bug#65068: 29.1; xkb-interception interaction causes problems with key combinations

["Alexander Prähauser" <alexander.praehauser@gmx.at>]

It does seem to register the correct level shifts now, but it 
seems to evaluate each key immediately after having been
pressed. For instance, when I press Space-CapsLock-U, this should 
evaluate to C-; and it does, but this is the output
I'm receiving in-between:

<Control_R> is undefined
C-' is undefined
C-; is undefined

Po Lu [2023-08-24 Thu 17:56] wrote:

> Alexander Prähauser <alexander.praehauser@gmx.at> writes:
>
>> The significance of the U-key is that when I press 
>> Space-CapsLock-U
>> (or any other key, I think), it doesn't send the signal
>> that is configured by xkb but the corresponding key of the 
>> default
>> English keymap (u for the U-key), which really
>> shouldn't happen. Anyway, here the output after pressing the 
>> sequence
>> you wanted:
>>
>> keycode: 48, keysym: 65027, 0
>> keycode: 192, keysym: 269025093, 0
>> keycode: 37, keysym: 65507, 0
>> keycode: 37, keysym: 65507, 0 <- Control_R
>> keycode: 48, keysym: 65027, 0 <- AC11, ISO_Level3_Shift
>> keycode: 192, keysym: 269025093, <- XF86Launch5
>
> This attests to Emacs registering the keysyms you meant it to. 
> If
> ISO_Level3_Shift is registered as a modifier key, it should not 
> be
> translated into keyboard input afterwards, let alone the 
> original
> apostrophe symbol.
>
> Does this patch resolve your problems with typing level 3 
> characters
> coupled with Control?
>
> diff --git a/src/xterm.c b/src/xterm.c
> index 6a1642ff56e..7391041ea0c 100644
> --- a/src/xterm.c
> +++ b/src/xterm.c
> @@ -23734,6 +23734,7 @@ handle_one_xevent (struct x_display_info 
> *dpyinfo,
>  	      ptrdiff_t i;
>  	      unsigned int old_state;
>  	      struct xi_device_t *device, *source;
> +	      bool is_modifier_key;
>  
>  	      coding = Qlatin_1;
>  
> @@ -24175,17 +24176,7 @@ handle_one_xevent (struct 
> x_display_info *dpyinfo,
>  		       /* Any "vendor-specific" key is ok.  */
>  		       || (keysym & (1 << 28))
>  		       || (keysym != NoSymbol && nbytes == 0))
> -		      && ! (IsModifierKey (keysym)
> -			    /* The symbols from XK_ISO_Lock
> -			       to XK_ISO_Last_Group_Lock
> -			       don't have real modifiers but
> -			       should be treated similarly to
> -			       Mode_switch by Emacs. */
> -#if defined XK_ISO_Lock && defined XK_ISO_Last_Group_Lock
> -			    || (XK_ISO_Lock <= keysym
> -				&& keysym <= 
> XK_ISO_Last_Group_Lock)
> -#endif
> -			    ))
> +		      && !is_modifier_key)
>  		    {
>  		      STORE_KEYSYM_FOR_DEBUG (keysym);
>  		      /* make_lispy_event will convert this to a 
>  symbolic
> @@ -24204,7 +24195,11 @@ handle_one_xevent (struct 
> x_display_info *dpyinfo,
>  		      STORE_KEYSYM_FOR_DEBUG (copy_bufptr[i]);
>  		    }
>  
> -		  if (nbytes)
> +		  /* Mind that NBYTES can be set even if KEYSYM
> +		     represents a modifier key, but that no 
> character
> +		     events should be sent in that case.  */
> +
> +		  if (nbytes && !is_modifier_key)
>  		    {
>  		      inev.ie.kind = 
>  MULTIBYTE_CHAR_KEYSTROKE_EVENT;
>  		      inev.ie.arg = make_unibyte_string 
>  (copy_bufptr, nbytes);