Mailing-List: contact river-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: river-dev@incubator.apache.org
Received-SPF: neutral (nike.apache.org: 84.2.40.5 is neither permitted nor
 denied by domain of la.tinca@gmail.com)
Date: Mon, 15 Nov 2010 21:05:33 +0100
From: Zsolt =?UTF-8?B?S8O6dGk=?= <la.tinca@gmail.com>
To: river-dev@incubator.apache.org
Subject: Re: test framework migration
Message-ID: <20101115210533.425bb16e@brain>
In-Reply-To: 
 <77F1E32F67C8D5479858C0C7E93EB46503B1507D@WAL-MAIL.global.avidww.com>
References: <2101938159.1941288776689346.JavaMail.hudson@aegis>
	<1987111689.2571288839742596.JavaMail.hudson@aegis>
	<AANLkTinLUtHaTbjjb5HxRBtVCFk+-WkogxUuYCis9K67@mail.gmail.com>
	<4CD3B6C4.4080903@zeus.net.au>
	<AANLkTi=CNA96bRLRrDpH3zQ1ajweOASrrGnQf73yfBMb@mail.gmail.com>
	<4CD4EEB9.1030300@zeus.net.au>
	<AANLkTi=ko5Nt491Mbir5A8rY7MX3wFRCnjpYJodwzGk8@mail.gmail.com>
	<20101107115412.22d31db1@brain>
	<4CD6AB27.1020409@acm.org>
	<4CDC9294.2070100@zeus.net.au>
	<77F1E32F67C8D5479858C0C7E93EB46503AA43A5@WAL-MAIL.global.avidww.com>
	<20101113142908.09f59ace@brain>
	<77F1E32F67C8D5479858C0C7E93EB46503B1507D@WAL-MAIL.global.avidww.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

On Mon, 15 Nov 2010 12:38:06 -0500
"Christopher Dolan" <christopher.dolan@avid.com> wrote:

> While useful, I recommend against that feature in general.
> Philosophically, it's bad practice to connect the unit tests -- they
> should be as simple as possible and stand alone.  Technologically,
> when the first test fails, then the second one fails with a really
> obscure error.  I can't remember the specifics, but it was bad enough
> to convince me to rewrite all of my "dependsOnMethod" tests.

I agree in general :-)
In some cases test setup can take long or a resource cannot be reused
quickly enough and (test) class level setup is in order. For the former
doing an exhaustive testing of database functionality can be an
example. I also had a case involving extended XML trees for which
initializations would have been cumbersome and would have made more
hardly followable code. (Admittedly it is still a trade-off considering
the brittleness you mention.) For the latter I have only a gut feeling
that some use of the resources by River might fall into that category.

Zsolt

Ps.: It should be noted that TestNG runs test cases parallely by
default.