadblockplus-android/file/tip/libadblockplus-and....

[sergei, 2016/08/30 10:07:31]

Even better without affecting address arithmetic
return JniPtrToLong(new
AdblockPlus::FilterPtr(*JniLongToTypePtr<AdblockPlus::FilterPtr>(jsValue)));
or even
return jsValue;

[anton, 2016/08/30 10:57:29]

Done.

   }
   CATCH_THROW_AND_RETURN(env, 0)
 }
 
 static jobject JNICALL JniGetType(JNIEnv* env, jclass clazz, jlong ptr)
 {
   AdblockPlus::Filter::Type type;
   try