Message-Id:	<200003070250.VAA19151@qnx.com>
Subject:	Re: DJGPP innovations ?????
To:	eliz AT is DOT elta DOT co DOT il (Eli Zaretskii)
Date:	Mon, 6 Mar 2000 21:50:25 -0500 (EST)
From:	"Alain Magloire" <alain AT qnx DOT com>
Cc:	neldredge AT hmc DOT edu (Nate Eldredge), djgpp-workers AT delorie DOT com
In-Reply-To:	<Pine.SUN.3.91.1000306095426.9921L-100000@is> from "Eli Zaretskii" at Mar 06, 2000 09:55:22 AM
X-Mailer:	ELM [version 2.5 PL0b1]
MIME-Version:	1.0
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	dj-admin AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

> 
> 
> On Sun, 5 Mar 2000, Alain Magloire wrote:
> 
> > The Daemon thing is another step nothing special about it, you start a cron
> > job or any shell scripts.
> 
> Running a program in background is easy, but finding files that are
> identical (efficiently) is not.

find $1 -xdev -type f -printf '%p %s\n' | \
 sort -nk1 | tee candidates | \
 uniq -f1 >uniquefiles && \
 comm -3 candidates uniquefiles >redundant && \
 join -1 2 -2 2 -o 2.1 1.1 redundant uniquefiles | xargs -n2 ln -f  

This was proposed by a coworker here.  I did not try it, but it's just a proof
that this is something that sh/awk/perl gurus can probably come up with in
one afternoon.

> > > But you'd also need something like copy-on-write to deal with one of the
> > > files being modified, if you wanted it to be transparent.
> > 
> > Why do you want COW (Copy On Write)?  I would expect any modifications
> > to be reflected on all the files.
> 
> I think you forget that the links were not created by the user, they
> were created by a program which didn't have any clue whether the user
> did or didn't want these files to be linked.  Since the user didn't
> link the files, she might be mightily surprised when changing one of
> them changes the other(s) as well.

Agreed.  Nate also remind me of this.

> 
> For example, imagine that you have 5 different versions of the Grep
> distribution on your disk ;-).  Suppose that the ``link daemon'' finds

8-) [laughs] actually I just finish the count and I have 6 lying around.

> out that all 5 versions have the same dfa.c in them, and decides to
> link them together.  Then you modify dfa.c in one of the directories.
> Then you want to diff it against the old version, to send diffs to
> your pretesters.  Imagine your surprise when Diff reports identical
> files...

How shocking !! ;-)

Agreed, on the example.

We have the same thing here, it is call the filesystem package
and not the "Single Instance Store" and I've been using it for quite
sometimes on QNX/Neutrino.

-- 
au revoir, alain
----
Aussi haut que l'on soit assis, on est toujours assis que sur son cul !!!

- Raw text -