3

I'm finding this particular bit of code quite difficult (Not least of which because I only started playing with C a week ago).

I've been trying hard to find the right syntax to correctly create a java string array in C (i.e., an array of jstring objects, i.e. an object which represents an array of jstring objects). I've been using the following resources and from them I've constructed code which compiles. I'm not sure if the error which occurs afterwards is due to the syntax being incorrect or because of a completely separate reason. Since the code is mostly in isolation I'm assuming the syntax is incorrect.

(Suns Native Programming Documentation & Suns JNI documentation)

The code compiles but after passing the "FindClass" line of code a SIGSEGV signal is sent which kills the C process:

jint size = 5;
jclass StringObject = (*env)->FindClass(env, "java/lang/String");
jobjectArray sampleMessage = (*env)->NewObjectArray(env, size, StringObject, NULL);
jobjectArray returnArray = (jobjectArray) (*env)->NewObjectArray(env, messageCount, &sampleMessage, 0);

Could anyone point me to a useful resource for this? Or confirm the syntax is correct.

EDIT

I large part of my problem was that debugging this code caused the problem. I don't have time to narrow down the reproducing factor but stepping over JNI code in a gdb-client through eclipse DOESN'T work.

Volo
  • 28,673
  • 12
  • 97
  • 125
Graeme
  • 25,714
  • 24
  • 124
  • 186

1 Answers1

17

To get a jclass for the row type, you can call GetObjectClass() on one of the rows. This works:

Main.java

public class Main {

    static {
        System.loadLibrary("mynative");
    }

    private static native String[][] getStringArrays();

    public static void main(String[] args) {
        for (String[]  array : getStringArrays())
            for (String s : array)
                System.out.println(s);
    }
}

mynative.c

static jobjectArray make_row(JNIEnv *env, jsize count, const char* elements[])
{
    jclass stringClass = (*env)->FindClass(env, "java/lang/String");
    jobjectArray row = (*env)->NewObjectArray(env, count, stringClass, 0);
    jsize i;

    for (i = 0; i < count; ++i) {
        (*env)->SetObjectArrayElement(env, row, i, (*env)->NewStringUTF(env, elements[i]));
    }
    return row;
}

JNIEXPORT jobjectArray JNICALL Java_Main_getStringArrays(JNIEnv *env, jclass klass)
{
    const jsize NumColumns = 4;
    const jsize NumRows = 2;

    const char* beatles[] = { "John", "Paul", "George", "Ringo" };
    jobjectArray jbeatles = make_row(env, NumColumns, beatles);

    const char* turtles[] = { "Leonardo", "Raphael", "Michaelangelo", "Donatello" };
    jobjectArray jturtles = make_row(env, NumColumns, turtles);

    jobjectArray rows = (*env)->NewObjectArray(env, NumRows, (*env)->GetObjectClass(env, jbeatles), 0);

    (*env)->SetObjectArrayElement(env, rows, 0, jbeatles);
    (*env)->SetObjectArrayElement(env, rows, 1, jturtles);
    return rows;
}

Building, error handling omitted for clarity.

fizzer
  • 13,551
  • 9
  • 39
  • 61
  • A variation of this code works great in my test harness. However, when slotted into my application the line `jclass stringClass = (*env)->FindClass(env, "java/lang/String");` throws a `SIGSEGV`. From all of my checks both the framework and app are setup identically. Not only that i've even made the c file generic enough to copy between the harness and code and it still breaks. Does anyone know what `jclass stringClass = (*env)->FindClass(env, "java/lang/String");` is dependant on? – Graeme May 23 '11 at 14:56
  • Probably http://stackoverflow.com/questions/5991615/unable-to-get-jnienv-value-in-arbitrary-context – fizzer May 23 '11 at 16:47
  • When the code fails, is it being called from a native context as opposed to from Java? If so, see linked question. Where you you get your env pointer? – fizzer May 24 '11 at 10:37
  • Code is being run as native code after being called from Java. *env is passed through the JNI call to method. – Graeme May 24 '11 at 13:11