I am trying to subclass the currently focused window on a Windows system using a global CBT hook. This is related to what happens in this question, but the bug is different.
What happens when this subclassing is in effect, is that Opera's (version 10.50) main window is prevented from displaying. Opera has a "splash screen" where you are required to click "Start" for the main window to show that appears after Opera has not shut down properly. Whenever this window pops up, Opera's main window won't show. If Opera was shut down properly, and this splash screen does not show, the main window displays as it should.
HHOOK hHook;
HWND hWndSubclass = 0;
void SubclassWindow(HWND hWnd)
{
Unsubclass();
FARPROC lpfnOldWndProc = (FARPROC)SetWindowLongPtr(hWnd, GWLP_WNDPROC, (LPARAM)SubClassFunc);
SetProp(hWnd, L"PROP_OLDWNDPROC", lpfnOldWndProc);
hWndSubclass = hWnd;
}
void Unsubclass()
{
if (hWndSubclass != 0 && IsWindow(hWndSubclass))
{
FARPROC lpfnOldWndProc = (FARPROC)GetProp(hWndSubclass, L"PROP_OLDWNDPROC");
RemoveProp(hWndSubclass, L"PROP_OLDWNDPROC");
SetWindowLongPtr(hWndSubclass, GWLP_WNDPROC, (LPARAM)lpfnOldWndProc);
hWndSubclass = 0;
}
}
static LRESULT CALLBACK SubClassFunc(HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam)
{
if (message == WM_MOVING)
{
// do something irrelevant
}
else if (message == WM_DESTROY)
{
Unsubclass();
}
FARPROC lpfnOldWndProc = (FARPROC)GetProp(hWndSubclass, L"PROP_OLDWNDPROC");
return CallWindowProc((WNDPROC)lpfnOldWndProc, hWndSubclass, message, wParam, lParam);
}
static LRESULT CALLBACK CBTProc(int nCode, WPARAM wParam, LPARAM lParam)
{
if (nCode == HCBT_SETFOCUS && hWndServer != NULL)
{
SubclassWindow((HWND)wParam);
}
if (nCode < 0)
{
return CallNextHookEx(hHook, nCode, wParam, lParam);
}
return 0;
}
BOOL APIENTRY DllMain( HINSTANCE hInstance,
DWORD Reason,
LPVOID Reserved
)
{
switch(Reason)
{
case DLL_PROCESS_ATTACH:
hInst = hInstance;
return TRUE;
case DLL_PROCESS_DETACH:
Unsubclass();
return TRUE;
}
return TRUE;
}
My suspicion is that Opera's main window is somehow already subclassed. I imagine the following is happening:
The window is created with it's own basic WndProc, and is given focus
My application subclasses the window, storing the original WndProc
Opera subclasses its own window
When the window loses focus, I restore the original WndProc, thus ignoring the second WndProc
Can this really be the case? Are there any other explanations?
This can happen, as Raymond Chen writes:
Consider what would happen if somebody else had subclassed the window during the "... do stuff ..." section. When we unsubclassed the window, we would have removed two subclasses, the one we installed, and the one that was installed after us. If the other subclass allocated memory (which is very common), then that memory got leaked, in addition to the subclass failing to do whatever it was trying to do.
He continues with a solution:
This is quite a cumbersome process, so the shell team wrote some helper functions to do all this for you. The SetWindowSubclass function does all the grunt work of installing a subclass procedure, remembering the previous one, and passing reference data to the subclass procedure you provide. You use the DefSubclassProc function to forward the message to the previous subclass procedure, and when you're done, you use the RemoveWindowSubclass function to remove yourself from the chain. RemoveWindowSubclass does all the work to do the right thing if you are not the window procerure at the top of the chain.
Related
I am using the Windows API on Windows7 with Visual Studio 2019.
I'm trying to set a low level keyboard hook and read the key input to the console, to do so I use the function: HHOOK SetWindowsHookExA( [in] int idHook, [in] HOOKPROC lpfn, [in] HINSTANCE hmod, [in] DWORD dwThreadId )
with the argument: WH_KEYBOARD_LL so I can monitor keyboard events.
This is my main function:
int main()
{
...
HHOOK keyBoardHook = SetWindowsHookExW(WH_KEYBOARD_LL, keyboardHook, NULL, 0);
MSG message;
while (true)
{
while(PeekMessage(&message, NULL, 0, 0, PM_REMOVE) != 0)
{
DispatchMessage(&message);
}
}
...
UnhookWindowsHookEx(keyBoardHook);
return 0;
}
And this is the callback implementation:
LRESULT CALLBACK keyboardHook(int nCode, WPARAM wParam, LPARAM lParam)
{
if (nCode == HC_ACTION)
{
KBDLLHOOKSTRUCT* keyData = (KBDLLHOOKSTRUCT*)lParam;
if (wParam == WM_KEYDOWN || wParam == WM_SYSKEYDOWN)
{
fileHandler.addStr(keyHandler.translate(keyData,true,topWindowThreadId));
}
else if (wParam == WM_KEYUP || wParam == WM_SYSKEYUP)
{
fileHandler.addStr(keyHandler.translate(keyData,false,topWindowThreadId));
}
}
return CallNextHookEx(NULL, nCode, wParam, lParam);
}
How both variables work is not the issue here but just know that
in the callback keyHandler translates the message and fileHandler prints the result to a file.
Here comes the issue. The callback receives:
a key down event when I press a key
a key up event when I release
a repeated key down event when I hold a key
a control key up event VK_LCONTROL when focused window changes
Why does that happen ? This behavior is never mentioned in the MSDN documentation as far as I know. If I switch from console window or any window to any other window this event is generated and a [left control up] message is printed by my keyHandler !
Now to make sure it's not a problem with my code I changed the Callback to this:
LRESULT CALLBACK keyboardHook(int nCode, WPARAM wParam, LPARAM lParam)
{
if (nCode == HC_ACTION)
{
KBDLLHOOKSTRUCT* keyData = (KBDLLHOOKSTRUCT*)lParam;
if (keyData->vkCode == VK_LCONTROL)
{
printf("CTRL");
}
}
return CallNextHookEx(NULL, nCode, wParam, lParam);
}
And still CTRL is printed to the consol whenever I change the focused window, Note that only VK_LCONTROL is generated and wParam is always WM_KEYUP
This is not supposed to happen!
I tried analyzing and modifying my code, tried removing everything from the main function until it looked just like in the example above even rebooted the computer to check if it's an internal issue, nothing helps. and when searching the web for a similar post with a similar issue I found absolutely none, which makes it even more confusing.
Update
I tried compiling someone else's keyboard-hook related code from GitHub, same issue.. is this normal ? Why is it not documented ?
According to the description of the SetWindowsHookExA in the official Microsoft documentation.
The global hooks are a shared resource, and installing one affects all applications in the same desktop as the calling thread. All global hook functions must be in libraries. Global hooks should be restricted to special-purpose applications or to use as a development aid during application debugging. Libraries that no longer need a hook should remove its hook procedure.
HHOOK SetWindowsHookExA(
[in] int idHook,
[in] HOOKPROC lpfn,
[in] HINSTANCE hmod,
[in] DWORD dwThreadId
);
For desktop apps, if this parameter(dwThreadId) is zero, the hook procedure is associated with all existing threads running in the same desktop as the calling thread. That's why after you changed the focused window, CTRL still printed to the console. So it is recommended that you use hook in DLL.
I am having issues with creating a modeless dialog from a DLL file. My dialog has nothing special on it, just an OK button and an edit box. I have looked at this Microsoft KB Article (http://support.microsoft.com/kb/233263) and have implemented its solution to create a window hook to grab and process messages.
The method provided by Microsoft solves the tab key problem, however, it creates another problem. When I type into the edit box on the dialog, whatever I press is duplicated 4 times. For example, if I press 'a' on the keyboard, 'aaaa' will show up in the edit box.
If I disable the Window Hook, then the edit box works correctly and only displays one 'a'.
What do I need to do to the Window Hook procedure to solve this problem?
Any help is greatly appreciated.
- - EDIT - -
As per request, my Window Hook Procedure Code: (It's the same as the KB article)
LRESULT FAR PASCAL GetMsgProc(int nCode, WPARAM wParam, LPARAM lParam) {
LPMSG lpMsg = (LPMSG) lParam;
if (nCode >= 0 && PM_REMOVE == wParam) {
// Don't translate non-input events.
if ((lpMsg->message >= WM_KEYFIRST && lpMsg->message <= WM_KEYLAST)) {
if (IsDialogMessage(hwndDllDlg, lpMsg)) {
// The value returned from this hookproc is ignored,
// and it cannot be used to tell Windows the message has been handled.
// To avoid further processing, convert the message to WM_NULL
// before returning.
lpMsg->message = WM_NULL;
lpMsg->lParam = 0;
lpMsg->wParam = 0;
}
}
}
return CallNextHookEx(NULL, nCode, wParam, lParam);
}
My Dialog Callback Procedure:
BOOL CALLBACK DllDlgProc(HWND hwndDlg, UINT uMsg, WPARAM wParam, LPARAM lParam) {
switch (uMsg) {
case WM_INITDIALOG:
hHook = SetWindowsHookEx(WH_GETMESSAGE, GetMsgProc, NULL, GetCurrentThreadId());
return TRUE;
case WM_COMMAND:
if (LOWORD(wParam) == IDOK || LOWORD(wParam) == IDCANCEL) {
DestroyWindow(hwndDlg);
hwndDllDlg = NULL;
}
return TRUE;
case WM_DESTROY:
UnhookWindowsHookEx(hHook);
return FALSE;
}
return FALSE;
}
}
Both hHook and hwndDllDlg are defined as HHOOK and HWND respectively.
HHOOK hHook;
HWND hwndDllDlg = CreateDialog(0, MAKEINTRESOURCE(DLG_MAIN), 0, DllDlgProc);
I looked at the KB article. It sounds reasonable. There is some point where you was not enough accurate while following the instructions from KB. Post your code. This may help.
If you have control over the message pump of the executable and can add IsDialogMessage there, then you do not need any hook. Code from the dll is part of the code of the process. Window handles are in the common space either.
Other approach is starting your own UI thread. If you create your dialog on this thread, then you will have your own message pump. The hook will not be needed in this case either.
Well, this is more of a question to author of the post..
I have the tab key issue and am trying to understand the microsoft article better.
So my dialog is shipped out as Dll and the application which I don't have access to is launching dialog from my dll.
HWND hwndDllDlg = CreateDialog(0, MAKEINTRESOURCE(DLG_MAIN), 0, DllDlgProc);
I don't understand what dialog the code refers to when they said hwndDllDlg in the article. Should I point my dialog creation to this variable ?
I'll go ahead and give a summary to this, how can I use a dialog procedure that is a member of a class? I am creating a window wrapper class, but CreateDialogParam needs a global dialog procedure, so I tried this workaround:
I have done a bit of searching on this topic. I am making a Dialog class which I am subclassing to make a CMainWnd and then instantiating that. In the Dialog class I have a member function defined as INT_PTR CALLBACK Dialog::cb_proc(HWND,UINT,WPARAM,LPARAM). Now, I know that windows must have a global function as a callback procedure.
So I made a std::map<HWND,Dialog*> DlgProcs map to associate the dialogs window handle with its Dialog class pointer.
And a INT_PTR CALLBACK DlgMainProc(HWND,UINT,WPARAM,LPARAM) so I could pass that to CreateDialogParam(). In the body of DlgMainProc(...) I search the map for using the hWnd parameter to find the Dialog* and return its cb_proc(..) member.
My problem is that none of the messages get processed, this is because the member procedure in my Dialog class never gets called. Even though when I put a MessageBox() in DlgMainProc inside a if (DlgProcs.find(hWnd) != DlgProcs.end()) { statement, the messagebox is displayed, over and over again until I have to abort the program from Visual Studio 2008. Which tells me that it is finding the hWnd in my map. The weird thing is it also does this if I put it in the else statement after that, which contradictingly tells me it is NOT finding the hWnd in the map.
If I put a messagebox in the cb_proc member function it does not get displayed at all. But during this I never get any compiler, linker, or runtime errors. When I remove the messagebox from it (as to not have to abort the program, it was just for debugging purposes) the program runs but no messages get processed, the X button does not close the program, button clicks do nothing.
Here is the PasteBin code: http://pastebin.com/GsGUBpZU
Btw, I have no problem subclassing this, my window is created fine, just no messages are processed, cb_proc just never gets called.
EDIT: Here is the relevant parts of the code
map<HWND,Dialog*> g_DlgProcs;
INT_PTR CALLBACK g_MainDlgProc(HWND hWnd, UINT msg, WPARAM wParam, LPARAM lParam) {
if (g_DlgProcs.find(hWnd) != g_DlgProcs.end()) {
Alert("blah"); // Gets executed repeatedly
return g_DlgProcs[hWnd]->cb_proc(hWnd, msg, wParam, lParam);
} else {
Alert("blah"); // Removing the above alert, this gets
// executed repeatedly, erm, as well.. O.o strange
return FALSE;
}
}
Dialog::Dialog(int id, HWND parent /* = HWND_DESKTOP */) {
_id = id;
_parent = parent;
// Tried this before CreateDialogParam
g_DlgProcs.insert(make_pair(_handle, this));
_handle = CreateDialogParam(
(HINSTANCE)GetModuleHandle(NULL),
MAKEINTRESOURCE(id), _parent,
(DLGPROC)g_MainDlgProc, NULL
);
// Then tried it after CreateDialogParam
g_DlgProcs.insert(make_pair(_handle, this));
}
INT_PTR CALLBACK Dialog::cb_proc(HWND hWnd, UINT msg, WPARAM wParam, LPARAM lParam) {
Alert("blah"); // Never gets executed
bool handled = true;
switch (msg)
{
case WM_INITDIALOG:
OnInitialize();
break;
case WM_COMMAND:
if (HIWORD(wParam) == 0 || HIWORD(wParam) == 1) {
OnMenuCommand((HIWORD(wParam) == 1), (int)LOWORD(wParam));
} else {
OnCtrlCommand((int)HIWORD(wParam), (int)LOWORD(wParam), (HWND)lParam);
}
break;
case WM_NOTIFY:
{
LPNMHDR head = (LPNMHDR)lParam;
OnNotification(head->code, head->idFrom, head->hwndFrom);
}
break;
case WM_CLOSE:
OnClose(); // DestroyWindow(_handle)
break;
case WM_DESTROY:
OnDestroy(); // PostQuitMessage(0)
default:
handled = ProcessMsg(msg, wParam, lParam);
}
// Convert bool to Windows BOOL enum
return ((handled == true) ? TRUE : FALSE);
}
Does anybody know why it never gets called? Or maybe just guide me to another way to use a member function as a DLGPROC?
The standard solution is to pass your this pointer as the last parameter to Create,DialogParam, stash it in DWLP_USER in your WM_INITDIALOG handler, and retrieve it from DWLP_USER thereafter. Basically you use DWLP_USER as your map.
I tried your code and it worked: cb_proc gets called. You will miss any messages (e.g. WM_INITDIALOG) that get sent before CreateDialogParam returns.
You can fix the latter problem by adding the window handle and the object to the map in g_MainDlgProc. If you get a message for an unknown window, you know it belongs to the window you're creating; put the object in a global and you can add the handle/object to the map.
I'm just adding this here in case someone finds it useful; using the magic of C++11 lambdas and templates you can have a simple wrapper template that means you don't have to continually rewrite the boiler-plate code for saving and loading userdata in window and dialog box procedures.
Here's an example for the DialogBoxParam function, but the same technique can be applied to CreateDialogParam and CreateWindowEx as well.
template <typename T, INT_PTR (T::*P)(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam)>
INT_PTR DialogBoxThis(T* pThis, HINSTANCE hInstance, LPCWSTR lpTemplateName, HWND hWndParent)
{
return ::DialogBoxParam(hInstance, lpTemplateName, hWndParent, [](HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam) -> INT_PTR {
if (uMsg == WM_INITDIALOG) SetWindowLongPtr(hWnd, DWLP_USER, lParam);
T* pThis = reinterpret_cast<T*>(GetWindowLongPtr(hWnd, DWLP_USER));
return pThis ? (pThis->*P)(hWnd, uMsg, wParam, lParam) : FALSE;
}, reinterpret_cast<LPARAM>(pThis));
}
You would use it like this:
class MyClass
{
INT_PTR MyDlgProc(HWND hWnd, UINT uMsg, WPARAM wParam, LPARAM lParam);
};
// from inside MyClass, we can show a dialog that uses member function MyDlgProc as the dialog procedure
// note it is NOT a static function
DialogBoxThis<MyClass, &MyClass::MyDlgProc>(this, hInstance,
MAKEINTRESOURCE(IDD_MYDIALOG), hWndParent);
I am using ATL do build my software. I create a main frame window which contain serviral childen windows. the one of the childen windows contain some chinden windows also.
Now, my app will alway get the failed of assert:
virtual ~CWindowImplRoot()
{
#ifdef _DEBUG
if(m_hWnd != NULL) // should be cleared in WindowProc
{
ATLTRACE(atlTraceWindowing, 0, _T("ERROR - Object deleted before window was destroyed\n"));
ATLASSERT(FALSE);
}
#endif //_DEBUG
}
I defined WM_NCDESTROY handler for any window and set bHanded=FALSE, the code can arrived there.
Anymore, I add OnFinalMessage override function. unfortunatelly, no all of the functions can be called. Actually, the top-level window's OnFinalMessage will be called, the middle-level windows' OnFinalMessage may not be called, the bottom-level windows' OnFinalMessage will never be called.
I oberve the source code of ATL:
template <class TBase, class TWinTraits>
LRESULT CALLBACK CWindowImplBaseT< TBase, TWinTraits >::WindowProc(
_In_ HWND hWnd,
_In_ UINT uMsg,
_In_ WPARAM wParam,
_In_ LPARAM lParam)
{
CWindowImplBaseT< TBase, TWinTraits >* pThis = (CWindowImplBaseT< TBase, TWinTraits >*)hWnd;
// set a ptr to this message and save the old value
_ATL_MSG msg(pThis->m_hWnd, uMsg, wParam, lParam);
const _ATL_MSG* pOldMsg = pThis->m_pCurrentMsg;
pThis->m_pCurrentMsg = &msg;
// pass to the message map to process
LRESULT lRes = 0;
BOOL bRet = pThis->ProcessWindowMessage(pThis->m_hWnd, uMsg, wParam, lParam, lRes, 0);
// restore saved value for the current message
ATLASSERT(pThis->m_pCurrentMsg == &msg);
// do the default processing if message was not handled
if(!bRet)
{
if(uMsg != WM_NCDESTROY)
lRes = pThis->DefWindowProc(uMsg, wParam, lParam);
else
{
// unsubclass, if needed
LONG_PTR pfnWndProc = ::GetWindowLongPtr(pThis->m_hWnd, GWLP_WNDPROC);
lRes = pThis->DefWindowProc(uMsg, wParam, lParam);
if(pThis->m_pfnSuperWindowProc != ::DefWindowProc && ::GetWindowLongPtr(pThis->m_hWnd, GWLP_WNDPROC) == pfnWndProc)
::SetWindowLongPtr(pThis->m_hWnd, GWLP_WNDPROC, (LONG_PTR)pThis->m_pfnSuperWindowProc);
// mark window as destryed
pThis->m_dwState |= WINSTATE_DESTROYED;
}
}
if((pThis->m_dwState & WINSTATE_DESTROYED) && pOldMsg== NULL)
{
// clear out window handle
HWND hWndThis = pThis->m_hWnd;
pThis->m_hWnd = NULL;
pThis->m_dwState &= ~WINSTATE_DESTROYED;
// clean up after window is destroyed
pThis->m_pCurrentMsg = pOldMsg;
pThis->OnFinalMessage(hWndThis);
}else {
pThis->m_pCurrentMsg = pOldMsg;
}
return lRes;
}
In above code, the purpose of the variable "pOldMsg" is to indicate the call stack level. if "pOldMsg" not equal to NULL, it's means that the call stack is not returned. when it's NULL, means the the call stack is outermost level, this moment we can call OnFinalMessage function now.
In my case, it's seems that sometimes "pOldMsg" will never equal to NULL.
chat:
MainFrame V
|----- ViewDevices V
|----- ViewContainer X
|----- ViewBooks X
|----- ViewFiles V
After testing it for many times, I find that, if I didn't add any content to the ViewBooks(ListView) window, every things is OK. Once I add just a line, this issue will repro. I don't know whether the sending message by crossing thread cause the issue.
Can anybody give me some suggestions?
I am writing a Windows Api wrapper and I have run into a problem. I am abstracting the Window Procedure (WndProc) and wrote a static MsgHandler which would call the WndProc of my WinHandler class. Here is the code:
LRESULT CALLBACK WinHandle::MsgHandler(HWND hwnd,UINT msg,WPARAM wParam,LPARAM lParam)
{
WinHandle* wnd = 0;
if (msg == WM_NCCREATE)
{
::SetWindowLong(hwnd,GWL_USERDATA,
long((LPCREATESTRUCT(lParam))->lpCreateParams));
}
wnd = (WinHandle*) (::GetWindowLong(hwnd,GWL_USERDATA));
if (wnd)
return wnd->WndProc(msg,wParam,lParam);
else
return ::DefWindowProc(hwnd,msg,wParam,lParam);
}
Since this functions is static , there is no this pointer and it works perfectly with WNDCLASSEX, but for some reason it never calls wnd->WndProc() and always returns the default window procedure. Need Help. What is wrong? Does anyone know a better approach?
Call SetWindowLong when you receive WM_CREATE, not WM_NCCREATE. I don't think lpCreateParams is valid in WM_NCCREATE. That is:
if (msg == WM_CREATE)
{
::SetWindowLong(hwnd,GWL_USERDATA,
long((LPCREATESTRUCT(lParam))->lpCreateParams));
}