since the keycode for a tap dance process gets process only after the
TAPPING_TERM timeout, you really only have ONESHOT_TIMEOUT -
TAPPING_TERM time to tap or double tap on the key. This fix save the
oneshot_mods into the action.state structure and applies the mods with
the keycode when it's registered. It also unregisters the mod when the
the tap dance process gets reset.
Scenario:
Locking the KC_LSHIFT, and then using a tap dance key that registers a
S(KC_9) will unregister the KC_LSHIFT.
The tap dance or any keycode that is registered should not have the
side effect of cancelling a locked moditifier. We should be using a
similar logic as the TMK codes in tmk_core/comman/action.c:158.
In register_code16 and unregister_code16 we call register_code and
unregister_code twice, once for the mods and once for the keycode.
The (un)register_code have many check to see that keycode we have sent
however because we know that we are sending it a mods key, why not
just skip all of it and call (un)register_mods instead. This will skip
alot of checks and should speedup the loop a little.
Since we can't read the real_mods and oneshot_mods static variable
directly within the update_user_visualizer_state
function (Threading and serial link). We are know storing the mods
states in the visualizer_keyboard_status_t structure. We can now
display the status of the modifier keys on the LCD display.
Fix memory leaks by using stack instead of malloc
Reduce memory usage by having less temporary bufffers
Remove warnings by adding includes
Decrease code size by 608 bytes (mostly due to not linking malloc)
More robust handling of buffer overflows
Unlike the arduino functions, these don't take abstract pin numbers,
they take pin labels like `B0`. Also, rather than taking very
generic parameter names, these take slightly more descriptive
enum values.
These improve the clarity of code that would otherwise be inscrutable
bit manipulation in tersely named port register names.
