[Bug 56582] vb3 combobox regression: single click scrolls twice

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56582

--- Comment #11 from Fabian Maurer  ---
Problem occurs in this part of the code:
> do
>     {
>         if (!NtUserGetMessage( &msg, 0, 0, 0 )) break;
>         if (NtUserCallMsgFilter( &msg, MSGF_SCROLLBAR )) continue;
>         if (msg.message == WM_LBUTTONUP ||
>             msg.message == WM_MOUSEMOVE ||
>             msg.message == WM_MOUSELEAVE ||
>             msg.message == WM_NCMOUSEMOVE ||
>             msg.message == WM_NCMOUSELEAVE ||
>             (msg.message == WM_SYSTIMER && msg.wParam == SCROLL_TIMER))
>         {
>             pt.x = (short)LOWORD( msg.lParam ) - rect.left;
>             pt.y = (short)HIWORD( msg.lParam ) - rect.top;
>             handle_scroll_event( hwnd, scrollbar, msg.message, pt );
>         }

There is two delays:
> /* Delay (in ms) before first repetition when holding the button down */
> #define SCROLL_FIRST_DELAY   200
>
> /* Delay (in ms) between scroll repetitions */
> #define SCROLL_REPEAT_DELAY  50

The timer is set like follows:
> NtUserSetSystemTimer( hwnd, SCROLL_TIMER,
>                                   msg == WM_LBUTTONDOWN ? SCROLL_FIRST_DELAY : SCROLL_REPEAT_DELAY );

So, when we get a WM_MOUSEMOVE, the timer will be set to the shorter repeat
relay.
Now, due to the regression, we get that WM_MOUSEMOVE instantly. Ir comes from:
> BOOL release_capture(void)
> {
>     HWND previous = NULL;
>     BOOL ret;
> ret = set_capture_window( 0, 0, &previous );
>
> /* Somebody may have missed some mouse movements */
> if (ret && previous)
> {
>     INPUT input = { .type = INPUT_MOUSE };
>     input.mi.dwFlags = MOUSEEVENTF_MOVE;
>     NtUserSendInput( 1, &input, sizeof(input) );
> }

This is injected intentionally. Now it breaks the scrolling logic though.