Comments by Maxim Kartavenkov (Top 155 by date)

5.
One note: the casting to the reference raises runtime exception if cast not available, if exceptions not handled in application then better to use casting into pointer type:

auto pAB = dynamic_cast<AB *>(ab);
if (pAB) { // if casting succeeded then pointer is not null
   pAB->show();
}

You can use .NET and C++/CLI or C# for easy formatting output

Yes, but in C/C++ language you need to specify what you want to print it can be string or pointer or single character.

in your code you also do not transform the string because '%' character in FormatMessage API have different meaning. Use printf function format specification depending on what's your input arguments format: Format specification syntax: printf and wprintf functions
For example if you want to set integer value at the place you have "%" then use "%d".

Thanks

What happening if you setup explicitly platform while you building your application? Try x64 or x86 instead of "Any CPU" in the configuration.

5.

I see, but he will be asking why your solution not working for his code.

5.

You didn't look at the code in the question:

 A *a1 = new A();
*(a1->a1) = 99;

Your code will crash on assignment.
Just make you note for improving your solution to become perfect instead of posting my own which is not needed here.

Add Set(0) in default constructor -to allocate a1 pointer and set 0 value as in example he creates default object and assign value to a1 after.
Also add destructor to show how to free memory.

That link is related to DWM you are not using DWM in your code. From my point you have to many painting things: So you should review all of them, first disable erase background - you should skip it by returning TRUE, as the default handler cause entire window to be invalidated and it invalidated with the brush you specified during window creation. In your painting you fill the background with the color compare to brush in created window and on sizing you call InvalidateRect with erasing background plus have the open ogl painting. The way doing rendering is better to have no default painting for the client area of the window you rendering plus disable background erasing on resizing resize the render target.

You should disable drawing for the child window in the window dispatch routine: childProc and skip erasing background. On Sizing message necessary to resize render target, but not sure if that require for OpenGL.

If you don't include gdiplus.h file directly then in VS Open project properties -> C/C++ ->Advanced -> Show Includes - this way during build you will see in output the sequence of included headers so you can figure out who request the gdiplus things

Yes you can set it with linker option: with project properties: Linker -> Input -> Ignore Specific Default Libraries or with #pragma comment( lib, "/nodefaultlib:<name>")

Your calculation was fine based on the array type, I just explain from the bytes base so he will be less confused.

600 is the address of the value - **matrix gives you the actual value - you mix the correct and non correct ideas.

Yes the int* p - single pointer and the *p - gives you the value, int ** pp, double pointer or pointer to pointer, so *pp gives the pointer and **pp gives you the value.
which example you mean is incorrect?

*(matrix + 0) gave you value same as matrix[0] which is address as your matrix is int**, matrix[n] will be have type int*, matrix[n][m] will have type int.
printf("%p", ptr) display you the address variable var = which equals to &var.
You don't have anywhere 4 bytes pointers in your examples.
To retrieve amount of pointer bytes you can use sizeof(int*)
sizeof(int*) and sizeof(int**) are the same size and those size of pointers compare to sizeof(int) is the size of the value.
Seems you asking the same questions all the time without trying to understand the answers.
I remember that I explaining that to you already 3 times.

Just to clarify some points: the matrix[0] is allocated and it has the type of (int *). so the matrix[1] value will be shifted by the columns * sizeof(int) number of bytes which is 20 bytes (5 * 4) as you have matrix typed pointer of int, the matrix[2] will be shifted by 40 bytes from matrix[0] and by 20 bytes from the matrix[1].

Yes that is correct. The same example I already gave you in your in the comments to the previous question and even explain to you how it works.

matrix[0] at the address of 500 have the value of 600 and at this address written the array of integer with size of 20 bytes = columns * sizeof(int)
matrix[1] at the address of 504 has the value of 620 which means matrix[0] + columns * sizeof(int);
matrix[1] - matrix[0] has the value of 4 bcs it is for x86 platform which has pointer size of 4 bytes

In second time you just access second integer value in the array which is 4 bytes long

0x55eff083e2d4 - 0x55eff083e2c0 yep you right - this is the difference between matrix[1] and matrix[0] so in your example its equals to columns * sizeof(int);

The value of 0x55ae7f302013 seems was not from the code you specify but from some of your experiment.
arr + 1 means arr[1] which is pointer
*(arr + 1) means arr[1][0] which is character
&arr + 1 means arr + sizeof(arr) which has invalid address

what you try to get by the &arr + i ?
0x7fffcac41540 - 0x7fffcac41528 = 0x18 which is equals to sizeof(arr);
0x7fffcac41528 - 0x7fffcac41510 = 0x18 - same

0x7fffcac41520 - 0x7fffcac41518 = 0x08
0x7fffcac41518 - 0x7fffcac41510 = 0x08

on x86 platform pointer 4 bytes long, on x64 platform it is 8 bytes

In the example you mentioned they showing the addresses for the x86 platform there size of pointer is 4 bytes long
as you have an array of pointers to another array so you at the top have an array of pointers.
Size of each pointer is the 4 bytes so if the address of the first element is 2000 then next array element is 2004
As you have pointer type then at the address of 2000 you contains the pointer to your string but not the actual string data.
This pointer is the address of the first character in the string. So at the address 2000 - first array element may be stored the value 2020 address of the string.
So at the 2020 you will have 'g' at 2021 - 'f' 2022 - 'g' and at 2023 - '\0'

Just remember that I also face such thing long time ago. But this works fine with own allocated arrays as alignment will be same in that case:

int mat[3][3] = { {0,1,2},{3,4,5},{6,7,8} };
int rows = 3;
int cols = 3;
int ** matrix = (int**)malloc(rows * sizeof(int*)+ rows * cols * sizeof(int));
matrix[0] = (int*)(((unsigned char*)matrix) + rows * sizeof(int*));
for (int i = 1; i < rows; i++)
    matrix[i] = (int*)((unsigned char*)matrix[0] + i * cols * sizeof(int));
for (int i = 0; i < rows; i++)
    for (int j = 0; j < cols; j++)
        matrix[i][j] = mat[i][j];
int **ptr = (int **)matrix;
printf("%d\n", mat[1][1]);
printf("%d\n", matrix[1][1]);
printf("%d\n", ptr[1][1]);

The array of characters its the string
you can define it as char text[100] - this array can contains up to 100 characters
most of the common library strings functions operates with the zero-ending strings, this means that the last character in the string is the '\0'
so our string may have 99 characters as we should keep one character.
In the string you can access the each character by the index.
In the example we iterate trough all characters in the string and display them.

char text[100] = "Some text";
for (int i = 0; i <  strlen(text); i++) { printf("%c\n",text[i]); } 

As you can see we uses the "%c" in format string which display the single character.
Now compare it to the next code.

char text[100] = "Some text";
for (int i = 0; i <  strlen(text); i++) { printf("%s\n",&text[i]); } 

Here we are using the "%s" format which output the zero-ending string.
As this format expect not single character we specify ampersand to get reference of the character.
In your example you have array of array of characters or array of strings:
char* arr[3] = { "geek", "Geeks", "Geeksfor" };
See that you have type of array "char *" - this means that pointer to character.
with such initialization it point to zero ending string which is hardcoded.
in your case arr[0] - character pointer to the "geek\0", arr[1] - "Geeks\0"
The '\0' character added automatically with such definition.
As in your example it uses the "%s" format - it display whole string.
To display single character you can use the "%c" format but it need you you pass character instead of array of characters, so you should use the array index or "*" to get data from the address

char* arr[3] = { "geek", "Geeks", "Geeksfor" };
// display entire string
for (int i = 0; i <  3; i++) { printf("%s\n",arr[i]); } 
// displays the first character
for (int i = 0; i <  3; i++) { printf("%c\n",*arr[i]); } 
// also displays first character
for (int i = 0; i <  3; i++) { printf("%c\n",arr[i][0]); } 
// displays second character
for (int i = 0; i <  3; i++) { printf("%c\n",arr[i][1]); } 
// Compare that with the previous output
for (int i = 0; i < 3; i++) { 
    printf("-----\n");
    for (int j = 0; j < strlen(arr[i]); j++)
          printf("%c\n", arr[i][j]);
}

In next example you allocate memory for the single element in the second dimension of the array.
so accessing arr[0][1] - cause crash:

int* arr[5];
for(int i=0; i<5; i++) {
   arr[i] = (int*)malloc(sizeof(int));
   *arr[i] = i;
}

In that example you have array of the pointers to int with the 5 elements
you initialize each pointer by allocating memory with size of int - which is single element.
But you can access that one element with arr[0..4][0].
To access more elements you should allocate larger memory

int nb_elements = 20;
arr[0] = (int*)malloc(sizeof(int) * nb_elements);
for (int i = 0; i < nb_elements; i++) arr[0][i] = i;

The arr[5][1] same as arr[5] is out of bound of the array so you got the crash as index should be in range from 0 to 4 inclusive.
The operators "*" and "&"- used with typed pointers or values only.
The operator "*" used to retrieve value at the specified typed pointer.
The operator "&" used to retrieve address of the specified typed variable.

// Variable in stack
int a = 20;
// Pointer o that variable
int * p = &a;
printf("%d\n",*p); // display the value
printf("%d\n",p[0]); // display the value
int ** pp = &p; // Pointer to pointer - now "pp" contains address of the variable "p"
// to access the variable "a" we should use the two times "*" 
printf("%d\n",*(*pp)); // display the value
printf("%d\n",(*pp)[0]); // display the value
printf("%d\n",pp[0][0]); // display the value
printf("%d\n",*(pp[0])); // display the value

Regards,
Maxim

Good point - that is correct for static defined array, but nobody saying that you can access all elements of the fixed multidimensional array with casting to int** in same way - this way you are pointing to the first line only. Accessing all elements this way is possible if the memory allocate as the single block and/or without alignment - for example int** will work on x86 platform, as on x64 you have different pointer size, this way you can access the ptr as one dimension array from 0 to 9.
In your case we don't know the array bounds for the resulted ptr value and the alignment for the element or for the row will be different.
Btw there is a bug in your code: int **ptr = (int **)&mat; should be int **ptr = (int **)mat;

Yep, that’s why I suggested to change type of “y” to unsigned to see the magic. It is also possible to print output as hex with “%x” format.

👍

MSVC does not supports inline __asm directive on x64 platform build - so he build code on x86 platform this means there is no 64 bit registers available

He is on msvc

make the y type as unsigned int...

It installs also fine. Worked fine and able to be attached with the local debugger. You need to look at the EvtIoDeviceControl implementation VirtualSerial2\queue.c it already contains the IOCTL handlers, see switch statement, but you should define your own codes: see IOCTL_SERIAL_SET_BAUD_RATE, use other value of the device type instead of FILE_DEVICE_SERIAL_PORT - you can define your own value here, and use property index value which is next argument in CTL_CODE macro, other values are related to how you handling data transfer from/to user space, just use the same way as in that sample.

I just checked, the VirtulSerial2 is the driver based on UMDF Version 2, that's why it looks like the KMDF. I had build it and the result is the DLL output. I build without changing anything in the code.

I suggest to use VirtualSerial instead of VirtualSerial2 as first one is UMDF and second is KMDF - UMDF easy to debug and implement. You can find it in this commit
https://github.com/microsoft/Windows-driver-samples/commit/97cf5197

In my I uses snippet from your - to be sure it works similar. There is only changes I mention. Test code is fully workable - yes you can use it as a basis, just make sure it works for you after build - so the problem can be in something else.

Ok, then if it does not work compare example which I made with your code. You can build this example as console application. The destination bitmap updated every repaint call with random color. Also in the output buffer you may not specify alpha value, but initialize render target which check for alpha, so your data output is fully transparency, so set alpha value to 0xff - like I do in random color preparing.

I see, you probably forgot to call triggering window redraw such as InvalidateRect or InvalidateRgn or RedrawWindow once your new frame comes up, as you mentioned that drawing called from WM_PAINT. You also can draw outside of WM_PAINT once the new frame comes and buffer updated. But on WM_PAINT just draw the latest frame in case if streaming stopped

You can look how I design managed/unmanaged callback linkage here: https://www.codeproject.com/Articles/5337264/NET-Wrapper-of-FFmpeg-Libraries#Delegates%20and%20callbacks
the sources are available in github

As I understand your code is already half managed as you are use the C++/CLI. Right, your callback should be managed, but from that managed part you can call unmanaged things.

pDirect3DDevice->GetBackBuffer(0, 0, D3DBACKBUFFER_TYPE_MONO, &pDestTarget ) ;

hr = pDirect3DDevice->GetRenderTargetData(pRenderTarget,pDestTarget);

From application you can use IQualProp Interface and get_AvgFrameRate method query this interface from video renderer
https://msdn.microsoft.com/en-us/library/windows/desktop/dd376916(v=vs.85).aspx

Sure your code will not work.
You are even not even read mine answer, but asking the same!!!
You are not copy back buffer data to your surface in your code.

1. Create Device with the ARGB32 back buffer format.
2. Create offscreen surface not with the resolution which you take from GDI but from the back buffer resolution (they can be not the same)
3. Get back buffer surface and copy to your surface
4. Save surface into a file.

Regards.
Maxim.

Create destination surface in system memory pool instead of scratch.
Maxim.

If you not able to build filter graph then you not have codecs that required to split the TS files. VLC have it embedded but for DirectShow you need the splitter. Try to download K Lite codecs pack - in there Haali splitter and LAV splitter which allows to handle TS files.
Maxim.

The ffdshow filter internally can query for application name and according to that enable/disable some features according blacklist or white list.Also ffdshow have multi instancing property, check for the filter settings.
Another thing that Probably GraphStudioNext is x64 and you creates x86 application - so the filters are loading are different and have different settings loads from registry.

Maxim.

The user account from which you currently launching the service does not allows you to perform operation you trying to do in the service, that's why you getting access denied exception. Change the user account from which service will be started as I described.

Initialize ServiceProcessInstaller object properties: Account, Password and Username.

http://en.wikipedia.org/wiki/Covariance
Isn't complexy of implementation in C#.

then use memcmp(&ip->ip_dst,&ip->ip_src,sizeof(ip->ip_dst)) == 0 instead

inet_ntoa - returns the string of IPv4 in format "a.b.c.d" so you should use comparing the strings in this case like I mention. But you can compare directly the addresses without converting then into the strings: ip->ip_dst == ip->ip_src

You seems uses the COM wrapper. register the object on the new system with regsvr32 tool and after use it. Your error means that Object with specified guid not registered in system.

Open ffdshow property page in graphedit or graphstudio tool in there you can find about tab - on that tab the version information will be.

wrong, solution 3 and 4 are correct.

Actully, the best way from performance and sync position it is to use the allocator presenters with same Direct3D device and multiple render targets. But it require brief Direct3D knowledge and DirectShow.

Yep, that links can be useful.

Once you have WDK installed you can find samples as the followed path DDK\version\src\filesys\miniFilter

had you try google search?

What is necessary under XP is to specify property sheet guid (or guids). Under Vista+ as I remember necessary to specify provider (which will map your file properties to standard metadata values) and also property sheet guid (or guids) to show your properties via right click. Whole such information you can find in registry and articles abt shell extension should helps you figure out what properties to use.

USHORT usValue = 123;
double dValue = (double)usValue;

Try to call WM_CAP_DLG_VIDEOFORMAT and specify one of resolution value from that dialog in your code.

Are you sure that Width and Height is accepted by the capture driver?

Hi Alex,
Not sure that will be usefull to him. He starting with the drivers and with IFS. In IFS drivers there is no access to any WinAPI includig CryptoAPI as it totally kernel driver type.
Maxim.

Let's start again. You can download the ffdshow source code.
In there you can find how to configure it.
You can even build it and debug from GraphEdit or your application.
I'm not see any issues and problems here.

You have source code which you can build, in there you have interface declarations and guids,and even see how the properties changes, what else you need?

So you even have a wiki.

You can download FFDShow sources and see the all methods

Sure that will works as it is an automation, but integrating simple callback is much easy and faster to implement and integrate, plus the COM plumbing will require additional knowledge of COM itself implementing COM objects in C++, probably ATL and it classes and their architecture connection points, dispinterfaces and so on. Compare to implementing callback it just require add the one function to set callback and make to call callback routine from existing code without much modify existing C++ or delphi code.

I didn't mean only my answer. And my answer is well explained. for you it is not bcs you even not understand the question as you not understand the scope of the task it require to solve.
My answer is one which have code example. And you even can't display code of your point. So how can you vote? You just making the abuse. I saw couple of your anwsers there you also answering without understanding the scope - just as it should be and all.

I no need the discussion. I need you to be responsible of what you are saying, that's it. Let's deal next way:
1. You showing the code of the way you mention - to show that can be responsibe of your words.
2. You revote on answers and questions there you vote of 1, or provide description of the reason there you vote this way.
3. You stop insulting user questions and become commenting and answering on questioins there you really sure that your answer is correct and useful.
Otherwise I'll do just putting reference to that discussion next to your answer or comment there you are not correct, so users will see your background.

Same thing - just the words and nothing more. I responsible - but you - not. As expected you can't prove your words. But why you are insulting newble who asked the questions? Mostly saying "Why you asking that..." or "Why you are doing this way..." along downwoting questions and answers which are correct. So new user registred to asking something and you saying to him that he dumb. But you have no reason on that, bcs you are not responsible for your words as you even can't prove simple implementation which you are saying easy way and tooks less time than mine?

Look, I professional COM developer, who have tons of certificates in different programming languages and technologies. And I sure that I know much more abt IDL and COM than you that's why I asking you to show me the code if you telling that such solution will be easy to use and easy to integrate - my code sample I implement within 10-15 minutes. Are you ahve 10-15 minutes to implement stuff which you telling is much easy?

That doesn't matter I only describe to you what was the meaning by "messaging". Mostly looks like that you sitting here and just posting comments or some answers with just url with google searching without understanding what actually that for and how that works. I showed the code which is very easy and describe 2 solutions. You posting nothing and telling me what I not understanding something. So show what you have behind that, otherwise that just a words without background as well as most of your answers.

Show what you did? As I see just couple of words which have nothing behind that.

That mean the application and the DLL. Application is written in Delphi and the DLL in C++ and it is one process.

Well, it was just "mean" or you can show the code of that implementation?

No, abt messages,instead of directly callback function you can pass HWND of Delphi form into the DLL and use of particular SendMessage routine with that handle and receive and process that messages in application in window routine.

#pragma once - should be in .h file - had you seen example in my solution?

in case of exporting classes - then yes different compilers cause an issue, but that can be easy solved to make C functions export wrap around classes, actually that is mostly done in particular API just the classes pointers are output as a HANDLE for example.

No, I wasn't mean IPC. I doesn't even have mentioned that the better way to use IPC here. I describe how to simplify modify existing dll to make callback from one appllication to another see the code below. You probably not read proper the question which was asked or my answer.

I'm not suggested shared memory, I even post the whole code how it should looks. And it not require C++ Builder, it not require to be in same calling conversion bcs delphi allows to specify calling conversion without problems. What for necessary here the GCC and MSVC? here is the delphi. but I'm sure if you proper bind you dll interface you will never get your issue with GCC and MSVC.

If he doing DLL from scratch the COM will be fine and fair enought to make whole system, but if he already have dll and to modify it for COM along with whole application will be very painfull and requres a lot of time, so simple callback it is much easy. Regarding interfaces in object pascal - delphi without problem allows to import typelibrary - so no necessary to writing interfaces in pascal in case you have proper typelibrary in your COM.

Your solution just copy of mine, but there is require to have dll to be compiled in C++ Builder.

What for necessary to define COM interfaces. What for necessary in there re-creating the DLL itself? For programmer who asking such question, reimplementing DLL in C++ will be brain crashing.
Ofcouse you are here just to show your knowledge, if you are know COM that is good for you but re-creating all methods into interfaces binding connection points and events (hope you understand what it is or you have top of COM knowledge) - and imagine how much time it tooks compare to implement simple callback and integrate it in DLL? For me it tooks 15-20 minutes including examples on Delphi and C++:
C++ DLL: "MyDll.dll"
/////////////////////////////////////////////////////
typedef HRESULT (CALLBACK * CALLBACKPROC)(DWORD wEvent,LONG_PTR arg,PVOID pParam);
/////////////////////////////////////////////////////
CALLBACKPROC g_pCallback = NULL;
LPVOID g_pCallbackParam = NULL;
/////////////////////////////////////////////////////
EXTERN_C VOID __declspec(dllexport) WINAPIV SetCallback(CALLBACKPROC _proc,LPVOID pParam)
{
g_pCallback = _proc;
g_pCallbackParam = pParam;
}
/////////////////////////////////////////////////////
#define SOMEDLLMETHOD_EVENT_ID 0x01
/////////////////////////////////////////////////////
EXTERN_C VOID __declspec(dllexport) WINAPIV SomeDllMethod()
{
if (g_pCallback) g_pCallback(SOMEDLLMETHOD_EVENT_ID,(LONG_PTR)"some argument",g_pCallbackParam);
}
/////////////////////////////////////////////////////
Delphi wrapper with dynamic DLL linkage:
/////////////////////////////////////////////////////
type
TDllWrapper = class
protected
FOnDllEvent: TNotifyEvent;
m_hDll: HMODULE;
protected
function OnCallbackEvent(wEvent:DWORD;arg:Pointer):HRESULT; virtual;
public
procedure CallSomeDllMethod();
public
constructor Create;
destructor Destroy;
published
property OnDllEvent: TNotifyEvent read FOnDllEvent write FOnDllEvent;
end;

type
TCALLBACKPROC = function(wEvent:DWORD;arg:Pointer;pParam:Pointer):HRESULT stdcall;
type
TSetCallbackFunc = procedure(_proc:TCALLBACKPROC;pParam:Pointer) cdecl;

const SOMEDLLMETHOD_EVENT_ID = $01;

function CallbackProc(wEvent:DWORD;arg:Pointer;pParam:Pointer):HRESULT stdcall;
begin
result := 0;
if (pParam <> nil) then result := TDllWrapper(pParam).OnCallbackEvent(wEvent,arg);
end;

constructor TDllWrapper.Create;
var
fnSetCallback:TSetCallbackFunc;
begin
inherited Create;
m_hDll := LoadLibrary('MyDll.dll');
fnSetCallback := GetProcAddress(m_hDll,'SetCallback');
fnSetCallback(CallbackProc,self);
end;

destructor TDllWrapper.Destroy;
var
fnSetCallback:TSetCallbackFunc;
begin
fnSetCallback := GetProcAddress(m_hDll,'SetCallback');
fnSetCallback(nil,nil);
FreeLibrary(m_hDll);
inherited Destroy;
end;

function TDllWrapper.OnCallbackEvent(wEvent:DWORD;arg:Pointer):HRESULT;
begin
result := 0;
if (wEvent = SOMEDLLMETHOD_EVENT_ID) then
begin
if Assigned(FOnDllEvent) then FOnDllEvent(self);
end;
end;

procedure TDllWrapper.CallSomeDllMethod();
var
fnSomeDllMethod:procedure() cdecl;
begin
fnSomeDllMethod := GetProcAddress(m_hDll,'SomeDllMethod');
fnSomeDllMethod();
end;
/////////////////////////////////////////////////////
Example of usage:
/////////////////////////////////////////////////////
type
TMainForm = class(TForm)
m_MyDll:TDllWrapper;
procedure OnDllEventCallback(Sender: TObject);
procedure FormCreate(Sender: TObject);
procedure FormClose(Sender: TObject; var Action: TCloseAction);

end;

procedure TMainForm.OnDllEventCallback(Sender: TObject);
begin
OutputDebugString('Our callback event called');
end;

procedure TMainForm.FormCreate(Sender: TObject);
begin
m_MyDll:= TDllWrapper.Create;
m_MyDll.OnDllEvent := OnDllEventCallback;
m_MyDll.CallSomeDllMethod;
end;

procedure TMainForm.FormClose(Sender: TObject; var Action: TCloseAction);
begin
m_MyDll.Free;
end;

Maybe your Dialog have style that it is main application window but you call it from other application?

Had you try to debug and check what g_hInstance value? Does your code get to the DialogBoxParam function once you call it from .NET (bcs CoInitialize already called in case you call dll from .NET and may return failure)? Had you try to call LoadResource and SizeOfResource to check proper resource ID once you call it from .NET?

You should read SDK documentation for the functions I mentioned.
The code you make totally wrong. You get backbuffer and after you should copy that data into your surface, but you are using same variable for all.

I think he just at beginning of the cryptography, this way in compression called "vlc" or Variable-Length code, two numbers you may use this way but more numbers are coded with prefixes. Mostly uses in video/audio compression. http://en.wikipedia.org/wiki/Variable-length_code
So, he is correct that he calls that as compression.

You are waiting till playback complete.

I know what your code do, and it were never plays the video before - this bcs you just copy it from somethere without understanding - and due that some code parts are at wrong location.

You modify the code without understanding it - you should research your self before pasting here code which you just cut from somethere.

In C++ it will be the same logic: create LPDIRECT3DSWAPCHAIN9 object with resolution you need to render. Perform rendering your scene onto that swap chain and call present method, once you resizing the window - just recreate swap chain object with new resolution.

maybe you forgot to initialize gdi+ library?

// C#
[DllImport("Dll.dll", CharSet=CharSet.Unicode)]
public static extern void RenderFile(string strFilename);
// C++
extern "C" __declspec(dllexport) void RenderFile(WCHAR *strFilename)
{
IGraphBuilder *pGraph = NULL;
IMediaControl *pControl = NULL;
IMediaEvent *pEvent = NULL;
HRESULT hr = S_OK;

// Create the filter graph manager and query for interfaces.
hr = CoCreateInstance(CLSID_FilterGraph, NULL, CLSCTX_INPROC_SERVER,
IID_IGraphBuilder, (void **)&pGraph);
if (FAILED(hr))
{
printf("ERROR - Could not create the Filter Graph Manager.");
return;
}

hr = pGraph->QueryInterface(IID_IMediaControl, (void **)&pControl);
hr = pGraph->QueryInterface(IID_IMediaEvent, (void **)&pEvent);

// Build the graph. IMPORTANT: Change this string to a file on your system.
hr = pGraph->RenderFile(strFilename, NULL);
if (SUCCEEDED(hr))
{
// Run the graph.
hr = pControl->Run();
if (SUCCEEDED(hr))
{
// Wait for completion.
long evCode;
pEvent->WaitForCompletion(INFINITE, &evCode);

// Note: Do not use INFINITE in a real application, because it
// can block indefinitely.
}
}
pControl->Release();
pEvent->Release();
pGraph->Release();
}

1. remove "HRESULT hr = CoInitialize(NULL);"
2. remove "CoUninitialize();"
3. trash - "hr = pGraph->RenderFile((LPCWSTR)strFilename, NULL);" - if you use ANSI then use Conversion, otherwise use UNICODE.

here you can find sample:
http://www.codeproject.com/Articles/421167/Pure-NET-DirectShow-Filters-in-Csharp

Same

Change filter merit so it will be selected by default for video playback, be sure that your filter available to output RGB24 and RGB32 - that you can check in GraphEdit tool.

try to look at that posts: maybe you will find some useful things: http://www.google.ru/search?hl=en&q=Run-time+error+%27438%27

You also can try retrieve methods at runtime from IDispatch and invoke them directly without binding.

Why are you doing conversion from C++ into C# if you not much familar with C#? - use C++ code. I already wrote you - there is no equivalent, I already gave you the URL which you should read.

Sure bcs C# and C++ it is different languages and I told you to review C# specification of what you are planing to do.

Sure, but had you asking the question to yourself: "what for it is needed?"

Yes, I wrote you note regarding that.

NOTE: those DSP may work under vista and above only - but not sure - I didn't try to port them into XP

You should register DMO classes example for resizer:
WINDIR\System32\vidreszr.dll

IBaseFilter * _filter;
hr = CoCreateInstance(CLSID_DMOWrapperFilter,NULL,CLSCTX_ALL,IID_IBaseFilter,(void**)&_filter);
ASSERT(hr == S_OK);
IDMOWrapperFilter * _wrapper;
hr = _filter->QueryInterface(IID_IDMOWrapperFilter(void**)&_wrapper);
if (S_OK == hr)
{
hr = _wrapper->Init(__uuidof(CResizerDMO),DMOCATEGORY_VIDEO_EFFECT);
ASSERT(hr == S_OK);
_wrapper->Release();
}
// Use DMO wrapper filter: _filter
...

IMediaObject * _object;
hr = CoCreateInstance(__uuidof(CResizerDMO),NULL,CLSCTX_ALL,IID_IMediaObject,(void**)&_object);
ASSERT(hr == S_OK);
// Use DMO object : _object
...

Welcome

Type library it is description of what your COM dll is exports - so all applications can use your interfaces and declarations, but if you want this information to be private and only you will be know how to work with your objects then you can hide that info and do not export type library.

had it find pin properly, what an HRESULT it returns during render stream call?

and what's the trouble?

show me code and there you have trouble

I gave you the code which solve your question. If you not understand the code what for you asking that? Maybe you should read books to understand?

For that you can use shaders, OpenCL, CUDA or OpenMP

Forgot to mention that GDI+ also allows to modify the image:
http://www.c-sharpcorner.com/UploadFile/mahesh/Transformations0512192005050129AM/Transformations05.aspx
Here is some additional information:
http://www.gamedev.net/page/resources/_/technical/graphics-programming-and-theory/an-introduction-to-digital-image-processing-r2007
Once you download CUDA Toolkit in there you can find some samples of image processing.
Maxim.

That's all depends on what for you need getpixel.

Create your class for pin inherit it from SourceStream implement GetMediaType DecideBufferSize and FillBuffer.
My classes are much similar to native:
http://msdn.microsoft.com/en-us/library/windows/desktop/dd375198(v=vs.85).aspx

Regards,
Maxim.

Thanks

Solution described in there should works fine.
You need to create interface with your method to export so you can query it, otherwise you should look what is imported from type library. In there should be generated interface for your object and your coclass (You can also use OleView tool and open resulted tlb). If you have no tlb - you have to export typelibrary with regasm tool. Along with it auto generated interface is always inherited from IDispatch - and you can figure out it's methods in runtime. You may use CoCreateInstance function instead of smart pointers.

wprintf in my code just an example what field in variant you should use. Sure you need an array of strings or as you using .NET use any collection objects.

Audio works same way, but you should takes care about output format - notify the source filter with format you sending. For testing you can hardcode output format, or if you planing to transfer PCM data you can do resampling on input and send fixed format so all sources will knows it. Along with that you need to make buffering of received data, and delivery without drops.

Maxim.

Create dll project in Visual Studio add your h file, create c or cpp file, make functions decalrations with exports and call in there your functions in h file, link with your lib file and build.

You can build a dll using lib and export functions from that lib. But in C# you can access the dll functions or COM objects. One note for dll's and C# DLL should be compiled for same platform as target PC, or you should specify fixed platform in C# app settings and build DLL for it.

Maxim.

Welcome

You are mixing my code and yours without understanding what I wrote.

Full code - what goes before while?

You should enumerate interfaces inside enumeration of the devices. You need to get the device from the list and only after enumerate interfaces on it. so your "do" loop should be inside 2. in my list, and the SP_INFO_DATA structure which you got on device enumeration should be specified as the second argument in your SetupDiEnumDeviceInterfaces call. Follow?

Maybe help viewer executable is missed, are you available to open that chm file by clicking on it?
Try the next staements instead of yours:

Assembly _assembly = Assembly.GetExecutingAssembly();
string _path = Path.GetDirectoryName(_assembly.Location);
Directory.SetCurrentDirectory(_path);
Help.ShowHelp(this, _path + "\\" + "HelpFile.chm");

Why it is not method? bcs it is not described as an article? You should specify that is related to mine stuff or even just make the comment but you instead specify the solution.
Not personal - but it looks not good from your side.

Welcome

then you should declare you object in Form1.h as extern: "extern WebcamShow::CamShow test;" - that means object declaration located somethere but not here. In Form1.cpp declare that object: "WebcamShow::CamShow test;" - same line without "extern" and then include your Form1.h there you need your class. - follow?

Wow, you found an article which is described same stuff as I did in my solution. Are you reading that article or you just google and then copy link here, bcs why you downwove mine solution if it describes the same way? You are only found the proof that my solution is works fine. For you seems posting solution means to put any link in answer. I always describing the solution based on my experience and only if that's not enought explained I put the links.

This one is also seems contains similar stuff
http://www.codeproject.com/Articles/21330/Easy-Rendering-with-XNA-Inside-a-Windows-Form

Yes, I also uses the XNA 4.0.
Try this link:
http://xbox.create.msdn.com/en-US/education/catalog/sample/winforms_series_1

Maxim.

Solution is not based on that article - it is describes that article does not handled all aspects which can be with the OLE/COM Controls in .NET and what is missed in .NET for that. If you didn't implement that things from scratch in C# you can't understand the issue.
You had downvote one of my solution without any reason, and even have no reply on comment I posted. I check questions you had answered and found few which are totally wrong I improve them and downvote one with the reason I described.

No, I'm not thinking that the "Solution 2" is the solution. I post my solution here with brief overview in there the issue can be and how it can be solved. I downvote your anwser bcs of next: 1. Actually, solution with words "not possible" "not allowed" and similar can't be dedicated as "Solution". 2. You are not reseraching in actual problems which are possible to be in there which are makes your answer unusable. 3. As in question he mentioned that he available to select his control but it can't be created due error information he specified - than article u mean is useless.

Actually not the solution, reference to an article without understanding of the issue itself, and the article does not handle all aspects of embedding COM/OLE .NET control in any types of application requirements, just basic approach.

So, I have voted your answer as #1.

Maxim.

Paint.NET not offer source code any more - I didn't mention what it's not alive.
And if you are read proper his question - he mention "like in adobe photoshop".
Had you work with such things in photoshop before posting here you questions in comments?
So your comment seems not at right place. Plus making the vote of 1 my solution which can really helps, you just showing that you even not reading proper the question.

WDK require for kernel development and it includes native windows headers and libs (as you also can build user mode applications with WDK) and kernel headers and libs which are only can be used inside kernel (in case if you are creating the driver). You can't mixing usage of kernel and non-kernel routines as most of kernel routines not works in user mode. - That's was meaned in previous comment.

Forgot to mention - your code seems correct.

What an actual issue you have? Is your device enumerated in list you got?


Maxim.

That doesn't matter for directshow: there you getting the data from is it real time data or file data or data from somethehere else - so I think that you should read the documentation before making words "Not actual" - and rephase to "I haven't knowledge". For playback PCM data there are a lot of DirectSound Samples in SDK so I think there will be enougth. For decoding mp3 you can check the ffmpeg or lame - they are open source.

Welcome.

Check in GraphEdit graphs on both machines. Try with the same decoders. NOTE on XP you can add graph to the ROT and connect to it remotely from GraphEdit

Same way, but all images/mpeg4/h264 packets will be in a single stream

in your camera specification described and URL which you can use for reading data.

You should read the specification of your camera, after enable streaming over protocol you want' to use on camera (most cameras can stream over HTTP) open HTTP stream in your application, perform reading, decoding and preview.

Had you try the Dependency Walker Tool to see what Dll's are missing?