Load and call dependent C# DLLs in Jupyter Notebook

I'm setting up a Jupyter Notebook within which a C# DLL is called. This works fine when calling a single self contained DLL, but fails when the DLL depends on another C# DLL.

The call to the DLL is done using ctypes and the UnmanagedExports as explained here: https://stackoverflow.com/a/29854281/11588296

The entry DLL's source code is:

using RGiesecke.DllExport;
using System.Runtime.InteropServices;

namespace SimpleCSharpPython
{
    public class Test
    {
        [DllExport("addDirect", CallingConvention = CallingConvention.Cdecl)]
        public static int TestExport1(int left, int right)
        {
            return left + right;
        }

        [DllExport("addExtern", CallingConvention = CallingConvention.Cdecl)]
        public static int TestExport2(int left, int right)
        {
            return UnderlyingLib.UnderlyingFunc.TestExternExport(left, right);
        }
    }
}

The underlying DLL is:

namespace UnderlyingLib
{
    public class UnderlyingFunc
    {
        public static int TestExternExport(int left, int right)
        {
            return left + right;
        }
    }
}

Finally, the Jupyter Notebook is:

import ctypes
a= ctypes.cdll.LoadLibrary("SimpleCSharpPython.dll")
b= ctypes.cdll.LoadLibrary("UnderlyingLib.dll")
a.addDirect(3, 5)
a.addExtern(3, 4)

The output of the first call (a.addDirect(3, 5)) is 8, as expected. The output of the second call is the following error

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
<ipython-input-4-c4df87cc7b32> in <module>
----> 1 a.addExtern(3,4)

OSError: [WinError -532462766] Windows Error 0xe0434352

Note the notebook and the 2 DLLs are in the same folder, and the 2 DLLs run fine when called by a C# executable in Visual Studio. I've tried multiple build, path, ctypes calls, but no luck.

Any idea would be greatly appreciated. Let me know if any additional information is required.

I am not an expert but i will try this:

• remove the loading of the underlying dll in Jupyter ;
• and ensure the other dll is compiled with a link to the underlying dll so that when Jupyter uses the dll, it does need the underlying dll as only the main dll needs to know it...