From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
cc: Michael Paquier <michael@paquier.xyz>,
        Alvaro Herrera <alvherre@2ndquadrant.com>,
        Thomas Munro <thomas.munro@enterprisedb.com>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Allowing printf("%m") only where it actually works
In-reply-to: <20180926185734.sdfewkgpheoktae5@alap3.anarazel.de>
References: <20180925154829.rbjxmmcpd5z7ueyq@alvherre.pgsql>
 <28378.1537891542@sss.pgh.pa.us> <20180925224223.GB1659@paquier.xyz>
 <3034.1537974599@sss.pgh.pa.us>
 <20180926185734.sdfewkgpheoktae5@alap3.anarazel.de>
Comments: In-reply-to Andres Freund <andres@anarazel.de>
	message dated "Wed, 26 Sep 2018 11:57:34 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <10876.1537993279.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Wed, 26 Sep 2018 16:21:19 -0400
Message-ID: <10877.1537993279@sss.pgh.pa.us>
Precedence: bulk

Andres Freund <andres@anarazel.de> writes:
> The strerror push, I assume it's that at least, broke something on icc:
> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=3Dfulmar&dt=3D20=
18-09-26%2018%3A00%3A16

Yeah.  It looks like the problem is that configure's test for strerror_r's
return type does not work on icc:

onfigure:10784: checking whether strerror_r returns int
configure:10805: icc -std=3Dgnu99 -c  -mp1 -fno-strict-aliasing -g -O2 -pt=
hread -D_REENTRANT -D_THREAD_SAFE -D_POSIX_PTHREAD_SEMANTICS  -D_GNU_SOURC=
E -I/usr/include/libxml2  conftest.c >&5
conftest.c(45): warning #159: declaration is incompatible with previous "s=
trerror_r" (declared at line 438 of "/usr/include/string.h")
  int strerror_r(int, char *, size_t);
      ^

configure:10805: $? =3D 0
configure:10812: result: yes

Configure is expecting this to throw a hard error (if the platform
declares strerror_r to return char*) but icc only makes it a warning.

What I am not quite figuring out here is how the code seemed to work
before.  Before this patch, it only mattered in libpq because that was the
only place using pqStrerror.  Maybe the regression tests don't ever
produce a strerror output from libpq?

			regards, tom lane