From: Dmitry Kurochkin <dmitry.kurochkin@gmail.com>
To: David Bremner <david@tethera.net>
Subject: Re: [PATCH 3/3] test: add emacs test for hiding a message following
	an HTML part
In-Reply-To: <87d3ehylon.fsf@zancas.localnet>
References: <1309744743-8556-1-git-send-email-dmitry.kurochkin@gmail.com>
	<1309744743-8556-3-git-send-email-dmitry.kurochkin@gmail.com>
	<87zkhrlg8z.fsf@zancas.localnet> <874nzz2p3f.fsf@gmail.com>
	<87d3ehylon.fsf@zancas.localnet>
User-Agent: Notmuch/0.8-50-gd528fd9 (http://notmuchmail.org) Emacs/23.3.1
	(x86_64-pc-linux-gnu)
Date: Sun, 02 Oct 2011 05:45:53 +0400
Message-ID: <87hb3s414e.fsf@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: notmuch@notmuchmail.org
Precedence: list

Hi David.

On Sat, 01 Oct 2011 08:51:20 -0300, David Bremner <david@tethera.net> wrote:
> On Mon, 26 Sep 2011 21:25:08 +0400, Dmitry Kurochkin <dmitry.kurochkin@gmail.com> wrote:
> > Most likely the test passes because emacs is run in server mode and
> > visibility stuff works differently.  I sent a patch series [1] to run
> > emacs in screen exactly for this reason.  Please consider pushing it.
> > Then the test should fail as expected.
> 
> > [1] id:"1309496122-4965-1-git-send-email-dmitry.kurochkin@gmail.com"
> > [2] id:"874o2germq.fsf@gmail.com"
> 

I was wrong about it.  I am moving to a new system.  And I see the test
pass on one system and fail on another.  So it is not related to running
emacs tests inside screen.  At the moment, I have no idea why it behaves
differently on the two systems.  Both are Debian unstable (one is
current another 2 weeks old).  Could it be some extra emacs packages
installed?  I know notmuch disables system and user init files, but
perhaps it is not enough?

> I'd like to hear more feedback from people on the list about using
> screen in the tests. Is this a portability issue for non-linux users?
> 

Nevertheless, I believe running emacs tests inside screen is a
requirement for doing complex UI tests.  For example, `window-end'
function behaves differently when there is no emacs frame.  Of course,
it does not have to be screen, we may use any other similar tool
(e.g. tmux, dtach).  We may even use a dummy X server.  But screen seems
to be the best option.

I see no problem with not running emacs tests on systems which do not
have screen.  After all, emacs is a UI.  Core tests would run fine as
before.

BTW there were patches to support tmux (or maybe it was dtach).  IIRC
the consensus was that supporting it does not worth complicating test
lib.  I think this can be reconsidered if it turns out that many users
do not have screen, but do have tmux.  But you never know if the
patches are not applied.  They have been around for several months, I
doubt you would suddenly get feedback now.

Regards,
  Dmitry

> d