Mailing-List: contact cygwin-help@sourceware.cygnus.com; run by ezmlm
Sender: cygwin-owner@sources.redhat.com
Message-ID: <397D6B99.77361071@cygnus.com>
Date: Tue, 25 Jul 2000 12:27:37 +0200
From: Corinna Vinschen <vinschen@cygnus.com>
Reply-To: cygwin <cygwin@sources.redhat.com>
MIME-Version: 1.0
To: Andrej Borsenkow <Andrej.Borsenkow@mow.siemens.ru>
CC: Cygwin@Sourceware.Cygnus.Com
Subject: Re: memory leak in setenv()?
References: <001201bff612$88df19f0$21c9ca95@mow.siemens.ru>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Andrej Borsenkow wrote:
> 
> If existing value is longer than new, old value is simply overwritten.
> But if old value is shorter, new memory chunk is allocated via malloc(),
> but old one is never freed. Initially all environment strings are
> allocated using strdup() in environ_init(), so who is responsible for
> freeing unused memory in this case?

Nobody in case of setenv() and unsetenv(). It's a known problem
of the environment implementation that freeing environment
variables is not easy. The library can't decide if a string is
currently in use by the application.

As another example take a look into the implementation of
setenv/unsetenv in glibc. It does never free environment strings
by itself so each call to setenv(NAME, VAL, 1) creates a new
memory leak unless it is a known string value once given to
setenv. In that case, glibc tries to recycle strings to use
memory more efficient.

Corinna

-- 
Corinna Vinschen
Cygwin Developer
Cygnus Solutions, a Red Hat company

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com