Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!hc!beta!cmcl2!phri!roy
From: roy@phri.UUCP (Roy Smith)
Newsgroups: comp.unix.wizards
Subject: impossible problem for find(1)?
Message-ID: <2645@phri.UUCP>
Date: Thu, 23-Apr-87 20:57:02 EST
Article-I.D.: phri.2645
Posted: Thu Apr 23 20:57:02 1987
Date-Received: Sat, 25-Apr-87 19:02:02 EST
Reply-To: roy@phri.UUCP (Roy Smith)
Organization: Public Health Research Inst. (NY, NY)
Lines: 61


	A user posed a problem which I can't solve.  She wants to find all
the files named *refs which are newer than a file called INDEX in the same
directory.  Thus she wants to know if any of ./x/foorefs, ./x/barrefs or
./x/bazrefs is newer than ./x/INDEX, and if ./y/z/refs is newer than
./y/z/INDEX, etc.  I can't think of how to do this with find, or even any
combination of find and any other standard tools.  We're running MtXinu
4.3BSD/NFS on a Vax and Sun-3.1 on some Sun-3's.

	The obvious

% find . "*refs" -newer INDEX -print

finds *refs files which are newer than INDEX in the directory in which I
ran the find.  This is not, unfortunately, what I need.

	The Sun version of make has a -q option which I could use to do
something like

% find . -name INDEX -exec make -q INDEX ';'

but this will only work on the Suns, and there is still the problem of how
to phrase the Makefiles and where to put them.  Of course, if the MtXinu
NFS support worked better I would contemplate running this on a Sun CPU
mounting the Vax disk, but it doesn't, so I won't.

	I could envision something horrible along the lines of

% find . -type d -exec find '{}' -type d -prune -name '"*refs"' \
-newer INDEX -exec pwd -print '";"' ';'

but I no longer do the type of drugs required to seriously contemplate
doing that.  I havn't wasted the time to get the exact syntax of the above
correct so please don't bother flaming me about messing up the double layer
of quotes or some other silly detail.  It was meant as a general idea, not
a finished product.  Note that you need the "-exec pwd" in the inner-level
find because all "-print" will say is "./foorefs" which isn't useful.

	Another hair-brained (a cross between hairy and brain-damaged?)
idea centers around doing "ls -lt INDEX *refs" and checking to see that the
first output line contains (or actually doesn't contain) INDEX, perhaps by
piping it through "head -1 | grep INDEX".  This gives (roughly)

% find . -type d \
-exec sh -c 'ls -lt {}/INDEX {}/*refs | head -1 | egrep -vs INDEX' ';' \
-print

which is also too sick to think about doing, let alone explaining to a
non-wizard.  Besides, I wouldn't count on the "sh -c" to properly pass back
the exit status of the pipeline (earlier Berkeley versions of sh did this
wrong; I have no idea if the 4.3 version is fixed or not), and you run into
problems if the sh dies because the *refs expansion doesn't work in a
particular directory.

	Is there really no easy way to do this?
-- 
Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016

"you can't spell deoxyribonucleic without unix!"