Re: [m17n-list] Modifier Keys for fcitx-m17n?

[Richard Wordingham]

On Thu, 16 Nov 2017 22:37:57 +0900
handa <address@hidden> wrote:

> Actually "FcitxKeyState_ScrollLock" is very mysterious, but it seems
> that fcitx-m17n somehow tries to generate an input key symbol modified
> by AltGr for m17n-lib.  At the moment I can't try it because I don't
> know how to generate "AltGr"ed key on my Japanese keyboard.

It seems that fctix-m17n is innocent.  The culprits seems to be higher
up in fcitx.  For X, function XIMProcessKey() in ximhandler.c contains
the devastating sequence

    kev = (XKeyEvent *) & call_data->event;
    memset(strbuf, 0, STRBUFLEN);
    keyCount = XLookupString(kev, strbuf, STRBUFLEN, &originsym, NULL);

    const uint32_t originstate = kev->state;
    state = kev->state - (kev->state & FcitxKeyState_NumLock) -
    (kev->state & FcitxKeyState_CapsLock) - (kev->state &
    FcitxKeyState_ScrollLock); 

With a fairly common modifier mapping in X of:

shift       Shift_L (0x32),  Shift_R (0x3e)
lock        Caps_Lock (0x42)
control     Control_L (0x25),  Control_R (0x69)
mod1        Alt_L (0x40),  Meta_L (0xcd)
mod2        Num_Lock (0x4d)
mod3      
mod4        Super_L (0x85),  Super_R (0x86),  Super_L (0xce),  Hyper_L
(0xcf)
mod5        ISO_Level3_Shift (0x5c),  Mode_switch (0xcb)

I am left with shift, control, alt and super.  The application takes
alt, the desktop takes super (i.e. left and right Windows keys).  Ctrl
is limited, partly by fcitx-m17n, and partly by something higher up.

If the characters are handled via the D-bus, then the assignment 

    state = originstate & FcitxKeyState_SimpleMask;

in function IPCProcessKey() in file ipc.c of fcitx does the damage.
The mask is defined as

         FcitxKeyState_SimpleMask =
FcitxKeyState_Ctrl_Alt_Shift | FcitxKeyState_ Super |
FcitxKeyState_Super2 | FcitxKeyState_Hyper | FcitxKeyState_Meta,

My keyboards don't have physical keys for Super2, Hyper or Meta.

I haven't yet confirmed that state bit 7 (mysteriously called 'scroll
lock') gets through to XIMProcessKey() and IPAProcessKey(), but they
seem to be where the problem lies.

Richard.