[immodule-qt] Completing comments and documents

[LiuCougar]

Hi,

I just noticed when I was implementing scim-qtimm that Qt does not
send the keyrelease event for Alt/Shift/Ctrl, if you press any two of
them and release one, then the other, you won't get the keyrelease
event for the second key.

Do you think this can be fixed?

Regards,
Cougar

On Tue, 13 Jul 2004 01:14:32 +0900, YamaKen <yamaken at bp.iij4u.or.jp> wrote:
> Hi all,
> 
> I've written following important comments in the source. Please
> review it. The comments are design decisions about key
> filtering.
> 
> Index: src/kernel/qapplication_x11.cpp
> ===================================================================
> --- src/kernel/qapplication_x11.cpp     (revision 16)
> +++ src/kernel/qapplication_x11.cpp     (revision 17)
> @@ -3152,6 +3152,14 @@
>     }
> 
> #ifndef QT_NO_IM
> +    // Filtering input events by the input context. It has to be taken
> +    // place before any other key event consumers such as eventfilters
> +    // and accelerators because some input methods require quite
> +    // various key combination and sequences. It often conflicts with
> +    // accelerators and so on, so we must give the input context the
> +    // filtering opportunity first to ensure all input methods work
> +    // properly regardless of application design.
> +
>     if( keywidget && keywidget->isEnabled() && keywidget->isInputMethodEnabled() ) {
>        if( ( event->type==XKeyPress || event->type==XKeyRelease ) &&
>            sm_blockUserInput ) // block user interaction during session management
> @@ -3180,6 +3188,8 @@
>                                                  state, ascii, code, type,
>                                                  FALSE, FALSE );
> 
> +           // both key press/release is required for some complex
> +           // input methods. don't eliminate anything.
>            QKeyEvent keyevent( type, code, ascii, state, text, FALSE, count );
> 
>            if( qic && qic->filterEvent( &keyevent ) )
> @@ -5293,9 +5303,25 @@
>     // compress keys
>     if ( !text.isEmpty() && testWState(WState_CompressKeys) &&
> #ifndef QT_NO_IM
> -        // input methods need discrete key events
> -         // UPDATED COMMENT REQUIRED -- 2004-07-08 YamaKen
> -        // describe design decision
> +        // Ordinary input methods require discrete key events to work
> +        // properly, so key compression has to be disabled when input
> +        // context exists.
> +        //
> +        // And further consideration, some complex input method
> +        // require all key press/release events discretely even if
> +        // the input method awares of key compression and compressed
> +        // keys are ordinary alphabets. For example, the uim project
> +        // is planning to implement "combinational shift" feature for
> +        // a Japanese input method, uim-skk. It will work as follows.
> +        //
> +        // 1. press "r"
> +        // 2. press "u"
> +        // 3. release both "r" and "u" in arbitrary order
> +        // 4. above key sequence generates "Ru"
> +        //
> +        // Of course further consideration about other participants
> +        // such as key repeat mechanism is required to implement such
> +        // feature.
>         ! qic &&
> #endif // QT_NO_IM
>         // do not compress keys if the key event we just got above matches
> 
> -------------------------------
> YamaKen  yamaken at bp.iij4u.or.jp
> 
> _______________________________________________
> Immodule-qt mailing list
> Immodule-qt at freedesktop.org
> http://freedesktop.org/mailman/listinfo/immodule-qt
>