views:

1629

answers:

7

I am using DllImport in my solution.
My problem is that I have two versions of the same DLL one built for 32 bit and another for 64 bit.

They both expose the same functions with identical names and identical signatures. My problem is that I have to use two static methods which expose these and then at run time use IntPtr size to determine the correct one to invoke.

private static class Ccf_32
{
    [DllImport(myDllName32)]
    public static extern int func1();
}

private static class Ccf_64
{
    [DllImport(myDllName64)]
    public static extern int func1();
}

I have to do this because myDllName32 and myDllName64 must be constant and I have not found a way to set it at run time.

Does anyone have an elegant solution for this so I could get rid of the code duplication and the constant IntPtr size checking.

If I could set the file name, I would only have to check once and I could get rid of a ton of repeated code.

+8  A: 

You can probably achieve this with the #if keyword. If you define a conditional compiler symbol called win32, the following code will use the win32-block, if you remove it it will use the other block:

#if win32
    private static class ccf_32
    {
        [DllImport(myDllName32)]
        public static extern int func1();
    }
#else    
    private static class ccf_64
    {
        [DllImport(myDllName64)]
        public static extern int func1();
    }
#endif

This probably means that you can remove the class wrapping that you have now:

    private static class ccf
    {
#if win32
        [DllImport(myDllName32)]
        public static extern int func1();
#else    
        [DllImport(myDllName64)]
        public static extern int func1();
#endif
    }

For convenience, I guess you could create build configurations for controlling the compilation symbol.

Fredrik Mörk
Yeah, but I'd like to keep the 'Any CPU' option as opposed to have 32 and 64 bit versions.
Matt
But you aren't compiling for Any CPU anymore, you now need to target one or the other. This is a good way to do it, as you'll inevitably want to make more of these decisions as you split over CPU's.
Noon Silk
+2  A: 

Why not wrap them into a method?

private static class ccf_32_64
{
    private static class ccf_32
    {
        [DllImport(myDllName32)]
        private static extern int func1();
    }

    private static class ccf_64
    {
        [DllImport(myDllName64)]
        private static extern int func1();
    }

    public static int func1()
    {
        if (32bit)
        {
            return ccf_32.func1();
        }
        else
        {
            return ccf_64.func1();
        }
    }
}
ChaosPandion
That's essentially what I have now :-)
Matt
Well once you wrap it up you don't have to worry about it.
ChaosPandion
+1  A: 

One alternative option is to have both the 32- and 64-bit versions of the unmanaged DLL have the same name, but have them live in separate folders in your build output (say, x86\ and x64\).

Then, your installer or however else you're distributing this is updated so it knows to install the proper DLL for the platform it's installing on.

Zack Elan
A: 

You can't do this the way you want. You need to think of the DllImport attribute as metadata that is used at compile time. As a result you can't change the DLL it is importing dynamically.

If you want to keep your managed code targeted to "Any CPU" then you either need to import both the 32-bit and 64-bit libraries wrapped as two different functions that you can call depending on the runtime environment or use some additional Win32 API calls to late-load the correct version of the unmanaged assembly at runtime and additional Win32 calls to execute the required methods. The drawback there is that you won't have compile time support for any of that type of code for type safety, etc.

Scott Dorman
A: 

Hmm, I'm wondering if you could create an interface and then a class with the methods based on the 32 bit and 64 bit dlls.

I'm not sure if there is an explicit method to determine if you are running 64 bit, but the following might work: allow unsafe code and have an unsafe function that gets a pointer to some address and then determine whether the pointer is 4 or 8 bytes in size. Based on the result determine which implementation of the interface to create.

Brent Scriver
A: 

You can determine whether you are running 64Bits or not by checking the size of the IntPtr type (which is called native int anyways). Then you can load the approriate DLL using an imported LoadLibraryW call, get the function pointer using GetProcAddress, and then, check out Marshal.GetDelegateForFunctionPointer

This not nearly as complicated as it might look like. You have to DllImport both LoadLibraryW and GetProcAddress.

Robert Giesecke
+7  A: 

I prefer to do this by using the LoadLibrary call from kernel32.dll to force a particular DLL to load from a specific path.

If you name your 32-bit and 64-bit DLLs the same but placed them in different paths, you could then use the following code to load the correct based on the version of Windows you are running. All you need to do is call ExampleDllLoader.LoadDll() BEFORE any code referencing the ccf class is referenced:

private static class ccf
{
    [DllImport("myDllName")]
    public static extern int func1();
}

public static class ExampleDllLoader
{
    [DllImport("kernel32", CharSet = CharSet.Unicode, SetLastError = true)]
    private extern static IntPtr LoadLibrary(string librayName);

    public static void LoadDll()
    {
        String path;

        //IntPtr.Size will be 4 in 32-bit processes, 8 in 64-bit processes 
        if (IntPtr.Size == 4)
            path = "c:/example32bitpath/myDllName.dll";
        else
            path = "c:/example64bitpath/myDllName.dll";

        LoadLibrary(path);
    }
}
Josh Sklare
It looks like this solution requires that the user has admin rights. Can someone confirm?
Julio Garcia