«

My application is using the jni interface to link java with c.

but, there is problem to use it. that is,

in side of the jni, I set a callback function to catch errors and
print it. this callback function tasks error informations.
and additionally, I can't expect the time this callback function wil
be called.

this is the simple structure of my code.

java:

ret = SetErrorHandle();

private native SetErrorHandle();

c:
static int generateException(void *aArg)
{
JNIEnv *sEnv = (JNIEnv *)aArg;
jclass sClass = (*aEnv)->FindClass("MyException");

sClass->ThrowNew(sEnv, "Error is here");
}

JNIEXPORT jint SetErrorHandle(JNIEnv *aEnv, jclass aClass)
{
/* this function will be called if there is an error */
/* first argument : function pointer
second argument : argument for function pointer
*/
setCallback(generateException, aEnv)
}

in this code, there are critical problems.
first one is sEnv variable pointer is not available in
generateException function.
whenever this function is called, it generated a segmentation fault
because the pointer is not AVAILABLE.

so my idea is to create new JNIEnv variable dynamically in
generateException function.

how do u think? if it is possible, how APIs I can reference?

thanks u.

On 27/04/2010 3:02 PM, junyoung wrote:
> private native SetErrorHandle();

This isn't legal Java as there is no return type, but from this evidence
it's not static.

> sClass->ThrowNew(sEnv, "Error is here");

That's not the correct syntax, and it's also a terrible error message.
Surely the system you're interfacing to is providing some information of
its own?

> JNIEXPORT jint SetErrorHandle(JNIEnv *aEnv, jclass aClass)

If this method is non-static this is the wrong signature for it. Have
you changed that without re-running javah? Also it's defined as 'jint',
which means that the native method returns 'int', but that's not what
you posted above.

> setCallback(generateException, aEnv)

You can't do that. A JNIEnv* is only valid within the JNI function it is
supplied to. Specifically it is not valid across thread boundaries.

> in this code, there are critical problems.

Yep.

> first one is sEnv variable pointer is not available in
> generateException function.
> whenever this function is called, it generated a segmentation fault
> because the pointer is not AVAILABLE.

.... whatever AVAILABLE means. The real reason is that the JNIEnv* value
you are arranging to supply to it is no longer valid - see above.

> so my idea is to create new JNIEnv variable dynamically in
> generateException function.

You can do that with AttachCurrentThread().