gint/TODO

Bugs to fix:
-  A few key hits ignored after leaving the application  (could not reproduce)
-  Lost keyboard control at startup                      (could not reproduce)
-  Alignment of ALL .data / .rodata files is required to ensure converted data
   is properly aligned

Things to do before 1.0:
-  bopti:	Test partial transparency
*  core:	Allow return to menu
-  demo:	Try 284x124 at (-60, -28) (all disadvantages)
-  project:	Check size of *all* library structures
-  project:	Clean headers that have some internal definitions
-  project:	Unify this hellish mess of register access!
-  time:	Compute CLOCKS_PER_SEC

Things to do later:
-  bopti:	Implement blending modes for monochrome bitmaps
-  clock:	Handle overclock (relaunch clocks when overclocking)
-  core:	Change interrupt priority using the gint API
-  core:	Register more interrupts (and understand their parameters)
-  core:	Remove redundant code linked to environment saves
-  core:	Review interrupt system (again) - this one is too slow
-  errno:	Introduce errno and use it more or less everywhere
-  esper:	Cleaner playback, synthesizing
-  events:	Allow customization of keyboard event system (option to return
   |		events with modifiers, etc)
-  events:	Generate keyboard events on-the-fly by reading state arrays,
   |		allowing both a faster interrupt and avoiding supressing other
   |		events inside getkey() and multigetkey()
-  serial:	Implement a driver
-  stdio:	More serious formatted printing functions
-  string:	Use cmp/str to implement memchr() (assembler examples)
-  string:	Do some tests for memcmp() and memcpy()
-  usb:		Implement a driver

Things to investigate:
-  Packed bit fields alignment
-  Registers that may need to be saved within setjmp()
-  Optimizing core/gint.c leads to raising of an illegal slot exception when
   running the interrupt handler, although it ends on rte; lds.l @r15+, mach,
   which is totally not an illegal slot.