From mboxrd@z Thu Jan  1 00:00:00 1970
From: erik quanstrom <quanstro@labs.coraid.com>
Date: Thu, 11 Jul 2013 10:05:18 -0400
To: 9fans@9fans.net
Message-ID: <24ccd68343090eaea81aabf6e539c2b1@coraid.com>
In-Reply-To: <18706FFE-E7CD-4A93-B69E-604AC2370B1E@ar.aichi-u.ac.jp>
References: <18706FFE-E7CD-4A93-B69E-604AC2370B1E@ar.aichi-u.ac.jp>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: Re: [9fans] grep bug?
Topicbox-Message-UUID: 6bba428e-ead8-11e9-9d60-3106f5b1d025

On Thu Jul 11 09:13:10 EDT 2013, arisawa@ar.aichi-u.ac.jp wrote:
> Hello,
>
> It seems f option of grep is buggy.
> or any limitations in using the RE?
>
> term% wc MD5dir
>    4584    9168  388756 MD5dir
> term% wc x
>    4582    4582  151206 x
> term% grep -f x MD5dir | wc
>    4580    9160  388463
> term%
> term% grep e54272690d513f8b2403568a7574b1ba MD5dir
> e54272690d513f8b2403568a7574b1ba /usr/arisawa/src/taskfs/Q/task.387.a/
> term% grep e54272690d513f8b2403568a7574b1ba x
> e54272690d513f8b2403568a7574b1ba
> term% grep -v -f x MD5dir
> 7b6d7ae369226b6d0195ac3fe4487ce7 /usr/arisawa/src/elnfs/WWW/
> d44d788ad1237311d8282bbabca65977 /usr/arisawa/src/hg/python-2.5.1-ape/Modules/_ctypes/libffi/src/darwin/
> e54272690d513f8b2403568a7574b1ba /usr/arisawa/src/taskfs/Q/task.387.a/
> 84a0f83f5020f16d0b277e8b19407791 /usr/arisawa/src/trans
> term%

a trick i often use for many fixed strings is sort + uniq.
(internally, grep/comp.c:/^increment does O(n^2)
qsorts on the patterns) perhaps it could be used to
double-check.

to find the md5 hashes that only appear in one file or the other
(only the first field is considered by uniq),

	cat x MD5dir | sort | uniq -c | sed '/^ *2 /d'

to count the fields that appear in both

	cat x MD5dir | sort | uniq -c | grep '^ *2 ' | wc -l
or
			...	| awk '$1==2{n++}END{print n}'

can you find a smaller test case that has the same issue.  this
should be fixed

- erik