Re: C++ ActiveX project worked with VB6, broken in later versions - C / C++ / MFC Discussion Boards

IDirect3DSurface9 Problem after Windows 10 Update Creator

002comp24-Jul-17 19:42

24-Jul-17 19:42

Hello Guys

I am creating a direct3D surface for further creation of OffScreenplainSurface for video sample.
And This Code was working fine before Windows10 Update.
After Update, I found that IDirect3DSurface not returning the right D3DSURFACE_DESC.format which I am using to createOffScreenPlainSurface.

Here is the Code

C++

CComPtr<IDirect3DSurface9> sampleSurface;
MFGetService(mediaBuffer, MR_BUFFER_SERVICE, __uuidof(IDirect3DSurface9), (void**)&sampleSurface);

D3DSURFACE_DESC surfaceDesc;
sampleSurface->GetDesc(&surfaceDesc);

CComPtr<IDirect3DSurface9> sharedSampleSurface;
HANDLE shareHandleSurface = 0;
device3D->CreateOffscreenPlainSurface(
			surfaceDesc.Width,
			surfaceDesc.Height,
			surfaceDesc.Format,
			D3DPOOL_DEFAULT,
			&sharedSampleSurface,
			&shareHandleSurface
			);

Now, after windows10 Update surfaceDesc.format = 842094158 not the D3DFMT_YUY2

And I realized that sometimes after restart of Machine, it returns D3DFMT_YUY2 which looks like some services in windows10 update obstructing MFGetService.
I found one service sppsvc which automatic delayed start and Trigger start but I am not able to get much of it.

Any Help will be Appreciated.

Thank You.

Re: IDirect3DSurface9 Problem after Windows 10 Update Creator

Jochen Arndt24-Jul-17 21:19

Jochen Arndt

24-Jul-17 21:19

842094158 is 0x3231564E which is MAKEFOURCC('N', 'V', '1', '2').

So your mediaBuffer seems to contain frames in that format.

I don't know much about Direct3D but I guess that the format has changed with the video driver installed with Windows 10.

Re: IDirect3DSurface9 Problem after Windows 10 Update Creator

002comp24-Jul-17 23:03

002comp

24-Jul-17 23:03

Thanks Jochen for your Reply.

It makes a sense that As I read on many formuns that latest windows10 update had problem with Graphic drivers beacause it is not behaving correctly with the updates so some of them choose to roll back.

Again Thanks a Lot.

Re: IDirect3DSurface9 Problem after Windows 10 Update Creator

002comp25-Jul-17 19:36

002comp

25-Jul-17 19:36

Hello
Is that possible that I get different Format from particular one machine and unchanged hardware?
As after Windows 10 Update, sometimes I am getting NV12 surface format and sometimes YUY2.

Thank you.

Re: IDirect3DSurface9 Problem after Windows 10 Update Creator

Jochen Arndt25-Jul-17 20:57

Jochen Arndt

25-Jul-17 20:57

As I wrote, I don't know much about Direct3D. But there are some components like drivers and decoders involved which also have their own configuration.

A quick research shows that Windows is providing decompressed video streams in the YUY2 or NV12 format. So an application should probably support both formats.

Re: IDirect3DSurface9 Problem after Windows 10 Update Creator

002comp25-Jul-17 21:19

002comp

25-Jul-17 21:19

Thanks Jochen for your reply.

I also found that Windows Supports both formats YUY2 or NV12.
But, I am struggling why it is randomly showing either format which proves that there are some setting that needs to be set As I did not get any theory/description on getting random format from these two formats.

Thanks Again for your Reply.

C++ ActiveX project worked with VB6, broken in later versions

Burl Rice22-Jul-17 10:30

Burl Rice

22-Jul-17 10:30

I have an ActiveX control (written originally in VC6, and later upgraded to VC 2010). When we first released it, I included a VB6 demo program to show how to use it. This all still works.

Recently, a customer requested sample code in VB 2017, and I have found that it crashes in this environment. I have also checked VB 2015 and VB 2010, and they also crash in the same place.

Here is a simplified overview of the ActiveX library:

library BoxWriter
{
    interface ILine : IDispatch 
    {
        [propget, id(8), helpstring("property m_lCount")] HRESULT m_lCount([out, retval] long *pVal);
    };
    interface IPrinter : IDispatch
    {
        [propget, id(1), helpstring("property m_line")] HRESULT m_line([optional, defaultvalue ("")] BSTR strLine, [out, retval] ILine* *pVal);
    };
};

Here is a VB code snippet that causes the crash:

Dim printer As New BoxWriter.PRINTER
printer.Connect("192.168.1.3")
MsgBox(m_printer.m_line.m_lCount)    ‘ <—crashes here

This is the error it gives:

An unhandled exception of type 'System.AccessViolationException' occurred in demo.exe
Additional information: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.

The reference returned by CPrinter::get_m_line seems to be what it is complaining about. When I debug, I see the reference is set correctly by CPrinter::get_m_line, but after the function returns, something is calling the destructor for the CLine class.

STDMETHODIMP CPrinter::get_m_line(BSTR bstrLine, ILine **pVal)
{ 
    AFX_MANAGE_STATE(AfxGetStaticModuleState())

    CString strLine = (LPCTSTR)_bstr_t (bstrLine);
    CComObject <CLine>* pLine = GetLine (strLine);

    if (!pLine) {
        return E_INVALIDARG;
    }

    * pVal = pLine;

    return S_OK;
}

I have tried creating a new ActiveX project in VC2010, and it also exhibits the same behavior. I tried returning the reference in several different ways (property vs method), and still have not been able to solve this. Surely I am not the first person to encounter this. Can anyone suggest how to fix this?

Re: C++ ActiveX project worked with VB6, broken in later versions

Richard MacCutchan22-Jul-17 20:29

Richard MacCutchan

22-Jul-17 20:29

At a guess your object is going out of scope and getting destroyed. You need to allocate a fixed block of memory (malloc or new) and copy the data to that block which you can then return. But that also means that the calling method will need to clean up after itself.

Re: C++ ActiveX project worked with VB6, broken in later versions

Burl Rice23-Jul-17 2:32

Burl Rice

23-Jul-17 2:32

I thought that was what CComObject::AddRef was supposed to do?

Re: C++ ActiveX project worked with VB6, broken in later versions

Richard MacCutchan23-Jul-17 2:54

Richard MacCutchan

23-Jul-17 2:54

Quite possibly but it is not clear in your code where that occurs, and if it applies to the pointer you are passing back.

Re: C++ ActiveX project worked with VB6, broken in later versions

Burl Rice23-Jul-17 9:12

Burl Rice

23-Jul-17 9:12

After some experimentation (mostly due to this article: CComObject Class[^]), I have made some progress towards a fix. It seems this line, which was OK in VB6, is no longer appropriate:

CComObject <CLine> * pLine = new CComObject <CLine> ();

I replaced it with this logic, and my VB 2017 app no longer crashes:

CComObject <CLine> * pLine;

HRESULT hRes = CComObject<CLine>::CreateInstance(&pLine);
ATLASSERT(SUCCEEDED(hRes));
pLine->AddRef ();

Class Shape

kumn1d22-Jul-17 9:05

kumn1d

22-Jul-17 9:05

This is the output I am supposed to get but I am getting errors please help me

C++

int main()
{
	int size = 5;
	
	Shape** shapes = new Shape*[size];

	Point* centre = new Point(5,5);
	Rectangle* rectangle1 = new Rectangle(*centre, 4, 3);
	shapes[0] = rectangle1;
	
	shapes[1] = new Rectangle (*new Point(10, -10), 4, 4);

	Rectangle* rectangle3 = new Square(*new Point(-8.5,-10), 2.5);
	shapes[2] = rectangle3;

	Square* square2 = new Square (*new Point(-1,-1), 1.5);
	shapes[3] = square2;
	
	Circle* circle = new Circle(*new Point(100,100), 25.4);
	shapes[4] = circle;

	cout << "\n11111111111111111111111111\n";
 
	for (int i = 0; i < size; ++i)
		shapes[i] -> display();

	cout << "\n22222222222222222222222222\n";

	cout << "\nthe area of the first rectangle is: ";
	cout << rectangle1->area() << endl;

	cout << "It is a square, the area is: ";
	Square* square = dynamic_cast<Square*> (rectangle3);
	cout << square->area() << endl;	

	cout << "the perimeter of the circle is: ";
	cout << circle->perimeter() << endl;	

	cout << "\n33333333333333333333333333\n";

	cout << "\nThe four vertices of the first rectangle are: (clockwise from top-left)\n";
	rectangle1->printFourVertices();

	cout << "\nThe four vertices of the third rectangle are:\n";
	rectangle3->printFourVertices();

	cout << "\n44444444444444444444444444\n";

	rectangle1->reflectX();

	shapes[2] -> translate(1.5, 3);

	shapes[4] -> translate(-100,-100);

	cout << "\nAfter reflection and translation, here are the moved shapes:\n" ;
	for (int i = 0; i < size; ++i)
		shapes[i] -> display();
	
	cout << "\n55555555555555555555555555\n";
 
	cout << "\nNow, the four vertices of the first rectangle are:\n";
	rectangle1->printFourVertices();
 
	cout << "\nNow, the four vertices of the third rectangle are:\n";
	rectangle3->printFourVertices();

	cout << "\n66666666666666666666666666\n\n";
 
	for (int i = 0; i < size; ++i)
		cout << shapes[i] -> area() << endl;

	// From the following statements, understand how setLength() and setWidth()
	// work for rectangles and squares.

	cout << "\n777777777777777777777777777\n";

	shapes[1] -> display();
	dynamic_cast<Rectangle*> (shapes[1]) -> setLength(8.2);
	cout << "\nAfter setLength(8.2), the rectangle is: ";
	shapes[1] -> display();
	dynamic_cast<Rectangle*> (shapes[1]) -> setWidth(5);
	cout << "\nAfter setWidth(5), the rectangle is: ";
	shapes[1] -> display();

	square2 -> display();
	square2 -> setLength(8.2);
	cout << "\nAfter setLength(8.2), the rectangle is: ";
	square2 -> display();
	square2 -> setWidth(5);
	cout << "\nAfter setWidth(5), the rectangle is: ";
	square2 -> display();

	system("pause");
	return 0;
}


/*
11111111111111111111111111

Rectangle:
        Centre: (5, 5)
        Length: 4
        Width: 3

Rectangle:
        Centre: (10, -10)
        Length: 4
        Width: 4

Square:
        Centre: (-8.5, -10)
        Side Length: 2.5

Square:
        Centre: (-1, -1)
        Side Length: 1.5

Circle:
        Centre: (100, 100)
        Radius: 25.4

22222222222222222222222222

the area of the first rectangle is: 12
It is a square, the area is: 6.25
the perimeter of the circle is: 159.512

33333333333333333333333333

The four vertices of the first rectangle are: (clockwise from top-left)
(3, 6.5)
(7, 6.5)
(7, 3.5)
(3, 3.5)

The four vertices of the third rectangle are:
(-9.75, -8.75)
(-7.25, -8.75)
(-7.25, -11.25)
(-9.75, -11.25)

44444444444444444444444444

After reflection and translation, here are the moved shapes:

Rectangle:
        Centre: (5, -5)
        Length: 4
        Width: 3

Rectangle:
        Centre: (10, -10)
        Length: 4
        Width: 4

Square:
        Centre: (-7, -7)
        Side Length: 2.5

Square:
        Centre: (-1, -1)
        Side Length: 1.5

Circle:
        Centre: (0, 0)
        Radius: 25.4

55555555555555555555555555

Now, the four vertices of the first rectangle are:
(3, -3.5)
(7, -3.5)
(7, -6.5)
(3, -6.5)

Now, the four vertices of the third rectangle are:
(-8.25, -5.75)
(-5.75, -5.75)
(-5.75, -8.25)
(-8.25, -8.25)

66666666666666666666666666

12
16
6.25
2.25
2025.8

777777777777777777777777777

Rectangle:
        Centre: (10, -10)
        Length: 4
        Width: 4

After setLength(8.2), the rectangle is:
Rectangle:
        Centre: (10, -10)
        Length: 8.2
        Width: 4

After setWidth(5), the rectangle is:
Rectangle:
        Centre: (10, -10)
        Length: 8.2
        Width: 5

Square:
        Centre: (-1, -1)
        Side Length: 1.5

After setLength(8.2), the rectangle is:
Square:
        Centre: (-1, -1)
        Side Length: 8.2

After setWidth(5), the rectangle is:
Square:
        Centre: (-1, -1)
        Side Length: 5
*/

This is what I have tried

C++

// jianger.cpp : Defines the entry point for the console application.
//

#include "stdafx.h"
# include<iostream>
# include<cmath>
# include<string>

using namespace std;
class Point
{
private:
	double x;
	double y;
public:
	Point();
	Point(const double &x,const double &y);
	void setX(const double &x);
	void setY(const double &y);
	double getX()const;
	double getY()const;
	double DisFromOrigin()const;
	virtual void translate(const double &xDis, const double &yDis);
	void reflectX(); // chnage my private members
	void reflectY();
	virtual void PrintName()const;
};
Point::Point()
{
	x=0;
	y=0;
}
Point::	Point(const double &x,const double &y)
{
	this->x=x;
	this->y=y;
}
void  Point::setX(const double &x)
{
	this->x=x;
	this->y=y;
}
void Point:: setY(const double &y)
{
	this->y =y;
}
double Point:: getX()const
{
	return x;
}
double Point:: getY()const
{
	return y;
}
double  Point::DisFromOrigin()const
{
	return (sqrt(x*x + y*y));
}
void Point:: translate(const double &xDis, const double &yDis)
{
	double a;
	double b;
	a = this->x + xDis;
	b= this->y + yDis;
	this->x= a;
	this->y=b;
}
void Point::reflectX()
{
	this->y = -1 *y;
}
void Point::reflectY()
{
	this->x= -1*x;
}
void Point::PrintName()const
{
	cout<<"Point ("<<getX()<<" , "<<getY()<<")"<<endl;
}
class Shape
{
public:
	virtual double Peri()const=0;
	virtual void PrintName()const=0;
	virtual void translate(double x1,double y1)=0;
	virtual void reflectX()=0;
	virtual void reflectY()=0;
	virtual double getArea()const=0;
	virtual double getPeri()=0;
	virtual string getName()const=0;
}
class Circle:public Shape
{

public:
	Circle();
	Circle(const Point &p, const double &r);
	double getR()const;
	void setR(const double &r);
	virtual double getPeri()const;
    virtual void PrintName()const;
	virtual void translate(const double &x1, const double  &y1);
	virtual void reflectX();
    virtual void reflectY();
    virtual double getArea()const;
   virtual string getName()const;
protected:
	Point center;
	string name;
	double radius;
	 

};
Circle::Circle()
{
	name = "Circle";
	center = Point(0,0);
	radius = 1;
}
Circle::Circle(const Point &p, const double &r)
{
	name = "Circle";
	this->center= p;
	this->radius=r;
}

double Circle::getPeri()const
{
	return (2*3.14*getR());
}
double Circle::getR()const
{
	return radius;
}
void Circle::setR(const double &r)
{
	this->radius = r;
}
 void Circle:: PrintName()const
 {
	 cout<<"Circle"<<endl;
 }
 void Circle::translate(const double &x1, const double &y1)
 {
	 this->center.translate(x1 , y1);
 }
void Circle::reflectX()
{
	this->center.reflectX();
}
void Circle::reflectY()
{
	this->center.reflectY();
}
double Circle::getArea()const
{
	return (3.14*this->getR()*this->getR());
}
string Circle::getName()const
{
	return this->name;
}
class Rectangle:public Shape
{
private:
	double len;
	double width;
	Point center;
	string name;
public:
	Rectangle();
	Rectangle(const Point &p, const double &len , const double &width);
	void setLen(double len);
	void setWid(double width);
	double getLen()const;
	double getWid()const;
	virtual double getPeri()const;
	virtual void PrintName()const;
	virtual void translate(const double &x1, const double  &y1);
	virtual void reflectX();
	virtual void reflectY();
	virtual double getArea()const;
	virtual string getName()const;
};
Rectangle::Rectangle()
{
	len =1;
	width=1;
	this->name = "Rectangle";
	center = Point(0,0);
}

Rectangle::	Rectangle(const Point &p, const double &len , const double &width)
{
	this->name ="Rectangle";
	this->center = p;
	this->len = len;
	this->width = width;
}
double Rectangle:: getPeri()const
{
	return (2*getLen() + 2*getWid());
}
void Rectangle::PrintName()const
{
	cout<<"Rectangle"<<endl;
}
void Rectangle::translate(const double &x1, const double  &y1)
{
	this->center.translate(x1, y1);
}
void Rectangle::reflectX()
{
	this->center.reflectX();
}
void Rectangle::reflectY()
{
	this->center.reflectY();
}
double Rectangle::getArea()const
{
	return(getLen()*getWid());
}
string Rectangle::getName()const 
{
	return name;
}
void Rectangle::setLen(double len)
{
	this->len = len;
}
void Rectangle::setWid(double width)
{
	this->width = width;
}
double Rectangle::getLen()const
{
	return len;
}
double Rectangle::getWid()const
{
	return width;
}
class Square:public Rectangle
{
private:
	Point center;
	string name;
	double side;
public:
	Square();
	Square(const double &side);
	void setSide(const double &side);
	double getSide()const;
	virtual double getPeri()const;
	virtual void PrintName()const;
	virtual void translate(const double &x1, const double  &y1);
	virtual void reflectX();
	virtual void reflectY();
	virtual double getArea()const;
	virtual string getName()const;
};
Square::Square()
{
	side=0;
}
Square::Square(const double &side)
{
	this->side = side;
}
void Square::setSide(const double &side)
{
	this->side = side;
	this->Rectangle::setLen(side);
	this->Rectangle::setWid(side);
}
double Square::getSide()const
{
	return this->Rectangle::getLen();
}
double Square:: getPeri()const
{
	return(this->Rectangle::getLen()* 4);
}
void Square::PrintName()const
{
	cout<<"Square"<<endl;
}
void Square::translate(const double &x1, const double  &y1)
{
	this->Rectangle::translate(x1, y1);
}
void Square::reflectX()
{
	this->Rectangle::reflectX();
}
void Square::reflectY()
{
	this->Rectangle::reflectY();
}
double Square::getArea()const
{
	return(this->Rectangle::getLen()* this->Rectangle::getLen());
}
string Square::getName()const
{
	return name;
}

int _tmain(int argc, _TCHAR* argv[])
{
	int size = 5;
	
	Shape** shapes = new Shape*[size];

	Point* centre = new Point(5,5);
	Rectangle* rectangle1 = new Rectangle(*centre, 4, 3);// it means that this can not be abstract class!!! coz it is calling object of itself so no virtual functions
	shapes[0] = rectangle1;
	
	shapes[1] = new Rectangle (*new Point(10, -10), 4, 4);

	Rectangle* rectangle3 = new Square(*new Point(-8.5,-10), 2.5);
	shapes[2] = rectangle3;

	Square* square2 = new Square (*new Point(-1,-1), 1.5);
	shapes[3] = square2;
	
	Circle* circle = new Circle(*new Point(100,100), 25.4);
	shapes[4] = circle;

	cout << "\n11111111111111111111111111\n";
 
	for (int i = 0; i < size; ++i)
		shapes[i] -> display();

	return 0;
}

modified 22-Jul-17 15:46pm.

Re: Class Shape

Afzaal Ahmad Zeeshan22-Jul-17 10:00

Afzaal Ahmad Zeeshan

22-Jul-17 10:00

Do not repost questions, you have already posted the same question here, Class shape and related functions. If someone is active to answer your question, they will do it. If there is no one, reposting the question doesn't help at all.

The sh*t I complain about
It's like there ain't a cloud in the sky and it's raining out - Eminem
~! Firewall !~

Re: Class Shape

_Flaviu23-Jul-17 23:42

_Flaviu

23-Jul-17 23:42

What kind of errors do you get ?

Re: Class Shape

Richard MacCutchan23-Jul-17 23:58

Richard MacCutchan

23-Jul-17 23:58

See Afzaal's reply above.

Re: Class Shape

Victor Nijegorodov24-Jul-17 22:38

Victor Nijegorodov

24-Jul-17 22:38

kumn1d wrote:
This is the output I am supposed to get but I am getting errors please help me

What error? The compiler ones? - Then show us the errors.
Or you meant the incorrect output? - Then try to debug your code.

Re: Class Shape

kumn1d25-Jul-17 8:52

kumn1d

25-Jul-17 8:52

I corrected the entire code , but the only thing now I dont know is how to implement this Print four Vertices function . 
I want to know how to implement anf declare this function . Beside that , I corrected the rest of code

Re: Class Shape

David Crow25-Jul-17 2:23

David Crow

25-Jul-17 2:23

There does not seem to be a Rectangle::printFourVertices() method. There is a redundant assignment operator in the Point::setX() method.

Short of that, what's the issue? "I am getting errors" is meaningless.

"One man's wage rise is another man's price increase." - Harold Wilson

"Fireproof doesn't mean the fire will never come. It means when the fire comes that you will be able to withstand it." - Michael Simmons

"You can easily judge the character of a man by how he treats those who can do nothing for him." - James D. Miles

Re: Class Shape

kumn1d25-Jul-17 8:51

kumn1d

25-Jul-17 8:51

I corrected the entire code , but the only thing now I dont know is how to implement this Print four Vertices function .
I want to know how to implement anf declare this function . Beside that , I corrected the rest of code

Re: Class Shape

David Crow25-Jul-17 10:15

David Crow

25-Jul-17 10:15

How would you do it using pencil and paper? If you coded up everything that calculates said vertices, printing them should be a non-issue, yes?

Educational: my buffer scroll C++ working code.

Member 132416522-Jul-17 0:31

Member 1324165

22-Jul-17 0:31

Hello. It my first message here, I'll try to do it the right way.

The programming is my hobby (MinGW 32bit). About topic. I would like to represent my C++ code for buffer scroll. Perhaps I designing another bike, but still... What I mean about scrolling. For example, I have some buffer in which data starts somewhere in the middle:

C++

char * buffer = "World!Hello, ";

But I need:

C++

char * buffer = "Hello, World!";

So I need to cut out data at the end, move the rest of data to the end of buffer and paste cutted data to the beginning. This is "lossless buffer scrolling" of what I am talking about.

This is a trivial job if ones use temporary buffer not less than half of the *buffer's length, few memcpy() statements and so on. But this is not my case.

Condition 1. Source buffer size is very large (1/2 or 2/3 of total physical ram). So it is not good idea to allocate temporary buffer.
Condition 2. I can not use data storage like HDD drive. I can not use any kind of swap techniques. (It will produces temporary system freeses.)
Condition 3. I am not what to use virtual addressing techniques to virtually scroll buffer.
Condition 4. The buffer size and scroll distance is arbitrary (sets as function parameters).

So I created a scrolling buffer by permutation, which source code I would like to public. It is quite tested by brutte-force method (buffer size and scrolling distance were alternated, the result is compared with alternative scrolling techniques). In my opinion, the implementation of the method is unusual (or, at least, interesting).

Here it is:

C++

namespace CommonTools
  {
 /// Code by Shepel' Nikolay E., Moscow, Russia, published at 22 July 2017, http://www.codeproject.com.
#ifdef DEBUG
  uint32_t scroll_raw_buffer_last_rounds = 0;
#endif

  void scroll_raw_buffer (void *data , const uint32_t buffer_size, uint32_t distance)
  {
#ifdef DEBUG
    scroll_raw_buffer_last_rounds = 0; // Statistics about rounds...
#endif
    // Для реализации скроллинга необходим двойной запас по размеру буффера,
    // т.е. buffer_size<(UINT_MAX/2) если использовать в функции 32х разрядные
    // переменные, иначе можем выскачить за их максимальный размер, если
    // надо будет буффер сместить влево на 1.

    // WARNING!!!
    // We need double reserve for addressing buffer elements (buffer_size<(UINT_MAX/2)).
    // Otherwise we can get register overflow while scrolling buffer with
    // size>=INT_MAX to the left with distance==1
    if (buffer_size >= INT_MAX)
      throw CommonTools::Errors::Simple (CommonTools::Errors::StE_Overflow, __PLACE__);

    if ( (!data) || (buffer_size < 2)) return;
    distance %= buffer_size;
    if (!distance) return;
    TRAP_in_DEBUG (distance >= buffer_size, msg_interror)

    uint8_t *buf = (uint8_t *) data;
    const uint32_t shift_value = distance;



#ifdef DEBUG
    static const size_t max_stack_buffer = 256;
#else
    static const size_t max_stack_buffer = 1024;
#endif

    // If buffer size is quite small we can use the simplest way...
    if (buffer_size <= max_stack_buffer)
      {
        uint8_t *tmp_buf = (uint8_t *) alloca (buffer_size);
        memcpy (tmp_buf, buf, buffer_size);

        memcpy (buf + shift_value, tmp_buf, buffer_size - shift_value);
        memcpy (buf, tmp_buf + buffer_size - shift_value, shift_value);
        return;
      }

#ifdef DEBUG
    scroll_raw_buffer_last_rounds = 1;
#endif

    uint32_t bytes_left = buffer_size, bytes_per_round = 0;
    uint32_t total_rounds = 0;
    uint32_t round_number = 1; // [1...total_rounds]
    uint8_t  *next_data_buffer = 0;
    uint32_t dest_index;
#ifdef DEBUG
    const uint8_t * const out_of_buffer = buf + buffer_size;
#endif

    do // Starting "merry-go-round"
      {
        {
          const uint32_t round_start_offset = round_number - 1;
          dest_index = round_start_offset;
          if (dest_index >= buffer_size) dest_index -= buffer_size;
          uint8_t next_data = buf[dest_index];
          do
            {
              dest_index += shift_value; if (dest_index >= buffer_size) dest_index -= buffer_size;
              register uint8_t &destination = buf [dest_index];
              swap (destination, next_data);
              bytes_left--;
            }
          while (dest_index != round_start_offset);
        }

        if (!bytes_left) // If it was last round (or the single one) we finish here...
          break;
        if (round_number == 1)
          {
            bytes_per_round = buffer_size - bytes_left;
            total_rounds = buffer_size / bytes_per_round;
#ifdef DEBUG
            scroll_raw_buffer_last_rounds = total_rounds;
#endif
            // The residue of division must be zero !!!
            TRAP_in_DEBUG (buffer_size % bytes_per_round, msg_bogus_state)
          }
        round_number++;

        // If we come here so multiple rounds are required  for scroll operation
        // I think it is because buffer length and distance have common dividors... (I'm not strong in mathematics)
        // We can use some sort of optimisation in this case. We can move blocks of data.
        // The size of such blocks are limited by common dividors mensioned before...

        if (!next_data_buffer) // The temporary buffer for blocks operation.
          next_data_buffer = (uint8_t *) alloca (max_stack_buffer);

        // The code in cycle below is usefull if block_size>1.
        // We check this indirectly by means of condition in "for()" statement below.
        // Otherwise this cycle will be skipped and upper cycle
        // will be used with single byte code.
        for (;round_number  < total_rounds;) //the same condition as if we use "(total_rounds-round_number+1)>1" but less operations are required
          {
            const uint32_t rounds_left = total_rounds - round_number + 1;
            // Activating block move
            const size_t block_size = min_ct (max_stack_buffer, rounds_left);
            const uint32_t round_start_offset = round_number - 1;
            dest_index = round_start_offset;
            if (dest_index >= buffer_size) dest_index -= buffer_size;

            TRAP_in_DEBUG (buf + dest_index + block_size >= out_of_buffer, msg_interror)
            memcpy (next_data_buffer, buf + dest_index, block_size);
            TRAP_in_DEBUG (bytes_left < bytes_per_round*block_size, msg_interror)
            do // block move DO
              {
                dest_index += shift_value; if (dest_index >= buffer_size) dest_index -= buffer_size;
                register uint8_t *destination = buf + dest_index;

/**
        If region to copy overlaps the end of main buffer, special copy
        routine is required (see below).
        But this is never happens, because after the end of the buffer
        the beggining  starts. The beggining element is moved at the
        firstest step in upper cycle. The algo is designed that it is
        impossible if some byte is moved twice.
        So I left this commented (disabled) block (I am not sure that
        it is works good because it is never tested):
**/
/*
                if (destination + block_size > out_of_buffer)
                  {
                    const uint32_t part1 = out_of_buffer-destination;
                    TRAP_in_DEBUG (part1>block_size,msg_interror)
                    memswap (destination,next_data_buffer,part1);
                    const uint32_t part2 = block_size-part1;
                    memswap (buf,next_data_buffer+part1,part2);
                  }
                else
*/
              memswap (destination, next_data_buffer, block_size);
              } // EOF block move DO
            while (dest_index != round_start_offset);
            bytes_left -= bytes_per_round * block_size;
            round_number += block_size;
          } // EOF for (;round_number  < total_rounds;)
      } // Do "merry-go-round"
    while (bytes_left > 0);

    return;
  }

} // namespace CommonTools

Important notes:
1. If ones need negative direction you can use distance=buffer_size-1 as -1 direction.
2. swap() is just variables swapping, min_ct() function returns the minimum value from two arguments.
3. TRAP_in_DEBUG() or TRAP() are macroses to terminate program with message if first parameter is true (message is in 2nd parameter, like ASSERT() but little different way).
4. Function has size limit of scrolling buffer (up to INT_MAX, about 2 Gb in my case).
5. If someone would like to use this code, I will be pleasured if my name is mentioned somewhere in credits..

Echo7

modified 22-Jul-17 11:27am.

Re: Educational: my buffer scroll C++ working code.

Richard MacCutchan22-Jul-17 0:57

Richard MacCutchan

22-Jul-17 0:57

This looks more like a Tip than a question and would be better posted at Submit a new Article[^]. But please read the posting guidelines[^] first.

Re: Educational: my buffer scroll C++ working code.

Member 132416522-Jul-17 5:19

Member 1324165

22-Jul-17 5:19

Quote:
This looks more like a Tip than a question and would be better posted at Submit a new Article[^].

I'll do it, when I have read about rules.

Echo7

Re: Educational: my buffer scroll C++ working code.

leon de boer22-Jul-17 1:14

leon de boer

22-Jul-17 1:14

yes it belongs in tips but I also have a little tip for you.
memmove from the C/C++ standards library can deal with source and destination memory overlapping and it can do it with no extra buffers at all.
So you really only need the string search parts of your code and then use memmove to do the shuffle replacing the memcpy's

memmove[^]
memmove ensures that the original source bytes in the overlapping region are copied before being overwritten.

It was probably a good learning excercise and you can rate how you went against a standards library and here is the code
public source Implementation of memcopy[^]

It's never wasted time to learn the standards libraries Smile | :)

In vino veritas

modified 22-Jul-17 7:22am.

Re: Educational: my buffer scroll C++ working code.

Member 132416522-Jul-17 5:09

Member 1324165

22-Jul-17 5:09

Quote:
memmove from the C/C++ standards library can deal with source and destination memory overlapping and it can do it with no extra buffers at all

I know it. But I can not implement memmove() to my task. memmove() will ovewrite some part of my buffer (or I have to use temporary buffer to store that part) so I'll loose some data. Perhaps there is a way to use it, but I do not yet know how.

Quote:
you really only need the string search parts of your code

C++
char * buffer = "World!Hello, ";

It was only the small example to illustrate what I meaning under "buffer scrolling" term. In real code I have more that 100Mb binary (non-text) raw data which I have got from external device.

I do not know (in advance) the size, and I do not know the move distance.

Echo7

modified 22-Jul-17 11:21am.

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.