GNU bug report logs - #10390
Bug with Test Command

Previous Next

Full log

Message #12 received at 10390-done <at> debbugs.gnu.org (full text, mbox):

From: Eric Blake <eblake <at> redhat.com>
To: "Mathi.C" <mathiazhaganc <at> gmail.com>
Cc: 10390-done <at> debbugs.gnu.org
Subject: Re: bug#10390: Bug with Test Command
Date: Wed, 28 Dec 2011 06:28:06 -0700

[Message part 1 (text/plain, inline)]

tag 10390 notabug
thanks

On 12/27/2011 11:12 PM, Mathi.C wrote:
> Hi,
> 
> I have come across a scenario where test command resulted wrong result than
> expected.
> 
> Please follow the below steps to simulate the scenario.
> 1. Define a variable to blank/null
>      export a=
> 2. Test the parameter with test command on file exists
>      test -f $a

There's your problem.  You failed to properly quote things.  Per shell
expansion rules, unquoted $a is elided, rather than an empty argument,
which means you passed test exactly one argument ("-f"), and since that
one argument is not empty, it is always true.

If you want to cope with an explicit empty string, you must use quoting:

test -f "$a"

> 3. Echo the result of the last executed command
>     echo $?

As this behavior matches POSIX requirements on test, there is nothing to
fix here.  Also, be aware that you were probably testing your shell's
built-in test, rather than coreutils' test (running the latter usually
requires that you do things like 'env test' or '/usr/bin/test' to bypass
shell built-ins).

-- 
Eric Blake   eblake <at> redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

[signature.asc (application/pgp-signature, attachment)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.