A strange problem with regions

**Overlord** · 05-09-2008

Hi!

I'm trying to loop through pixels of a bitmap, check for their RGB values and add to a region if they match with a specified color.

Code:

	for(WORD y=0; y<wBmpHeight; y++)
	{
		for(WORD x=0; x<wBmpWidth; x++)
		{
			for(std::map<char*, CMould>::iterator it=moulds.begin(); it!=moulds.end(); it++) //this map holds 26 different color values and region handles
			{
				if(RGB(pPixels[p+2], pPixels[p+1], pPixels[p]) == it->second.trnsphRGB)
				{ //if the colors match
					int iRes = CombineRgn(it->second.regionHandle, 
                                                            it->second.regionHandle, 
                                                            CreateRectRgn(x, y, x+1, y+1), RGN_XOR); //add this pixel to the region
				}
			}
			p+=3;
		}
		itoa(y, pOut, 10); //just to display the current y line
		SetWindowText(hwnd, pOut);
	}

All goes well and CombineRgn returns COMPLEXREGION (success) until line 138, pixel 514. The function returns ERROR on this and further pixels. I found out with GetLastError() that the error code is 6 (invalid handle). I don't get why it->second.regionHandle is invalid, because the debugger shows the exact same handle address for both- successful and unsuccessful calls to CombineRgn.

The CMould class is nothing special, regionHandle is initialized in the constructor.

Header:

Code:

#ifndef MOULD_H
#define MOULD_H

class CMould
{
public:
	CMould(COLORREF color, char** shouts);
	CMould(){}
	HRGN regionHandle;
	COLORREF trnsphRGB;
	char** pShouts;
};

#endif

CPP:

Code:

#include "stdafx.h"

CMould::CMould(COLORREF color, char** shouts)
{
	trnsphRGB=color;
	memcpy(pShouts, shouts, sizeof shouts);
	regionHandle = CreateRectRgn(0, 0, 1, 1);
}

I can't think of what might be causing it.
Thanks for any help.

**brewbuck** · 05-09-2008

I suspect you are just creating a region too complicated for Windows to deal with.

**~~master5001~~** · 05-09-2008

Regions are a screwy thing in the API, my friend. They are actually a very resource consuming object. Thus memory leaking is always lurking around the corner.

Rasterization is a good way around this (in my experience). Instead of adding each pixel to your region at once, add code that does something like:

Example:

Code:

// Sorry for not being up on my MFC... but you should be able to MFC-ify this.
RECT line = {0,0,0,0};
bool inside = false

for(int y = 0; y < wBmpHeight; ++y)
{
    line.top = y;
    line.bottom = y+1;

    for(int x = 0; x < wBmpWidth; ++x)
    {
        // I am switching more to pseudo-coding now, but you get the idea :)
        if(RGB(pixel(y*wBmpWidth+x)) == myTransparentColor)
        {
            if(!inside)
              line.left = x;
        } else
        {
            if(inside)
            {
               line.right = x-1;   // This makes it go back to the last pixel which was good.
               inside = false;

               HRGN extra = CreateRectRgnIndirect(&line);
               CombineRgn(second, second, extra, RGN_XOR);
               DeleteObject(extra); // this is actually probably what killed your code, btw.
            }
        }
    }
   
    // Make sure to finish a line too
    if(inside)
    {
        line.right = x-1;   // This makes it go back to the last pixel which was good.
        inside = false;

        HRGN extra = CreateRectRgnIndirect(&line);
        CombineRgn(second, second, extra, RGN_XOR);
        DeleteObject(extra); // again, we don't want memory leaks.
    }
}

**~~master5001~~** · 05-09-2008

Originally Posted by brewbuck

I suspect you are just creating a region too complicated for Windows to deal with.

hahaha! You are officially my favorite member.

**Overlord** · 05-09-2008

Originally Posted by master5001

Code:

               DeleteObject(extra); // this is actually probably what killed your code, btw.

You are MY favorite member now

Thanks, it works flawlessly now, hadn't thought about deleting GDI objects before posting.

Thanks again!

**~~master5001~~** · 05-10-2008

No problemo. I have run into issues like that before and they are so painfully difficult to debug.

Thread: A strange problem with regions

Thread Tools

Search Thread

Display

A strange problem with regions

Similar Threads

Strange problem with GETLINE

Strange problem

Strange problem with classes in header files

Strange problem