[PATCH v2 1/2] winex11.drv: Activate window when restoring from iconic state.

[Zhiyi Zhang]

On 5/24/19 11:47 PM, Dmitry Timoshkov wrote:
> Zhiyi Zhang <zzhang at codeweavers.com> wrote:
>
>>>> #2 The reason why Mirror's Edge doesn't handle WM_SYSCOMMAND SC_RESTORE
>>>>
>>>> Disassembly of the game win proc shows that the win proc return 0 directly
>>>> when handling WM_SYSCOMMAND SC_RESTORE in fullscreen mode, instead of
>>>> handing it to DefWindowProc or do any sort of processing. So this means
>>>> it rely on the WM_ACTIVATE message before WM_SYSCOMMAND.
>>> It should be possible to add a message test that replicates this behaviour.
>> I couldn't find a way to do it automatically. It seem an interactive test is required if we really need a test.
> In Windows UI practically everything could be reproduced with a set of
> API calls, I don't see a need for an interactive test either since you
> can't then reliably record and test a message sequence.
>
>>>> @@ -1266,6 +1266,8 @@ static void handle_wm_state_notify( HWND hwnd, XPropertyEvent *event, BOOL updat
>>>>              {
>>>>                  TRACE( "restoring win %p/%lx\n", data->hwnd, data->whole_window );
>>>>                  release_win_data( data );
>>>> +                if ((style & (WS_MINIMIZE | WS_VISIBLE)) == (WS_MINIMIZE | WS_VISIBLE))
>>>> +                    SetActiveWindow( hwnd );
>>>>                  SendMessageW( hwnd, WM_SYSCOMMAND, SC_RESTORE, 0 );
>>>>                  return;
>>> According to your spy++ logs activating a minimized window before sending
>>> WM_SYSCOMMAND/SC_RESTORE is the result of a SetWindowPos call. A message
>>> test should help to figure it out its exact flags, in any case it's wrong
>>> to call SetActiveWindow directly.
>>>
>> I don't think that specific WM_ACTIVATE message is the result of SetWindowPos because
>> WM_ACTIVATE message is sent after WM_WINDOWPOSCHANGED.
> That could be a defect in your testing, a message case would show that
> for sure.
>
I tested it multiple times with multiple apps. So it's unlikely to be a defect.
Anyway, I will try to write a test for this. If you could give some suggestions
on how to do that, I would appreciated it.

Thanks,
Zhiyi