Re: [PATCH 2/2] user32: Set SWP_NOSIZE for windows with SWP_NOREDRAW.

[Alexandre Julliard]

"Lei Zhang"  writes:
> @@ -1808,6 +1808,10 @@ static BOOL fixup_flags( WINDOWPOS *winpos )
>      if ((wndPtr->rectWindow.right - wndPtr->rectWindow.left == winpos->cx) &&
>          (wndPtr->rectWindow.bottom - wndPtr->rectWindow.top == winpos->cy))
>          winpos->flags |= SWP_NOSIZE;    /* Already the right size */
> - else if (((winpos->flags & (SWP_NOZORDER|SWP_NOREDRAW|SWP_FRAMECHANGED|SWP_NOACTIVATE)) == SWP_NOREDRAW) &&
> -
>    (wndPtr->rectWindow.right - wndPtr->rectWindow.left == origSize.x) &&
> -
>    (wndPtr->rectWindow.bottom - wndPtr->rectWindow.top == origSize.y))
> -
>    winpos->flags |= SWP_NOSIZE;    /* Already the right size */

It makes no sense to compare the window rect against the original size
since the window rect hasn't been changed yet, it will always
match. Also I fail to see why this would have anything to do with
SWP_NOREDRAW. I think you are on the wrong track with this bug.

[PATCH 2/2] user32: Set SWP_NOSIZE for windows with SWP_NOREDRAW.

[Lei Zhang]

On Tue, Jul 8, 2008 at 2:10 AM, Alexandre Julliard  wrote:
> "Lei Zhang"  writes:
> > @@ -1808,6 +1808,10 @@ static BOOL fixup_flags( WINDOWPOS *winpos )
> >      if ((wndPtr->rectWindow.right - wndPtr->rectWindow.left == winpos->cx) &&
> >          (wndPtr->rectWindow.bottom - wndPtr->rectWindow.top == winpos->cy))
> >          winpos->flags |= SWP_NOSIZE;    /* Already the right size */
> > - else if (((winpos->flags & (SWP_NOZORDER|SWP_NOREDRAW|SWP_FRAMECHANGED|SWP_NOACTIVATE)) == SWP_NOREDRAW) &&
> > -
> >    (wndPtr->rectWindow.right - wndPtr->rectWindow.left == origSize.x) &&
> > -
> >    (wndPtr->rectWindow.bottom - wndPtr->rectWindow.top == origSize.y))
> > -
> >    winpos->flags |= SWP_NOSIZE;    /* Already the right size */
>
> It makes no sense to compare the window rect against the original size
> since the window rect hasn't been changed yet, it will always
> match. Also I fail to see why this would have anything to do with
> SWP_NOREDRAW. I think you are on the wrong track with this bug.

You're right, the problem required more investigation and I'll have a
better patch soon. Though while running the user32 tests, I did see
situations where those comparisons do not return true.