Re: Parsing regression with gmime-2.6?

Subject: Re: Parsing regression with gmime-2.6?

Date: Thu, 08 Mar 2012 10:27:21 -0800

To: David Bremner, Austin Clements, notmuch@notmuchmail.org

Cc: Daniel Kahn Gillmor

From: Jameson Graef Rollins


On Tue, 06 Mar 2012 18:04:50 -0400, David Bremner <david@tethera.net> wrote:
> There seems to be something weird going on with gmime-2.6; maybe we
> didn't catch some api change? 
> 
> I noticed a surprising number of messages failing to parse, so I wrote a
> script to take a random sample of 1000 messages and run notmuch show on
> them. A shocking 650 to 700 of them fail to parse with gmime-2.6. No
> failures are reported with gmime-2.4.

This *is* shockingly high, and very confusing.  Somehow all of the tests
pass?  That's very strange, particularly since there are enough real
messages used in the tests that we should experience at least one of
these failures, right?  Is there really nothing special about the
messages that are failing to parse?  Your later patches seem to indicate
that this has something to do with mbox, although you don't mention that
hear.

I thought I had been using gmime-2.6 for the last month, so I was very
skeptical of this issue.  Unfortunately I just found a bug in the
configure script [0] that meant I had actually been using gmime-2.4.
However, since fixing things and actually using gmime-2.6 now, I still
don't see any problems.

If the problem is specific to mbox, which we explicitly don't support, I
don't think this is tragic enough to warrant removing gmime-2.6 as a
dependency satisfier for 0.12 in Debian.  In an event, a more explicit
description of the problem you're seeing would be helpful.

jamie.

[0] id:"1331225101-24385-1-git-send-email-jrollins@finestructure.net"
part-000.sig (application/pgp-signature)

Thread: