Bug 67167

Output:

keysyms [ egrave           ] unicode [ è ] group [ French (Switzerland) (0) ] mods [ Lock ] leds [ Caps Lock ] 
keysyms [ eacute           ] unicode [ é ] group [ French (Switzerland) (0) ] mods [ Lock ] leds [ Caps Lock ] 
keysyms [ A                ] unicode [ A ] group [ French (Switzerland) (0) ] mods [ -Lock ] leds [ Caps Lock ] 
keysyms [ S                ] unicode [ S ] group [ French (Switzerland) (0) ] mods [ -Lock ] leds [ Caps Lock ]

Hmm, this is this thing here:
http://www.x.org/releases/current/doc/kbproto/xkbproto.html#default_symbol_transformations

And the Xlib implementation:
http://cgit.freedesktop.org/xorg/lib/libX11/tree/src/xkb/XKBBind.c?id=6d926088d80a08e13e6d6c4ff207b81ad52e667f#n572

In short: if Caps Lock is active and was not consumed, the keysym needs to be capitalized... how unfortunate.

I was under the impression that this was not an issue, since all of the capitalization rules were already present in the xkeyboard-config keymaps themselves (i.e. written in a level the layout). But I guess not.

Note that the spec allows two ways of doing this transformation: locate sensitive and locale insensitive. We already have the tables for the latter (in keysym.c) as the are given in the spec; but Xlib does the former as far as I can tell.

Also note that Xlib separates the two things (at least in some functions), i.e. you can get the untransformed keysym with e.g.
    XkbLookupKeysym, XkbTranslateKeyCode
and then you can transform them separately (you need to pass the mods though), e.g. with
    XkbTranslateKeySym.

I'm not entirely sure what best to do here..

- Use the built-in tables or do some locale sensitive lookups with glibc or whatever?

- Always do the transformation, or with a keymap flag, or another function? (note that xkb_keymap_key_get_syms returns a const array so it'd be hard to change that).

- Or we can punt this entirely to the application. Expose xkb_keysym_to_upper/lower. And then the logic can be implemented outside with

   if (mod_index_is_active(CAPS) && !mod_index_is_consumed(CAPS))
        keysym = xkb_keysym_to_upper(keysym)
   etc.

- Or we can do nothing and ask that all of these rules be put directly in the xkeyboard-config symbols files.

Daniel, what do you think?

Since we can't touch the return value of xkb_state_key_get_syms() (because it returns a pointer to const data), and this functionality doesn't make much sense for multiple-keysyms anyway, one option would be:

- xkb_state_key_get_one_sym() does the transformation.

- xkb_state_key_get_syms() doesn't do the transformation, i.e. it returns "raw" keysyms (hopefully, hmm, when mutliple-keysyms become useful, the keymap authors would have the good sense to not rely on the CapsLock "feature").

- xkb_state_mod_index_is_consumed() - to me it makes sense to report Caps consumed when the keysym is changed, but Xlib doesn't, and there's also the Control thing. So probably should stay unconsumed.

Of course we'd need to document this, since then _get_one_sym() becomes more than just a convenience.

Created attachment 84022 [details] [review]
0001-keysym-add-xkb_keysym_to_-lower-upper.patch

Here are patches along the comment above. I'll apply them unless there's a problem.

Gatis, QT uses xkb_state_key_get_one_sym(), if I remember correctly, so that should be fine. Other people who use xkb_state_key_get_syms() and want this functionality would have to adapt I'm afraid.

Created attachment 84023 [details] [review]
0002-test-interactive-use-xkb_state_key_get_one_sym-when-.patch

Created attachment 84024 [details] [review]
0003-state-apply-capitalization-transformation-on-keysyms.patch

These are now in master.