-11.4.99
	tested libc. not yet threadsafe. 

	made drivers
	drivers/tpu
	drivers/i2c, just code of i2ctest with changed semaphores.
	
	splitted global_def.h into global_def.h (asm-compat)
	and cubeos.h: 
	includes global_def.h
	contains read&wirte macros
	contains semaphores & mutexes

	made semaphores posix-compatible 
	sem_signal as equivalent to sem_post for thomas

	
-29.4.

	new: drivers/sai.h/c
	     drivers/sai_cube.h/c
	     and implementation for TPU PAC and PWM. Works.

	new: enable() and disable() for interrupts.
	     used in io_duart.c and i2c.c
	new: ptable[x].irq to save irq level for threads.
	     init to 0x2...

	new: basic/list.h
	new: basic/iobuf.h


-4.5.
	new: basic/rdio.h
             inlined functions for rx control and iobuf management
	changed: basic/cubeos.h inlined enable() and disable() functions
	         basic/cubeos.h now #includes sys_var.h and schedule.h
        new: support for radio link interface in basic/io_duart.c
	changed: basic/io_duart.c init now uses iobuf's functions
	fixed: bug in io_duart.c that hung cubeos by forgetting to enable()
	interrupts after detecting a kernel output buffer overflow

-6.5
	new: sai support for i2c 
	i2cc: complex devices which require a control word to be sent first
	i2ca: simple i2c devices which just respond to read and write
              on the i2c bus
	new initialisations for addr[0] to addr[2] for the fuel gauge on
	i2c b. Should not stay there. Maybe a dynamic allocation scheme
	based on lists?
	
-7.5.
	cleaned up sai. now, each driver has it's on sai file
	and sai_cube only uses them.

	wrote sai's documentation

-9.6.
	found bug in scheduler, libcproblem solved.
	
	drivers/icontrol for the qdec motorcoltroller with potential
	fields (not yet finished)

	generic handler for IVtable basic/ivtab.c
	handler for TPU irq handlers 
	driver for qdec in tpud.c/.h

	chanred SAI to use pointers instead of entries struct sai.
	Dynamic creation of sai_cube's content.

	made test with minimal system without libc

----- release 0.4.0 ---------

fixed sincos.c
	- new busy indicator while building tables

changed icontrol.c 
	- icontrol_set now normalized values -10000 to 10000
	- icontrol_set reversed left channel 

changed icontrol.h 
	- New KICK_ for third motor
	- Changed LEFT and RIGHT to reflect usage in old RoboCup Controller

changed Makefile
	- make clean cleans drivers/icontrol/

changed hardware/tpu.h
	- changed setCISR and getCISR
	
changed make.in
	- appended .tw as new binary download file

changed tty interface, basic/ttyio.h,.c basic/io_sci.h/.c

New i2c interface below sai for analog and digital i/o
	Adresses of digital and analog chips are scanned. Chips are inserted
	in ascending order of adresses and bus numbers into array ad_data
	and bin_data. see drivers/i2c/digital.[c,h] drivers/i2c/analog.[c,h]

--- 0.4.9

removed minimal test stuff

added netcode in /net/rsm.h 
	the new netcode has a simple statemachine for reception (/net/rsm.c)
	and a userlevel transmit that uses the DUART channel A. The 
	radiomodule is controlled via the rdio.h functions from /basic/rdio.h.
	At the moment, only broadcast is possible

added support for cube3 (#define for hardware.h)

added support for the cube3 LED into cubeos.h. Does not hurt on cube1.

added softreset() by using external watchdog to reset

changed e_handle to also execute the softreset code

added ctrl-c reset trigger on duart channel B 

--- 0.4.10

changed basic/ into kernel/ and cube/ into libc/

added support code for bdm gcc into ivtable.S. if GOBACKGROUND is defined,
cubeos invokes the BDM on every exception vector. This needs a nice
integration into the makefile. maybe a config.h?

changed the struct _reent initialisation, moved it into its own file, 
libc/cubereent.c

changed make.in to use -m68332 instead of -mcpu32 -msoft-float 

added support for C++ into make.in, Makefile, cubeos.ld_sunc++

cubeos now compiles both on gcc2.8.1 newlib 1.8.1 and gcc2.95.2, newlib1.8.2

changed .ld files. for gcc2.95.2, libraries are now cpu32/lib[c,m,gcc,stdc++].a

changed crt0.s to call IVTab_init();

--- 0.4.11 (internal)

added XDR support in net/xdr. It generates its own libcubexdr.a, can be distributed
separately and should not have any problems with copyright.

drivers/hardware does hardware-specific init for RCJ and RG support boards

changed many API names. Separated code into components. All functions
carry component name as prefix, _ is used for internal things
Low-Level-Drivers:
DUART QSM FBIN TPU
Bus & Com Drivers:
I2C TTY LIST RSM NET
Hardware Init Drivers:
RCJ RG
rest:
KERN KERN_IVTAB

started BSD network code embedding. Anybody want to help?

Driver for CubeCAM as test module. Works fine. Still needs JPEG encoder.

Separated semaphores (thread aware) from simple semaphores. I2C still uses
simple semaphores, it's not yet clear how to wake up a thread from
an ISR. Question: can we force all ISRs to call the scheduler on request?
Otherwise, each transfer takes at least one quantum. Maybe that's too long.
p.s. I need a decent test for the semaphores. 

Documentation system
	started sewing-based comment to tex doc system. Not happy with it.


---- 0.4.12 (internal)

new tty driver, hw-independent, hooks for break and input char, hooks
for RTS/CTS depending on buffer level, high watermark in #define

JPEG works, uuencoding of data works, both in camtest, maybe move into cubeos.

New documentation system based on doxygen, sewing abandoned. 

New dynamic configuration system in /config , can configure dynamically by 
linking config.c or statically via hardware class define (CUBE1 CUBE2 RG etc.)

rewritten scheduler. Now two implementations: old priorityless RR,
new priority-aware RR

rewritten delta list handler. still API missing

makefile now system-independent, just set SYS variable in make.in

---- 0.4.90 (experimental)
