Hello all

Following the StringArrays, example I created the following class:

class LongArrays extends Memory { private final List<Memory> buffers;

public LongArrays(final long[][] arrs) { super(arrs.length * Pointer.SIZE); this.buffers = new ArrayList<Memory>(arrs.length); for (int i = 0; i < arrs.length; i++) { Memory buf = new Memory(arrs[i].length * LONG_BYTES); buf.getByteBuffer(0, buf.getSize()).asLongBuffer().put(arrs[i]); buffers.add(buf); setPointer(i * Pointer.SIZE, buf); } } }

and I load my library as follows:

Map<String, TypeMapper> options = new HashMap<String, TypeMapper>(); DefaultTypeMapper mapper = new DefaultTypeMapper(); mapper.addToNativeConverter(long[][].class, new ToNativeConverter() { public Class<LongArrays> nativeType() { return LongArrays.class; } public Object toNative(final Object obj, final ToNativeContext context) { return new LongArrays((long[][]) obj); } }); options.put(Library.OPTION_TYPE_MAPPER, mapper); MyLib lib = (MyLib) Native.loadLibrary("mylib", MyLib.class, options);

I then call my mapped function that takes a long[][] argument.

From my debugging I can see that the type mapper kicks in and does the conversion.

I can also see this parameter is handled by the (*env)->IsInstanceOf(env, arg, classPointer) case in dispatch() in dispatch.c.

If I print the native address here, I can see that it corresponds to the address of my LongArrays instance.

However, when I arrive in the native function I'm calling, this argument has become null, so it would seem the libffi might be misbehaving. At this point, I don't really know what went wrong, since I can't properly debug the JNA code using Visual Studio, since GCC doesn't generate symbols that this debugger understands.

Any thoughts? I might try to build JNA's native code using Visual Studio so that I can use the debugger all the way through...

Regards,

Albert