Managed C++/CLI

13-Dec-02 9:49

I feel your pain, since there is so little info on how __gc and __nogc stuff interacts in MC++. One thing is that any __gc class you create, should inherit from IDisposable when there are any heap (gc heap or regular heap) members to deallocate. Regular C++ destructors are not treated quite the same in __gc classes. For one thing, they are considered a last resort, and synonymous with Finalize methods defined in .Net. They are supposed to be declared protected to hide them from users, and should themselves just call the Dispose method that IDisposable descendants are supposed to implement.

It is actually in the Dispose method where deallocation is supposed to occur (this all seems complicated, I know). Here is a simplified snippet from one of my __gc classes that implements IDisposable. I left in some comments from the .Net docs:

public __gc class GcClass : public IDisposable {<br />
<br />
   char *m_pBuf;<br />
<br />
public:<br />
   GcClass() {<br />
      m_pBuf = new char __nogc[BUF_SIZE];<br />
   }<br />
<br />
   void Dispose() {   //required by IDisposable to be made avail to users<br />
      Dispose(true);  //true indicates user-called<br />
       // Take ourselves off of the Finalization queue to prevent<br />
       // finalization code for this object from executing a second time.<br />
      GC::SuppressFinalize(this);<br />
   }<br />
<br />
protected:<br />
   ~GcClass() {        //CLR dtors are supposed to be protected<br />
      Dispose(false);  //DO NOT delete[] m_pBuf here!<br />
   }<br />
<br />
   // Dispose(bool disposing) executes in two distinct scenarios.<br />
   // If disposing equals true, the method has been called directly<br />
   // or indirectly by a user's code. Managed and unmanaged resources<br />
   // can be disposed.<br />
   // If disposing equals false, the method has been called by the <br />
   // runtime from inside the finalizer and you should not reference <br />
   // other objects. Only unmanaged resources can be disposed.<br />
   virtual void Dispose(bool bWeAreDisposing) {<br />
<br />
      if(this->m_pBuf) {          //if Dispose has not been called yet<br />
         if(bWeAreDisposing) {    //If disposing, dispose all gc resources. <br />
            //only release managed resources here<br />
         }<br />
          //Release unmanaged resources. If bWeAreDisposing is false (called by runtime), <br />
          // only the following code is executed.<br />
          //Note that this is not thread safe. Another thread could start<br />
          // disposing the object after the managed resources are disposed,<br />
          // but before the disposed flag is set to true (same as m_pBuf set NULL in this class).<br />
         delete[] m_pBuf;<br />
         m_pBuf = NULL;     //serves to flag that we are disposed of<br />
      }<br />
   }<br />
//...<br />
};

The __nogc syntax is not required; I just leave it in to remind myself what's not gc-collected. Lookup Dispose and IDisposable in the docs to see more (though rather scant Frown | :(

) info on this stuff.

I imagine your code should still work, since destructors would eventually get called by the framework, but are getting invoked very late. I gather from the docs that Dispose is called earlier (where delete[] should be), and thus might avoid the leak reportings you are currently getting. Nevertheless, __gc code might always look leaky to regular C++ detectors. I hope this helps...

Cheers

Conversion problem

Anthony_Yio12-Dec-02 15:12

Anthony_Yio

12-Dec-02 15:12

Hello all,

Does anyone know how to convert System::String to C++ language char * ?
I found a few functions but it convert to wchar_t which is not what i want.
Any hints?

Secondly, i have a unmanaged c++ function in which prototype in such.
BOOL MImImage::Combine(MImImage *pImImage);

why is that the parameter appear as class NamespaceXX.MImImage in ILDASM? I want a pointer parameter so that i could pass in an array of MImImage.

Thanks in advance.

Re: Conversion problem

Jeff J12-Dec-02 22:12

12-Dec-02 22:12

If you haven't already, you can use Marshal::StringToHGlobalAnsi() in the System::Runtime::InteropServices namespace. That will allocate to regular heap which needs to be freed by calling FreeHGlobal().

If you want to copy the bytes to your own memory space/buffer, you can do it with a little extra work. First, pin the String on the GC heap (tell the GC not to move it), and then convert like any wchar_t array. Roughly, given CLR String "Str":

inline wchar_t * PtrToStringArray(System::String *s) {<br />
   System::String __pin*pps = s;  //pin to avoid 1-instruction GC hole in reinterpret_cast<br />
   System::Byte __pin*bp = reinterpret_cast<System::Byte *>(s);<br />
   if( bp != 0 )<br />
    	bp += System::Runtime::CompilerServices::RuntimeHelpers::OffsetToStringData;<br />
   return reinterpret_cast<wchar_t*>(bp);<br />
};<br />
<br />
wchar_t *pStr = PtrToStringArray(Str);  //pin and access array<br />
<br />
int iChrs = WideCharToMultiByte(CP_ACP, 0, pStr, Str->Length, pYourBuf, iBufLen, NULL, NULL);<br />
<br />
//do whatever with ANSI string in pYourBuf

PtrToStringArray() is my version of PtrToStringData() from vcclr.h, which keeps a String pinned, and gets a pointer to its internal wchar_t array. I use it often in MC++ wrappers. The pinning will release when the block that contains PtrToStringArray() exits.

As to your second question, all __gc objects are passed around as pointers, and there is no assumption about pointers possibly being to an array of objects. To do so, you need to declare the param as a gc array:

BOOL MImImage::Combine(MImImage *ImImageArray __gc[]);

The CLR has a strange syntax for declaring arrays, but that's it.

Cheers

Re: Conversion problem

Anthony_Yio12-Dec-02 22:28

Anthony_Yio

12-Dec-02 22:28

Thanx. Your reply really comes in at a good time.
Smile | :)

Re: Conversion problem

Jeff J13-Dec-02 8:57

Object * to (void *,size_t) and back

13-Dec-02 8:57

My pleasure. However, I forgot there were embedded < and > characters in my last post, so the code was clipped. Here is the correct code displayed as non-HTML:

inline wchar_t * PtrToStringArray(System::String *s) {
System::String __pin*pps = s; //pin to avoid 1-instruction GC hole in reinterpret_cast
System::Byte __pin*bp = reinterpret_cast<System::Byte *>(s);
if( bp != 0 )
bp += System::Runtime::CompilerServices::RuntimeHelpers::OffsetToStringData;
return reinterpret_cast<wchar_t*>(bp);
};

wchar_t *pStr = PtrToStringArray(Str); //pin and access array

int iChrs = WideCharToMultiByte(CP_ACP, 0, pStr, Str->Length, pYourBuf, iBufLen, NULL, NULL);

//do whatever with ANSI string in pYourBuf

Cheers

Help!!!!!

dyerstein11-Dec-02 9:20

dyerstein

11-Dec-02 9:20

Porting problem???

I have a set of defines:
#define PROT_NACK_NONE 0
#define PROT_NACK_BYE 1
…
…
…
#define PROT_NACK_BAD_PROT 3
________________________________

And I also have a typedef structure:
typedef struct reject_s
{
short int rejectCode;
const char *message;
} reject_t;

reject_t *m_rejectTable;
_________________________________
I than implement my function which uses the above:

void CMessageProtocol::setRejectTable( )
{//private:
static reject_t reject[]=
{
{ PROT_NACK_NONE, ""},
{ PROT_NACK_BYE, "Bye"},
{ PROT_NACK_NO_MORE_CON, "No More Connection availiable"},
{ PROT_NACK_BAD_PROT, "Protocol version not supported"},
{ PROT_NACK_BAD_ENC, "Encryption version not supported"},
{ PROT_NACK_BAD_APP, "Application refused"},
{ PROT_NACK_BAD_LOG_PASS, "Invalid Login or password"},
{ PROT_NACK_BAD_MSG, "Invalid Message format"},
{ PROT_NACK_DEFAULT, "Error"},
{ 0, 0}
};

m_rejectTable = reject;
}

The above has been complied and runs under Microsoft Visual Studio 6

I am now attempting to implement this into a .NET library, that has a bunch of __gc classes, when I now attempt to convert using the following syntax:

typedef __nogc struct reject_s
{
const char message;
short int RejectCode;
} reject_t;
reject_t* m_rejectTable;

typedef __nogc struct encrypting_s
{
char crypto;
char * encryptingName;
unsigned char cypher[2];
} encrypting_t;
encrypting_t* m_encryptingTable;

when I attempting to implement the __gc class method that use’s the above:

void GEMBrokerMessengerProtocol::CGEMMessageProtocol::setRejectTable()
{
static reject_t reject[]=
{
{ PROT_NACK_NONE, ""},
{ PROT_NACK_BYE, "Bye"},
{ PROT_NACK_NO_MORE_CON, "No More Connection availiable"},
{ PROT_NACK_BAD_PROT, "Protocol version not supported"},
{ PROT_NACK_BAD_ENC, "Encryption version not supported"},
{ PROT_NACK_BAD_APP, "Application refused"},
{ PROT_NACK_BAD_LOG_PASS, "Invalid Login or password"},
{ PROT_NACK_BAD_MSG, "Invalid Message format"},
{ PROT_NACK_DEFAULT, "Error"},
{ 0, 0}
};

m_rejectTable = reject;
}

I get the following error under Visual Studio C++.NET

c:\Documents and Settings\mdyer\My Documents\Visual Studio Projects\CGEMBroker\GEMBrokerMessengerProtocol.cpp(22): error C2440: 'initializing' : cannot convert from 'int' to 'GEMBrokerMessengerProtocol::CGEMMessageProtocol::reject_t'

c:\Documents and Settings\mdyer\My Documents\Visual Studio Projects\CGEMBroker\GEMBrokerMessengerProtocol.cpp(22): error C2689: Initializing 'reject' : this form of non-aggregate initialization requires a unary constructor

I am very curious as to why I am having these compile errors under .NET and not under my previous compile of Visual Studio 6???? Very lost here????

Anonymous11-Dec-02 4:55

Anonymous

11-Dec-02 4:55

Class MX is a managed wrapper for the unmanaged (and unmodifiable) class X.
If you look the code below you'll see the main question: how do I get a void * pointer (to a size_t memory array) from an Object obj? And then, how do I come back from data and size to obj?
The only restriction I can admit on obj is that it has to be serializable.
I suppose I could use a MemoryStream and a BinaryFormatter but I can't figure how.

Thanks to anyone who can solve this problem.

Rose | [Rose]

public __gc class MX {<br />
private:<br />
 X * x;<br />
<br />
public:<br />
 MX(Object * obj) {<br />
  void * data;<br />
  size_t size;<br />
  <br />
  /*Problem 1*/<br />
<br />
  x = new X(data,size);<br />
 }<br />
<br />
 Object * get_Object() {<br />
  void * data = x->get_data();<br />
  size_t size = x->get_size();<br />
  Object * obj;<br />
<br />
  /*Problem 2*/<br />
<br />
  return obj;<br />
 }<br />
}

Re: Object * to (void *,size_t) and back

Anonymous15-Dec-02 22:50

Anonymous

15-Dec-02 22:50

I've a similar problem and I don't know how to solve it. I hope some Guru here knows the solution.

Jenny.

C.H.Y.S.P.R.xxx

MFC app - porting

MattG10-Dec-02 8:35

MattG

10-Dec-02 8:35

Hi All

I have come to the conclusion that i should ditch MFC and go the .NET way.

I have a dialog based MFC project 25% complete. Is there a way of been able to use these classes and forms from .NET (managed C++ or C#) so i don't have to rewrite everything.

Thanks

Matt

Matt G

Maximilien10-Dec-02 8:50

Maximilien

10-Dec-02 8:50

Harggg!!!

I don't think so!

You need to re-program the UI ( no builder ), and if your current MFC app is very UI intensive, it can take some times.

Max.

MattG10-Dec-02 9:28

MattG

10-Dec-02 9:28

Thanks Max,

Do you agree its probably better to move it now rather than the whole project later? As i get the feeling we'll all be going the C# way.

Cheers

Matt G

Maximilien10-Dec-02 13:43

Maximilien

10-Dec-02 13:43

There are 2 possibilities :

1- move now :
pros : less code to port; chance to learn a new technology.
cons : learning curve and false expectations from management.

2- move later :
pros : not much pros here, except maybe a better knowledge of the application domain.
cons : lot's of re-programming; maybe hard to convince your boos to pay to redo something that's already working.

Whatever the choice you take, I'd suggest trying to separate the UI from the core as much as possible.

If I was in your position, I'd move now to .net.

Max.

MattG10-Dec-02 14:03

MattG

10-Dec-02 14:03

From a personnel point of view (and maybe project), i think i'm heading out to buy my first C# book.

Thanks for your advice Max

Matt G

Linker Errors

Paul Ingles8-Dec-02 12:23

Paul Ingles

8-Dec-02 12:23

I'm trying to interoperate with an external library (the C interface for K -- mentioned in the lounge a short while ago -- www.kx.com[^])

I've added the include statements, and added the .lib to the list of files to link with, however, I still get:

KdbTest1 error LNK2001: unresolved external symbol "int __cdecl cd(struct k0 *)" (?cd@@$$FYAHPAUk0@@@Z)
KdbTest1 error LNK2001: unresolved external symbol "struct k0 * __cdecl gi(int)" (?gi@@$$FYAPAUk0@@H@Z)
KdbTest1 error LNK2001: unresolved external symbol "struct k0 * __cdecl gnk(int,...)" (?gnk@@$$FYAPAUk0@@HZZ)
KdbTest1 error LNK2001: unresolved external symbol "struct k0 * __cdecl ksk(char *,struct k0 *)" (?ksk@@$$FYAPAUk0@@PADPAU1@@Z)
KdbTest1 fatal error LNK1120: 4 unresolved externals

I'm trying to build a small Managed C++ console application to try interoperability with Managed C++. All the files can be obtained from http://www.kx.com/a/k/connect/c/[^]

Any suggestions would be very much welcomed.

--
Paul
"If you can keep your head when all around you have lost theirs, then you probably haven't understood the seriousness of the situation."
- David Brent, from "The Office"

MS Messenger: paul@oobaloo.co.uk
Sonork: 100.22446

Re: Linker Errors

Nick Parker9-Dec-02 15:17

Nick Parker

9-Dec-02 15:17

Paul Ingles wrote:
Any suggestions would be very much welcomed.

Ok, after a little looking I came across this[^] knowledge base article, however you may look into this[^] KB article as well, hope this points you in the right direction. Smile | :)

Nick Parker

Not everything that can be counted counts, and not everything that counts can be counted. - Albert Einstein

ans required?

imran_rafique7-Dec-02 9:44

imran_rafique

7-Dec-02 9:44

Q1-if a system is in loggoff state and i call
GetUserName(..).
is the function fails,and return value is zero?
and how did i know that the system is in loggoff state?
//////////////////////////////////////////////////////////

Q2-how i will trap the WlxDialogbox and where msgina asks for the userid,password
and fill in the fields by my self?
can any body help me in this regard?
//////////////////////////////////////////////////////////

Q3-how did i find the system32 path programatically?
//////////////////////////////////////////////////////////

Q4-if i call BlockInput(true)
and then i want to use postmessage or sendmessages for keyboard and mouse
will these messages works?
/////////////////////////////////////////////////////

r00d0034@yahoo.com

r00d0034@yahoo.com

C++ vs. Managed C++

SanShou6-Dec-02 4:50

SanShou

6-Dec-02 4:50

I know I risk getting a huge flame Cry | :((

but, what the heck its winter and I could use some heat to spice up my life...

I admit I have been out of it for a while, but I am struggling with the differences (benefits and analysis) between the different technologies out there. I understand and use Visual C++ (presumably unmanaged since my company doesn't want to spring for Visual Studio .Net or training... bahh). However I am seeing a lot of traffic on the managed C++ front. I am currently trying to catch up and learn the new technology (I had to buy Visual Studio.net for myself for Christmas...) and I was wondering, can anyone explain some things?

First what are the benefits for Manage C++ versus the regular C++? Things like speed, code usage, MFC usage vs. Win 32 usage, syntax? Also what is the difference between these and C#? Let's say I am writing a high performance application like a game, which of the three new language technologies are worth it: Unmanaged C++, Managed C++, or Big Grin | :-D

C#?

Second, how hard is it to convert over from my old unmanaged habits to the managed version? I am trying to decide whether it is worth making the change, even though I cannot do it where I work presumably...

I guess I could ask what other technologies are out there that I should concentrate on learning... I already want to learn XML (and yes I do not know it yet... OMG | :OMG:

), but I figure I am going to get flamed enough.

I thank anyone for their help, and I am attempting to research it myself, but between working on finishing up the project that will keep 20 some people employed going in to the new year, deciding that I have had enough of training for international level Kung Fu competitions (my hips, knees, and shoulders really could use a warm winter Mad | :mad:

not to mention they are they only keeping me awake while I work late into the night Sleepy | :zzz:

), and dealing with these Christmas things such as buying presents.... I am sure everyone knows how it goes, but I could use the help focusing my learning and growing, not to mention actually understanding...

Thanks,
Brian

Now assembling the pyre... waiting for the first spark... Cool | :cool:

If you start a fire for a man, he will be warm for a day. If you start that same man on fire, he will be warm for the rest of his life.

Jeff J6-Dec-02 12:55

6-Dec-02 12:55

I do not flame others, so you'll have to get your heat elsewhere Smile | :)

Managed C++ is essentially an extension to the .Net framework stuff (the Common Language Runtime, etc.), in C++ syntax. It is possible to put regular C++ and MC++ in the same project (even the same file). MC++ exposes the same runtime library as all other CLR-based languages, such as C#. Thus, much of the difference between MC++ and C#, is that MC++ exposes the existence of pointers, and C# calls them "reference objects" and bypasses the pointer syntax.

Picking up C# with a C++ background is really easy, and even easier if you have experience with Java. Several new keywords, but most of the stuff is the same. I cannot call using MC++ really changing, as it's more adapting to using some CLR library objects in a familiar syntax. Any class you define as a __gc class, you can use CLR objects in. There is no MFC, ATL, or Win32 API directly in the CLR itself, though you can write wrapper declarations around external DLL functions, and call them from within any .Net code. The CLR is comprehensive, so you won't often have to anyway.

Performance wise, there is little difference between the CLR-base languages, though MC++ allows integration with regular C++, which Microsoft freely admits will remain the speed demon for many things. Well, COM objects still suffer weight problems and circular references, and traditional heap allocation is not as fast as .Net's garbage collected (GC) heap. .Net objects are lightweight and 'new' allocation is really fast, although low-level manipulation is harder with .Net. However, it is possible in C# (research the unsafe and fixed keywords), and MC++ has the best access, although pinning GC memory (keeping it from moving) has a slight penalty. Thus, MC++ has some advantages, but C# is the cleanest syntax.

Some main benefits to CLR-based stuff is rapid development, greater platform independence, a good GUI designer, a good security model, web integration, and the like. Considering such a wide scope, the performance is pretty good, but not higher performance overall than C++. The CLR languages do not replace C++, they are simply better suited to certain tasks and projects.

Naturally, I expect to be flamed as well. Anyone can find an excuse to argue. If every man who's been put on fire stays warm for the rest of his life, then call me Mr. Toasty...

Cheers

Nitron7-Dec-02 12:49

Nitron

7-Dec-02 12:49

Jeff J wrote:
If every man who's been put on fire stays warm for the rest of his life, then call me Mr. Toasty...

lol!

But seriously, I am exploring the same origional question. Nicely put.

- Nitron

"Those that say a task is impossible shouldn't interrupt the ones who are doing it." - Chinese Proverb

SanShou9-Dec-02 11:02

SanShou

9-Dec-02 11:02

Thank you,

That was very imformative and very helpful. It gives me a good basis on what the differences are and where to start learning. I am very interested in picking up additional knowledge so I am glad people are willing to share, especially since nobody at work is available to discuss things with. The benefits of the very small company work place may be appealing but there are disadvantages to knowledge growth.

Anyway thanks again, I really appreciate it.

Brian

If you start a fire for a man, he will be warm for a day. If you start that same man on fire, he will be warm for the rest of his life.

Nick Parker7-Dec-02 5:08

Nick Parker

7-Dec-02 5:08

First off I would like to point out that Jeff did an excellent job covering the main jist of the languages, as you asked a very important question. Here are a few links to review that may help:

Here is a nice analysis between I whole slew of Microsoft languages, both managed and not:
A statistical analysis of the performance variations of assorted managed and unmanaged languages[^]

SanShou wrote:
which of the three new language technologies are worth it: Unmanaged C++, Managed C++, or C#?

SanShou wrote:
I already want to learn XML

The best way for me to answer this is to say that you don't have to particularly learn only one of the languages, they work together almost flawlessly. Take a look at my most recent article to point this out:
Cross Language Web Service Implementation[^]

Hope this helps. Smile | :)

Nick Parker

Not everything that can be counted counts, and not everything that counts can be counted. - Albert Einstein

SanShou9-Dec-02 11:05

SanShou

9-Dec-02 11:05

Thanks.

Yes it did help. Those articles were very interesting and provided a good look in to the architecture and the speeds, not that most of my code is tightly looped for optimization myself, but it was defitinely enlightening. Of course I also need to read them more thorougly when I get a chance.

They also pointed to a few new buzz words and technologies I am going to have to reasearch such as ngen'd. However that is the fun in trying to catch up and learn new things... Big Grin | :-D

Thanks,
Brian

If you start a fire for a man, he will be warm for a day. If you start that same man on fire, he will be warm for the rest of his life.