If I have the following C# DllImport which is importing a non-C# DLL and I want to port it over to Go, how would I go about doing that?
[DllImport("my.dll", EntryPoint = "Greet", CallingConvention = CallingConvention.Cdecl)]
public static extern int Greet(IntPtr name, ref IntPtr greetings);
I've run into problems figuring out how to pass a pointer to a pointer which is needed for the greetings parameter (I assume since the type is ref IntPtr, I'm not that familiar at all with C#). The dll function will populate the memory pointed to by the pointer that I provide which I'll use in subsequent syscalls. Here's what I've got so far,
package main
import (
"fmt"
"syscall"
"unsafe"
)
var (
MyDll = syscall.MustLoadDLL("my.dll")
greet = MyDll.MustFindProc("Greet")
)
func Greet(name string) error {
nameP, err := syscall.UTF16PtrFromString(name)
if err != nil {
return err
}
// I need to provide a pointer to a pointer for greetings. How can I allocate some memory here
// then pass a pointer to its pointer? I tried this: create a handle with a zero-value, then
// take a pointer to it, then pass a pointer to the pointer as the second parameter to Call but
// the pointer becomes nil after the Call.
handle := syscall.Handle(0)
handleP := &handle
r1, _, _ := greet.Call(uintptr(unsafe.Pointer(nameP)), uintptr(unsafe.Pointer(&handleP)))
if r1 == 0 {
return fmt.Errorf("there was an error")
}
return nil
}
I'm open to any and all suggestions including links and resources that might help me get a better grasp on this syscall and unsafe stuff. Thanks!
Firstly, it would help if you could show how the C# Greet method is used. A method in isolation its quite hard to understand especially when the parameter is the equivalent of void ** which means anything can go in.
TL;DR
ref IntPtr is probably just a **struct{} where you don't have to allocate any struct. The library will simply manage the memory for you. You just need to give it a pointer to "*MyStruct" so that it can change you "*MyStruct" to actually point to the internal resource.
REF
The C# ref keyword is quite well explained in the docs. Basically it allows a pass by reference for any type.
The following declaration in C#
void Increment(ref value) { ... }
int counter = 10;
Increment(ref counter);
// counter is now 11
Increment(counter); // won't compile must be passed with 'ref'
Increment(null); // won't compile
is equivalent to C++
void Increment(int& value) { ... }
int counter;
Increment(counter);
// counter is now 11
Increment(null); // won't compile
A reference which should not be null.
IntPtr
IntPtr is usually used to represent pointers and allows for interop between native and CLR (C#) programs.
If a C program has the following signature
void Increment(int* value);
a C# program could call it in one of a few ways
[DllImport("example.dll")]
static unsafe extern void Increment(int* value); // this way allows null to be passed in
unsafe {
int counter = 10;
Increment(&10);
}
,
[DllImport("example.dll")]
static extern void Increment(ref int value);
int counter = 10;
Increment(ref counter);
If a C program has the following signature
void AllocateStruct(struct MyStruct** ppStruct);
void IncrementStruct(struct MyStruct* pStruct);
then
[DllImport("example.dll")]
static extern void AllocateStruct(ref IntPtr ppStruct);
// static unsafe extern void AllocateStruct(MyStruct** ppStruct)
[DllImport("example.dll")]
static extern void IncrementStruct(IntPtr pStruct);
// static unsafe extern void IncrementStruct(MyStruct* pStruct);
IntPtr pMyStruct;
AllocateStruct(ref pMyStruct);
IncrementStruct(pMyStruct);
// Free My Struct
// If you need to get inside the struct then
// MyStruct myStruct = Marshal.StructureToPtr<MyStruct>(pMyStruct)
// Often you don't (need to) or (should not) manipulate the struct directly so keeping it as IntPtr is perfectly acceptable.
From the example above you can see MyStruct is more of a token/reference than anything else. ref IntPtr allows you to pass a reference to the location which you will use to store your token/reference after the library allocates it on your behalf. Then all the other methods will usually just use the reference IntPtr to perform subsequent manipulations on it. Basically Object Orientated Programming without classes.
"Real" Life Example with ref IntPtr
It is a bit quick and dirty and error handling leave a lot to be desired.
It shows the C, C# and Go versions which call the same GetSecurityInfo
and LookupAccountSid Win32 library functions.
The only real use case I can find for ref IntPtr is type**/void** so that the library can allocate the memory for you or give you a pointer to memory it has already allocated.
C
#include <stdio.h>
#include <windows.h>
#include <tchar.h>
#include "accctrl.h"
#include "aclapi.h"
#pragma comment(lib, "advapi32.lib")
int _tmain(int argc, _TCHAR* argv[])
{
// --- get the executing program's file path and handle -----
LPTSTR executablePath = argv[0];
_tprintf(TEXT("Opening File %s\n"), executablePath);
HANDLE hFile = CreateFile(
executablePath,
GENERIC_READ,
FILE_SHARE_READ,
NULL,
OPEN_EXISTING,
FILE_ATTRIBUTE_NORMAL,
NULL);
if (hFile == INVALID_HANDLE_VALUE) return EXIT_FAILURE;
// -------------------------------------------------
// --------- Get the owner SID of the file ---------
PSID pSidOwner = NULL;
PSECURITY_DESCRIPTOR pSD = NULL;
DWORD dwRtnCode = GetSecurityInfo(
hFile,
SE_FILE_OBJECT,
OWNER_SECURITY_INFORMATION,
&pSidOwner,
NULL,
NULL,
NULL,
&pSD);
if (dwRtnCode != ERROR_SUCCESS) return EXIT_FAILURE;
// -------------------------------------------------
// -------
TCHAR AcctName[MAX_PATH];
DWORD dwAcctName = MAX_PATH;
TCHAR DomainName[MAX_PATH];
DWORD dwDomainName = MAX_PATH;
SID_NAME_USE eUse = SidTypeUnknown;
BOOL bRtnBool = LookupAccountSid(
NULL, // local computer
pSidOwner,
&AcctName,
&dwAcctName,
DomainName,
&dwDomainName,
&eUse);
if (bRtnBool == FALSE) return EXIT_FAILURE;
_tprintf(TEXT("Account Owner = %s\n"), AcctName);
_tprintf(TEXT("Account Owner's Domain = %s\n"), DomainName);
return 0;
}
C#
using System;
using System.IO;
using System.Runtime.InteropServices;
using System.Text;
public class Example
{
[DllImport("advapi32.dll", SetLastError = true, CallingConvention = CallingConvention.Winapi)]
static extern uint GetSecurityInfo(
IntPtr handle,
uint ObjectType,
uint SecurityInfo,
ref IntPtr ppsidOwner, // <-- HERE
IntPtr ppsidGroup, // bit hacky (in safe C# you must "pass a reference" in C you can pass a pointer to a pointer or null)
IntPtr ppDacl,
IntPtr ppSacl,
ref IntPtr ppSecurityDescriptor // <-- HERE
);
[DllImport("advapi32.dll", CharSet = CharSet.Auto, SetLastError = true)]
static extern bool LookupAccountSid(
string lpSystemName,
IntPtr Sid,
StringBuilder lpName,
ref uint cchName,
StringBuilder ReferencedDomainName,
ref uint cchReferencedDomainName,
out uint peUse);
const uint ERROR_SUCCESS = 0;
const uint OWNER_SECURITY_INFORMATION = 0x00000001;
const uint SE_FILE_OBJECT = 1;
public static void Main()
{
// get the executing program's file path and handle
string executablePath = Environment.GetCommandLineArgs().GetValue(0).ToString();
IntPtr hFile = File.Open(executablePath, FileMode.Open, FileAccess.Read, FileShare.Read)
.SafeFileHandle.DangerousGetHandle();
IntPtr pSidOwner = IntPtr.Zero; // some internal struct you shouldn't allocate or modify (acts like a token)
IntPtr pSD = IntPtr.Zero; // some internal struct you shouldn't allocate or modify (acts like a token)
// vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
uint dwRtnCode = GetSecurityInfo(
hFile,
SE_FILE_OBJECT,
OWNER_SECURITY_INFORMATION,
ref pSidOwner, // <-- HERE
IntPtr.Zero,
IntPtr.Zero,
IntPtr.Zero,
ref pSD // <-- HERE
);
// ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
if (dwRtnCode != ERROR_SUCCESS) throw new InvalidOperationException("GetSecurityInfo Failed");
StringBuilder name = new StringBuilder(50);
uint cchName = (uint)name.Capacity;
StringBuilder domainName = new StringBuilder(50);
uint cchDomainName = (uint)domainName.Capacity;
uint sidUse;
LookupAccountSid(
null,
pSidOwner,
name,
ref cchName,
domainName,
ref cchDomainName,
out sidUse);
Console.WriteLine("Account Owner = {0}", name);
Console.WriteLine("Account Owner's Domain = {0}", domainName);
// PLEASE FREE pSD once done
}
}
Go
my second Go program I have ever written so there are probably some blaring mistakes (other than the lack of error checking)
package main
import (
"fmt"
"syscall"
"unsafe"
"os"
)
var (
advapi32, _ = syscall.LoadLibrary("advapi32.dll")
kernel32, _ = syscall.LoadLibrary("kernel32.dll")
createFileW, _ = syscall.GetProcAddress(kernel32, "CreateFileW")
getSecurityInfo, _ = syscall.GetProcAddress(advapi32, "GetSecurityInfo")
lookupAccountSidW, _ = syscall.GetProcAddress(advapi32, "LookupAccountSidW")
)
type SE_OBJECT_TYPE uint32
const (SE_FILE_OBJECT = 1)
type SECURITY_INFORMATION uint32
const (OWNER_SECURITY_INFORMATION = 0x00000001)
const (
GENERIC_READ = 0x80000000
FILE_SHARE_READ = 0x00000001
OPEN_EXISTING = 0x00000003
FILE_ATTRIBUTE_NORMAL = 0x00000080
)
type Handle uintptr
func CreateFile(
name string,
access uint32,
mode uint32,
sa *uint, // *SecurityAttributes,
createmode uint32,
attrs uint32,
templatefile *uint,
) (handle Handle, err error) {
utf16name, _ := syscall.UTF16PtrFromString(name)
r0, _, _ := syscall.Syscall9(
uintptr(createFileW), 7,
uintptr(unsafe.Pointer(utf16name)),
uintptr(access),
uintptr(mode),
uintptr(unsafe.Pointer(sa)),
uintptr(createmode),
uintptr(attrs),
uintptr(unsafe.Pointer(templatefile)),
0, 0)
handle = Handle(r0)
return
}
func GetSecurityInfo(
handle Handle,
objectType SE_OBJECT_TYPE,
securityInformation SECURITY_INFORMATION,
owner **struct{},
group **struct{},
dacl **struct{},
sacl **struct{},
sd **struct{}, //**SECURITY_DESCRIPTOR,
) (ret error) {
r0, _, _ := syscall.Syscall9(
uintptr(getSecurityInfo), 8,
uintptr(handle),
uintptr(objectType),
uintptr(securityInformation),
uintptr(unsafe.Pointer(owner)),
uintptr(unsafe.Pointer(group)),
uintptr(unsafe.Pointer(dacl)),
uintptr(unsafe.Pointer(sacl)),
uintptr(unsafe.Pointer(sd)),
0)
if r0 != 0 {
ret = syscall.Errno(r0)
}
return
}
func LookupAccountSid(
systemName *uint16,
sid *struct{}, // *SID,
name *uint16,
nameLen *uint32,
refdDomainName *uint16,
refdDomainNameLen *uint32,
use *uint32,
) (err error) {
r, _, e := syscall.Syscall9(
uintptr(lookupAccountSidW), 7,
uintptr(unsafe.Pointer(systemName)),
uintptr(unsafe.Pointer(sid)),
uintptr(unsafe.Pointer(name)),
uintptr(unsafe.Pointer(nameLen)),
uintptr(unsafe.Pointer(refdDomainName)),
uintptr(unsafe.Pointer(refdDomainNameLen)),
uintptr(unsafe.Pointer(use)),
0, 0)
if r == 0 {
err = e
}
return
}
func main() {
defer syscall.FreeLibrary(advapi32)
defer syscall.FreeLibrary(kernel32)
// get the executing program's file path and handle
var hFile, _ = CreateFile(os.Args[0], GENERIC_READ, FILE_SHARE_READ, nil, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, nil);
// defer LocalFree(Handle(unsafe.Pointer(pSD))) // PLEASE FREE pSD once done
// vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
var pSD *struct{} //*SECURITY_DESCRIPTOR
var pSidOwner *struct{}
GetSecurityInfo(hFile, SE_FILE_OBJECT, OWNER_SECURITY_INFORMATION, &pSidOwner, nil, nil, nil, &pSD)
// ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
nameLen := uint32(50)
name := make([]uint16, nameLen)
domainLen := uint32(50)
domainName := make([]uint16, domainLen)
var sidUse uint32
LookupAccountSid(nil, pSidOwner, &name[0], &nameLen, &domainName[0], &domainLen, &sidUse)
var n = syscall.UTF16ToString(name)
var dn = syscall.UTF16ToString(domainName)
fmt.Printf("Account Owner = %s\n", n)
fmt.Printf("Account Owner's Domain = %s\n", dn)
}
I have a function in a C++ DLL that takes one input. I'm trying to have that input be used as an output to the C# call.
Here is my C++ function:
MYAPI int testStuff3(unsigned char* str)
{
printf("%s\n", str);
str = (unsigned char*)malloc(9);
str[0] = 'G';
str[1] = 'o';
str[2] = 'o';
str[3] = 'd';
str[4] = 'b';
str[5] = 'y';
str[6] = 'e';
str[7] = '!';
str[8] = '\0';
return 1;
}
Here is the C# code:
public class Program
{
[DllImport("NativeLib.dll")]
private static extern int testStuff3([In, Out] IntPtr str);
static void Main(string[] args)
{
IntPtr junk3 = IntPtr.Zero;
int ret = testStuff3(junk3);
Byte[] stuff3 = new byte[9];
Marshal.Copy(junk3, stuff3, 0, 9);
}
}
When the Marshal.Copy is called, it gives an error saying that the source (junk3) can not be null.
Will this not work, sending a null pointer to C++ DLL from C# and having the DLL allocate the memory and store something inside and return it to the caller? I want to keep it an IntPtr and not a StringBuilder because the data won't necessarily be a string in the final code. Just an unsigned char array in C++ and I want the IntPtr to point to it.
I've tried different variations of [In, Out], [Out], out and ref for the IntPtr passing.
Never ever allow memory allocations to cross a DLL boundary. That way lies madness, and/or Sparta.
(For the pedantic: you can allocate memory and then pass a pointer across, as long as you either pass ownership back to free it, or guarantee that the same allocator is used as part of a contract. But it's still something to avoid when possible.)
Typically to use a string output parameter you should pass a StringBuilder as the argument, setting its capacity to the maximum expected length. Then in the native code you simply fill this existing buffer.
See the "Fixed length string buffers" section here for an example.
Thanks for the help!
Here's what I ended up with.
C++ function:
MYAPI int testStuff4(wchar_t* str)
{
unsigned char* stuff = (unsigned char*)malloc(10);
stuff[0] = 'G';
stuff[1] = 'o';
stuff[2] = 'o';
stuff[3] = 'd';
stuff[4] = 'b';
stuff[5] = 'y';
stuff[6] = 'e';
stuff[7] = '!';
stuff[8] = '\0';
mbstowcs(str, (const char*)stuff, 1024);
free(stuff);
return 1;
}
C# function:
public class Program
{
[DllImport("NativeLib.dll")]
private static extern int testStuff4(IntPtr str);
static void Main(string[] args)
{
IntPtr junk4 = Marshal.AllocHGlobal(1024);
int ret = testStuff4(junk4);
string junkString = Marshal.PtrToStringUni(junk4);
Console.WriteLine(junkString);
Marshal.FreeHGlobal(junk4);
}
}
Your C++ function doesn’t modify the passed string. It allocates a new one with malloc, stores it in a local variable forgetting the passed value, then returns leaking the memory.
If for some reason you want to do manual marshalling, you probably want something like this (assuming this is for Windows):
MYAPI BOOL __stdcall testStuff3( char** pp )
{
if( nullptr == pp )
return FALSE; // null pointer
if( nullptr != *pp )
{ // Print & release an old string
printf( "%s\n", *pp );
CoTaskMemFree( *pp );
*pp = nullptr;
}
// Allocate a new one
const char* const str = CoTaskMemAlloc( 9 );
if( nullptr == str ) return FALSE;
strncpy( str, "Goodbye!", 9 );
*pp = str;
return TRUE;
}
C#:
public class Program
{
[DllImport( "NativeLib.dll" )]
private static extern bool testStuff3( [In, Out] ref IntPtr str );
static void Main( string[] args )
{
IntPtr ptr = IntPtr.Zero;
if( testStuff3( ref ptr ) )
{
Console.WriteLine( Marshal.PtrToStringAnsi( ptr ) );
Marshal.FreeCoTaskMem( ptr );
}
}
}
However, this is not something I recommend doing unless you have very good reasons. In most cases automatic marshalling is better. For C# -> C++ way it’s trivially simple, const char* or const wchar_t* in C++, string (with correct attributes) in C#. For C++ -> C# you can allocate a StringBuilder in C#, pass char* or wchar_t* to C++, and buffer length in another argument.
I'm trying to create a PInvoke for FbwfFindFirst and am struggling with the struct FbwfCacheDetail.
In short, I'm not sure how to marshal WCHAR fileName[1]; seeing as it's a variable length array and a non-null terminated.
Any help would be welcomed
Since the whole structure is of variable size, one way to do this is like this (I can't test it because I don't have this dll on my system):
string volume = "";
int size = 0;
// ask for whole structure size
FbwfFindFirst(volume, IntPtr.Zero, ref size); // this call should return ERROR_MORE_DATA which is ok
// allocate for the structure
var ptr = Marshal.AllocHGlobal(size);
try
{
FbwfFindFirst(volume, ptr, ref size); // should not return error
// get the easy part
var detail = Marshal.PtrToStructure<FbwfCacheDetail>(ptr);
// compute filename offset and get the string
// file name length is in bytes, per documentation
var fileName = Marshal.PtrToStringUni(ptr + Marshal.OffsetOf<FbwfCacheDetail>("fileName").ToInt32(), detail.fileNameLength / 2);
}
finally
{
Marshal.FreeHGlobal(ptr);
}
[DllImport("fbwflib", CharSet = CharSet.Unicode)]
static extern int FbwfFindFirst(string volume, IntPtr cacheDetail, ref int size);
[StructLayout(LayoutKind.Sequential)]
struct FbwfCacheDetail
{
public int cacheSize;
public int openHandleCount;
public int fileNameLength;
byte fileName; // don't use this member
}
Simon Mourier's answer is 99% correct, and with normal APIs it would have definitely worked, but it appears as if this particular API doesn't follow "normal rules", whatever that might be ;). As such, I needed to modify a couple things and this is what worked for me:
const int ERROR_INSUFFICIENT_BUFFER = 122;
const int ERROR_MORE_DATA = 234;
var volume = "C:";
var fileName = string.Empty;
var size = 0;
while (true)
{
var ptr = Marshal.AllocHGlobal(size); // FbwfFindFirst fails if given IntPtr.Zero - regardless of what the value of size is.
try
{
var result = FbwfFindFirst(volume, ptr, ref size);
// Despite documentation saying otherwise, it can return either of these
if (result == ERROR_MORE_DATA || result == ERROR_INSUFFICIENT_BUFFER)
{
continue;
}
if (result != 0)
{
throw new Exception($"Failed with {result}");
}
// get the easy part
var detail = (FbwfCacheDetail) Marshal.PtrToStructure(ptr, typeof(FbwfCacheDetail));
// compute filename offset and get the string
// file name length is in bytes, per documentation
fileName = Marshal.PtrToStringUni(ptr + Marshal.OffsetOf(typeof(FbwfCacheDetail), "fileName").ToInt32(), detail.fileNameLength / 2);
break;
}
finally
{
Marshal.FreeHGlobal(ptr);
}
}
EDIT
Forgot to say that the pinvoke for FbwfFindFirst needs to be as follows, else it will return ERROR_INVALID_PARAMETER:
[DllImport("fbwflib.dll")]
public static extern uint FbwfFindFirst(
[MarshalAs(UnmanagedType.LPWStr)]
string volume,
IntPtr cacheDetail,
ref int size
);
I'd like to interpret the output of the following unmanaged function:
afc_error_t afc_get_device_info (afc_client_t client, char ***device_information)
I import the dll with the code:
[DllImport("libimobiledevice.dll", CallingConvention = CallingConvention.Cdecl)]
internal static extern short afc_get_device_info(IntPtr client, out IntPtr info);
As long as I only needed to convert response to string Marshal.PtrToStringAnsi was okay. However I have no idea how to convert that IntPtr back to char array.
It should be something like:
IntPtr di;
int result = afc_read_directory(client, #"C:\", out di);
if (di == IntPtr.Zero)
{
throw new Exception();
}
IntPtr di2 = di;
while (true)
{
IntPtr ptr = Marshal.ReadIntPtr(di2);
if (ptr == IntPtr.Zero)
{
break;
}
string str = Marshal.PtrToStringAnsi(ptr);
if (str == string.Empty)
{
break;
}
di2 = di2 + IntPtr.Size;
}
Try if it works, then I'll explain how...
important you are leaking memory here...
I've found this example in C:
char **dirs = NULL;
afc_read_directory(afc, "/eafaedf", &dirs);
if (!dirs)
afc_read_directory(afc, "/", &dirs);
printf("Directory time.\n");
for (i = 0; dirs[i]; i++) {
printf("/%s\n", dirs[i]);
free(dirs[i]);
}
if (dirs)
free(dirs);
you are responsible for freeing the memory (see the free inside the cycle and the final free?). In this case (and for other methods that return arrays of C-strings you can use afc_dictionary_free. Note that other methods like afc_receive_data that return a single block of memory you can't use it.
From a c# program I want to use WM_COPYDATA with SendMessage to communicate with a legacy c++/cli MFC application.
I want to pass a managed struct containing string objects.
I can find the handle to the c++ application for use with SendMessage fine.
The bit I don't know about is how the struct and it's strings can be marshalled and read at the other end. Especially as it contains non-blittables.
Do people think this is feasible?
I'll continue to work on it, but would apprecite someone who's done this sort of thing telling me if it just isn't going to work.
Here is some demo code if it was a c++/cli program and it's not difficult to get it working.
However, I'd like this to be in a .Net class library so it can easily be re-used.
//Quick demonstation code only, not correctly styled
int WINAPI WinMain(HINSTANCE hInstance,
HINSTANCE hPrevInstance,
LPSTR lpCmdLine,
int nCmdShow)
{
struct MessageInfo
{
int nVersion;
char szTest[ 10 ];
};
MessageInfo sMessageInfo;
sMessageInfo.nVersion = 100;
strcpy( sMessageInfo.szTest, "TEST");
COPYDATASTRUCT CDS;
CDS.dwData = 1; //just for test
CDS.cbData = sizeof( sMessageInfo );
CDS.lpData = &sMessageInfo;
//find running processes and send them a message
//can't just search for "MYAPP.exe" as will be called "MYAPP.exe *32" on a 64bit machine
array<System::Diagnostics::Process^>^allProcesses = System::Diagnostics::Process::GetProcesses();
for each (System::Diagnostics::Process^ targetProcess in allProcesses)
{
if (targetProcess->ProcessName->StartsWith("MYAPP", System::StringComparison::OrdinalIgnoreCase))
{
HWND handle = static_cast<HWND>(targetProcess->MainWindowHandle.ToPointer());
BOOL bReturnValue = SendMessage( handle, WM_COPYDATA, (WPARAM)0, (LPARAM)&CDS ) == TRUE;
}
}
return 0;
}
I have it working.
A simple approach is to serialize the struct to a single string and transfer a string.
The swhistlesoft blog was helpful http://www.swhistlesoft.com/blog/2011/11/19/1636-wm_copydata-with-net-and-c
This may be enough to provide the simple messaging.
The struct can be re-constructed at the other end if necessary.
If a struct with any number of strings is to be marshalled as-is then it must be a fixed size, that's the main thing I wasn't getting.
The
MarshalAs(System.Runtime.InteropServices.UnmanagedType.ByValTStr, SizeConst = 9)
basically sets the size to match the c++ size which in our case is a TCHAR szTest[ 9 ];
In order to transfer a .Net struct via WM_COPYDATA from c# to c++(/cli) I had to do as follows:
[System.Runtime.InteropServices.DllImport("user32.dll", CharSet = System.Runtime.InteropServices.CharSet.Auto)]
static extern IntPtr SendMessage(IntPtr hWnd, UInt32 Msg, IntPtr wParam, IntPtr lParam);
[System.Runtime.InteropServices.DllImport("user32.dll", CharSet = System.Runtime.InteropServices.CharSet.Auto)]
static extern bool SetForegroundWindow(IntPtr hWnd);
public static uint WM_COPYDATA = 74;
//from swhistlesoft
public static IntPtr IntPtrAlloc<T>(T param)
{
IntPtr retval = System.Runtime.InteropServices.Marshal.AllocHGlobal(System.Runtime.InteropServices.Marshal.SizeOf(param));
System.Runtime.InteropServices.Marshal.StructureToPtr(param, retval, false);
return (retval);
}
//from swhistlesoft
public static void IntPtrFree(IntPtr preAllocated)
{
if (IntPtr.Zero == preAllocated) throw (new Exception("Go Home"));
System.Runtime.InteropServices.Marshal.FreeHGlobal(preAllocated);
preAllocated = IntPtr.Zero;
}
[System.Runtime.InteropServices.StructLayout(System.Runtime.InteropServices.LayoutKind.Sequential)]
struct COPYDATASTRUCT
{
public uint dwData;
public int cbData;
public IntPtr lpData;
}
/// <summary>
/// Dot net version of AppInfo structure. Any changes to the structure needs reflecting here.
/// struct must be a fixed size for marshalling to work, hence the SizeConst entries
/// </summary>
[System.Runtime.InteropServices.StructLayout(System.Runtime.InteropServices.LayoutKind.Sequential, Pack = 1)]
struct AppInfoDotNet
{
public int nVersion;
[System.Runtime.InteropServices.MarshalAs(System.Runtime.InteropServices.UnmanagedType.ByValTStr, SizeConst = 9)]
public string test;
};
To send a string:
COPYDATASTRUCT cd = new COPYDATASTRUCT();
cd.dwData = 2;
cd.cbData = parameters.Length + 1;
cd.lpData = System.Runtime.InteropServices.Marshal.StringToHGlobalAnsi(parameters);
IntPtr cdBuffer = IntPtrAlloc(cd);
messageReceived = ((int)SendMessage(targetProcess.MainWindowHandle, WM_COPYDATA, IntPtr.Zero, cdBuffer)) != 0;
To receive string in C++:
else if(pCDS->dwData == 2)
{
//copydata message
CString csMessage = (LPCTSTR)pCDS->lpData;
OutputDebugString("Copydata message received: " + csMessage);
}
To send struct:
AppInfoDotNet appInfo = new AppInfoDotNet();
appInfo.test = "a test";
COPYDATASTRUCT cds3;
cds3.dwData = 1;
cds3.cbData = System.Runtime.InteropServices.Marshal.SizeOf(appInfo);
IntPtr structPtr = System.Runtime.InteropServices.Marshal.AllocCoTaskMem(System.Runtime.InteropServices.Marshal.SizeOf(appInfo));
System.Runtime.InteropServices.Marshal.StructureToPtr(appInfo, structPtr, false);
cds3.lpData = structPtr;
IntPtr iPtr = System.Runtime.InteropServices.Marshal.AllocCoTaskMem(System.Runtime.InteropServices.Marshal.SizeOf(cds3));
System.Runtime.InteropServices.Marshal.StructureToPtr(cds3, iPtr, false);
messageReceived = ((int)SendMessage(targetProcess.MainWindowHandle, WM_COPYDATA, IntPtr.Zero, iPtr)) != 0;
System.Runtime.InteropServices.Marshal.FreeCoTaskMem(iPtr);
System.Runtime.InteropServices.Marshal.FreeCoTaskMem(structPtr);
To receive struct in C++:
LRESULT CMainFrame::OnCopyData( WPARAM wParam, LPARAM lParam )
{
LRESULT lResult = FALSE;
COPYDATASTRUCT *pCDS = (COPYDATASTRUCT*)lParam;
//Matching message type for struct
if(pCDS->dwData == 1)
{
AppInfo *pAppInfo = (AppInfo*)pCDS->lpData
lResult = true;
}
Please note this is demo code and needs work in terms of styling, exception handling etc, etc...
From the documentation:
The data being passed must not contain pointers or other references to objects not accessible to the application receiving the data.
So you need to pack your string into COPYDATASTRUCT.lpData. If you have a max length for each string then you can embed it in a fixed length structure
typedef struct tagMYDATA
{
char s1[80];
char s2[120];
} MYDATA;
If you have only one variable length string you can put the string at the end and use a header followed by string data
typedef struct tagMYDATA
{
int value1;
float value2;
int stringLen;
} MYDATAHEADER;
MyCDS.cbData = sizeof(MYDATAHEADER)+(int)stringData.size();
MyCDS.lpData = new BYTE[MyCDS.cbData];
memcpy(MyCDS.lpData,&dataHeader,sizeof*(MYDATAHEADER);
StringCbCopyA (
((BYTE*)MyCDS.lpData)+sizeof*(MYDATAHEADER)
,stringData.size()
,stringData.c_str());
If you have multiple variable length strings you can still use a header and allocate more spaces for every strings plus a double null terminator, or serialize everything into one XML string.