Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!lll-winken!xanth!mcnc!ecsvax!dukeac!bet
From: bet@dukeac.UUCP (Bennett Todd)
Newsgroups: comp.lang.c
Subject: Re: A self-referential challenge
Message-ID: <1323@dukeac.UUCP>
Date: 27 Mar 89 22:59:21 GMT
References: <6219@bsu-cs.UUCP>
Reply-To: bet@orion.mc.duke.edu
Organization: Radiology, Duke Med. Center, Durham, NC
Lines: 35


[
	Asks for suggestions on how to make 

		CRC-computer * >CRC-output

	contain correct CRC values for all files -- including
	"CRC-output"
]

That's easy, just iterate looking for the fixed point, like so:

	touch .CRC-output
	finis=0
	while test $finis -eq 0
	do
		CRC-computer * >CRC-output
		diff CRC-output >/dev/null && finis=1
		cp CRC-output .CRC-output
	done
	rm .CRC-output

But please, not on my machine:-). Seriously, without an in-depth analysis of
how CRCs behave (which is way, way beyond my abilities) there is no reason
to believe that a fixed point exists; it might fall into a cycle instead
(given any finite number of bits in the CRC, it has to do one or the other).
Unless some mathematician steps forward and waves her magic wand, I expect the
best answer will be to fall back 10 and punt:

	CRC-computer * | egrep -v CRC-output >CRC-output

I really liked the question, though.

-Bennett
bet@orion.mc.duke.edu