Message-ID: <1380209888.17366.68.camel@joe-AO722>
Subject: Re: [PATCH 3/4] SunRPC: Use no_printk() for the null dprintk() and
 dfprintk()
From: Joe Perches <joe@perches.com>
To: David Howells <dhowells@redhat.com>
Cc: bfields@fieldses.org, Trond.Myklebust@netapp.com, olof@lixom.net,
        linux-nfs@vger.kernel.org, linux-kernel@vger.kernel.org
Date: Thu, 26 Sep 2013 08:38:08 -0700
In-Reply-To: <30597.1380209743@warthog.procyon.org.uk>
References: <1380209415.17366.66.camel@joe-AO722>
	 <20130926144502.29424.21633.stgit@warthog.procyon.org.uk>
	 <20130926144525.29424.11130.stgit@warthog.procyon.org.uk>
	 <30597.1380209743@warthog.procyon.org.uk>
Content-Type: text/plain; charset="ISO-8859-1"
Mime-Version: 1.0
Sender: linux-nfs-owner@vger.kernel.org

On Thu, 2013-09-26 at 16:35 +0100, David Howells wrote:
> Joe Perches <joe@perches.com> wrote:
> 
> > no_printk doesn't prevent any argument side-effects
> > from being optimized away by the compiler.
> > 
> > ie:
> > 	dprintk("%d", func())
> > func is now always called when before it wasn't.
> 
> Yes, I know.  There are half a dozen places where this is the case.  Those
> I've wrapped in ifdebug(FACILITY) { ... } in the code.  It's not the nicest,
> but at least the compiler always gets to see everything, rather than bits of
> it getting hidden by the preprocessor - which means the call points will be
> less likely to bit rot over time.

No code is eliminated by the preprocessor
with the #define I suggest.