repainting again...

I want to handle WM_ERASEBKGND message by myself, because when I let Windows handle it, the whole application window is erased, including all child windows. If I send WM_ERASEBKGND message manually, I must get the context device and send it with the message in WPARAM. I retreive such a context device using GetDCEx() with DCX_CLIPCHILDREN flag and the clipping region is set properly (does not contaion child windows). Then in WM_ERASEBKGND part of message handling procedure, I get this context device, get the clipping rectangle using GetClipBox() and fill it with color simulating "the erasing of a window". It may be like this:

Code:

  case WM_COMMAND:
    {
      // ...
      // in response to e.g. toolbar button click
      HDC hDC = GetDCEx(hWnd, NULL, DCX_CLIPCHILDREN | DCX_CACHE);
      SendMessage(WM_ERASEBKGND, (WPARAM)hDC, 0);
      // some drawing using hDC...
      ReleaseDC(hWnd, hDC);
    }
    break;

    // ...

  case WM_ERASEBKGND:
    {
      RECT rcWin;
      GetClipBox((HDC)wParam, &rcWin);  // GetUpdRect() returns a free rectangle here!
      FillRect(hDC, &rcWin, hBrush);  // hBrush can be obtained by calling GetWindowLong()
    }
    break;

This works fine. But only when I send WM_ERASEBKGND by myself. If it's sent by Windows, for example as a result of calling BeginPaint() in WM_PAINT message:

Code:

  case WM_PAINT:
    PAINTSTRUCT ps;
    HDC hDC = BeginPaint(hWnd, &ps);
    // now Windows send a WM_ERASEBKGND message and return here
    // some drawing...
   EndPaint(hWnd, &ps);  
   // I don't remember the exact syntax of BeginPaint() and EndPaint(), it doesn't matter now

the clipping and even update rectangle is free. So I need to handle it like this:

Code:

  case WM_ERASEBKGND:
    {
      RECT rcWin;
      GetClipBox((HDC)wParam, &rcWin);
      if (IsEmptyRect(rcWin)) GetClientRect(hWnd, &rcWin);
      FillRect(hDC, &rcWin, hBrush);
    }
    break;

But in this case, I there is a whole window in rcWin so I'm back where I began with problem. I'm redrawing whole window. This flickers on slow computers.
My question is, how to get the proper update (or clipping) rectangle when WM_ERASEBKGND is sent automatically by Windows?

I'm not sure what you mean by "the clipping and even update rectangle is free". But you have to return a non-zero value from a WM_ERASEBKGND message to indicate that all erasing has been done. Have you read -

http://msdn.microsoft.com/library/de...dn_flicker.asp

Take a close look at the PAINTSTRUCT that you pass to BeginPaint
in your WM_PAINT handler - specifically the rcPaint parameter. This (and the rest) is filled by the call to BeginPaint.

Use InvalidateRect(hwnd, &rc, 1); if you want to re-paint with erase of the rectangle given by rc. Although sending messages explicitly can be made to work it is a lot simple using InvalidateRect.

Look too at InvalidateRgn, UpdateWindow and RedrawWindow for more discussion and solutions.

I've had some problem with logging in, so I posted this thread as an Unregistered. There is a correction of the last article.

But in this case, there is a whole window in rcWin so I'm back where I began with problem. I'm redrawing whole window. This flickers on slow computers.
My question is, how to get the proper update (or clipping) rectangle when WM_ERASEBKGND is sent automatically by Windows?

Now for the answers.

I'm not sure what you mean by "the clipping and even update rectangle is free". But you have to return a non-zero value from a WM_ERASEBKGND message to indicate that all erasing has been done.

I meant that both GetUpdateRect() and GetClipBox() return empty rectangles. As for the return value, I just forgot to write it, but I know it should be there. I usually return TRUE from WM_ERASEBKGND.

I don't want to invalidate the window, because it would cause another WM_PAINT message to be sent and the rcPaint rectangle is empty or contains whole client window. I want to redraw the part of main window that doesn't contain any child windows. As I do when WM_ERASEBKGND is sent from WM_COMMAND part.

Sorensen, thanx for the link...