dll loaded from system32 instead of local folder - c++

I noticed some trouble with a dll in my software when deployed. I usually include all required 3rd party dlls in the folder of my executable. However, I noticed a dll not being found on some machines even though it is right there. On my development machine, I found out via Dependencies that the dll is taken in C:/windows/system32/. So from that I infer that on other machines it may be searched for in system32 directory and since it's not there, it's not loaded. I'd like to add that the dll is NOT a known dll, i.e. it is NOT present in the registry at HKML\SYSTEM\CurrentControlSet\Control\Session Manager\KnownDLLs. The library I'm talking about is in this case kinect10.dll and I link my code with #pragma comment( lib, "kinect10.lib" ). So my question is why is the local dll (the one next to the exe file) ignored in this case?

Related

C++, Visual Studio 2012, LIBs, DLLs, Includes, Source and Linker

I'm trying to understand what exactly all of these are and how they relate to each other (and most importantly, how to install them).
From what I've read, LIBs are libraries linked during the compilation of my project and DLLs are libraries linked during the runtime of my project.
So for me to use a LIB, I have to have the actual .LIB file somewhere in my computer, go to Project -> Properties -> VC++ Directories and add the path to the file in the Library Directories, and after this I have to go to Linker -> Input -> Additional Dependencies add the .lib name in there, and finally I need to type #include in my code, right?
So, some questions:
When I finish and build the release of my program, will the .exe only run if the target platform has the .lib installed in their PC as well? If yes, what steps do I need to do to make sure the .lib goes with the .exe?
When I get the source of a open source project, if I add them (using Add Existing Item...) to my project, can I use them just by using #include as if the files were mine and it would be the same as having the .lib installed? Or do I need to install the .lib file and still use these source files?
I have a project using OpenGL and I linked to glew32.lib, but I don't have the lib or any new directory added in the VC++ Directories, so I think this means I must've installed the .lib in the system folder or somewhere where the Visual Studio won't ask for another directory, should I worry about this when releasing a project?
How the above questions relate to DLLs and is there any reason why should I use DLLs over LIBs or the other way around?
I'm starting to use more and more libraries and I noticed I just dragged, copied and included it everywhere so I could use them but never really understood how they "fit" in the project. Especially those open source libraries where they provide so many files and I don't really know what to do with them...
You don't need to have LIB files along with your EXE file for running in another computer, LIB files are static files and DLL files are dynamic. So when you compile all static codes will be included in your EXE file, but DLL files will be loaded and used dynamically in runtime, so you just need to have your DLL files with your EXE file. This way, your code will work and run properly in other computers.
Just adding another project is not enough, you need to compile them and generate LIB files out of them. Then you add the generated LIB file to your final project and include external projects in your final binary. If you are compiling multiple projects together in a solution, you'll need to set project build order in solution properties in VS.
No, that's OK. It seems you've put LIB files in right folder and you don't need to have LIB file with your EXE file to run it in other computers.
DLLs are dynamic libraries, so you need to have them with your application. Installers usually install EXE files with DLL files in the same folder, so your app will run properly, but no need to include LIB files at all.
Also you can include LIB files like this:
#pragma comment(lib, "glew32.lib")
So you don't need to do it in project settings, but assuming you have your LIB file in "Library Directories" path.
Using DLL files can be done in two ways:
One is linking your application to DLL file and having DLL file's function entry in your EXE file's import table:
like using
#include <windows.h>
then
GetWindowsDirectory(windir, MAX_PATH);
So you'll have GetWindowsDirectory API entry in your EXE file's Import Table.
Also you can do it dynamically:
hinstDLL = LoadLibrary("kernel32.dll");
if (hinstDLL != NULL)
{
func_GetWindir = (DLLPROC) GetProcAddress(hinstDLL, "GetWindowsDirectoryA");
...
There is not much difference, only difference is:
In first method, as it's in your EXE file's Import Table, if there was no kernel32.dll or there was no GetWindowsDirectory entry in kernel32.dll, your EXE will not run at all, it will show a critical error and will not run. But in dynamic way (second way), your app will run, but as soon as your code try to use GetWindowsDirectoryA API, it will fail. You will have 0x00 in func_GetWindir. If you attempt to call it, then program will crash.

DLL that depends on other DLLs?

Static linking is not an option.
Let's say that I have an executable that relies on a DLL. One solution is to ship the DLL in the same folder as the executable. Now let's say that I need to inject a DLL into a process that relies on a DLL. Because the DLL is injected, where would I put the DLLs that it relies on? In the same folder? Or in the directory of the process that is injected to?
DLLs depending on one or more other DLLs is not something special. Even a trivial DLL will have dependencies on Windows shared components which are residing in other DLLs. A good example of these "shared components" would be Kernel32.dll and the CRT DLL such as MSVCR80.DLL etc.
You can find out exactly which other DLLs your DLL or EXE requires on by invoking the Dependency Walker. To do that, just run depends.exe from a Visual Studio Command Prompt and drag-and-drop the DLL of interest into the Window that appears. In case you don't have dependency walker available, you can download it from the above link.
I'm not sure of the DLL injection stuff, but it should generally be sufficient if you place all your (other DLL) dependencies in the same folder as your DLL, which would be the folder in which the EXE loading these DLLs resides.
Eg: If C:\test\foo.exe requires bar.dll (which in turn requires baz.dll, assuming baz.dll is not a standard windows shared component), then you would place both bar.dll and baz.dll in C:\test.
There is a lot more to how the OS determines which DLL to load, since multiple versions of the same DLL may exist at various locations and MSDN has a helpful article on the search order for dynamic linked libraries.

How can I automatically load DLLs from a subdirectory?

In Visual Studio, you create a .dll project and it creates a .dll and .lib files. You link statically to the .lib, and if the .dll is in the same folder as the .exe, everything works.
I suspect that everything would also work if the .dll was in System32 or any other PATH folder (confirm or correct, please).
But here's the question: I want my exe to find the .dll in ./DLLS/ folder, that is, if my exe is in ....../MyApp/MyApp.exe then it should look for the .dll in ...../MyApp/DLLS/MyDll.dll. I DO NOT want to include the DLLS folder in path. Is there any way I can do this?
Please note that I do not wish to use LoadLibrary explicitly, so I can't specify the path there.
Thanks in advance for any help.
Here is the default sequence which Win32 applications go through when looking for a DLL:
http://msdn.microsoft.com/en-us/library/7d83bc18(VS.80).aspx
So according to this, another approach might be to call SetCurrentDirectory or SetDllDirectory. But in order for this to work you have to use the Delay Loaded Library functionality (you specify that in Project Settings in Visual Studio). Delay loaded library means that the DLL is loaded only at the moment when the program needs it, not automatically at the programs' startup.
You could use SetDllDirectory for this. The loader will use the additional directory you specify when loading libraries. There can only be one additional directory, however, so you need to make sure that there aren't other calls to this at a later point in your application, otherwise the directory you specify will be ignored.
If that API does not allow relative directories (I'm not sure), you could always call GetModuleFileName with a NULL first parameter to get the file name of the currently executing program. With a little string manipulation, you can get the absolute path to your DLLs folder.

Infernal Libraries (aka DLL Hell)

In a Project of mine, I use a Delphi Application which dynamically loads a wrapper DLL (exporting C-Style functions) which in turn is statically link against a bunch of 3rd party DLLs.
It works fine on my test machines, but on my customers computer it failed to initialize with an error Message like "Couldn't find entrypoint _somefunction#4AKKZ in TMYlibrary.dll".
After some investigation with sysinternal's process monitor, I realized that Windows would look fror DLLs in windows/sytem32 first, so if a DLL named similar to my DLL was present in system32, windows would pick that one and try to find my function entry points in it - which would fail.
Do you know of a possiblity to change windows' DLL the searching behaviour?
Additional Information
[Update] The .exe file is located on the top level of the application's folder tree.
The Wrapper and the 3rd-party-DLLs ar e both located in the Subfolder /bin of my apps Folder
Dev platform is windows XP/7, using VS2008 for the dlll and Delphi 2010 for the application
I found another solution myself:
SetDllDirectory adds an additional search path to the list of locations to look at.
From http://msdn.microsoft.com/en-us/library/ms686203%28v=VS.85%29.aspx
After calling SetDllDirectory, the DLL search path is:
The directory from which the application loaded.
The directory specified by the lpPathName parameter.
The system directory. Use the GetSystemDirectory function to get the
path of this directory. The name of
this directory is System32.
The 16-bit system directory. There is no function that obtains the
path of this directory, but it is
searched. The name of this directory
is System.
The Windows directory. Use the GetWindowsDirectory function to get
the path of this directory.
The directories that are listed in the PATH environment variable.
(maybe i should do my googling before I post on SO ;)
Ship the DLL in your program's folder. (same as the exe file).
Then Windows should try your version first.

How to "add reference" in C++

I'm new to C++ and there's something I just completely don't get. In C#, if I want to use an external library, log4net for example, I just add a reference to the log4net DLL and its members are automatically available to me (and in IntelliSense). How do I do that in non-managed C++?
Often, the library comes with 1) a header file (.h) and 2) a .lib file in addition to the .dll.
The header file is #include'ed in your code, to give you access to the type and function declarations in the library.
The .lib is linked into your application (project properties -> linker -> input, additional dependencies).
The .lib file usually contains simple stubs that automatically load the dll and forward function calls to it.
If you don't have a .lib file, you'll instead have to use the LoadLibrary function to dynamically load the DLL.
The basic concept is the following:
There are 2 types of libraries: static & dynamic. The difference between them is that static libraries, during the linking build step, embed their compiled code in your executable (or dll); dynamic libs just embed pointers to the functions and instructions that some dll should be loaded when program is going to be loaded. This is realized for you by the linker.
Now you can decide which of those two you are going to use. DLLs have many advantages and disadvantages. If developing a huge application it might be worthy to consider using DLLs with delay loading instead of static lib's. Some libs are simply delivered to you as DLLs and you have no choice. Anyway the easiest way for a beginner would be to use static libraries. That would make your deployment and test much easier, since, when dealing with DLL you have to ensure that they are found at runtime (even when using debugger), this involves either copying everything in one directory or dealing with path variables.
Usually a DLL provider (if it is intended that you should be able to deal with the library) delivers you a header file(s) and a .lib which contains the calls into the desired DLL. Some vendors (e.g. boost) only require you to include the header file and the lib is automatically linked to your executable (can be achieved through compiler prorietary pragma directive). If it is not the case you must go into the project settings of the C++ project (project properties/Configuration Properties/Linker/Input) and enter the lib file name into the "Additional Dependencies" row, e.g. iced.lib; iceutild.lib. You can also put fully qualified path names there. Be aware that you have to enter the lib file names for both configurations (Debug, Release). This is the procedure you do with static libraries and Dll equally. The only difference that DLL will require a DLL lib to be either in you app-directory or in one of the path-directories.
After that step, you still might get compiler errors if you try to link incompatible libraries. There are many reasons, why they can be incompatible. But try to first link the lib this way and see if works. If not, post again your errors here ;)
Include file(s) is(are) used to be included in places, where you would like to use smth. from the lib. Just include it and the compiler will know that the symbols must come either from another (compiled) compilation unit (compiled cpp-file=>object file) or the .lib. It will make the look up and notify you if the required symbols are not found.
Good Luck,
Ovanes
P.S. This might be hard in the beginning, but when you get used to it, it will be easy.
C++ doesn't have libraries in the sense you're thinking of. It has header files that you #include, and it has things called libraries that the linker deals with, which contain the compiled code. You need to add the libraries (.LIB files) to the linker settings.
On Windows if you're using a DLL, ideally you should have a .LIB file to go with it that is called the Import Library for the DLL, and you add that .LIB file to your linker settings.
The first thing you need to do is to #include the header file that describes the functions that are available in that library.
The actual code for the library will be in one of 2 places:
A static library (.lib)
A dll (.dll)
Depending on how the library's code is given to you (as .lib files, or as a .dll), you'll have to either:
#pragma comment( lib, "libraryname.lib" ) if its a .lib
LoadLibrary if its a .dll
Sometimes a package comes with BOTH a .lib file that you need to link to, and a .dll file. In this case you don't need to call LoadLibrary, you only need to #pragma comment( lib, "libaryfile.lib" ) because in this case the .lib links you into the .dll.
A very important detail is to put the DLL where your application can find it. Charles Petzold says:
When Windows needs to load a DLL module before running a program that requires it, the library file must be stored in the directory containing the .EXE program, the current directory, the Windows system directory, the Windows directory, or a directory accessible through the PATH string in the MS-DOS environment. (The directories are searched in that order.)
Programming windows, 5th ed
MSDN
I don't recommend using the project properties menu to link because it isn't as visible what libraries you're linking to.
See also