From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>
cc: Jacob Champion <pchampion@vmware.com>,
        "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>,
        Daniel Gustafsson <daniel@yesql.se>
Subject: Re: Preventing abort() and exit() calls in libpq
In-reply-to: <340320.1624912475@sss.pgh.pa.us>
References: <3128896.1624742969@sss.pgh.pa.us>
 <340320.1624912475@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Mon, 28 Jun 2021 16:34:35 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <451277.1624986930.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Tue, 29 Jun 2021 13:15:30 -0400
Message-ID: <451278.1624986930@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/451278.1624986930%40sss.pgh.pa.us>
Precedence: bulk

I wrote:
>> This relies on "nm" being able to work on shlibs, which it's not
>> required to by POSIX.  However, it seems to behave as desired even
>> on my oldest dinosaurs.  In any case, if "nm" doesn't work then
>> we'll just not detect such problems on that platform, which should
>> be OK as long as the test does work on common platforms.

So I pushed that, and not very surprisingly, it's run into some
portability problems.  gombessa (recent OpenBSD) reports

! nm -A -g -u libpq.so.5.15 2>/dev/null | grep -v '_eprintf\\.o:' | grep -=
e abort -e exit
libpq.so.5.15:__cxa_atexit

So far as I can find, __cxa_atexit is a C++ support routine, so
I wondered what the heck libpq.so is doing calling it.  I managed
to reproduce the failure here using an OpenBSD installation I had
at hand, and confirmed that __cxa_atexit is *not* referenced by any
of the .o files in src/port, src/common, or src/interfaces/libpq.
So apparently it's being injected at some fairly low level of the
shlib support on that platform.

Probably the thing to do is adjust the grep filter to exclude
__cxa_atexit, but I want to wait awhile and see whether any
other buildfarm animals report this.  morepork at least will
be pretty interesting, since it's a slightly older OpenBSD
vintage.

			regards, tom lane