From: Dave Chinner <david@fromorbit.com>
Subject: Re: [PATCH] ext4/004: add dump/restore test
Date: Sun, 7 Dec 2014 10:44:44 +1100
Message-ID: <20141206234444.GL18131@dastard>
References: <1416910869-28538-1-git-send-email-wangxg.fnst@cn.fujitsu.com>
 <20141206214040.GA27484@thunk.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: Xiaoguang Wang <wangxg.fnst@cn.fujitsu.com>,
	fstests@vger.kernel.org, linux-ext4@vger.kernel.org
To: Theodore Ts'o <tytso@mit.edu>
Content-Disposition: inline
In-Reply-To: <20141206214040.GA27484@thunk.org>
Sender: linux-ext4-owner@vger.kernel.org

On Sat, Dec 06, 2014 at 04:40:40PM -0500, Theodore Ts'o wrote:
> On Tue, Nov 25, 2014 at 06:21:09PM +0800, Xiaoguang Wang wrote:
> > This test case will first use fsstress to fill a file system, then
> > dump it to standard output and restore it from standard input, finally
> > check that the original contents and the new contents generated by
> > restore tool will be same.
> > 
> > Signed-off-by: Xiaoguang Wang <wangxg.fnst@cn.fujitsu.com>
> 
> One question --- what is the intent of this test?  Is it to test the
> kernel, or the dump/restore program?  I have not bothered putting
> regression tests for e2fsprogs in xfstests, because if I'm developing
> e2fsprogs, it actually makes much more sense to put the regression
> tests in the e2fsprogs git tree.

xfstests is for testing the entire suite of filesystem software,
from userspace to kernel code. It is not a "kernel only" regression
test suite - we test all the XFS userspace tools in the xfs subdir, 
including xfsdump/xfsrestore...

> If this is because it's more convenient to put this in xfsprogs
> because it has fsstress, maybe we should adjust the groups that it is
> in so that it's not in auto or quick, but some other group?

Why? If it's testing that something works properly and runs with
hardware that everyone has available, we always want to run it as
part of the auto group. And we already have a "dump" group to
trigger running xfsdump/xfsrestore tests, so it would make sense to
use that as well.

> to some group like "userspace" so I can exclude it when I'm mostly
> interested in testing development kernels?

Which means you have no idea whether the kernel changes you are
testing breaks userspace tools or vice versa?

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com