Invocation API (JNI) completely broken

[Martin Kahlert <martin dot kahlert at infineon dot com>]

Hi!
Once upon a time JNI worked for me in 3.1.

Testcase:

#include <jni.h>
#include <stdio.h>

#define PATH_SEPARATOR ':'

#define USER_CLASSPATH "." /* where Prog.class is */

int main()
{
 JNIEnv *env;
 JavaVM *jvm;
 JavaVMInitArgs vm_args;
 jint res;
 jclass cls;
 jmethodID mid;
 jstring jstr;
 jobjectArray args;
 char classpath[1024];

 vm_args.version = JNI_VERSION_1_2;

 JNI_GetDefaultJavaVMInitArgs(&vm_args);

 /* Create the Java VM */
 res = JNI_CreateJavaVM(&jvm, (void**)&env, &vm_args);
 if (res < 0)
    {
     fprintf(stderr, "Can't create Java VM\n");
     exit(1);
    }

 cls = (*env)->FindClass(env, "hello");
 if (cls == 0) {
     fprintf(stderr, "Can't find hello class\n");
     exit(1);
 }

 mid = (*env)->GetStaticMethodID(env, cls, "main", "([Ljava/lang/String;)V");
 if (mid == 0) {
     fprintf(stderr, "Can't find Prog.main\n");
     exit(1);
 }

 jstr = (*env)->NewStringUTF(env, " from C!");
 if (jstr == 0) {
     fprintf(stderr, "Out of memory\n");
     exit(1);
 }
 args = (*env)->NewObjectArray(env, 1, 
         (*env)->FindClass(env, "java/lang/String"), jstr);
 if (args == 0) {
     fprintf(stderr, "Out of memory\n");
     exit(1);
 }
 (*env)->CallStaticVoidMethod(env, cls, mid, args);

 (*jvm)->DestroyJavaVM(jvm);
}

Now it bombs over and over:

The first problem was
clinit_name = init_name = finit_name == NULL
I assume that got in with this change:
2001-10-16  Bryce McKinlay  <bryce@waitaki.otago.ac.nz>

        * name-finder.cc (_Jv_name_finder::lookup): Check for NULL dli_sname.
        
        Eliminate use of C++ static constructors.               
        * interpret.cc: Remove static Utf8Consts. Use namespace gcj.
        * jni.cc: Likewise. 
        * resolve.cc: Likewise.
        * defineclass.cc: Likewise.
        (_Jv_ClassReader::handleClassBegin): Synchronize call to 
        _Jv_RegisterClass.
        * include/jvm.h (void_signature, clinit_name, init_name, finit_name):
        Declare in namespace gcj.
        * java/lang/Class.h (Class): Remove initialization for primitive
        types.
        (friend void _Jv_InitPrimClass): This is in prims.cc.
        * prims.cc (_Jv_InitPrimClass): Do primitive type initialization
        here instead.
        (void_signature, clinit_name, init_name, finit_name): Define in
        namespace gcj.
        (_Jv_CreateJavaVM): Call _Jv_InitThreads, _Jv_InitGC, and
        _Jv_InitializeSyncMutex from here. Initialize Utf8 constants. 
        Initialize primitive types.
        * java/lang/natClassLoader.cc (_Jv_RegisterClasses): Don't call
        initialization routines. Don't synchronize.
        * java/lang/natRuntime.cc (_load): Synchronize on java.lang.Class
        across dlopen call.

I tried to correct that with 
clinit_name = _Jv_makeUtf8Const ("<clinit>", 8);
init_name = _Jv_makeUtf8Const ("<init>", 6);
finit_name = _Jv_makeUtf8Const ("finit$", 6);
before the call to _Jv_JNI_AttachCurrentThread in JNI_CreateJavaVM (jni.cc).
(i copied these lines from prims.cc).

Then the garbage collector bombed from deeply inside _Jv_makeUtf8Const. So i
inserted an _Jv_InitGC() before these lines - cannot heart i assume.
Then the thread creator bombed and i added a _Jv_InitThread() before that.
And because i wanted to avoid another compilation i added the
_Jv_InitializeSyncMutex () call too.

After that the JNI_CreateJavaVM() got through, but _Jv_NewPrimArray() bombs deeply
from inside _Jv_JNI_FindClass().

All that doesn't look good to me.
I am not that confident, that my 'fixes' are correct either, so i leave this
for the more knowledgable people as an exercise ;-)

Thanks for any help,
Martin.

-- 
The early bird catches the worm. If you want something else for       
breakfast, get up later.