How to Callback Java Using JNI From A Global Mouse/Keyboard Hook

SetWindowsHookEx() doesnt fail, as I previously stated, "MessageBox() is always called so I know I have the global hook setup correctly."

The problem is here;

        if (jvm != NULL) {
                if (jvm->AttachCurrentThread((void **)&env, NULL) >= 0) {
                        env->CallVoidMethod(hookObj, processKeyID, (jint)(wParam), true);
                }
        }

This is not executed properly by any process other than java, which means hooked events are only passed on to java when the java app is active, but MessageBox() is called regardless of the active application.

>>"MessageBox() is always called so I know I have the global hook setup
>>correctly."

Err, 'MessageBox()' is called when 'if (hkb == NULL)' evalueates to 'true', thus indicating that 'SetWindowsHookEx()' has failed. See http://msdn2.microsoft.com/en-us/library/ms644990.aspx ("SetWindowsHookEx"):

Return Value

If the function succeeds, the return value is the handle to the hook procedure.
If the function fails, the return value is NULL.

My apologies, we are both referring to different MessageBox() calls, I should have my myself more clear.

JNIEXPORT LRESULT CALLBACK HookKeyboardProc(INT nCode, WPARAM wParam, LPARAM lParam) {
    if (nCode < 0)  // do not process message
                return CallNextHookEx(hkb, nCode, wParam, lParam);
 
        MessageBox(NULL,"key event","Hook", MB_OK | MB_ICONEXCLAMATION);

This is always called regardless of the active window.

To clarify, I am not refering to any code inside the ' if (hkb == NULL) { ' block.

>>This is always called regardless of the active window.

Ah, now I get it, sorry. Since you are establishing your hook as 'global', that is the correct behaviour. If you don't want a global hook, you have to specify a thread ID as the last parameter to 'SetWindowsHookEx()', e.g.

DWORD dwTID = GetCurrentThreadId();

or

// pass the window you want to hook here
DWORD dwTID = GetWindowThreadProcessId(hwnd,NULL);

    hkb = SetWindowsHookEx(WH_KEYBOARD, (HOOKPROC)HookKeyboardProc, hinstDLL, dwTID);

I'm afraid you're missing the point of the question entirely jkr. I do indeed want a global hook, but the current code I have is only passing the events back to java while the java window is the active application. If I were to change focus to say, notepad and type, then MessageBox() is called (so I know I have a global hook, which I want) but ' env->CallVoidMethod(hookObj, processKeyID, (jint)(wParam), true); ' is not called, so java has no idea that the event has occured.

My current understanding is that the global jvm variable is not being shared throughout all processes, but I could be wrong.

Disregard the last comment, I completely missed you here - the problem is that

static JavaVM *jvm = NULL;

cannot be placed in a shared data section, since poiters are only valid inside the address space of the process that initializes them. You will have to either initialize that pointer in each process or think of an alternative method to communicate the messages.

Yes, I indeed did miss the point, but hopefully got back on track again ;o)

You will have to either initialize that pointer in each process or think of an alternative method to communicate the messages.

*ah-ha* right, so,
Question 1: How do I go about initializing the point in each process (as you can probably tell already I am rather weak on c++)

Incidentally I just removed the #pragma data_seg(".HOOKDATA") and I expected to no longer have MessageBox() called due to the hkb variable no longer being defined as shared? But I eperianced exactly the same behaviour as when the directive was there. Im using Bloodshed's Dev C++, v4.9.9.2.

Question 2: Could Dev c++ be ignoring the directive?

I believe 'env->GetJavaVM(&jvm);' is assigning a value to it inside the Java_inputhooks_PollThread_AttachKeyboardHook function.

Visual C++ Express is that free?

According to http://en.wikipedia.org/wiki/Java_Native_Interface you can obtain a handle by calling

JNIEnv *env;
(*g_vm)->AttachCurrentThread (g_vm, (void **) &env, NULL);

for each process, but that in turn would mean to create a JVM for each process your hook would be mapped into, and that for sure is not what you want...

And yes, VC++ Express is free.

Hmm, not an ideal solution, dont particularly want 200 virtual machines fired up hehe.

I'll install VC++ Express and try my previous code with an alternate way of passing the info back to java but would have failed if the #pragma directive was indeed being ignored.

Haha, hours of heartache over a 'non-standard but widely supported preprocessor directive', for the record and anyone else that finds this question, Dev C++, v4.9.9.2 does NOT support #pragma

So the moral of the story is clearly to pick the right tool for the job, hence my choice of accepted solution :)

Also, for anyone interested, the correct solution to this problem can be found here;

http://forum.java.sun.com/thread.jspa?threadID=632369

Good that you got that sorted out! ;o)