From: Jameson Graef Rollins <jrollins@finestructure.net>
To: Daniel Kahn Gillmor <dkg@fifthhorseman.net>,
	David Bremner <david@tethera.net>,
	Austin Clements <amdragon@MIT.EDU>, notmuch@notmuchmail.org
Subject: Re: Parsing regression with gmime-2.6?
In-Reply-To: <87vcme3kf6.fsf@pip.fifthhorseman.net>
References: <87d38w2e7h.fsf@zancas.localnet>
	<1331058417-13776-1-git-send-email-amdragon@mit.edu>
	<87wr6xmlml.fsf@zancas.localnet>
	<87vcme3kf6.fsf@pip.fifthhorseman.net>
User-Agent: Notmuch/0.11.1+266~gadb05cc (http://notmuchmail.org) Emacs/23.3.1
	(x86_64-pc-linux-gnu)
Date: Thu, 08 Mar 2012 13:32:44 -0800
Message-ID: <874ntybwxv.fsf@servo.finestructure.net>
MIME-Version: 1.0
Content-Type: multipart/signed; boundary="=-=-=";
	micalg=pgp-sha256; protocol="application/pgp-signature"
Precedence: list

--=-=-=
Content-Transfer-Encoding: quoted-printable

On Thu, 08 Mar 2012 15:30:21 -0500, Daniel Kahn Gillmor <dkg@fifthhorseman.=
net> wrote:
> This does seem to be a regression in gmime 2.6.  I've reported the bug
> upstream, along with a simplified (non-notmuch) demonstration:
>=20
>  https://bugzilla.gnome.org/show_bug.cgi?id=3D671680
>=20
> We'll see what gmime's upstream has to say about it.

Thanks so much for following up on this, Daniel.

> As a devil's advocate, i could argue that a message in a maildir that
> starts with a "From " line isn't a proper e-mail message in the first
> place, and therefore gmime 2.6 is being more rigorously correct about
> what it accepts.  In particular, if a user were to place a multi-message
> mbox file in their notmuch message store, i think that notmuch linked
> against 2.4 would happily index only the first message of it, and the
> rest of the message would be "hidden", whereas gmime 2.6 allows us to
> detect these failures and avoid indexing them directly.

I actually agree with this position.  mbox files are not proper email
messages, so if gmime does not explicitly support parsing them then we
really can't expect it to parse them.  We *can* expect gmime to return
some sort of proper error message/return code/etc and not fail in a bad
way, but beyond that I think the burden is on us.

> That said, i understand that this is probably not an entirely rare
> situation, and i lean toward the idea that gmime 2.4's behavior was
> actually the Right Thing.  Also, I haven't been able to find any
> explicit documentation to indicate that the behavior change was
> a deliberate one.

And I think I don't agree with this one: I not think that handling mbox
files in the way you've outlined above is the Right Thing.  Only
partially parsing an mbox file, by indexing only the first message for
instance, seems like a bad solution to me, and one that's likely to lead
to a lot of confusion (e.g. "where are the rest of the messages that
were in my mbox?").

If notmuch encounters an mbox file in the store it should just skip the
message and continue, while indicating that it is skipping the message
because it's not a proper email message, as it does with other non-email
files.

jamie.

--=-=-=
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
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=Flnt
-----END PGP SIGNATURE-----
--=-=-=--