Re: [PATCH v2] user32: Also remove WM_CHAR from queue when a menu gets VK_ESCAPE and add tests

[Alexandre Julliard]

Fabian Maurer  writes:
> @@ -3186,8 +3186,12 @@ static BOOL MENU_TrackMenu( HMENU hmenu, UINT wFlags, INT x, INT y,
>    	    break;
> case VK_ESCAPE:
> -
>            {
> -
>                MSG msg_dummy;
>                fEndMenu = MENU_KeyEscape(&mt, wFlags);
> -
>     break;
> -
>                PeekMessageW(&msg_dummy, 0, WM_CHAR, WM_CHAR, PM_REMOVE);
> -
>                break;
> -
>            }

It would be cleaner to avoid TranslateMessage instead of removing the
message (possibly getting the wrong one).

[PATCH v2] user32: Also remove WM_CHAR from queue when a menu gets VK_ESCAPE and add tests

[Fabian Maurer]

> It would be cleaner to avoid TranslateMessage instead of removing the
> message (possibly getting the wrong one).

Sorry, I don't really see the problem. TranslateMessage should always generate 
a WM_CHAR, how could we get a wrong message? Shouldn't be possible, except for 
a bug in TranslateMessage, or I'm missing something.

I could check for message/wParam to skip TranslateMessage, but I found it 
cleaner to put it into the handler for VK_ESCAPE.

Regards,
Fabian Maurer

[PATCH v2] user32: Also remove WM_CHAR from queue when a menu gets VK_ESCAPE and add tests

[Alexandre Julliard]

Fabian Maurer  writes:
> > It would be cleaner to avoid TranslateMessage instead of removing the
> > message (possibly getting the wrong one).
>
> Sorry, I don't really see the problem. TranslateMessage should always generate 
> a WM_CHAR, how could we get a wrong message? Shouldn't be possible, except for 
> a bug in TranslateMessage, or I'm missing something.

TranslateMessage will put a WM_CHAR in the queue, but there may be
another one already in there, which would then get removed.

[PATCH v2] user32: Also remove WM_CHAR from queue when a menu gets VK_ESCAPE and add tests

[Fabian Maurer]

> TranslateMessage will put a WM_CHAR in the queue, but there may be
> another one already in there, which would then get removed.

Right, I missed that. Thanks, I sent a corrected version.

Regards,
Fabian Maurer