DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=mime-version:sender:in-reply-to:references:date
	:x-google-sender-auth:message-id:subject:from:to:content-type
	:content-transfer-encoding;
	b=AyfDRys8ZmzsRqS3uDjo63JTzVM8nY7yCwjRyktdUD7JFW2LUScg4r0c5SJIcUxqK9
	oE5bXqwzjZ6iYbRGb9ynfVv/NjyeGQfS5iFb9fxNtOYOlT/9NytO49y1LFKgNUvXGmH4
	/bvGDTs/Q4o9UlEmTkq81xY9Y76V+y2zCj3hM=
MIME-Version: 1.0
Sender: amdragon@gmail.com
In-Reply-To: <4E0091AE.5070609@fifthhorseman.net>
References: <20110610073208.GA74787@codecafe.com>
	<4E0091AE.5070609@fifthhorseman.net>
Date: Tue, 21 Jun 2011 12:01:17 -0400
Message-ID: <BANLkTik_o6d03g7Jsy-sUbjCp9HR7Q=rkw@mail.gmail.com>
Subject: Re: [PATCH] notmuch-new.c infinite recursion symlink bug
From: Austin Clements <amdragon@mit.edu>
To: notmuch <notmuch@notmuchmail.org>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Precedence: list

On Tue, Jun 21, 2011 at 8:42 AM, Daniel Kahn Gillmor
<dkg@fifthhorseman.net> wrote:
> My point is: there are lots of ways to get infinite recursions via
> symlinks; =A0hard-coding a check for one specific way seems like a
> sub-optimal approach, because it leaves the other paths still present,
> and introduces an unexpected/surprising asymmetry.
>
> I'm not sure what the specific right way is to solve the problem you
> identified, though.

A simple solution to this problem would be to record the inode numbers
of each visited directory, probably in a hash table in
add_files_state_t.  At the beginning of add_files_recursive, right
after it stat's the directory, check if st.st_ino is in this hash
table; if it is, return immediately, otherwise add it to the hash
table and proceed as usual.

Alternatively, because of folder search, it might be better to keep a
stack of inode numbers to eliminate loops while retaining notmuch's
current approach of repeatedly indexing mail that's symlinked in
multiple folders.