Develop Reference

Initializing a C# IntPtr to accept data from unmanaged C++ DLL?

I have an exported function in unmanaged C++ code that expects a pointer to a BStr, where it will write some text data (258 bytes, max)
extern "C" __declspec(dllexport)
int CppFunc(BSTR *data)
{ ... }
I want that data as a string.
This works
[DllImport( ... CallingConvention = CallingConvention.Cdecl)]
public static extern int CppFunc([MarshalAs(UnmanagedType.BStr)] ref string data);
but it creates a memory leak.
I assume what I should be doing is creating and passing an IntPtr, then Marshal out the Bstr as a string, then free the IntPtr:
IntPtr p = Marshal.AllocHGlobal(512);
CppFunction(p);
string data = Marshal.PtrToStringBSTR(p);
Marshal.FreeHGlobal(p) ;
The problem is, with that code, I get a System.AccessViolationException on the call into Marshal.PtrToStringBSTR(p).
What am I doing wrong?!

The first line of the Remarks for Marshal.PtrToStringBSTR is
Call this method only on strings that were allocated with the unmanaged SysAllocString and SysAllocStringLen functions.
Which is probably where your crash came from.
Add to this your C++ function expects BSTR* (effectively a pointer to a pointer to the first character of data in the string), but you pass it a pointer to data.
Remember that a BSTR has a special structure: it starts with 4 bytes of length, then data, then a null. The pointer points to the first character of data. So Marshal.PtrToStringBSTR is looking backwards from the pointer to find the length of the string - but that isn't memory which was allocated by Marshal.AllocHGlobal.
It could be that your C++ function does something like *data = ....AllocSysString(); - that is, it never reads the string it's given, but instead assigns the pointer to a string which it allocates.
In that case, you probably want something like:
[DllImport( ... CallingConvention = CallingConvention.Cdecl)]
public static extern int CppFunc(out IntPtr data);
...
CppFunc(out IntPtr p);
string data = Marshal.PtrToStringBSTR(p);
Marshal.FreeBSTR(p) ;
Here, we pass it a pointer to a pointer. The C++ function re-assigns the pointer to point to the first character of data in a BSTR, and we use that to deserialize the BSTR, and then free it (using a method which knows how to free BSTRs).
If this isn't the case, it's unclear why your C++ function takes a BSTR* (as opposed to a BSTR), and what it does with it. I think we need to see that before much else can be said.
If your C++ function took a BSTR instead (remember that BSTR is itself a pointer), then what you should be doing is using a StringBuilder (with a particular initial capacity) - the marshalling layer turns that into a pointer the C++ code can write to, and then you can turn the StringBuilder into a string.
[DllImport( ... CallingConvention = CallingConvention.Cdecl)]
public static extern int CppFunc([MarshalAs(UnmanagedType.BStr)] StringBuilder data);
...
var data = new StringBuilder(512);
CppFunction(data);
string result = data.ToString();

PInvoke method that takes char[] of fixed size as input

I have a third party lib that has the following signature:
int GetError(char *Message, int Length)
Here the Message has to be a buffer of at least size Length, otherwise the function fails. How would one PInvoke this? I have tried Stringbuilder for the message, but here I cannot specify the length and the function fails.
The Message parameter is declared as an input parameter to the function.

This is a routine scenario. You are asking the callee to populate a buffer allocated by the caller. That's the domain of StringBuilder.
The p/invoke declaration would be:
[DllImport(dllname, CharSet = CharSet.Ansi, CallingConvention = CallingConvention.???)]
public static extern int GetError(StringBuilder message, int length);
You presumably know what the calling convention is, and can fill it in.
You then call the function like this:
StringBuilder message = new StringBuilder(256);
int retval = GetError(message, message.Capacity);
// check retval for error conditions
string error = message.ToString();

.NET Interop: How to get returned string (not null terminated) from unmanaged DLL in C#

I defined a function in C DLL library.
__declspec(dllexport) void* GetText();
It will return a string which is dynamically allocated from heap memory (And GlobalAlloc is used here for allocating memory). Note that the returned string is not null-terminated.
Then at C# side I tried two methods to declare the function
[DllImport("D:\\ca\\TextAccessLibrary.dll", CallingConvention = CallingConvention.Cdecl)]
static extern String GetText();
When calling above method, the application will crash without any exception thrown.
[DllImport("D:\\ca\\TextAccessLibrary.dll", CallingConvention = CallingConvention.Cdecl)]
static extern IntPtr GetText();
ptr = GetText();
string text = Marshal.PtrToStringAuto(ptr, 1000);
And calling this method will return incorrect string. Checked the real bytes by using Marshal.Copy, I found the bytes value is not same as the value in DLL library. (I think it's caused by Virtual Memory, C# process cannot access memory space of the DLL directly)
(Don't mind the string length, I hard coded it to 1000 for ease)
This is the C++ code and the memory value of the string when debugging (It's a Console Application but not the original DLL, because Console Application is easy to debug. But the DLL code is same as this one except the logging part).
Following is the original DLL code
__declspec(dllexport) char* GetText(){
VTHDOC hDoc = NULL;
VTHTEXT hText = VTHDOC_INVALID;
DAERR da_err = NULL;
DAERR ta_err = NULL;
DAERR read_err = NULL;
char *buf = (char*)GlobalAlloc(GMEM_FIXED, 1000);
DWORD real_size;
DAInitEx(SCCOPT_INIT_NOTHREADS, OI_INIT_DEFAULT);
da_err = DAOpenDocument(&hDoc, 2, "D:\\1TB.doc", 0);
ta_err = TAOpenText(hDoc, &hText);
read_err = TAReadFirst(hText, (VTLPBYTE)buf, 1000, &real_size);
return buf;
}
But at C# side the bytes are not same as C++ side
You can see the first byte in C++ is 0, but it's 200 for C# (decimal)
Another thing to note: if I return a const string(e.g. "AASSDD") directly in DLL code, C# side will get the correct string

You can't do it that way. Marshaling of string works only for null-terminated strings (or for BSTR, if you specify some options). You can:
[DllImport("D:\\ca\\TextAccessLibrary.dll", CallingConvention = CallingConvention.Cdecl)]
static extern IntPtr GetText();
But from there, it isn't clear how the C# program should know the length of the string.
The various Marshal methods of C# handle BSTR (that have internally their length) or NUL terminated strings.

As already stated, it works for null-terminated strings only, in the following way:
C# part, declaration:
[DllImport("myDll.dll", EntryPoint = "myString", CallingConvention = CallingConvention.Cdecl, SetLastError = false)]
extern private static string myString(out int size);
C# part, usage:
int size;
string s = myString(out size);
C++ part:
char* myString(int* size)
{
*size = 20;
char* strg = (char*)::GlobalAlloc(GMEM_FIXED, *size);
memset(strg, 0x3f, *size); //preset with a questionmark
for (int i=0; i < 9; i++)
strg[i] = 0x40 + i;
strg[*size -1] = 0; //limit the maximum string length
return strg;
}
And the obtained C# string:
"#ABCDEFGH??????????", value of size: 20
A treatment of the issue may be found here

What type to PInvoke for a char**

Consider the following C function:
void get_lib_version(const char **ver_string);
How do I marshall this correctly with PInvoke? The documentation says it returns a pointer to a static string. I thought this would do it:
[DllImport(DllPath, CallingConvention = CallingConvention.Cdecl)]
public static extern int get_lib_version(StringBuilder version);
but all I get is gibberish.

The function returns a brand new C-string. The pinvoke marshaller always makes sure that the memory required to store a string that's returned by native code is released again. This will not come to a good end, surely the caller of this function is not supposed to release it. The const keyword is a strong hint that the native code will return a pointer to a string literal that's not allocated on the heap. Trying to release such a pointer will crash your program on later Windows versions, the kind that have a strict heap implementation (after XP).
You have to help to stop the marshaller from doing this. This requires you to declare the argument as a raw pointer, not a string:
[DllImport(DllPath, CallingConvention = CallingConvention.Cdecl)]
public static extern int get_lib_version(out IntPtr version);
And you have to make the extra step to convert the pointer to a string:
public string GetLibraryVersion() {
IntPtr strptr;
get_lib_version(out strptr);
return Marshal.PtrToStringAnsi(strptr);
}
Write a little test program to verify this assumption. Call GetLibraryVersion() a billion times. If the memory usage doesn't explode then you're good.

According to this answer, when you marshal something as string, PInvoke makes all sorts of assumptions about how it's supposed to get freed. Notice that this is a const char *; it's a constant string somewhere. It never needs to be deallocated!
Apparently the way to deal with this is
Marshall as IntPtr.
Use Marshall.PtrToStringAnsi() to copy the result into a C# string.
I managed to get this to work correctly:
[DllImport(DllPath, CallingConvention = CallingConvention.Cdecl)]
private static extern int get_lib_version(ref IntPtr version);
public static string GetLibVersion()
{
var ptrVersion = IntPtr.Zero;
get_lib_version(ref ptrVersion);
var version = Marshal.PtrToStringAnsi(ptrVersion);
return version;
}

Marshalling char arrays in the Compact Framework

On the compact framework (3.5) I have a pinvoke as follows:
[DllImport("VAMDll.dll", SetLastError = true, EntryPoint = "openLiteSpatial")]
public static extern int OpenLiteSpatial(byte[] mode, byte[] data, byte[] capture, IntPtr callback);
For the following native function sig:
int openLiteSpatial(char *mode, char *data, char *capture, void callBack(char *, char *));
Passing parameters inwards works fine like this:
Native.OpenLiteSpatial(Encoding.UTF8.GetBytes(Mode.FunctionalLocation), Encoding.UTF8.GetBytes(functionalLocation), Encoding.UTF8.GetBytes(capture), Marshal.GetFunctionPointerForDelegate(callback));
But in my callback I have to do some strange transformations on the callback parameter strings:
public delegate void LiteSpatialCallback(string attributeData, string geoCoordinate);
public static void Callback(string attributeData, string geoCoordinate)
{
byte[] decodedAttributeData = Encoding.Unicode.GetBytes(attributeData);
byte[] decodedGeoCoordinate = Encoding.Unicode.GetBytes(geoCoordinate);
attributeData = Encoding.UTF8.GetString(decodedAttributeData, 0, decodedAttributeData.Length);
geoCoordinate = Encoding.UTF8.GetString(decodedGeoCoordinate, 0, decodedGeoCoordinate.Length);
}
I have tried specifying the Charset on the DllImport attribute, and using byte[] arrays (which causes the callback to not get called at all), and specifying various MarshalAs options.
Is there a better way?

I'd be inclined to pass IntPtrs for the delegate parameters, then use the Memory windows to see exactly what's coming back in them. Is it a pointer to data? A pointer to a pointer? The data itself? Is it Unicode or ASCII? All these could be determined by walking the addresses returned. You could then either just manually copy from that target, or you could re-work the P/Invoke and/or delegate signature based on your findings.