Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!husc6!rice!sun-spots-request
From: klm@goon.cme.nbs.gov (Ken Manheimer)
Newsgroups: comp.sys.sun
Subject: 3.5, 4.0{,.1} join bug
Keywords: Software
Message-ID: <KLM.89Apr20152017@goon.cme.nbs.gov>
Date: 6 May 89 01:50:39 GMT
Sender: usenet@rice.edu
Organization: Nat'l Institute of Standards and Technology
Lines: 68
Approved: Sun-Spots@rice.edu
Original-Date: 20 Apr 89 19:20:17 GMT
X-Sun-Spots-Digest: Volume 7, Issue 269, message 8 of 23

Issue: Unix 'join' utility dumps core when applied to certain very simple data

			     Environment
			     -----------
OS:	Sun OS 3.5, 4.0, 4.0.1

CPUs: 	Apparently irrelevant - explicitly tested on 3/50, 3/180, 3/280

Severity: moderate

			 Problem Description
			 -------------------
Applied to specifically formed data in a specific mode, join works part
way and then dumps core.
			     Repeat By:
			     ----------
Given files:

"/tmp/base":
a b
b b b b b b b b b b b b b b b b b b b b
<EOF>

and "/tmp/add":
ac c
<EOF>

then:

% join -a3 /tmp/base /tmp/add
a b
Segmentation fault (core dumped)
% 

Some apparently crucial aspects of these files - a line in the first file
('/tmp/base'), other than the first line, must have exactly twenty fields
in it, and that line must come up for comparison against a line from the
other file ('/tmp/add').  I cooked down these files from a real situation
where i was merging two files of well over one thousand lines apiece.  The
situation is not artificial - the lines represent file-names followed by
dates in a registry for a backup system, where the new items are being
merged with the established entries.

			      Workaround
			      ----------
Use 'cat', 'sort', and 'awk' to accomplish the 'join -a3':

% cat /tmp/base /tmp/add | sort -b +0 -1 +n1 - | awk '\
				   BEGIN { PREVIOUS = ""}\
				   $1 == previous { printf " %s",$2 }\
				   $1 != previous { if(NR != 1) printf "\n"\
						    printf "%s", $0;\
						    previous = $1 }\
				   END {printf "%c",10} '
a b
ac c
b b b b b b b b b b b b b b b b b b b b
% 



Ken Manheimer		 	klm@cme.nbs.gov or ..!uunet!cme-durer!klm
National Institute of Standards and Technology
(Formerly "National Bureau of Standards")
CME Factory Automation Systems, Software Support

"For without the inner the outer loses its meaning; and without the outer,
 the inner loses its substance." - R.D.Laing _The Politics of Experience_