Thread: HMODULE from current DLL

Hi all,

I have a DLL which has a method (extractRcdata) which extracts a resorurce from itself. This works perfectly when it's run as a standalone binary, but not as a dll.

Code:

/**
 * 		@brief	Extracts given resource and returns it as a character array.
 *		@param	hMod			Module to extract resource from.
 *		@param	resourceId		ID of the resource to extract
 *		@return	Character array containing the resource.
 */
char* extractRCDATA( int resourceId )
{
	return extractRCDATA( GetModuleHandle(NULL), resourceId );
}
char* extractRCDATA( HMODULE hMod, int resourceId )
{
	HRSRC hRsrc;
	HGLOBAL hResource;
	char* cResource = new char;

	hRsrc = FindResource(hMod, MAKEINTRESOURCE( resourceId ), RT_RCDATA);

	logError( TEXT("FindResource") , GetLastError(), "Errors.log", false );

	hResource = LoadResource(hMod, hRsrc);
	logError( TEXT("LoadResource") , GetLastError(), "Errors.log", false );

	cResource = (char*)LockResource(hResource);
	logError( TEXT("LockResource") , GetLastError(), "Errors.log", false );

	return cResource;
}

When run as an exe, GetModuleHandle(NULL) is used for hMod, so the resource gets extracted just fine. When build as a DLL, GetModuleHandle(NULL) now returns the handle for the exe invoking the DLL, not the DLL itself.

My hope was to use a global variable in my dll wrapper which would be populated from the HINSTANCE of dllMain, like so:

Code:

HINSTANCE hinstDLL_;

/**
 *		@brief	DLL Entry point.
 */
BOOL WINAPI DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved)
{
    switch (fdwReason)
    {
        case DLL_PROCESS_ATTACH:
            // attach to process
            // return FALSE to fail DLL load
            hinstDLL_ = hinstDLL;
            break;
        case DLL_PROCESS_DETACH:
            // detach from process
            break;
        case DLL_THREAD_ATTACH:
            // attach to thread
            break;
        case DLL_THREAD_DETACH:
            // detach from thread
            break;
    }

    return TRUE; // successful
}

myClass myInstance( _hinstDLL );

hinstDLL_ then gets passed to the constructor of my class, and eventually passed to extractRCDATA(). However _hinstDLL seems to be null?

I have two questions;

1: Where am I going wrong with _hinstDLL? - Why is it null?

2: Is there a better way to go about this? - Using C++, no MFC/ATL.

Thanks guys, I've extensively goggled this over the weekend and have concluded that I must be missing something - it can't be as complex as I'm making it out to be

Thanks!

Originally Posted by execute

I'm not sure if I'm just imagining things but is this right?

Code:

HINSTANCE hinstDLL_;
myClass myInstance( _hinstDLL );

That was the only bit I typed out rather than copying & pasting, and I messed it up. The code actually reads

Code:

myClass myInstance( hinstDLL_ );

If I do the following, it works perfectly. But I'd rather not hardcode the filename.

Code:

myClass myInstance( GetModuleHandle("myfile.dll") );

GetModuleHandleEx

Code:

HMODULE GetCurrentModuleHandle()
{
    HMODULE hMod = NULL;
    GetModuleHandleExW(GET_MODULE_HANDLE_EX_FLAG_FROM_ADDRESS | GET_MODULE_HANDLE_EX_FLAG_UNCHANGED_REFCOUNT,
          reinterpret_cast<LPCWSTR>(&GetCurrentModuleHandle),
          &hMod);
     return hMod;
}

Originally Posted by adeyblue

GetModuleHandleEx

Thanks adeyblue. Currently my application runs on Win95+ (other than GetLastError which is 2k+), I'm not keen to introduce dependencies on this newer stuff unless I have to.

Can you think of any other options? I'm still unsure why passing hinstDLL from dllMain isn't working.

adeyblue,

I've given up looking for a decent alternative solution and am trying to use the code, however it errors as follows

file.cpp|75|error: invalid reinterpret_cast from type `HINSTANCE__*(MyNamespace::MyClass::*)()' to type `const WCHAR*'|

Code:

Code:

HMODULE MyClass::getCurrentModuleHandle()
{
	HMODULE hMod = NULL;
	GetModuleHandleExW(
		GET_MODULE_HANDLE_EX_FLAG_FROM_ADDRESS | GET_MODULE_HANDLE_EX_FLAG_UNCHANGED_REFCOUNT,
		reinterpret_cast<LPCWSTR>(&MyNamespace::MyClass::getCurrentModuleHandle),
		&hMod);

	return hMod;
}

Any thoughts? Thanks in advance.

Thanks codeplug, I understand why that code doesn't work now. I've read through the Microsoft document you referred to (I'd seen it before, but was worth re-reading). The document says the ideal DllMain should just be an empty stub (It's safe to say, mine's empty).

If myInstance is constructed before DllMain gets called, what's the best way for me to pass hInstance over to my constructor? Can I construct myInstance as part of DLL_PROCESS_ATTACH? Surely then the scope of myInstance would be local, and not accessible by the methods I have which make use of myInstance and are exported for the DLL?

What's the best way for me to achieve what I'm wanting to do?

Sorry for what are basic questions, I've worked my way through a few different books but haven't had chance to pickup a WIN32 book yet.

Originally Posted by Sebastiani

Maybe something like:

Code:

std::auto_ptr<myClass> myInstance;

BOOL WINAPI DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved)
{
    switch (fdwReason)
    {
        case DLL_PROCESS_ATTACH:
            // attach to process
            // return FALSE to fail DLL load
            myInstance = std::auto_ptr<myClass>(new myClass(hinstDLL));
            break;
        case DLL_PROCESS_DETACH:
            // detach from process
            break;
        case DLL_THREAD_ATTACH:
            // attach to thread
            break;
        case DLL_THREAD_DETACH:
            // detach from thread
            break;
    }

    return TRUE; // successful
}

Thanks, just tried that but the same problem ocured. I've even tried using std:fstream in DllMain to write out some info but nothing happens :/

If I manually do myClass myinstance(getModuleHandle("myDll.dll")) it works as expected, but that obviously means the DLL can never be renamed.

Nope, DLL is the latest version.

My DLL test app is currently c#, which makes debugging the usage a bit of a pain. I'll rewrite my test app in c++ using LoadLibrary rather than DllImport/PInvoke in c# which will hopefully make debugging using gdb a bit easier.

Thanks for all your help thus far guys. I've rewritten my test case to use C++ and get the same problem.

Dllmian is as follows:

Code:

std::auto_ptr<myClass> myInstance;
/**
 *		@brief		DLL Entry point.
 */
BOOL WINAPI DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved)
{
	std::ofstream debugLog("debug.log", std::ios::out | std::ios::app );
	debugLog << "DllMain" << std::endl;

    switch (fdwReason)
    {
        case DLL_PROCESS_ATTACH:
            // attach to process
            // return FALSE to fail DLL load
			myInstance = std::auto_ptr<myClass>(new myClass( GetModuleHandle("myFile.dll") ));

            break;
        case DLL_PROCESS_DETACH:
            // detach from process
            break;
        case DLL_THREAD_ATTACH:
            // attach to thread
			break;
        case DLL_THREAD_DETACH:
            // detach from thread
            break;
    }

    return TRUE; // succesful
}

debug.log never gets created, so DllMain never gets called. I've been looking online for tips on debugging DLLs with GDB but can't find much.

Any thoughts as to why DllMain isn't getting called?

Thanks Codeplug,

I've actually got a full Visual Studio 2008 license for some work that I do so will just install the c++ component and try and compile with VS2008 and see what I get.

Will let you know how I get on

Cheers