From: Austin Clements <amdragon@MIT.EDU>
To: Thomas Schwinge <thomas@schwinge.name>, Carl Worth <cworth@cworth.org>,
	notmuch@notmuchmail.org
Subject: Re: [PATCH 2/2] notmuch.el:notmuch-search-process-filter: Rewritten.
	Cope with incomplete lines.
In-Reply-To: <87pqgtmxfi.fsf@kepler.schwinge.homeip.net>
References: <87zkqeiffj.fsf@kepler.schwinge.homeip.net>
	<1296690999-16542-3-git-send-email-thomas@schwinge.name>
	<87r5ae8lem.fsf@yoom.home.cworth.org>
	<87pqgtmxfi.fsf@kepler.schwinge.homeip.net>
User-Agent: Notmuch/0.15.2+182~gd0bd88f (http://notmuchmail.org) Emacs/23.4.1
	(i486-pc-linux-gnu)
Date: Mon, 24 Jun 2013 18:43:41 -0400
Message-ID: <87d2rbyvs2.fsf@awakening.csail.mit.edu>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Precedence: list

On Tue, 15 Nov 2011, Thomas Schwinge <thomas@schwinge.name> wrote:
> Hi!
>
> On Thu, 10 Mar 2011 18:02:09 -0800, Carl Worth <cworth@cworth.org> wrote:
>> On Thu,  3 Feb 2011 00:56:39 +0100, Thomas Schwinge <thomas@schwinge.name> wrote:
>> > This issue has been lying in ambush as of 2009-11-24's commit
>> > 93af7b574598637c2766dd1f8ef343962c9a8efb.
>> 
>> Thanks very much for tracking down this bug, Thomas. What a nasty bug to
>> have in notmuch!
>> 
>> Your fix seems to drop the last thread from the search results
>> view. I've now committed a slightly modified fix that avoids that
>> problem. I also made the test case provide slightly cleaner results.
>> 
>> Let me know if you see any problems.
>
> That is much better, thanks!
>
> But we're not there yet...  %-| That is, today I hit another issue that
> appears to hide in the same elisp code.  See ``Error: Unexpected output
> From notmuch search''.  (And, thanks to eBay for long subject lines...)
>
> Unfortunately I'm totally out of time at the moment (final month of
> writing and wrapping up my diploma thesis), so I'm just dumping my state
> here, for now.

*snip*

I happened to notice that this is still marked as a bug.  I'm marking it
as fixed.  This code was completely rewritten to use the streaming JSON
parser (and now the streaming S-expression parser) and there's now a
test that feeds the process filter the search data one character at a
time to catch these sorts of buffer boundary bugs.