From: Carl Worth <cworth@cworth.org>
To: Jesse Rosenthal <jrosenthal@jhu.edu>,
	Sebastian Spaeth <Sebastian@SSpaeth.de>,
	Notmuch development list <notmuch@notmuchmail.org>
Subject: Re: sort order regression
In-Reply-To: <87tyr37e4p.fsf@jhu.edu>
References: <87bpdbmvj0.fsf@SSpaeth.de> <87wrvz7ex3.fsf@jhu.edu>
	<87vdbj7elk.fsf@jhu.edu> <87tyr37e4p.fsf@jhu.edu>
Date: Thu, 22 Apr 2010 14:10:48 -0700
Message-ID: <87ljcfqjl3.fsf@yoom.home.cworth.org>
MIME-Version: 1.0
Content-Type: multipart/signed; boundary="=-=-=";
	micalg=pgp-sha1; protocol="application/pgp-signature"
Precedence: list

--=-=-=
Content-Transfer-Encoding: quoted-printable

On Thu, 22 Apr 2010 16:06:27 +0200, "Sebastian Spaeth" <Sebastian@SSpaeth.d=
e> wrote:
>=20
> jkr and I noticed that patch series are shown in reverse order now, in
> fact threads seem to display messages at the same depth in reverse
> chronological order now.

My fault! Sorry about that.

On Thu, 22 Apr 2010 10:14:16 -0400, Jesse Rosenthal <jrosenthal@jhu.edu> wr=
ote:
> Just to follow up on this, it seems that the regression comes from the
> fix Carl introduced in 2a1a4f0551 to make his simplification of my patch

Indeed that was the problem. And your original code *did* pass the tests
I wrote. There are still a couple of cleanups I'd like to make, (like
assigning the subject only in _add_matched_message and not uselessly in
_add_message as well). But I can do those now without breaking the test
suite!

> Unfortunately, I'm not quite skilled enough at git to turn back some
> files to a certain state, tests to another state, and so on.

For now, since the test suite is just a single file, it's easy enough to
do:

	cp test/notmuch-test latest-notmuch-test
        git checkout HEAD~5	# or whatever
	make
	./latest-notmuch-test

Though that doesn't answer the question of how to do this with
git. Probably something like:

	git checkout HEAD~5
	git checkout master -- test
	make test

On Thu, 22 Apr 2010 10:31:18 -0400, Jesse Rosenthal <jrosenthal@jhu.edu> wr=
ote:
> Sorry, got that slightly wrong. The following commits need to be
> reverted:
>=20=20
> 36e4459a328b8449b3e9d510be81a332a9b35aaa
> f43990ce134d838cdb2cdd5d0752a602e81cfdd9
> 7fb56f9dc5d8e66f717f5e48ecbfbc11c8190182
>=20
> When I do this, search order is right, and all tests are passed.

I did this. And before doing it I added a new test to show the
regression, (which the reverts then make pass).

So in my defense, it wasn't my fault[*] I didn't notice the
regression. It was the test suite's fault for not testing "notmuch show"
at all.

=2DCarl

[*] Except that I'm the one that wrote the test suite and didn't put any
"notmuch show" tests into it. Oops!

--=-=-=
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iD8DBQFL0LtY6JDdNq8qSWgRAlMvAJ4lBPkehg9sWrJs/cVytNPYchG4UACfaSuF
JTJziwK1L+4TG9ldNg9KTFs=
=r8Y+
-----END PGP SIGNATURE-----
--=-=-=--