Implementation of hooks is incorrect if other software hooks functions after Windower

[Kaldaien]

It is unlikely that you can do anything about this, but I thought I'd point out a serious flaw in nearly all of the hooks:

Fenestra/core/src/hooks/user32.cpp

Lines 632 to 652 in 8bf697f

	::HHOOK WINAPI SetWindowsHookExW(
	int idHook, ::HOOKPROC lpfn, ::HINSTANCE hMod, ::DWORD dwThreadId) noexcept
	{
	if ((idHook == WH_KEYBOARD_LL || idHook == WH_KEYBOARD) &&
	windower::is_game_module(WINDOWER_RETURN_ADDRESS))
	{
	return std::bit_cast<::HHOOK>(
	std::bit_cast<std::intptr_t>(windower::windower_module()) +
	::InterlockedIncrement(&hhook_counter));
	}
	return hooks::SetWindowsHookExW(idHook, lpfn, hMod, dwThreadId);
	}
	::BOOL WINAPI UnhookWindowsHookEx(::HHOOK hhk) noexcept
	{
	if (windower::is_windower_module(hhk))
	{
	return TRUE;
	}
	return hooks::UnhookWindowsHookEx(hhk);
	}

windower::is_game_module(WINDOWER_RETURN_ADDRESS)

That assumes that the return address from your hook is always the thing that originally called the function. However, if another DLL hooks SetWindowsHookW after you do, then the return address is the other DLL. The return address of the other DLL's hook, assuming no other hooks, is the original caller.

You can't implement this thing this way, without properly walking the call stack to compensate for other hook libraries (i.e. ReShade), this is a disaster waiting to happen. Said disaster has occurred with ReShade 6.4.0, the mod no longer blocks keyboard input correctly.

Other things were broken before ReShade 6.4.0, but more subtly.

[alphaONE2]

Yeah, I was aware that this implementation wasn't correct when I wrote it. I just did not have the time to write a proper check, and getting the rest of the functionality implemented was a higher priority. I'm no longer active within the Windower dev team, and I don't want to speak for the rest of the team about their priorities. Having said that, we would certainly welcome a PR from you or anyone else.