From: Dmitry Kurochkin <dmitry.kurochkin@gmail.com>
To: Carl Worth <cworth@cworth.org>, Austin Clements <amdragon@MIT.EDU>
Subject: Re: [PATCH 2/2] test: use emacsclient(1) for Emacs tests
In-Reply-To: <87zkl190od.fsf@yoom.home.cworth.org>
References: <1309132379-24089-1-git-send-email-dmitry.kurochkin@gmail.com>
	<1309146846-30991-1-git-send-email-dmitry.kurochkin@gmail.com>
	<1309146846-30991-2-git-send-email-dmitry.kurochkin@gmail.com>
	<BANLkTi=nwHRndz367K5oh4=tPaU-KNeZ8Q@mail.gmail.com>
	<87hb7byqge.fsf@gmail.com> <87d3hyzs1c.fsf@gmail.com>
	<20110628034937.GB4120@mit.edu> <20110628162257.GD4120@mit.edu>
	<871uydzz3x.fsf@gmail.com> <87zkl190od.fsf@yoom.home.cworth.org>
User-Agent: Notmuch/0.5-275-g75c95e5 (http://notmuchmail.org) Emacs/23.3.1
	(x86_64-pc-linux-gnu)
Date: Wed, 29 Jun 2011 00:47:37 +0400
Message-ID: <87mxh1y97a.fsf@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: notmuch@notmuchmail.org
Precedence: list

On Tue, 28 Jun 2011 13:10:58 -0700, Carl Worth <cworth@cworth.org> wrote:
> On Tue, 28 Jun 2011 20:42:42 +0400, Dmitry Kurochkin <dmitry.kurochkin@gmail.com> wrote:
> > I would like to hear what other (Carl in particular) think about this.
> > If the consensus is for your approach, I would be happy to implement
> > it.
> 
> In general, I love the whole series, thanks! I'm looking forward to our
> future, faster test suite.
> 
> Even more, I love the constructive dialog that follows the original
> series and the attention being focused on getting things right.
> 
> As for the detail of whether to use emacsclient or Austin's look-alike,
> I don't have a strong attachment to either solution. I do appreciate
> concrete technical things like "robust against recycled PIDs",

This is an issue indeed.  We can take more care to better identify the
process.  But I think this is not really needed.  I am fine with leaving
a small (very tiny, really :)) possibility for leaving Emacs running for
longer than needed.  Note this only happens if test is terminated
abnormally.

> "more
> robust against leaving daemon's around for some reason", etc.
> 

Not sure I agree with this.  The watchdog is not perfect, but not less
reliable.  It reacts with a delay, but again I do not think this is so
important.

I think I found an issue with Austin's approach: if an error happens
(and is not catched), the loop would stop.  Currently, this would result
in non-working tests and also would leave the Emacs process behind.  I
guess we will have to add error catching in the loop.  Austin?

> Would any of this potentially interfere with my own usage of emacsclient
> and emacs server? I use them regularly and would be quite surprised (and
> likely frustrated) if the test suite got mixed up with my existing emacs
> server (or the other way around). Maybe that's already taken care of
> with either approach? (A quick skim of the emacsclient manual age didn't
> make it obvious to me how emacslcient finds its server.)
> 

Emacs uses `server-name' and emacsclient --socket-name to identify the
server.  Tests use "notmuch-test-suite-<TEST_SHELL_PID>" server name.
While it is not entirely impossible to mess your existing Emacs server,
it is unlikely enough IMO.

Regards,
  Dmitry

> -Carl
> 
> -- 
> carl.d.worth@intel.com