- <http://geekhack.org/index.php?topic=14290.msg1075201#msg1075201>

- <http://geekhack.org/index.php?topic=14290.msg1070604#msg1070604>

- Add delay between scans to prevent overlaod of ADB keyboard controllers
- <http://geekhack.org/index.php?topic=14290.msg1068919#msg1068919>

- Add action_util.c and remove action_oneshot.c
- Add oneshot_mods for MODS_ONESHOT
- Add weak_mods for MODS_KEY and MACRO
- weak_mods is cleared when layer switching

- See https://github.com/tmk/tmk_keyboard/issues/60
- **Except for modifiers** a key pressed before the tapping starts should be released immediately
- 'Mod-Tap key'(like shift-;) didn't work from this fix: 4d0b3aa Fix Tapping: release of a key pressed before tap
This key sequence should register ':', not ';'. With the fix Shift is
released before settlement of tapping, this registers ';'.
    Shift   ~~___~~~~~~~
    ;(Tap)  ~~~____~~~~

- Blocking occurs if built without option INTERRUPT_CONTROL_ENDPOINT

- Add print messages of USB intialize procesure.
- call out sei() in earlier stage. LUFA API doc of USB_Init() says:
interrupts must be enabled within 500ms ... to ensure that the host does
not time out whilst enumerating the device
- remove matrix_init() from suspend_wakeup_init(). It is not needed in
  fact and if matrix_init() takes long time to initialize it prevents
host emnumerating the device.

- Call of function pointer is compiled into 'icall' instruction.
It should use word address but it has used byte address :( It seems
jump has worked luckily by chance until now. why it worked?

- immediately process release event of a key pressed before tapping

- Makefile use LUFA now
- Change MCU ports PF{0,1} to PD{0,1}
- Add build option for keymap to Makefile: make keymap=hasu