We were having a similar problem. We ended up having to invalid the region of the parent window to get it to update when we got WM_VSCROLL. I tried to do as user demorge says here:
SetBkMode(hdc, TRANSPARENT) doesn't work
But our code doesn't use handles, we actually use the class CWnd, so we ended up doing this in the WindowProc instead:
switch(message)
{
...
case WM_VSCROLL:
case WM_HSCROLL:
LRESULT answer;
PAINTSTRUCT ps;
CDC* pdc;
CWnd* MyParentHWnd;
// We want the scroll to work the same way it has always worked for our
// ancestor class. Let them handle the scrolling and save off their
// return.
answer = AncestorClass::WindowProc(message, wParam, lParam);
pdc = BeginPaint(&ps);
// DO NOT change the assignement operator in the conditional below to an
// equality operator. We are actually trying to get the parent window and
// and storing locally, and then verifying that we didn't get back null.
// This is a purposeful design decision.
if (MyParentHWnd = GetParent()){
RECT MyRect;
GetClientRect(&MyRect);
ClientToScreen(&MyRect);
MyParentHWnd->ScreenToClient(&MyRect);
MyParentHWnd->InvalidateRect(&MyRect);
}
EndPaint(&ps);
return answer;
break;
...
}
Of course, I had to genericize it a little bit. I just wanted you to know that yes, there are other people that are seeing your problem, and we found how to fix it.
