Java FFI conversion of arrays is inconsistent between call and return types

[Mark Evenson]

In ​http://article.gmane.org/gmane.lisp.armedbear.devel/3085, Cyrus notes that:

I am trying to connect ABCL up with some java code that uses multidimensional arrays of the form:

        int[][] theArray = new int[4][4];

The problem is that when I call this code I get something like this in return:

	#(#<jarray [I <at> 2d983837 {6492BA43}> #<jarray [I <at> 5d02b84a {5D9131FA}> #<jarray [I <at> 67684413
{6E4D706E}> #<jarray [I <at> 1e107e55 {4CD60DDE}>) 

that is a lisp array that contains 4 java arrays. I can’t pass this off to a method like this:

    static public int[][] arrayTest2(int[][] oldArray) {
      …
    }

[Mark Evenson]

This behavior certainly seems wrong given that for class:

public class Array {
  public static int[][] getMultiDimensional() {
    return new int[4][4];
  }
  
  public static int[][] returnRef(int[][] ref) {
    return ref;
  }
}

Compiled with javac, (java:add-to-classpath “~/directory/“), then the following

CL-USER> (java:jnew-array "int" 4 4)
#<jarray [[I@2e9e32cc {141906FD}>
CL-USER> (jstatic "returnRef" "Array" *)
#(#<jarray [I@4147aa25 {47F3849E}> #<jarray [I@4487c5f9 {2DD68195}>
  #<jarray [I@5dd574b5 {68E14733}> #<jarray [I@1e099b10 {8A346D8}>)

shows that we are clearly getting a different sort of Lisp object than we pass into the Java FFI for what should be a symmetric reflection.

[Mark Evenson]

Patch to test ramifications of not converting multi-dimensional arrays added.

[Mark Evenson]

Closing as "wontfix" 'cuz it isn't broken.

Users who want the unadorned, non-translated result from a call info the Java Function Interface should use the JAVA:*-RAW or JSS's #0"*" methods.