Re: [PATCH RFC] fstests: allow running custom hooks

From: Eryu Guan <eguan@linux.alibaba.com>
To: Dave Chinner <david@fromorbit.com>
Cc: Theodore Ts'o <tytso@mit.edu>,
	Damien Le Moal <Damien.LeMoal@wdc.com>,
	Qu Wenruo <quwenruo.btrfs@gmx.com>, Qu Wenruo <wqu@suse.com>,
	"fstests@vger.kernel.org" <fstests@vger.kernel.org>,
	"linux-btrfs@vger.kernel.org" <linux-btrfs@vger.kernel.org>
Subject: Re: [PATCH RFC] fstests: allow running custom hooks
Date: Fri, 23 Jul 2021 12:32:23 +0800	[thread overview]
Message-ID: <20210723043223.GK60846@e18g06458.et15sqa> (raw)
In-Reply-To: <20210722222150.GE2112234@dread.disaster.area>

On Fri, Jul 23, 2021 at 08:21:50AM +1000, Dave Chinner wrote:
> On Thu, Jul 22, 2021 at 10:41:29AM -0400, Theodore Ts'o wrote:
> > On Thu, Jul 22, 2021 at 09:28:30AM +1000, Dave Chinner wrote:
> > > 
> > > I'm thinking that it is something relatively simple like this:
> > > 
> > > fstests/tests/hooks
> > > - directory containing library of hook scripts
> > 
> > I'd suggest fstests/common/hooks instead, since the hook scripts
> > aren't actually *tests* per so, but rather utility scripts, and common
> > would be a better place for it, I think.

Agreed, then we could avoid naming hooks dir as "Hooks".

> 
> True, but I don't think common/ is the right place, either, because
> that's for common test infrastructure. I only just looked, but
> there's a lib/ directory in fstests.  lib/hooks seems like the right
> place for this, and if I had of looked yesterday I would have put it
> there from the start. :/
> 
> Is that an acceptible location?

Right now the lib dir contains mainly c libs used by test binaries, it
looks like part of the build infrastructure to me. OTOH, I think
common/hooks is fine, as hooks is part of the test harness to me. Or
maybe tools/hooks is another choice?

Thanks,
Eryu

> 
> > > fstests/hooks/
> > > - directory containing symlinks to hook scripts
> > 
> > This might be a good default, but it might be better if the location
> > of the hook directory could be overridden via an environment variable.
> > In some cases, instead of having run-time configuration inside the
> > fstests directtory with .gitignore, it might be more convenient for it
> > if were made available externally (for example, via a 9p file system
> > in a case where tests are being run via KVM using a rootfs test image
> > with qmeu's snapshot mode so the hook directory could be supplied from
> > the host).
> 
> Yup, that's easy enough to do. We can do it exactly the same way we
> allow RESULT_BASE to point the results to a user defined directory.
> 
> Cheers,
> 
> Dave.
> -- 
> Dave Chinner
> david@fromorbit.com