Date: Wed, 11 Mar 2026 18:29:47 +0000
From: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Peter Eisentraut <peter@eisentraut.org>,
	Andres Freund <andres@anarazel.de>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: Defend against -ffast-math in meson builds
Message-ID: <abG0m6D8bhSYTT0w@ip-10-97-1-34.eu-west-3.compute.internal>
References: <abFXfKC8zR0Oclon@ip-10-97-1-34.eu-west-3.compute.internal>
 <vwnxbsap65km373ghtqhhvhwnc35suwraegb7z5op4okxyuwzu@kpaumampogpe>
 <a9d82d5f-58c5-4252-960f-f91cb5bc5a99@eisentraut.org>
 <3316.1773242385@sss.pgh.pa.us>
 <abGPko4lPv/Bj6Bt@ip-10-97-1-34.eu-west-3.compute.internal>
 <3583.1773247555@sss.pgh.pa.us>
 <abGsWdCgAhKRJI2e@ip-10-97-1-34.eu-west-3.compute.internal>
 <3757.1773252208@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <3757.1773252208@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/abG0m6D8bhSYTT0w%40ip-10-97-1-34.eu-west-3.compute.internal>
Precedence: bulk

Hi,

On Wed, Mar 11, 2026 at 02:03:28PM -0400, Tom Lane wrote:
> Bertrand Drouvot <bertranddrouvot.pg@gmail.com> writes:
> > But still, I'm not sure the code-level guard is enough for meson. I think we
> > need to put a guard in meson.build for the "oddity" described in [1]:
> 
> > it compiles fine with CFLAGS="-ffast-math" CC="gcc" meson setup meson_build but
> > produces issues during the regression tests.
> 
> > I just had a closer look and it looks like that the reason is that it's being
> > added at link time:
> 
> > $ cat meson_build/build.ninja | grep LINK_ARGS | grep -c "ffast-math"
> > 254
> 
> Oy.  Surely that is a meson bug?  AFAICS most of the effects of
> -ffast-math should occur at compile time.  Perhaps it also results
> in linking a different version of libm, but that's not everything.
> So just transposing it into link flags is not okay.

What I can see is that with CFLAGS="-ffast-math", I get:

$ objdump -t meson_build/src/backend/postgres | grep crtfast
0000000000000000 l    df *ABS*  0000000000000000              crtfastmath.c

but it's empty without it.

Regards,

-- 
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com