I am a beginner in C# and I have been asked to create a C# code which uses a C++ DLL.
Functions from C++ DLL need a void* parameter so I send a IntPtr parameter from the C# code.
But at the end of this function my pointer seems to be NULL. Am I missing something?
Here my C# code:
int somme = 0;
IntPtr deviceHandle = new IntPtr();
uint[] SpiConf = new uint[7];
somme = Opening(deviceHandle, SpiConf);
if (deviceHandle == IntPtr.Zero)
{
Console.WriteLine("deviceHandle is NULL"); //
}
And here is my function from the C++ DLL:
int Opening(void* deviceHandle, unsigned char SpiConf[])
{
wchar_t devPath;
unsigned long devPathsize = 0;
unsigned short VID = 0x4d8;
unsigned short PID = 0xde;
int res;
deviceHandle = Mcp2210_OpenByIndex(VID, PID, 0, &devPath, &devPathsize);
res = Mcp2210_GetLastError();
if (res != E_SUCCESS)
{
//Console.WriteLine("Failed to open connection");
return -1;
}
if (deviceHandle == NULL)
{
return -2;
}
return 0; // This function returns 0
}
Any help would be very appreciate.
Related
I have a C++ dll in 64-Bit, which contains a function that returns an LPStr. I would like to call this function in C#. The function declaration looks like this:
__declspec(dllexport) LPSTR __stdcall function(int16_t error_code);
In my C# code I have tried the following:
[DllImport(#"<PathToInterface.dll>", EntryPoint = "function")]
[return: MarshalAs(UnmanagedType.LPStr)]
public static extern string function(Int16 error_code);
And then in the program:
string ErrorMessage = "";
ErrorMessage = function(-10210);
I know that the function itself is good, as I can call it from another program (written in LabVIEW FWIW). But when I execute the C# Program, it just exits with error code 0x80000003, I can't even try, catch the exeption.
How do I call this function properly?
As a side node: I do have other functions in this dll, that use LPStr as parameters, which I can call without a problem. It is only two functions that return LPStr that make problems
How do I call this function properly?
As interop? you can't ... it is also error prone in plain C++
you should rather do it like
extern "C" __declspec(dllexport) int __stdcall function(int16_t error_code,
LPSTR buffer, size_t size)
{
LPCSTR format = "error: %i";
size_t req = _scprintf(format, error_code); // check for require size
if (req > size) //buffer size is too small
{
return req; //return required size
}
sprintf_s(buffer, size, format, error_code); //fill buffer
return 0;
}
And usage
class Program
{
static void Main(string[] args)
{
short error_code = -10210;
var ret = function(error_code, null, 0); // check for required size of the buffer
var sb = new StringBuilder(ret); // create large enough buffer
ret = function(error_code, sb, (uint)sb.Capacity + 1); //call again
var error_desc = sb.ToString(); //get results
Console.WriteLine(error_desc);
Console.ReadKey();
}
[DllImport("TestDll.dll", EntryPoint = "function", CharSet = CharSet.Ansi)]
public static extern int function(short error_code, StringBuilder sb, int size);
}
usage in C++
typedef int (__stdcall *function)(int16_t error_code, LPSTR buffer, size_t size);
int main()
{
auto handle = LoadLibrary(L"TestDll.dll");
auto proc = (function)GetProcAddress(handle, "_function#12");
// of course it can be done via linking
int16_t error_code = 333;
const int ret = proc(error_code, NULL, 0); // check for size
CHAR* buffer = new CHAR[ret + 1];
//CHAR buffer[200]; //eventually allocate on stack
//but size have to be constant value and may be too small
proc(error_code, buffer, ret+1); // call again
MessageBoxA(0, buffer, "Return", 0); //show result
delete[] buffer; //free buffer!
FreeLibrary(handle);
return 0;
}
What will be the C# declaration of the following C++ methods of an unmanaged dll:
long WINAPI MT123_GetVersion(HANDLE hComHandle, BYTE *Version, int &len);
long __stdcall MT123_GetStatus(HANDLE hComHandle, BYTE *Saddr, BYTE &status)
long __stdcall MT123_GetTagData(HANDLE hComHandle, BYTE *Saddr,BYTE &status,
BYTE *Data, int &len);
with DllImport attribute?
Generally this:
long WINAPI MT123_GetVersion(HANDLE hComHandle, BYTE *Version, int &len);
Becomes:
int MT123_GetVersion(IntPtr hComHandle, IntPtr pVersionBuffer, ref int len)
The pVersionBuffer parameter must be initialised using Marshal.AllocCoTaskMem. You should use Marshal.Copy to copy the data to a managed array, then free the buffer using Marshal.FreeCoTaskMem.
I suggest you write a wrapper function to manage the nitty-gritty. For example
how much memory do you need to allocate: Is this one where you call the function twice, first time passing null to retrieve the length? Or does it have a reasonable maximum size you can hard-code?
What's the meaning of the return value: Is it an error code you should turn into an exception?
E.g:
byte[] GetVersionWrapper(IntPtr hComHandle)
{
IntPtr pVersionBuffer = IntPtr.Zero;
try
{
int len = 0;
// First retrieve the length
int status = MT123_GetVersion(hComHandle, IntPtr.Zero, ref len);
if(status != 0) throw new Exception("message here");
if(len < 0 || len > 1024) throw new Exception("message here");
// Now allocate a buffer of the given length.
pVersionBuffer = Marshal.AllocCoTaskMem(len);
// Now retrieve the version information into the buffer
int status = MT123_GetVersion(hComHandle, pVersionBuffer, ref len);
if(status != 0) throw new Exception("message here");
// Now copy the version information to a managed array.
byte[] retVal = new byte[len];
Marshal.Copy(pVersionBuffer, retVal, 0, len);
// Return the managed array.
return retVal;
}
finally
{
// done in Finally in case anything above throws an exception.
Marshal.FreeCoTaskMem(pVersionBuffer);
}
}
I'm writing a DLL in C++ to search a subpicture in a Bitmap.
When the C++ Part is executed, the HBITMAP is not valid.
Here is my code:
C#:
[System.Runtime.InteropServices.DllImport("FindSubPicture.dll", EntryPoint = "FindSubPictures", CallingConvention = CallingConvention.Cdecl)]
private static extern System.IntPtr FindSubPicturesFct(System.IntPtr mImg, System.IntPtr sImg, int* nMatches);
public static System.Collections.Generic.List<TPoint> FindSubPictures(System.Drawing.Bitmap mImg, System.Drawing.Bitmap sImg)
{
TPoint* PStack = null;
int nMatches = 0;
System.Collections.Generic.List<TPoint> MyList = new System.Collections.Generic.List<TPoint>();
MyList.Clear();
PStack = (TPoint*)FindSubPicturesFct(mImg.GetHbitmap(), sImg.GetHbitmap(), &nMatches);
if(PStack == null) { return MyList;}
for (int i = 0; i < nMatches; i++) { MyList.Add(new TPoint(PStack[i].x[0], PStack[i].x[1])); }
try
{
System.Runtime.InteropServices.Marshal.FreeHGlobal((System.IntPtr)PStack);
}catch(System.Exception ex) { System.Console.WriteLine(ex.Message); }
return MyList;
}
C++:
#include "FindSubPictures.h"
#include <stdio.h>
#include <stdlib.h>
#include <Windows.h>
#include <omp.h>
struct TPoint
{
int x[2];
TPoint(int fx, int fy) {
x[0] = fx;
x[1] = fy;
}
void Reset() { x[0] = 0; x[1] = 0; }
void Set(int fx, int fy) { x[0] = fx; x[1] = fy; }
};
extern "C" __declspec(dllexport) TPoint* FindSubPictures(HBITMAP mImg, HBITMAP sImg, int* nMatches) {
int mImgWidth = -1, mImgHeight = -1, sImgWidth = -1, sImgHeight = -1;
TPoint* MyList = nullptr;
if (mImg == nullptr || sImg == nullptr || nMatches == nullptr) { return nullptr; }
return MyList;
}
Well, the C++ Library function is not doing anything until now. That's because the HBITMAP is not valid. In C#, the System.Drawing.Bitmap is valid.
When I enter "mIng.", no auto-completion is available.
Am I missing anything?
Obtaining a bitmap does not mean you can access it like a data field. You'll have to lock the bitmap data in memory first.
See also the answers in this thread here.
Bitmap.LockBits
Code would look as follows:
Rectangle rect = new Rectangle(0, 0, mImg.Width, mImg.Height);
System.Drawing.Imaging.BitmapData bmpData =
mImg.LockBits(rect, System.Drawing.Imaging.ImageLockMode.Read,
mImg.PixelFormat);
// Get the address of the first line.
IntPtr ptr2mImg = bmpData.Scan0;
Ok, I have the solution.
HBITMAP is only a handle "void *"
You first have to get the complete object by the function GetObject
Code:
extern "C" __declspec(dllexport) TPoint* FindSubPictures(HBITMAP mImg, HBITMAP sImg, int* nMatches) {
int mImgWidth = -1, mImgHeight = -1, sImgWidth = -1, sImgHeight = -1;
TPoint* MyList = nullptr;
if (mImg== nullptr || sImg== nullptr || nMatches == nullptr) { return nullptr; }
BITMAP mBMP, sBMP;
GetObject(mImg, sizeof(BITMAP), &mBMP);
GetObject(sImg, sizeof(BITMAP), &sBMP);
/*Now, BITMAP mBMP and sBMP are valid*/
return MyList;
}
I have a c++ method which creates, fills and returns SAFEARRAY:
SAFEARRAY* TestClass::GetResult(long& size)
{
return GetSafeArrayList(size);
}
How should I export that function in a DLL so that c# could take it
How should I write c# method signature?
I have in c++ something along these lines:
extern "C" __declspec(dllexport) void GetResult(SAFEARRAY*& data, long& size)
{
size = 0;
data = handle->GetResult(size);
}
Is it correct, isn't it?
Thanks for help!
EDIT:
c# call:
public static extern void GetResult(IntPtr handle, [MarshalAs(UnmanagedType.SafeArray, SafeArraySubType = VarEnum.VT_USERDEFINED)] TestStruct[] data, ref int size);
Full example of use of a SAFEARRAY(int) C#->C++->C# (so the array is initialized with some data in C#, passed to C++, modified there and returned to C#).
C++:
// For the various _t classes for handling BSTR and IUnknown
#include <comdef.h>
struct ManagedUDT
{
BSTR m_str01;
int m_int01;
~ManagedUDT()
{
::SysFreeString(m_str01);
m_str01 = NULL;
}
};
extern "C" __declspec(dllexport) void GetResult(SAFEARRAY*& data)
{
if (data != NULL)
{
// Begin print content of SAFEARRAY
VARTYPE vt;
HRESULT hr = SafeArrayGetVartype(data, &vt);
if (SUCCEEDED(hr))
{
// To make this code simple, we print only
// SAFEARRAY(VT_I4)
if (vt == VT_I4)
{
int *pVals;
hr = SafeArrayAccessData(data, (void**)&pVals); // direct access to SA memory
if (SUCCEEDED(hr))
{
long lowerBound, upperBound; // get array bounds
SafeArrayGetLBound(data, 1, &lowerBound);
SafeArrayGetUBound(data, 1, &upperBound);
long cnt_elements = upperBound - lowerBound + 1;
for (int i = 0; i < cnt_elements; i++) // iterate through returned values
{
int val = pVals[i];
printf("C++: %d\n", val);
}
SafeArrayUnaccessData(data);
}
else
{
// Error
}
}
}
else
{
// Error
}
// End print content of SAFEARRAY
// Delete the SAFEARRAY if already present
SafeArrayDestroy(data);
data = NULL;
}
{
// Creation of a new SAFEARRAY
SAFEARRAYBOUND bounds;
bounds.lLbound = 0;
bounds.cElements = 10;
data = SafeArrayCreate(VT_I4, 1, &bounds);
int *pVals;
HRESULT hr = SafeArrayAccessData(data, (void**)&pVals); // direct access to SA memory
if (SUCCEEDED(hr))
{
for (ULONG i = 0; i < bounds.cElements; i++)
{
pVals[i] = i + 100;
}
}
else
{
// Error
}
}
}
C#
[DllImport("NativeLibrary.dll", CallingConvention = CallingConvention.Cdecl)]
private static extern void GetResult([MarshalAs(UnmanagedType.SafeArray, SafeArraySubType = VarEnum.VT_I4)] ref int[] ar);
and
var data = new int[] { 1, 2, 3, 4, 5 };
GetResult(ref data);
if (data != null)
{
for (int i = 0; i < data.Length; i++)
{
Console.WriteLine("C#: {0}", data[i]);
}
}
else
{
Console.WriteLine("C#: data is null");
}
Code partially taken from https://stackoverflow.com/a/12484259/613130 and https://stackoverflow.com/a/3735438/613130
SAFEARRAY(VT_RECORD)
It is doable... Very hard... but doable. Please don't do it. You can't hate enough the world to do it. I do hope you don't!
C++:
// For the _com_util
#include <comdef.h>
extern "C"
{
__declspec(dllexport) void GetResultSafeArray(SAFEARRAY *&psa)
{
// All the various hr results should be checked!
HRESULT hr;
// Begin sanity checks
if (psa == NULL)
{
// Error
}
VARTYPE pvt;
hr = ::SafeArrayGetVartype(psa, &pvt);
if (pvt != VT_RECORD)
{
// Error
}
UINT size;
size = ::SafeArrayGetElemsize(psa);
if (size != sizeof(ManagedUDT))
{
// Error
}
// From tests done, it seems SafeArrayGetRecordInfo does a AddRef
_com_ptr_t<_com_IIID<IRecordInfo, NULL> > prinfo;
// The_com_ptr_t<>::operator& is overloaded
hr = ::SafeArrayGetRecordInfo(psa, &prinfo);
// From tests done, it seems GetName returns a new instance of the
// BSTR
// It is ok to use _bstr_t.GetAddress() here, see its description
_bstr_t name1;
hr = prinfo->GetName(name1.GetAddress());
const _bstr_t name2 = _bstr_t(L"ManagedUDT");
if (name1 != name2)
{
// Error
}
// End sanity checks
long lowerBound, upperBound; // get array bounds
hr = ::SafeArrayGetLBound(psa, 1, &lowerBound);
hr = ::SafeArrayGetUBound(psa, 1, &upperBound);
long cnt_elements = upperBound - lowerBound + 1;
// Begin print
ManagedUDT *pVals;
hr = ::SafeArrayAccessData(psa, (void**)&pVals);
printf("C++:\n");
for (int i = 0; i < cnt_elements; ++i)
{
ManagedUDT *pVal = pVals + i;
// If you are using a recent VisualC++, you can
// #include <memory>, and then
//std::unique_ptr<char[]> pstr(_com_util::ConvertBSTRToString(pVal->m_str01));
// and you don't need the char *pstr line and the delete[]
// line
char *pstr = _com_util::ConvertBSTRToString(pVal->m_str01);
printf("%s, %d\n", pstr, pVal->m_int01);
delete[] pstr;
}
hr = ::SafeArrayUnaccessData(psa);
// End print
// Begin free
SAFEARRAYBOUND sab;
sab.lLbound = 0;
sab.cElements = 0;
// SafeArrayRedim will call IRecordInfo::RecordClear
hr = ::SafeArrayRedim(psa, &sab);
// End Free
// Begin create
int numElements = 10;
sab.cElements = numElements;
hr = ::SafeArrayRedim(psa, &sab);
hr = ::SafeArrayAccessData(psa, (void**)&pVals);
for (int i = 0; i < numElements; i++)
{
ManagedUDT *pVal = pVals + i;
char pstr[100];
sprintf(pstr, "Element #%d", i);
pVal->m_str01 = _com_util::ConvertStringToBSTR(pstr);
pVal->m_int01 = 100 + i;
}
hr = ::SafeArrayUnaccessData(psa);
// End create
}
__declspec(dllexport) void GetResultSafeArrayOut(SAFEARRAY *&psa, ITypeInfo *itypeinfo)
{
// All the various hr results should be checked!
HRESULT hr;
// Begin sanity checks
if (psa != NULL)
{
// Begin free
// SafeArrayDestroy will call IRecordInfo::RecordClear
// if necessary
hr = ::SafeArrayDestroy(psa);
// End Free
}
// Begin create
int numElements = 10;
SAFEARRAYBOUND sab;
sab.lLbound = 0;
sab.cElements = numElements;
// The_com_ptr_t<>::operator& is overloaded
_com_ptr_t<_com_IIID<IRecordInfo, NULL> > prinfo;
hr = ::GetRecordInfoFromTypeInfo(itypeinfo, &prinfo);
psa = ::SafeArrayCreateVectorEx(VT_RECORD, 0, numElements, prinfo);
ManagedUDT *pVals;
hr = ::SafeArrayAccessData(psa, (void**)&pVals);
for (int i = 0; i < numElements; i++)
{
ManagedUDT *pVal = pVals + i;
char pstr[100];
sprintf(pstr, "Element #%d", i);
pVal->m_str01 = _com_util::ConvertStringToBSTR(pstr);
pVal->m_int01 = 100 + i;
}
hr = ::SafeArrayUnaccessData(psa);
// End create
}
}
C#:
[ComVisible(true)]
[Guid("BBFE1092-A90C-4b6d-B279-CBA28B9EDDFA")]
[StructLayout(LayoutKind.Sequential)]
public struct ManagedUDT
{
[MarshalAs(UnmanagedType.BStr)]
public string m_str01;
public Int32 m_int01;
}
[DllImport("NativeLibrary.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
static extern void GetResultSafeArray([MarshalAs(UnmanagedType.SafeArray)] ref ManagedUDT[] array);
[DllImport("NativeLibrary.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
static extern void GetResultSafeArrayOut([MarshalAs(UnmanagedType.SafeArray)] out ManagedUDT[] array, IntPtr itypeinfo);
[DllImport("NativeLibrary.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi, EntryPoint = "GetResultSafeArrayOut")]
static extern void GetResultSafeArrayRef([MarshalAs(UnmanagedType.SafeArray)] ref ManagedUDT[] array, IntPtr itypeinfo);
and
var arr = new[]
{
new ManagedUDT { m_str01 = "Foo", m_int01 = 1},
new ManagedUDT { m_str01 = "Bar", m_int01 = 2},
};
{
Console.WriteLine("C#:");
for (int i = 0; i < arr.Length; i++)
{
Console.WriteLine("{0}, {1}", arr[i].m_str01, arr[i].m_int01);
}
}
{
Console.WriteLine();
var arr2 = (ManagedUDT[])arr.Clone();
GetResultSafeArray(ref arr2);
Console.WriteLine();
Console.WriteLine("C#:");
for (int i = 0; i < arr2.Length; i++)
{
Console.WriteLine("{0}, {1}", arr2[i].m_str01, arr2[i].m_int01);
}
}
{
Console.WriteLine();
ManagedUDT[] arr2;
IntPtr itypeinfo = Marshal.GetITypeInfoForType(typeof(ManagedUDT));
GetResultSafeArrayOut(out arr2, itypeinfo);
Console.WriteLine();
Console.WriteLine("C#:");
for (int i = 0; i < arr2.Length; i++)
{
Console.WriteLine("{0}, {1}", arr2[i].m_str01, arr2[i].m_int01);
}
}
{
Console.WriteLine();
var arr2 = (ManagedUDT[])arr.Clone();
IntPtr itypeinfo = Marshal.GetITypeInfoForType(typeof(ManagedUDT));
GetResultSafeArrayRef(ref arr2, itypeinfo);
Console.WriteLine();
Console.WriteLine("C#:");
for (int i = 0; i < arr2.Length; i++)
{
Console.WriteLine("{0}, {1}", arr2[i].m_str01, arr2[i].m_int01);
}
}
There is a single big caveat for GetResultSafeArray: you must pass from C# at least an empty array (like a new ManagedUDT[0]). This because to create a SAFEARRAY(ManagedUDT) from nothing in C++ you would need a IRecordInfo object. I don't know how to retrieve it from C++. If you already have a SAFEARRAY(ManagedUDT) then clearly it has the IRecordInfo already set, so there is no problem. In the example given, in C++ there are first some sanity checks, then the passed array is printed, then it is emptied, then it is re-filled. The GetResultSafeArrayOut/GetResultSafeArrayRef "cheat": they receive from C# a ITypeInfo pointer (that is easy to retrieve in C#, with Marshal.GetITypeInfoForType()), and from taht the C++ can retrieve the IRecordInfo interface.
Some notes:
I wrote Ansi-charset-C++. Normally for myself I always write Unicode-ready C++ (or directy Unicode-C++, because all the Windows NT support Unicode), but I've noticed that I'm an exception... So in various parts of the code there are conversions BSTR->Ansi->BSTR.
I'm retrieving the HRESULT of all the function calls. They should be checked, and the failure handled.
The most complex thing in C++/COM is knowing when to free something... In general always free/Release() everything! (be it BSTR/IUnknown derived interfaces, ...)
Unless there is a bug, there is no support for this code. Consider it to be a proof of concept. I already lost various hours on it out of curiosity. You break it, you repair it.
Below I have a code snippet from c++.
I need to return array of pointers (to TempStruct).
The problem is that on c# side I get only one element. On the other I get AV.
**C++**
extern "C" __declspec(dllexport) void GetResult(TempStruct** outPtr, long *size)
{
*outPtr = (TempStruct*)new TempStruct*[2];
outPtr[0] = new TempStruct("sdf", 123);
outPtr[1] = new TempStruct("abc", 456);
*size = 2;
}
**C#**
[DllImport("test.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
public static extern void GetResult(out IntPtr outPtr, out int size);
IntPtr ptr = IntPtr.Zero;
int length;
GetResult(out ptr, out length);
int size = Marshal.SizeOf(typeof(TempStruct));
TempStruct[] someData2 = new TempStruct[length];
for (int i = 0; i < length; i++)
{
IntPtr wskptr = (IntPtr)(ptr.ToInt64() + (size * i));
someData2[i] = (TempStruct)Marshal.PtrToStructure(wskptr, typeof(TempStruct));
}
You are doing it wrong.
You are mixing pointer types.
By using the new TempStruct() you are creating an array of pointers to TempStruct. The example I gave you created an array of TempStruct. See the difference?
Now... TempStruct** outPtr should be TempStruct*** outPtr (because you want to return (*) an array (*) of pointers (*)... Or TempStruct**& if you prefer :-)
Change this line
someData2[i] = (TempStruct)Marshal.PtrToStructure(Marshal.ReadIntPtr(wskptr), typeof(TempStruct));
Because you must read the single pointers.
I do hope you are deleting the various TempStruct with delete and using the
delete[] ptr;
operator to delete the array of structures.
Full example:
C++:
struct TempStruct
{
char* str;
int num;
// Note the strdup. You don't know the source of str.
// For example if the source is "Foo", then you can't free it.
// Using strdup solves this problem.
TempStruct(const char *str, int num)
: str(strdup(str)), num(num)
{
}
~TempStruct()
{
free(str);
}
};
extern "C"
{
__declspec(dllexport) void GetResult(TempStruct ***outPtr, int *size)
{
*outPtr = new TempStruct*[2];
(*outPtr)[0] = new TempStruct("sdf", 123);
(*outPtr)[1] = new TempStruct("abc", 456);
*size = 2;
}
__declspec(dllexport) void FreeSomeData(TempStruct **ptr, int size)
{
for (int i = 0; i < size; i++)
{
delete ptr[i];
}
delete[] ptr;
}
}
C#:
[StructLayout(LayoutKind.Sequential, CharSet = CharSet.Ansi, Pack = 1), Serializable]
internal struct TempStruct
{
public string str;
public int num;
}
[DllImport("NativeLibrary.dll", CallingConvention = CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
static extern void GetResult(out IntPtr outPtr, out int numPtr);
[DllImport("NativeLibrary.dll", CallingConvention = CallingConvention.Cdecl)]
static extern void FreeSomeData(IntPtr ptr, int num);
// C++ will return its TempStruct array in ptr
IntPtr ptr;
int size;
GetResult(out ptr, out size);
TempStruct[] someData2 = new TempStruct[size];
for (int i = 0; i < size; i++)
{
IntPtr ptr2 = Marshal.ReadIntPtr(ptr, i * IntPtr.Size);
someData2[i] = (TempStruct)Marshal.PtrToStructure(ptr2, typeof(TempStruct));
}
// Important! We free the TempStruct allocated by C++. We let the
// C++ do it, because it knows how to do it.
FreeSomeData(ptr, size);
Note that you don't need [Serializable] and Pack=1 on the C# struct
More correct for the C++:
__declspec(dllexport) void GetResult(TempStruct **&outPtr, int &size)
{
outPtr = new TempStruct*[2];
outPtr[0] = new TempStruct("sdf", 123);
outPtr[1] = new TempStruct("abc", 456);
size = 2;
}
It is more correct because both outPtr and size can't be NULL. See https://stackoverflow.com/a/620634/613130 . The C# signature is the same.
The C++ code is wrong. It's returning an array of pointers to struct. The fact that you cast the value returned by new should have alerted you to the fact that you made a mistake. You want to return an array of struct.
It should be:
*outPtr = new TempStruct[2];
(*outPtr)[0].str = "sdf";
(*outPtr)[0].i = 123;
(*outPtr)[1].str = "abc";
(*outPtr)[1].i = 456;
*size = 2;