Comments by mbue (Top 200 by date)

hello AlwaysLearningNewStuff:
I would suggest you to stay in the message loop handler as short as possible.
If the thread signals its state by WM_THREAD_OK your message handler should test the thread state and return.

case WM_THREAD_OK: // thread sent this

if( IsThreadRunning() ) // just wait until it fully exits
{
PostMessage(hwnd,WM_THREAD_OK,0,0); // thread is already running post the message again
// alternative: use a timer to check the threads state
}
else
{
ThreadFinished(hwnd);
}

return TRUE;

case WM_COMMAND:
switch(LOWORD(wParam))
{
case IDC_BUTTON1:
StartThread(hwnd);
return 0;
}
break;



Encapsulate the event handling into functions. Show the user whats going on by setting the buttons state and text.

int IsThreadRunning()
{
if(threadHandle)
{
if(WAIT_TIMEOUT==WaitForSingleObject( threadHandle, 1 ))
return 1;

CloseHandle( threadHandle );
threadHandle = NULL;
}
return 0;
}

void ThreadFinished(HWND hDlg)
{
HWND hbtn = GetDlgItem( hDlg, IDC_BUTTON1 );

// show the user whats going on
EnableWindow( hbtn, 1 );
SetWindowText(hbtn,__TEXT("start thread"));
}

void ThreadStarted(HWND hDlg)
{
HWND hbtn = GetDlgItem( hDlg, IDC_BUTTON1 );

// show the user whats going on
EnableWindow( hbtn, 0 );
SetWindowText(hbtn,__TEXT("thread running"));
}

void StartThread(HWND hDlg)
{
if(!IsThreadRunning())
{
DWORD threadID;

td.hwnd = hDlg;
td.PrimaryKey = primaryKey;
threadHandle = CreateThread( NULL, 0, (LPTHREAD_START_ROUTINE)MyThread, (void*)&td, 0, &threadID );

if( !threadHandle )
{
MessageBox( hDlg, L"Error", L"Error", MB_ICONERROR );
DestroyWindow(hDlg);
}
else
{
ThreadStarted(hDlg);
}
}
}



Regards.

you should use a configurable server path. with that path you are independant from your current directory. otherwise your server will be very unsafe.
regards.

only 3. prevent memory fragmentation, extra calls and special realease functions.

float** alloc2D(const unsigned int x,const unsigned int y)
{
unsigned char* buff;
unsigned int need,ix;
float** ax;
float* ay;

ASSERT(0<x);
ASSERT(0<y);

need = (x * sizeof(float*)) + ((x*y) * sizeof(float));
buff = (unsigned char*)malloc(need);

ax = (float**)buff;
ay = (float*)(ax+x);
for(ix=0;ix<x;ix++)
{
ax[ix] = ay;
ay += y;
}
return ax;
}


why dont you examine the os version from system files version?
ps: remember not all registry informations are stored on disk!

what project you are talking about? is there a link?

*****: i love standards - everybody has another ;).

Yes. The compiler places the destructor at the end of scope. The end of scope is (;).
Regards.

Sorry i forgot to set the head->next ;(. Set after the newContact as follows if(head) head->next=newContact;.
Regards.

What about waiting until thread is ready?
WaitForSingleObject(mainHandle,INFINITE)

Little example to apply the bitmap to the control:


#pragma once
#define WIN32_LEAN_AND_MEAN
#include <windows.h>

// multibyte or unicode
#include <tchar.h>

// needed libraries
#pragma comment (lib,"user32.lib")
#pragma comment (lib,"gdi32.lib")

// 32 vs. 64 bit
#ifdef _WIN64
#define SET_DWL_USER(H,P) SetWindowLongPtr(H,DWLP_USER,(LONG_PTR)P)
#define GET_DWL_USER(H) GetWindowLongPtr(H,DWLP_USER)
#else
#define SET_DWL_USER(H,P) SetWindowLong(H,DWL_USER,(long)P)
#define GET_DWL_USER(H) GetWindowLong(H,DWL_USER)
#endif

/////////
// simple dialog class
class CMyDlg
{
public:
CMyDlg(HINSTANCE h);
~CMyDlg();

int DoModal(const int idd);

private:
static INT_PTR FAR PASCAL __dlg(HWND h,unsigned int m,WPARAM w,LPARAM l);

HINSTANCE _hinst;
HBITMAP _hbmp;
};

// dialog message handler
INT_PTR FAR PASCAL CMyDlg::__dlg(HWND h,unsigned int m,WPARAM w,LPARAM l)
{
CMyDlg* This = (CMyDlg*)GET_DWL_USER(h);
switch(m)
{
case WM_INITDIALOG:
{
// attach my this pointer to the dialog
This = (CMyDlg*)l; SET_DWL_USER(h,This);

// delete if already loaded
if(This->_hbmp) DeleteObject(This->_hbmp);
// load the image from file
This->_hbmp = (HBITMAP)LoadImage(0,__T("myimage.bmp"),IMAGE_BITMAP,0,0,LR_LOADFROMFILE);
if(This->_hbmp)
{
HWND hctrl;

hctrl = GetDlgItem(h,101);
if(IsWindow(hctrl))
{
// attach image to the control
SendMessage(hctrl,STM_SETIMAGE,IMAGE_BITMAP,(LPARAM)This->_hbmp);
}
hctrl = GetDlgItem(h,102);
if(IsWindow(hctrl))
{
// attach image to another control
SendMessage(hctrl,STM_SETIMAGE,IMAGE_BITMAP,(LPARAM)This->_hbmp);
}
}
}
break;
case WM_COMMAND:
switch(LOWORD(w))
{
// end the dialog and quit
case IDOK: EndDialog(h,LOWORD(w)); break;
case IDCANCEL: EndDialog(h,LOWORD(w)); break;
}
break;
}
return 0;
}

CMyDlg::CMyDlg(HINSTANCE h)
{
_hinst = h;
_hbmp = 0;
}

CMyDlg::~CMyDlg()
{
// finally delete object
if(_hbmp) DeleteObject(_hbmp);
}

int CMyDlg::DoModal(const int idd)
{
HRSRC hres = FindResource(_hinst,MAKEINTRESOURCE(idd),RT_DIALOG);
HGLOBAL hdlg = hres ? LoadResource(_hinst,hres):0;
// find resource handle and show dialog here
return hdlg?(int)DialogBoxIndirectParam(_hinst,(LPCDLGTEMPLATE)LockResource(hdlg),(HWND)0,__dlg,(LONG_PTR)this):0;
}

/////////
// program starts here
int FAR PASCAL _tWinMain(HINSTANCE h,HINSTANCE p,LPTSTR c,int sw)
{
CMyDlg dlg(h);
return dlg.DoModal(100);
}


If the CatchConsole function returns NULL there is possibly nothing written to output. Have you tried the example as it is (with: "cmd.exe dir *.*")? This should work. Example for MyApp.exe (this should be a console application): #pragma once #include <tchar.h>#include <iostream>int _tmain(int argc, _TCHAR* argv[]) { std::cout << "Hello world"; return 0; } The program should finally show an edit window displaying "Hello world". If you run in a debugger the call CreateProcess must return true. Writes your "MyApp.exe" any text to stdout the output is written to a memory file (CMemWrite). The memory handle is detached from the file and attached for displaying to an edit control. Best regards and good luck.
PS: sorry its really heavy to place preformatted code here :(

Your problem is the function Minor produces a matrix with smaller dimensions and calls the function Determinant always with the same dimensions.
The best way is to encapsulate the array into a class. Or you call all functions with matrix parameter with corresponding dimensions.
The exception is thrown because of invalid dimensions (this->row and this->col) are always the same but dimensions are changing.
Your code:
double** tempMtx=new double*[rows-1];
for (int i=0;i<rows-1;i++)
tempMtx[i]=new double[cols-1];
Best regards.

Where is the code?
Use the settings of additional header search paths in your project.

Nobody knows what happens with your exe. But the best way seems to use "copy'n'paste" into a dll project and do the exe-main-job on DllMain.
Regards.

This is a hint not a solution! Most common mistakes are the mixed usage of integer and floating point values. The most common errors base on rounding differences by a simple cast from double to int.
I see in your code alway on the left side: 11 or 12 or 9 these are int types. Most c compilers casts from right to left. Thats the function with cos and sin are always casted to int (left value). You should use 11.0*sin 12.0*sin and so on.
Regards.

what do you think about owner draw buttons?

Thanks for the candy ;)

Ok. To prevent the hooking mechanism to hook the events by your dummy interface, you should identify your own interface:
* create a new CLSID
* return your interface on QueryInterface call, so you know that you got your own implementation.
example:

static const GUID CLSID_ThisIsMyInterface = { 0x85d8c4a1, 0xb644, 0x4cf2, { 0xb7, 0x72, 0xb2, 0x7e, 0xb2, 0x27, 0x1, 0x3d } };

HRESULT iDummy::QueryInterface(REFIID riid,void** ppv)
{
if(CLSID_ThisIsMyInterface==riid) return *(iDummy**)ppv=this,AddRef(),S_OK;
// the other interfaces
}

// hook the old handler
static HRESULT iDummy::HookTheHandler(IDispatch* pOld,IDispatch** ppNew)
{
iDummy* isme;
if(S_OK==pOld->QueryInterface(CLSID_ThisIsMyInterface,(void**)&isme))
{
isme->Release();
return E_FAIL;
}
*ppNew = new iDummy(pOld);
return (*ppNew)?S_OK:E_OUTOFMEMORY;
}

iDummy::iDummy(IDispatch* pOld)
{
oldhandler = 0;
if(S_OK!=pOld->QueryInterface(IID_IDispatchEx,(void**)&oldhandler))
{
oldhandler = 0;
}
}

iDummy::~iDummy()
{
if(oldhandler) oldhandler->Release();
}

HRESULT iDummy::InvokeEx(DISPID id,LCID lcid,WORD wFlags,DISPPARAMS *pdp,VARIANT *pvarRes,EXCEPINFO *pei,IServiceProvider *pspCaller)
{
// forward the old handler
HRESULT hr = oldhandler ? oldhandler->InvokeEx(id,lcid,wFlags,pdp,pvarRes,pei,pspCaller):E_FAIL;
Beep(3000,100); // i am alive
return hr;
}

To call the old impelementation you can store the old IDispatchEx interface (if(S_OK!=handler->QueryInterface(IID_IDispatchEx,(void**)&oldhandler)) oldhandler=0;). Remember you should call Release in your destructor (if(oldhandler) oldhandler->Release();)
in the function InvokeEx of your implementation you can forward the old handler function. handler are always called with DISPID==0 and flags set to DISPATCH_METHOD.
Regards.

The event capturing on that code based on the replacement of the previous event handler by a dummy handler:
example
<a href="#" onclick="javascript:LinkClick();">link with handler</a>
the walk examines a handler LinkClick and replaces that with a new IDispatchEx handler (that beeps). internal html events such as submit i dont know if you can replace them.
To use an external script things would be a little bit difficult.
Therefore you have to load the external script file by using the IActiveScriptParse interface. Examine the function interfaces and replace the appropriate handlers by them from the external script.
This isnt really easy.
Regards.

Internal error: Too many numbers in question.

See my updated solution.
Please dont post your additional statements as solution, thanks.
Regards.

Broadcasted messages are only sent to windows, not to threads.
Regards.

Its always a good idea to have any kind of error handling. A simple message box to notify the user what happens when an error occures.
Regards.

Take
* this
* this
* this
* and this
and now tell me a reason to assume the op is talking about unix calls!
I jump on nobody - im not a rolling cat.
Perhaps the op should decide first what os he/her want to use.
Regards.

Beg your pardon, but take a look at the headline and the link the op posted above.

Nice ;). ShellExecute is definitely async. And the others have flags to control the execution (i.e. P_WAIT or P_NOWAIT). Hope that helps.
Regards.

Cannot understand the excitement nobody knows what color is. Assume it is:

class CColor
{
public:
CColor()
{
_c = _d = 0;
}
CColor(CColor& c)
{
_c = c._c; _d = c._d;
}

operator COLORREF& (){ return _c; }

template <class C> CColor& operator = (C c)
{
_d = c;
return *this;
}
template <class C> CColor(C c)
{
_c=0; _d=c;
}
template <class C> CColor& operator , (C c)
{
_d = _d<<8|(c&0xFF);
return *this;
}
operator bool ()
{
return _c==_d;
}
private:
COLORREF _c;
COLORREF _d;
};

void GetPixel(int x,int y,COLORREF& c){ c = RGB(0,128,255); }

int FAR PASCAL _tWinMain(HINSTANCE h,HINSTANCE p,LPTSTR c,int sw)
{
CColor col;
GetPixel(0,0,col);
if(col = 255,255,255)
{
Beep(3000,100);
}
else
{
Beep(1000,100);
}
return 0;
}

and it works! So far
Regards.

"hangup the main program untill it completes" who told you that? ShellExecute is completely ansync. You can obtain a process handle to wait for, if you use ShellExecuteEx.
Regards.

Indeed that was my first comment statement.
Ragards.

I think the op wants an explanation. There is no solution by mine.
Regards.

Thats because vista and w7 gdi functions dont draw directly to the primary display surface. On xp the window content beneath you transparent dialog is painted first and the dialog is rendered with transparency over the background.
Regards.

Not neccessarily, this depend on the things you want to do. For a single operation (i.e. greyscale) you can break the bitmap in some parts and process every part by a single thread.
i.e. your bitmap has 800 lines and every of your 8 threads processing 100 lines and put the 100 lines to the output bitmap.
Regards.

You can create a thread for each buffer. If you need the processing result - wait for the appropriate thread.
Reagards.

For the definition:
RGP == role playing game
Char == Avatar or NPC
Do you mean this?
Are you using COM-Interfaces? Or virtual classes for your chars?

Ive updated the solution above and marked the relevant code positions by comments.
Regards.

The times gives you the usage in user and kernel times for each process. In relation to the running time you get the utilisation. Or did i misunderstood - define your imagine of utilisation!
Regards.

Oh yes indeed - we need some additional functions:


template <class TSET>
int GetCodes( TSET& codes,const TSET set )
{
return 0;
}

template <class TSET>
int GetCodes0( TSET& codes,const TSET set )
{
_tprintf(__T(" set\r\n"));
codes = set;
return 0;
}

template <class TARR,class TSET>
int GetCodes0( TARR& codes,const TSET set )
{
return 0;
}

template <class TARR,class TSET>
int GetCodes( TARR& codes, const TSET set )
{
size_t i;
int dim = 0;
_tprintf(__T(" elem: %i\r\n"),(int)(sizeof(codes)/sizeof(codes[0])));
for(i=0;i<(sizeof(codes)/sizeof(codes[0]));i++)
{
dim = GetCodes(codes[i],set);
if(0==dim) GetCodes0(codes[i],set);
}
return 1+dim;
}


Thanks for your help.

Im afraid no. Cause the first template function marks the inner loop. And the inner loop must match the array type and the set type.
Regards.

Perhaps you have to set the duration for the output stream first (original duration + echo overhang).
Regards.

Look at this line:
m_output->write((char*)data, pcm_datalength+phase);
must be
m_output->write((char*)data, sizeof(short)*(pcm_datalength+phase));
Regards.

I see the problem: youre writing a short array with array size of bytes. you need to write the array length multiplied by sizeof(short), thats the sizeof the buffer in bytes. therefore your output is cutted to the half of its real size. your pcm data has a maximal size of 4096 and your phase is 4000 -> the last 96 samples contains the echo and the remaining echo is cutted.
Regards.

it seems to be a consens ;)

"Their lifetime does not exceed the statement they're used in"
No you fail: the destructor of the temporary object is called AFTER the function leaves. The compiler generates two variants:
void fn(Vector2D& x){}
and
void fn(){ Vector2D x; fn(x); }
you need not to declare two different function. Accept it or leave it. Compile and debug it you will see the result is exact what you want.
Once again:
thats the code:
<pre style="margin-top: 0pt;" lang="c++"><span class="code-preprocessor">#pragma</span> once
<span class="code-preprocessor">#include</span> <stdio.h>
<span class="code-preprocessor">#include</span> <tchar.h>

<span class="code-keyword">class</span> Vector2D
{
<span class="code-keyword">public</span>:
  double x;    double y;
  Vector2D(double val1=0.0, double val2=0.0):x(val1), y(val2)
  {
    printf(<span class="code-string">"constructor\r\n"</span>);
  }
  ~Vector2D()
  {
    printf(<span class="code-string">"destructor\r\n"</span>);
  }
};

bool CanShoot(Vector2D& Target=Vector2D())
{
  printf(<span class="code-string">"x: %lf, y: %lf\r\n"</span>,Target.x,Target.y);
  <span class="code-keyword">return</span> 1;
}

<span class="code-keyword">int</span> _tmain(<span class="code-keyword">int</span> argc, _TCHAR* argv[])
{
  CanShoot();
  _gettch();
  <span class="code-keyword">return</span> 0;
}


</pre>
thats the output:

constructor
x: 0.000000, y: 0.000000
destructor


Regards.

Swap the displayed image?

I mean:
int sum = pcm_databuffer[i] +(int)((double)pcm_databuffer[j]*damping);
sum = min(sum,35535);
sum = max(sum,-35535);
because if your sound is realy loud, you get short overflows.
Perhaps you can post your original code as "improve question" so i can see your normal (without echo) wave output function.
Regards.

You must send the resulting pcm-buffer to the output device inside this function. If you write it to a file use the correct buffer size. In the example above, i forgot, you have to check and adjust the sum for the 16-bit boundaries.
Regards.

Please, please use "improve question" or "have a question or comment" links to do them, why?: to enshure the answerer gets a response mail. Or for a better understanding of your question. Therefor you should never post a statement as solution.
Have a nice time on cp.
Regards.

its a pleasure. ;)

Should this be a joke? Everything on earth is temporarily. This sounds like "you can never use references except const"? Try to understand the temporary class on stack is fully accessible like other declarations, thats there is no reason to make a call to a const object. The objects lifetime is enshured during the function call.
If other compilers dont accept this construct - explain why (no links accepted)!
Regards.

This isnt illegal its the same behaviour like yours but inline.
The Vector2D() isnt a ctor rather than a constructor on the stack. The compiler produces the same result like your example and the two different function calls too.
Regards.

The code above is full working for both output (64/32-bit). The one logs the 32-bit applications and the other the 64-bit apps.
Remark: hook.cpp is for the dll and main.cpp for the exe.
Regards.

Perhaps you should search this site:
* Boyer-Moore
* string search
Most search algorithms using null terminated strings, you can use them if your string is null terminated or you have to clip the buffer range into a string. in this case you can use the other suggestions with strtok, strstr or others from string class libraries.
Take a look at msdn (strtok_s). There is an example too.
regards.

This isnt a fully boyer-moore, cause the skip value initialisation isnt complete. Where do you got the code from?
Ragards.

As Analyst point it. If you dont do that i assume you cant do. Books cannot be compiled. Im not interested on any further theoretical discussion.
Emilio Garavaglia said:
"...trust your expertise may cause him problems?"
thats quite right therfore if changed the code and spent more time i wanted. but there is no substancial part from the other subscribers, so the op can learn about. Its really simple for anybody to criticize whitout any useful hints. Hence, what is your part to show the op solving his problem, if there is a really interest to solve it?

1. yours: "things in discutable way." - do it.
2. tell us - what isnt robust in that code? everybody knows nothing can be perfect, especially in spent such less of time.
3. i assume youve readen the books - tell us what youve learned and we should be make better?

Regards.

Thats true. Threfore you need to implement a filesystem based on a file. Perhaps take a look at microsoft storage object (IStorage Interface). But there are tools to examine the content, but you can prevent them by any kind of encryption.
Regards.

Yes indeed:
result = 2<argc?atoi(argv[2]):0;
Regards.

Maybe, i dont know why the op wants to multiply two random digits. He only forgot that a digit in a string isnt equal to its numeric equivalent.

An accelerator table translates special key strokes into command messages. You can define ie. F2 to a command (WM_COMMAND) to do anything special.
Use the function names above for a further internet or msdn search.
Regards.

i know there is no validation of incoming values. the concept is: "do exactly what the op wants" - no more and no less.
Regards.

This problem ends up with an infinite loop to save itself (cause its a part of the computer).
Regards.

There is another case too (in visual studio 6). I dont know if the problem still exist:
if youre loading a dll (ie: x.dll) that has debug informations, and you load another dll with the same name (x.dll) without debug information (the dll will possibly contain only resources) the debugger will discard all informations from the first dll.
Regards.

Take a look to the modified solution.
Regards.

feel free to replace the scrap by your perfect solution. ;)

can you improve your question by showing the code for tracker.Draw? if you are painting the rect by DrawFocusRect or xor brush you should set the brushs origin by SetBrushOrg.
regards.

where is your 100% c++ compliant solution? im waiting...
nobody has yearned a 100% c++ solution - that was your imagination.
read that thread again and again and again, if you cant remember.
little hint: take a look at the ops question tags.
have a nice day.

i come along with you, but the question was too alluring for me.
There are three ways to handle that:
1. ignore it
2. use it.
3. or try to understand and make your own solution.
Regards.
ps: im shure i didnt spent more time to write that down, than others to comment the scrap.

;)

To deal with pointers is a difficult thing (especially in linked objects). without explicit copy contructor the pointers will have no owners. the default copy constructor cannot distinguish between owners (its a simple memcopy). so the copy constructor can duplicate the pointers if neccessary. default copy constructor cannot do that.
Regards.

using std is the real c++ skill (lol). show me anything in the code above that wont work. its on you to use std::vector, there is no reason to say malloc is obsolet. You say i cant do calling a default contructor: but i do that cause it worked all times. Take a look to the mfc-sources and you will see much more like this (ie. CArray element constructor and destructor). so far to your skill.
Regards.

congratulations. next step is to trace incoming values and the result of your operation (post it here). so we can see whats going on in your black box.

Most people having questions are telling us what they have done and what happens with there code. you are splashing a piece of code and asking whats wrong about that.
Have you ever used the debugger and debug messages like trace? in debugger you can examine the incoming values. so you can see whats wrong with your code.

He should start with a simple exercise. And should not spam many times a day the same question.
btw: this is a solution: "By you a book", "Check out simple examples", "Learn the basics", "Use the web to get examples"!
Regards.

Its a pleasure for me. ;)

Once upon a time we had: near, far and huge pointers. Im afraid i dont understand your question. Let do it the processor and the virtual address mapping these things for you.
Regards.

This should be nearly the same. But im afraid fprintf will not distinguish between ansi and unicode.
Regards.

[in addition]
This function is written so you can use it by a debug define (like trace):

#ifdef _DEBUG
#define LOG __loggi
#else
#define LOG
#endif

with a simple printf you cannot do that. On the other side you can extend the function inside to lock the access by mutex object (if necessary).

Compiler cannot cast implicit from (const T&)key to (T&)key. Otherwise you can change the constructor parameter to const T&.
Regards.

Please use 'Improve question' to tell us more about the OS and the compiler you use. And where the error message is thrown. Are you using an IDE or command line.
Regards.

Please dont post your response not as solution, rather as comment to the appropriate solution or comment.
Regards.

you have answerd your own question ;).

char *buffer = (char*)malloc(strlen(BODY) + strlen(HEADERS) + 256);
//...
closesocket(s);


Sorry:
v1.bstrVal = SysAllocString(L"111");

fine. i resume:

// old compiler
class A
{
public:
void Do(){}
void worker(void (A::*fn)())
{
(this->*fn)();
}
};

void main()
{
A a;
a.worker(A::Do);
}


// new compiler
void main()
{
A a;
a.worker(&A::Do);
}

You only have to add '&' before the function name as parameter.
Regards.

Enshure the function TranslateAccelerator is called. You can create the accelerator table manually by CreateAcceleratorTable. The command id must be (short) lower than 65536.
Regards.

class A
{
public: int i;
};
class B : public A
{
public: int i;
void set(int i)
{
A::i = i;
B::i = i;
}
};
This is the right way. What matters if the name of class A changes? ;)

Im afraid i dont understand exactly what you want.
Do you want to use 'MFCxxx.DLL' in your non mfc project?

No! Therefore the namespace operators are invented (MyClas::i) to resolve name conflicts.
Regards.

For the future ;) - use "Add comment" to my post. so i will be notified about your statement.
Regards.

.NET has nothing common with MFC. Regards.

I said:

LRESULT WndProc(HWND h,int m,WPARAM w,LPARAM l)
{
switch(m)
{
case WM_ACTIVATE:
if(w) SetFocus(GetWindow(h,GW_HWNDCHILD));
return 0;
}
return DefWindowProc(h,m,w,l);
}

Regards.

It woult be better to post a reply on my post so i will be notified about that.
There is indeed a possibility to create transparent windows by using non rectangular regions. that came up from win95. these windows are transparent an the part that is outside the region (for example the round corners of win-xp frames). to find such windows you should enumerate all windows (EnumWindows) and examine the windows region.
regards.

See my statement at solution 3. ;)

there is no universal solution in c.

Walk through the registry and look for the persist handler of the file types. You can use the IPersist (IPersistFile, IPersistStream) to load the objects. After loading you can ask (QueryInterface) the persist object for the IExtractImage interface. Another way is to use (i.e. ffmpeg) to extract a frame on a given timestamp.
Regards.

You cant find any because there arnt any images inside the wav, avi and mpeg formats. The thumbnails in the explorer view are created on the fly by supporting IExtractImage from the files persist handler.
Regards.

Whats your skype name?

Line 18 in error list is:
DWORD *a;//unsigned char *a;//pointer to the key value
If it isnt the right line tell us what line the error contains.
Regards.

No you didnt. You have to do this in that way:

int main()
{
{
x o;
o.show();
} // <-- destructor called
getch();
return 0;
}


Thanks. Formatting is a little bit tricky. ;)

Maybe ;). My template function works for every class that has a el member therfore there is more flexibility.
Regards.

What the hell you want? Do you really beleave anyone is reading your code!? ;)
I cant see any question in your post.
Regards.

Sorry you forgot the sort:

unsigned int findtwice(int* array,const unsigned int count)
{
enum{ NOTFOUND=-1 };
typedef struct { static int cmpi(void* a,void* b){ return *(int*)a-*(int*)b; } }_;
unsigned int i;
qsort(array,count,sizeof(int),_::cmpi);
for(i=1;(i<count)&&(array[i-1]!=array[i]);i++);
return i<count?i-1:NOTFOUND;
}

Regards.

Cant recognize the kind of the data encoding. The http header has not to be encoded - except the location (path and host when neccessary [1]) the data (urlencoded) has to be encoded in percent encoding [2].
Regards.

[1] http://www.ietf.org/rfc/rfc1738.txt
[2] http://en.wikipedia.org/wiki/Percent-encoding

A COM Server is starting by call CoCreateInstance, then the EXE will will wait for about 5000 ms (Microsoft COM server template in MFC) if the interfaces are not longer in use. Thats a thread is still waiting for a ref count becomes zero. If there are no further requests about the interfaces - the EXE server will shut down. To prevent the server from shutting down you have to AddRef one of the interfaces he provide. The are no external events or handles the exe is waiting for. The only thing to keep the COM-EXE alive is to hold an interface pointer. If there is no Release to that interfaces the server will never terminate (shutdown). So i had a problem with a product used by a C# environment their garbage collector never terminated my COM exe.
But be aware from usage of interfaces by loaded modules to prevent circular references. So you should define a common event object name the server can wait for until all loaded modules (DLL) and their threads are ready for unload. Usually COM uses the DllCanUnloadNow function to determine the module can be unloaded. Therefore you have to increment a global ref-count for critical objects or threads.
Regards.

Im afraid that i dont understand your intention. You can anytime- anwhere- from signal an event to another process. The event needs the neccessary access rigts to signal (see the naming conventions). You can tell your DLL the event name via function call parameter, so it can set the event on unloading (DLL_PROCESS_DETACH) the DLL. I tell it again: do not share any handle among processes.
Regards.

Cast operations are one of the fundamental operations in c. They exist a long time before microsoft invented their placebo cast keywords. The principle is that the programmer can cast anything to anything else. Therefore the programmer has to know what he is doing with a cast. That you should not cast a const value to a non const value isnt a evidence that this cannot be happen. Ive written this only for completeness and i never wrote that anybody should do that. But it can happen and a programmer should know that.
Thanks for attention.

You can give the root path as a parameter for the class constructor. to start at the corresponding tree node (HTREEITEM), you have to find it in the control and give it as paramter for the walk function. so you can walk from every starting point in your tree.
Regards.

In your code i cant see accessing a database. How do you access the database?
Regards.

I think the op has to learn more about the fundamentals of the c++ concept, before asking such questions. therefore i'm with you ;-)
Regards.

Its not my decision.
Usually virtual classes should be full proteced except the virtual members. Thats the background of COM or interfaces generally too. The pure virtual members are public and everyting else is protected. So the user can only call the virtual members.
Regards.

You should set the caret back to the first character position for the next search.
SendMessage(EM_SETSEL,0,0);
Regards.

Thats right. The variables i & j are declared 3 times. There is a potential possibility for buffer overflows without limit n & m to max 100 (buffer length). There must not be brackets for single statements, but this looks better. All that are not really errors but them make no sense at all.
Regards.

??? Why did you post my code again? Is there something wrong?

Sorry, i havent the mfcplus with cshelltree. this code works with CTreeCtrl this is a base class of cshelltree, so far as i know.
Regards

Why do you downgrade answers? Refine your question to get better answers. Nobody can understand what you want from that fuzzy code snippet. You have to tell more from your code or what you want to do.
Regards.

Well ive updated the code with comments. ive spent more time for the comments than for the code.
btw: i dont understand why the comments arent green.
Regards.

lol ;-) calculations for NeverTheSameColor (ntsc format).

You can add an offset to lpd:
lpd += bpld * (cyDest-cySource/2)/2;
Regards.

I dont understand: do you want to clip from the center of the original bitmap? which height and width you want to use? do you want to stretch (resize) the bitmap? i need some explanations for that.
if you want to define a clip rectange, the code needs some changes.
at this time the code should clip from top-left the half size of original.
Regards.

How did you create sConnect?
sConnect = socket(AF_INET,SOCK_STREAM,IPPROTO_TCP);??

Thats right. There are two things to pay attention:
1) 8-bit per pixel creates a palette based bitmap (windows default palette). if you wont that set the palette entries to 0. then bitmap bits are interpreted as grey values.
2) usually you should use the CreateDIB... function (CreateBitmap is obsolet). if you do that you get the bitmap bits and must not allocate the buffer. in this case you can write the bitmap bits whenever you want.
The op's given code sets indeed the uninitialized buffer to the bitmap.
Regards.

You wont understand!? Since windows 3.0 EVERY line of bitmap bits (palette or rgb) MUST be padded do DWORD length in bytes. i.e. try to load 299 pixel width (24-bit rgb) bitmap (that remains 1 byte) with "your" word padding or with dword (4-byte) padding. What will you see?
btw: i must not read any m$ statements its my experience!
There is no difference between the bitmaps cause the grey bitmap is created by the values of the original bitmap. take a look at the op's posted code.
Regards.

Congratulations, but this isnt true. Every bitmap beginnig from 1-bit color deepth is DWORD aligned.
Regards.

What a cryptic calculation for the grey value. Why you dont calc it by: (r+g+b)/3 (or predefined table - faster)?
You should not cast the rgb-values in the plane itselves to a COLORREF value. Every byte in the plane represents r,g, and b. shure only in 8-bit deepth color bitmaps (but its true windows doesnt support 16-bit color deepth).
Regards.

No! All scan lines are DWORD (4-bytes) aligned.
You cannot use a modified width - you you have to use bytes-per-line with 4-bytes boundary to get the next line.
Regards.

That would be right, therefore use binary operation they consume a few tick, a division needs about hundred. but its a kind of thinking. what about some million operations they need x% of speed?
The other aspect is the readability of code.
Regards.

If you want to have fast code: reduce the amount of comparisations. my code is most fast as possible. cause im possessed by speed.
Regards.

And i told him there is no macro (to push), its a simple overload - one for release and another for debug. and everybody can declare his own private overload for new.
Regards.

Some COM functions are using CoTaskMemAlloc resp. CoTaskMemFree to allocate strings.
Regards.

Thats the point - remember the headline is: "Trying to understand what a linker does with code in .lib files". The linked lib is only forwarding the linked calls. Is the needed directx version not installed the exe will fail.
Regards.

Imagine you have a full screen window. Now you start such an operation. A disabled window cannot be moved or minimized at all. The window now is blocking the entire screen. The same is when a modal dialog pops up. This is a mayor bug in windows design.
Regards.

This answer explains nothing about how libraries work. There is no piece of code 'putten' into the exe. Most system libraries contains only call forwards on system dll functions. The compatibility is only supported by the functions themselves.
Regards.

Quite right, but there are two things to distinquish: 1) loading inline 2) loading in thread. for 1) you dont need to enable/disable any window, because there is no interaction until the loading is ready. 2) you should disable the controls particular to quit the application on the regular way (by press the close button or menu item). otherwise using a progress window to terminate the loading.
Regards.

hDlg is the window handle of the browse dialog (first parameter). IDC_TREE is constant and the control id of the tree control that displays the folder items.
Regards.

substitute the TreeView-macros:

htir = (HTREEITEM)::SendMessage(htree,TVM_GETNEXTITEM,TVGN_ROOT,0);
htis = (HTREEITEM)::SendMessage(htree,TVM_GETNEXTITEM,TVGN_CARET,0);
htic = (HTREEITEM)::SendMessage(htree,TVM_GETNEXTITEM,TVGN_CHILD,(LPARAM)htir);
for(count=pos=0;htic;htic=(HTREEITEM)::SendMessage(htree,TVM_GETNEXTITEM,TVGN_NEXTVISIBLE,(LPARAM)htic),count++){ if(htis==htic) pos=count; }

Good luck.

As far as i know the transparency of the window affects all the children. You cannot make a window transparent for about 50% and the children for 80%. That wont work - the is no support from any windows function. If you want to do so, you have to paint all the controls by yourself into a bitmap and set it to the layered window. There never comes a paint message. To invalidate any part of your window means to repaint all into the bitmap again and set it to the layered window (see my statment above).

Summary:
1) you can make a windows transparent with all children with the same alpha.
2) you can make a window transparent with different alpha values - you have to paint all content of it by yourself into a bitmap with alpha-plane.
3) if you want a alpha transparent window that is click transparent region you have to create an underlying window with alpha-plane that is ws_ex_transparent.
4) there is no way to make windows controls as child of another window transparent with another alpha value than the parent.

Regards.

Can you post the code how you have implemented the callback function.
The This variable is a pointer to my class i've casted to from the parameter in the callback function. Put the PostMessage into the global scope (like ::PostMessage). To use the Tree-control macros you have to include #include <commctrl.h>.
Regards.

If your initialisation should anything do for you - use pointers or references. All parameters are allocated on stack and freed after return.
Regards

Beg your pardon, but the goal of the op seems to be impossible. A natural language is an interactive process, otherwise you can use an existing programming language. The speech recognition of microsoft is dead for a couple of years. The visual recognition (project natal) is unsatisfactory. It seems to me that IBM is a little step further (project watson and others).
Regards.

Please use Add comment to add a comment!
Winerror.h says: 6==ERROR_INVALID_HANDLE
Regards.

You have to include:

#include <tchar.h>
#include <conio.h>
</conio.h></tchar.h>
i assumed you can use the dev-studio help manual. that example don't work in pure c (only c++).
Regards.

That sounds good. Perhaps you can offer that to the op as solution.
Regards.

Sorry but your code skips 66716671 IP-addresses.

Dont worry ;-)

Thats impossible, because the function SetFocus(hwnd) calls directly SendMessage(hwnd,WM_SETFOCUS,...) there is no message between (so far as i know).
Regards.

CWnd::OnSetFocus or handle the message WM_SETFOCUS.
Regards.

What you want to do?
* The view should not gain the focus
* But the child can get input focus?
Is this correct?
Then call SetFocus(hChild) on CYourView::OnSettingFocus().
Regards.

The op didnt talk about a c-file. this could be a mp3 file or somthing else. The mp3 header (version1) is appended to the end of the file (i.e.).
Regards.

see my last comment. regards.

There are some important values i have set to constants:


lpc[0] = __alphatable_value(lpbits[0],0/*color of the back plane*/,lpbits[3]);
lpc[1] = __alphatable_value(lpbits[1],0/*color of the back plane*/,lpbits[3]);
lpc[2] = __alphatable_value(lpbits[2],0/*color of the back plane*/,lpbits[3]);


color of the back plane is in this example RGB(0,0,0).
you can combine any other bitmap with your originally bitmap.
black and your alpha==0 is fully transparent. every back color is click transparent but visible.

the other constant is:


BLENDFUNCTION blend = { AC_SRC_OVER, 0, 255/*blend alpha*/, AC_SRC_ALPHA };


blend alpha value tells the system how to merge the window surface with the existing (windows) background.
0==fully transparent 255==only the bitmap transparency. the surface is only click transparent on the fully transparent plane.
ive spent a lot of time to understand whats windows doing with all the combining values. its not easy to check out what kind of combination you need to get the right effect.

if you want to get a click transparent shadow you have to create a blurred grayscale bitmap and combine it with the original (color of the back plane, see above).

Good luck, regards.

You can search here at this site for some articles about COM. To use that browser object you dont need to know anything about the http-protocol. Its the same object used in internet explorer.
Regards.

Layered windows working like windows with region. The part of the window that is fully transparant is not processed by any window message (paint, mouse events and so on). If you make the whole window alpha plane full transparent this works like ShowWindow(SW_HIDE) - you only can kill the app to quit.

I dont know why so many peaople want to control other applications. They're looking like officials - control others but never get anything themselves in line.
Regards.

Correct.

Indeed! - sometimes there is a reason to open a file exclusive.
Regards.

Why do you want to hijack other windows? You can subclass another window by replace the windowproc but this shouldnt (i think so) work from an injected module.

Enshure to release all interfaces you have got (if(pBuffer) pBuffer->Release(); pBuffer=0;). If you dont, the interfaces will go into a zombie-state.

Thanks. This was a very expensive research - wait for the bill. ;-)

Where do you want to paint to? Bitmap or Screen (HDC)? You can paint shaded pixels only if you have the pixel beneath (i.e. to a bitmap).
Regards.

Your fail, you should to understand the intention of the op. hint: who's the owner of the desktop window?

brr, bubble sort ;)

Thats right.
@OP: its not fine to paint over other windows. Create a single window for each line, that would be the better solution.
Regards

1. Your original question was about ReadFile problem, right?
2. I told you to use these functions to use instead CreateFile, ReadFile, WriteFile and so on.
3. You can use the functions to open, read and write named devices too.
4. This is code from my problem (only as example for you), that solved my problem to read/write a named device.
Regards.

Good answer. This will be the best way to improve speed.
Regards.

Nice try, but very buggy code and produces tons of memory leaks.

You voter of 4! Do you think you have a faster and smaller solution? Thanks for your anonymous devaluation.
Regards.

Add HTTP, FTP or other protocols to your search phrase.
Regards.

Use typedef and everything is ok.

typedef int I5x6[5][6];

void func(I5x6& z)
{
unsigned int i,j;
for(i=0;i<5;i++)
{
for(j=0;j<6;j++)
{
TRACE(__TEXT("%i\r\n"),z[i][j]);
}
}
}
void main()
{
I5x6 ip;
unsigned int i,j;
for(i=0;i<5;i++)
{
for(j=0;j<6;j++)
{
ip[i][j] = i+j;
}
}
func(ip);
}


I think also its a problem with the calling conventions. Most callback functions or exported function use the PASCAL (__stdcall) calling conventions.
Regards.

You cant change the drawing behaviour of a common control. The frame will been painted by the control themselves. The only thing you can do is subclassing and draw it by your own. But you will loose all windows design behaviours, because you cant do any kind of mixed paint mode.

You have to recompile all they modules using your .lib. For better scaling of your code you should use COM.
Regards.

Tell us more about the C# & VB solutions. You can fulfill your work in a other folder and finally rename it to the name you want.

It seems your class is destroyed while the thread is running. If you're using COM, you should AddRef on the beginning of your thread and Release at the end. Otherwise you should enshure your class cannot been deleted while thread is running. You can do that in your destructor you mut wait until the thread execution is finished by WaitForSingleObject(hYourThread,INFINITE).
Regards.

The "mangled" names are so called "decorations", that make the compiler able to adjust the stack around imported functions.
regards.

On which circumstances you draw the frames (on OnPaint)? OnPaint you should get a DC you can paint in (thats the client dc).
btw: you should delete your brushes (after about 10k not deleted gdi-objects you cant create any gdi object anymore).
Regards.

you can't use colon within a file name! Thats all, see:
http://msdn.microsoft.com/en-us/library/aa368590%28VS.85%29.aspx
Regards.

this is an example how to do that. i only posting code thats fully working (no useless fragments). i think, if you're using mfc and import your dlls by import library all this dont work.
The best way to do this is:
* make your dlls
* export the needed functions
* better you make interfaces of your code
* load the dlls manually (as shown above)
why is this the best way - using interfaces or exported functions?
* you never must recompile the other code on changes
encapsulate the load library into a small class you can use everywhere. indeed you should not make the loading in winmain you can do it in InitInstance of your application. And freelibrary at ExitInstance - thats the same.

You can use the wizard to creaete a member of your ocx control. The wizard will build the whole class with members (methods and properties) your ocx has. the ocx must have a type library for that. the control site and their functions are as protected memebers encapsulated in CWnd.

the topic headline is about c++ not c#.

It depends on the kind of objects you use. some objects are marshalled by system rpc and become a zombie state on more releases than necessary. you should everytime call Release() corresponding to AddRef() or QueryInterface(). Have you made the call, you should discard the interface pointer (better you use CComPtr). again: the Release() function is not a normal function - you can call as often as you want. This function is called once to release the usage of that interface.

i cant do your work - you should be able to substitute a simple in/out class.
btw: internet has a search function :mad:

#pragma pack(push)
#pragma pack(2)
typedef struct
{
BYTE bWidth; // Width, in pixels, of the image
BYTE bHeight; // Height, in pixels, of the image
BYTE bColorCount; // Number of colors in image (0 if >=8bpp)
BYTE bReserved; // Reserved ( must be 0)
WORD wPlanes; // Color Planes
WORD wBitCount; // Bits per pixel
DWORD dwBytesInRes; // How many bytes in this resource?
DWORD dwImageOffset; // Where in the file is this image?

} ICONDIRENTRY, *LPICONDIRENTRY;

typedef struct
{
WORD idReserved; // Reserved (must be 0)
WORD idType; // Resource Type (1 for icons)
WORD idCount; // How many images?

// ICONDIRENTRY idEntries[1]; // An entry for each image (idCount of 'em)
} ICONDIR;

typedef struct
{
BITMAPINFOHEADER icHeader; // DIB header
// RGBQUAD icColors[1]; // Color table
// BYTE icXOR[1]; // DIB bits for XOR mask
// BYTE icAND[1]; // DIB bits for AND mask

} ICONIMAGE;

#pragma pack(pop)


Are dll and exe both in 64-bit?

I think his misunderstood is: when class A derived from IUnknown implements the function, another class B derived from A and IUnknown can/should use the functions from A. But the compiler doesn't - because the class B needs its own instantiation of IUnknown.
regards.

you have to do that for the call of GetClassName too. These functions need the count of characters in the buffer not the size of the buffer themselves. Run time errors can occur anywhere in your program they dont appear the next step of your code. this depends on usage of your heap.
regards.

sorry, codeproject auto format ;)

Oh sorry, yes indeed i've read the question vice versa. :(

what exactly you want to do?

you should wait in command until the running thread has finnished. start the thread by calling CreateThread to get a waitable object.