Re: [1/3] user32/tests: Comment-out PostQuitMessage (causes tests skip)

[Dmitry Timoshkov]

sir-lexa  wrote:
>  case WM_DESTROY:
> -
>    PostQuitMessage(0);
> -
>    /* PostQuitMessage(0);   It lets some tests to be skipped */
>    return TRUE;

This is not acceptable, there is much more behind the scene. You need
to figure out the real reason of skipping the tests.

[1/3] user32/tests: Comment-out PostQuitMessage (causes tests skip)

[Alexey Slepov]

I figured out now.
As described in wine sources, PostQuitMessage() sets a flag in the thread's message queue that signals it to generate
a WM_QUIT message when there are no other pending sent or posted messages in the queue.
The tests with skipped testcases are those which use WM_TIMER messages - test_timer_message() and my test.
In DIALOG_DoDialogBox() there is clearly defined that on receiving WM_QUIT message the window is destroyed. So timer messages just do not reach queue.

I thought about solution again and now I think that it would be better to remove at all lines about WM_DESTROY handling, such as we don't handle this message. And let it be handled by default. That code does nothing except posting WM_QUIT messages leading to test skipping. (test passed on WineBot)

What do you think?

28.05.2012, 10:32, "Dmitry Timoshkov" :
> sir-lexa  wrote:
> > case WM_DESTROY:
> >  -        PostQuitMessage(0);
> >  +        /* PostQuitMessage(0);   It lets some tests to be skipped */
> >           return TRUE;
>
> This is not acceptable, there is much more behind the scene. You need
> to figure out the real reason of skipping the tests.
>
> --
> Dmitry.

[1/3] user32/tests: Comment-out PostQuitMessage (causes tests skip)

[Alexey Slepov]

Guys, so how about this?

diff --git a/dlls/user32/tests/dialog.c b/dlls/user32/tests/dialog.c
index 6bd34b3..4301c8d 100644
--- a/dlls/user32/tests/dialog.c
+++ b/dlls/user32/tests/dialog.c
@@ -1027,9 +1027,6 @@ static INT_PTR CALLBACK DestroyOnCloseDlgWinProc (HWND hDlg, UINT uiMsg,
     case WM_CLOSE:
         DestroyWindow(hDlg);
         return TRUE;
-    case WM_DESTROY:
-        PostQuitMessage(0);
-        return TRUE;
     }
     return FALSE;
 }

29.05.2012, 01:16, "Alexey Slepov" :
> I figured out now.
>  As described in wine sources, PostQuitMessage() sets a flag in the thread's message queue that signals it to generate
>  a WM_QUIT message when there are no other pending sent or posted messages in the queue.
>  The tests with skipped testcases are those which use WM_TIMER messages - test_timer_message() and my test.
>  In DIALOG_DoDialogBox() there is clearly defined that on receiving WM_QUIT message the window is destroyed. So timer messages just do not reach queue.
>
> I thought about solution again and now I think that it would be better to remove at all lines about WM_DESTROY handling, such as we don't handle this message. And let it be handled by default. That code does nothing except posting WM_QUIT messages leading to test skipping. (test passed on WineBot)
>
> What do you think?
>
> 28.05.2012, 10:32, "Dmitry Timoshkov" :
> > sir-lexa  wrote:
> > > case WM_DESTROY:
> > >    -        PostQuitMessage(0);
> > >    +        /* PostQuitMessage(0);   It lets some tests to be skipped */
> > >             return TRUE;
> >
> > This is not acceptable, there is much more behind the scene. You need
> >   to figure out the real reason of skipping the tests.
> >
> > --
> >   Dmitry.