History:

0.1	some time in August 1998

	Original version, shows the same as asload.

0.2	20 September 1998

	The version that works and shows the average.
	I am looking for a better way to calculate
	average values so the 1.0 must wait.

1.0	22 September 1998

	Finally I got it all nice working. I could not
	figure an algorithm that would allow to adjust
	the number of samples without dynamic allocation
	of memory though. Probably does not exist ;-)

1.1	03 November 1998

	Oh, well, the whole thing rewritten almost
	from scratch again. It all started with the
	message from Cory Visi:

I have suggestion for your cool little Wharf app ascpu. I think that the CPU
meter on the side should have an interval independent of the graph display.
The CPU meter on the left is much for useful in real time (little or no
interval) while the graph is much more useful with a longer interval (like
10 seconds). So even if you hardcode a 0 or 1 interval for the CPU meter, I
think that would make it a lot nicer. I use an interval of 10 seconds, so
the CPU bar is pretty useless :)

	So, it runs in both conventional (as I dreamed it
	up) and the opposite (as Cory Visi dreamed it up)
	modes. The intervals are independent :)

1.2	02 January 1999

	Fixed the "-position" option to work with the 
	negative coordinates.

1.3	07 February 1999

	Thanks to Liang Tai-hwa <avatar@www.mmlab.cse.yzu.edu.tw>
	the ascpu now runs on FreeBSD systems as well.

1.4	12 April 1999

	Thanks to Stewart Gebbie <stewart@global.co.za> for fixing
	this embarrasing bug for me. He writes:
	"I use ascpu but noticed that the -user option didn't
	seem to work. Apon closer inspection I noticed that
	the code to check for the -user option was after
	the -u check and thus the 'if...else' statments
	picked up the -user as a -u.
	Attached is a patch for ascpu.c which fixes things."

1.5	28 April 1999

	Thanks to Evgeny Stambulchik <fnevgeny@plasma-gate.weizmann.ac.il>
	for fixing an annoying bug with "troughs" in the display.
	Apparently it is possible that the number of CPU cycles
	decreases by 1 under heavy load... Weird.

	Evgeny also helped me to set up the SMP support which
	goes out with this version.

	I added the -title option to be able to swallow a number
	of ascpu instances into the Wharf on an SMP machine.

1.6	06 May 1999

	Thanks to ii1 <ii1@grant.kharkov.ua> who sent in a patch
	ascpu can execute commands when clicked on now.

1.7	27 July 1999

	Stasinos Konstantopoulos <konstant@let.rug.nl> ported
	ascpu to HP-UX and kindly sent in the patch.

1.8	16 December 1999

	Changed the internal loop to react to the absolute
	value of the difference in time for the update.
	Suggested by Nicola Bernardini <nicb@axnet.it>

1.9	09 March 2000

	Fixes the security problem on the systems where ascpu
	must run with some kmem or root privileges (SUID)
	to access the system information (FreeBSD and now AIX).
	It was possible to execute arbitrary commands using
	the privilege. Now ascpu forks and the child drops
	privileges to execute the command.

	Kirill Bogachev <bogachev@mech.math.msu.su> ported
	ascpu to AIX on PowerPC:
	"I have ported it to AIX running on PowerPC. The port has been tested
	on AIX 4.1 but should works on AIX > 3.1.
	...
	In AIX "ascpu" should be installed setuid root to get access to
	kernel memory."

1.10	17 January 2004

	Patch by Mathieu Chouquet-Stringer <mathieu@newview.com>
	for Linux Kernel >= 2.5
	Patch by Jeff Brown <jabrown@caida.org> for script
	execution from "withdrawn" state.

1.11	10 October 2005

	Thanks to Kevin Chan-Henry <kch1@dsl.pipex.com> for
	pointing out that the Linux kernels after 2.6.10
	introduce an additional field in /proc/stat.
	If you never used the "-cpu" switch you will not be
	affected. But with multiple CPUs ascpu breaks.
	So this version should fix that.