From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mx1.redhat.com ([209.132.183.28]:48936 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934741AbdC3TcG (ORCPT ); Thu, 30 Mar 2017 15:32:06 -0400 From: David Howells In-Reply-To: References: <15267.1490891560@warthog.procyon.org.uk> Subject: Re: [PATCH] xfstests: Add first statx test MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-ID: <23330.1490902318.1@warthog.procyon.org.uk> Date: Thu, 30 Mar 2017 20:31:58 +0100 Message-ID: <23331.1490902318@warthog.procyon.org.uk> Sender: fstests-owner@vger.kernel.org To: Amir Goldstein Cc: dhowells@redhat.com, linux-xfs , Andreas Dilger , Christoph Hellwig , linux-fsdevel , Eric Sandeen , fstests List-ID: Amir Goldstein wrote: > > (2) Optionally compares the timestamps to see that they're sensibly > > ordered with respect to the saved clock time[**] and each other. > > I suggest that instead of comparing to absolute timestamp value > compare to a timestamp of the cmp file. > This will also solve you the problem of gettimeofday() vs. kernel_time > and will also open up the possibility of adding more interesting tests > later on (e.g. make sure that mtime of file A >= atime of file B). The whole point was to compare to a preceding time to make sure the timestamp didn't go backwards - exactly as it can do:-/. Comparing to a reference file doesn't prove that unless you've checked the reference file... Making use of more of a reference file's timestamps could be useful for other checks, though. > > +420 other > > Since your test takes a few seconds, you should add it to auto, quick groups > I don't know what this 'other' group is about. I don't know either. Ask the 'new' script. David