comctl32/toolbar: Potential fix for button redraw in TBSTYLE_FLAT toolbars

[Andrew de Quincey]

On Thursday 11 December 2003 23:00, Andrew de Quincey wrote:
> On Thursday 11 December 2003 22:45, Dimitrie O. Paun wrote:
> > On December 11, 2003 04:32 pm, Andrew de Quincey wrote:
> > > I have concluded that the WM_ERASEBKGND is sent from within WM_PAINT
> > > when the toolbar detects that the mouse has moved off a button and
> > > wants to de-highlight it.
> >
> > Well, it probably does a InvalidateRect() call. You can try an ugly
> >     InvaldateRect(hwnd, 0, TRUE);
> > but this will probably result in flicker. For added bonus, figure out
> > the button's rect, and pass it in:
> >
> >     /* figure out stuff in rect */
> >     InvalidateRect(hwnd, &rect, TRUE);
>
> Actually, that code is almost already in there... it does call
> InvalidateRect() already, but passes FALSE because TOOLBAR_HasText()
> returns false. This happens in the TOOLBAR_MouseMove() function.
>
> Originally, I had decided that this wasn't what windows does because in
> spyxx, the WM_ERASEBKGND occurs AFTER the WM_PAINT.
>
> However, I now see I have misread MSDN: The call to BeginPaint() inside the
> WM_PAINT issues the WM_ERASEBKGND because bErase was TRUE, so it would
> obviously occur after the WM_PAINT. This sounds much nicer.

On second thoughts, that method would need multiple changes in the file to 
ensure the correct value was passed to InvalidateRect(). Whereas the method I 
proposed just needs three lines.

Which do people prefer?