Date: Sun, 3 Feb 2019 19:12:17 -0800
From: Noah Misch <noah@leadboat.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Andrew Gierth <andrew@tao11.riddles.org.uk>,
	pgsql-hackers@postgresql.org
Subject: Re: Synchronize with imath upstream
Message-ID: <20190204031217.GE63707@rfd.leadboat.com>
References: <20190203035704.GA6226@rfd.leadboat.com>
 <87y36x76r7.fsf@news-spur.riddles.org.uk>
 <20190203062048.GB6226@rfd.leadboat.com>
 <5102.1549207886@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <5102.1549207886@sss.pgh.pa.us>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk

On Sun, Feb 03, 2019 at 10:31:26AM -0500, Tom Lane wrote:
> Noah Misch <noah@leadboat.com> writes:
> > The -Wno-declaration-after-statement approach takes eight lines of code, and
> > the filter-out approach takes one.  On the other hand, using $(filter-out)
> > changes any runs of whitespace to single spaces ("$(filter-out foo,a    b c)"
> > yields "a b c").  We do risk that with CPPFLAGS and LDFLAGS in a few places.
> > I don't want to proliferate that practice, because it changes semantics of
> > CFLAGS containing -DFOO="arbitrary    text".
> 
> I don't particularly buy that argument, because CPPFLAGS is where any -D
> switches ought to be put.  So we've already exposed ourselves to this
> risk, in the unlikely scenario where it's not hypothetical.

The $(filter-out) corruption is unlikely to matter, indeed.  The question is
whether to use eight lines of code to inject -Wno-declaration-after-statement
or one line to remove -Wdeclaration-after-statement using $(filter-out).  I
see negligible drawbacks on either side; both approaches are tolerable.  The
above-described hypothetical problem tips the scale in favor of
-Wno-declaration-after-statement.