Callback function translation in C-DLL to java JNA

Question

I am having a problem in mapping/translating a callback function from C-DLL to Java using JNA.

on C header file following callback function is written:

// ! callback function header whenever a data report is received from a device
typedef void (FR_callback_func)(Data_t frame);

The structure of the above Data_t is as follow:

// ! Carries information about one signal.
typedef struct
{
    unsigned char index;
    int isval;
    unsigned short val;
    int arr_Length;
    unsigned char array[8];
} Data_t ;

The function in which Data_t structure is getting called:

int getData(int val,Data_t *data);

Now I translated in my JAVA code which is as follows:

public interface device extends Library 
{
    public interface FR_callback_func extends Callback
    {
        void invoke(Data_t signal);
    }

    public class Data_t extends Structure implements com.sun.jna.Structure.ByReference 
    {
        public static class ByReference extends Data_t implements Structure.ByReference { }
        public byte index;
        public int isval;
        public  short val;
        public int arr_Length;
        public byte[] array = new byte[8];
        @Override
        protected java.util.List<java.lang.String> getFieldOrder()
        {    
            return Arrays.asList(new String[] {"index","isval","val","arr_Length","array"});
        }
    } 

    public int getData (int val,Data_t.ByReference data);
}

Then I tried to use it in my main function which is as follow:

public static void main(String[] args) throws IOException 
{
    Data_t .ByReference data_t = new Data_t .ByReference();
    int data = 0;
    int val = 0;

    device h = (device) Native.load("Library", device.class);

    data = h.getData (val, data_t);
}

My question is that am I translating the above C code correctly ? especially the callback function ? Since the C code can't be manipulated. Hence I have to translate the provided C-DLL code in JAVA.

Your advice will be highly appreciated.

Answer 1

Did a rough writeup of the needed bits you are still missing above:

• C example lib

// gcc -c -Wall -Werror -fPIC foo.c
// gcc -shared -o libfoo.so foo.o

#include <stdio.h>

typedef struct {
  int x;
  int y;
} Foo;

typedef Foo* (*FooCallback)(int, int);

void printFoo(Foo *foo) {
  printf("foo: %p x: %d y: %d\n", foo, foo->x, foo->y);
}

FooCallback callback = NULL;

void setCallback(FooCallback cb) {
  callback = cb;
}

void runCallback() {
  if (callback) {
    Foo *foo = callback(123, 456);
    printf("foo from callback: %p x: %d y: %d\n", foo, foo->x, foo->y);
  } else {
    printf("callback not set!\n");
  }
}

• java example code

// javac -classpath .:jna.jar Foo.java 
// java -classpath .:jna.jar Foo

import com.sun.jna.Library;
import com.sun.jna.Native;
import com.sun.jna.Structure;
import com.sun.jna.Callback;


public class Foo {
  public interface CFoo extends Library {
    public class SFoo extends Structure {
      public static class ByReference extends SFoo implements Structure.ByReference { }
      public static class ByValue extends SFoo implements Structure.ByValue { }
      public int x;
      public int y;
    }

    void printFoo(SFoo foo);

    public interface FooCallback extends Callback {
      SFoo invoke(int x, int y);
    }

    void setCallback(FooCallback cb);
    void runCallback();
  }

  public static void main(String[] args) {
    CFoo cfoo = (CFoo) Native.loadLibrary("libfoo.so", CFoo.class);

    // test construction of SFoo
    CFoo.SFoo foo = new CFoo.SFoo();
    foo.x = 23;
    foo.y = 42;

    // test auto ByReference
    System.out.println("foo pointer: " + foo.getPointer());
    cfoo.printFoo(foo);

    // callback test
    // no callback set yet
    cfoo.runCallback();

    // set callback
    // declare return value outside to "survive" callback scope
    // (this is needed for ByReference return values)
    CFoo.SFoo foo2 = new CFoo.SFoo();
    System.out.println("foo2 pointer: " + foo2.getPointer());
    CFoo.FooCallback cb = new CFoo.FooCallback() {
      public CFoo.SFoo invoke(int x, int y) {
        System.out.println("values from C x: " + x + " y: " + y);
        // Is it safe to initialize ByValue inside the callback?
        // CFoo.SFoo.ByValue foo2 = new CFoo.SFoo.ByValue();
        foo2.x = x;
        foo2.y = y;
        foo2.write();   // explicitly write to memory needed here
        return foo2;
      }
    };
    cfoo.setCallback(cb);

    // rerun with set callback
    cfoo.runCallback();

    // hack to avoid gc for cb and foo2
    assert cb != null : "Oops";
    assert foo2 != null : "Oops";
  }
}

Output

foo pointer: auto-allocated@0x7f5588284050 (8 bytes)
foo: 0x7f5588284050 x: 23 y: 42
callback not set!
foo2 pointer: auto-allocated@0x7f55882cfa50 (8 bytes)
values from C x: 123 y: 456
foo from callback: 0x7f55882cfa50 x: 123 y: 456

Note that the GC needs some special care if there are callbacks esp. with non primitive data types. Therefore I put in the example a callback returning a struct itself back to C (foo2). Since the object is returned by reference it must survive the C code dealing with it after the callback. I dont know whether JNA correctly moves structs returned by value prior any GC interaction can happen. You'd have to look that up in the JNA code.

Hope this helps.

Edit: On a sidenote - I kinda find JNI easier to work with in conjunction with C, mainly for being closer to the memory (more C friendly). Although it needs slightly more work initially (for the wrapper in C) it pays off later with better performance.