Re: [1/2] user32: Converted rcScrollBar to screen coordinates.

[Dmitry Timoshkov]

"Dylan Smith"  wrote:
> @@ -1252,6 +1253,9 @@ static BOOL SCROLL_GetScrollBarInfo(HWND hwnd, LONG idObject, LPSCROLLBARINFO in
>  SCROLL_GetScrollBarRect(hwnd, nBar, &info->rcScrollBar, &nDummy,
>                          &info->dxyLineButton, &info->xyThumbTop);
> - /* rcScrollBar needs to be in screen coordinates */
> - GetWindowRect(hwnd, &rect);
> - OffsetRect(&info->rcScrollBar, rect.left, rect.top);

You need to fix SCROLL_GetScrollBarRect for the SB_CTL case instead.
> +static void scrollbar_test4(void)
> +{
> - BOOL ret;
> - SCROLLBARINFO sbi;
> - RECT rect;
> -
> - /* Test GetScrollBarInfo to make sure it returns rcScrollBar in screen
> -
> * coordinates. */
> - sbi.cbSize = sizeof(sbi);
> - ret = GetScrollBarInfo( hScroll, OBJID_CLIENT, &sbi);
> - ok( ret, "The GetScrollBarInfo() call should not fail.\n" );
> - GetWindowRect( hScroll, &rect );
> - ok( ret, "The GetWindowRect() call should not fail.\n" );
> - ok( !(sbi.rgstate[0] & (STATE_SYSTEM_INVISIBLE|STATE_SYSTEM_OFFSCREEN)),
> -
>    "unexpected rgstate(0x%x)\n", sbi.rgstate[0]);
> - ok( !memcmp(&rect, &sbi.rcScrollBar, sizeof(RECT)),
> -
>    "WindowRect(%d, %d, %d, %d) != rcScrollBar(%d, %d, %d, %d)\n",

EqualRect is better than memcmp.

[1/2] user32: Converted rcScrollBar to screen coordinates.

[Dylan Smith]

On Mon, Oct 6, 2008 at 2:17 AM, Dmitry Timoshkov wrote:
> You need to fix SCROLL_GetScrollBarRect for the SB_CTL case instead.

SCROLL_GetScrollBarRect is used is used in multiple places where client
coordinates are expected, and it isn't just the SB_CTL case that needs to be
changed, in fact it was the SB_VERT case that I noticed first.

EqualRect is better than memcmp.
>

Yup.  I didn't know the function existed until now.

[1/2] user32: Converted rcScrollBar to screen coordinates.

[Dmitry Timoshkov]

"Dylan Smith"  wrote:
> > You need to fix SCROLL_GetScrollBarRect for the SB_CTL case instead.
>
> SCROLL_GetScrollBarRect is used is used in multiple places where client
> coordinates are expected, and it isn't just the SB_CTL case that needs to be
> changed, in fact it was the SB_VERT case that I noticed first.

Then you need to test the remaining cases as well.