Date: Fri, 1 Sep 2017 07:32:07 +1000
From: Dave Chinner <david@fromorbit.com>
To: Brian Foster <bfoster@redhat.com>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>,
        Stephen Rothwell <sfr@canb.auug.org.au>, linux-xfs@vger.kernel.org,
        Linux-Next Mailing List <linux-next@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: linux-next: build warning after merge of the xfs tree
Message-ID: <20170831213207.GP10621@dastard>
References: <20170831100703.16eeb69f@canb.auug.org.au>
 <20170831103040.GB21508@bfoster.bfoster>
 <20170831145752.GH3775@magnolia>
 <20170831152220.GF21939@bfoster.bfoster>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20170831152220.GF21939@bfoster.bfoster>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2855
Lines: 74

On Thu, Aug 31, 2017 at 11:22:20AM -0400, Brian Foster wrote:
> On Thu, Aug 31, 2017 at 07:57:52AM -0700, Darrick J. Wong wrote:
> > On Thu, Aug 31, 2017 at 06:30:41AM -0400, Brian Foster wrote:
> > > On Thu, Aug 31, 2017 at 10:07:03AM +1000, Stephen Rothwell wrote:
> > > > Hi all,
> > > > 
> > > > After merging the xfs tree, today's linux-next build (powerpc
> > > > ppc64_defconfig) produced this warning:
> > > > 
> > > > fs/xfs/xfs_buf_item.c: In function 'xfs_buf_item_unlock':
> > > > fs/xfs/xfs_buf_item.c:573:9: warning: unused variable 'ordered' [-Wunused-variable]
> > > >   bool   ordered = !!(bip->bli_flags & XFS_BLI_ORDERED);
> > > >          ^
> > > > 
> > > > Introduced by commit
> > > > 
> > > >   a097077ef708 ("xfs: remove unnecessary dirty bli format check for ordered bufs")
> > > > 
> > > 
> > > Ugh, this is due to the refactoring of this patch between v1 and v2. I
> > > specifically recall testing for this in v1 because I added the ordered
> > > bool purely to clean up the ASSERT(), then I apparently lost of track of
> > > it for v2.
> > > 
> > > Anyways.. Christoph, Darrick, preferences to clean this up..? I have no
> > > preference between the v1 or v2 factoring. Or if it's easier, we could
> > > always just drop something like the hunk below on top. Thoughts?
> > > 
> > > Brian
> > > 
> > > --- 8< ---
> > > 
> > > diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
> > > index ef2c137..f5d25f5 100644
> > > --- a/fs/xfs/xfs_buf_item.c
> > > +++ b/fs/xfs/xfs_buf_item.c
> > > @@ -567,10 +567,15 @@ xfs_buf_item_unlock(
> > >  {
> > >  	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
> > >  	struct xfs_buf		*bp = bip->bli_buf;
> > > -	bool			aborted = !!(lip->li_flags & XFS_LI_ABORTED);
> > > -	bool			hold = !!(bip->bli_flags & XFS_BLI_HOLD);
> > > -	bool			dirty = !!(bip->bli_flags & XFS_BLI_DIRTY);
> > > -	bool			ordered = !!(bip->bli_flags & XFS_BLI_ORDERED);
> > > +	bool			aborted;
> > > +	bool			hold;
> > > +	bool			dirty;
> > > +	bool			ordered;
> > > +
> > > +	aborted = !!(lip->li_flags & XFS_LI_ABORTED);
> > > +	hold = !!(bip->bli_flags & XFS_BLI_HOLD);
> > > +	dirty = !!(bip->bli_flags & XFS_BLI_DIRTY);
> > > +	ordered = !!(bip->bli_flags & XFS_BLI_ORDERED);
> > 
> > The trouble is, 'ordered' is still an unused variable on !DEBUG builds,
> > since the only user of ordered is that ASSERT.  So either we #ifdef
> > DEBUG the variable out of existence or employ one of those silly
> > 'ordered = ordered' constructions to shut up gcc, if that even still
> > works.
> > 
> 
> The warning goes away for me if we separate the initialization of
> ordered from the declaration. Do you observe otherwise?

Various versions of gcc will throw set-but-unused warnings on
this. Just #define it away or factor the debug code into another
function.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com