From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>,
        Thomas Munro <thomas.munro@gmail.com>
cc: pgsql-hackers@lists.postgresql.org
Subject: gokiburi versus the back branches
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2930302.1675726070.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 06 Feb 2023 18:27:50 -0500
Message-ID: <2930303.1675726070@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/2930303.1675726070%40sss.pgh.pa.us>
Precedence: bulk

I notice that Michael's new BF animal gokiburi is failing in
all the pre-v15 branches, though it's fine in v15 and HEAD.
It's evidently dying from ASLR effects because it's trying
to build with EXEC_BACKEND on Linux: there's lots of

2023-02-06 06:07:02.131 GMT [1503972] FATAL:  could not reattach to shared=
 memory (key=3D813803, addr=3D0xffff8c3a5000): Invalid argument
2023-02-06 06:07:02.131 GMT [1503971] FATAL:  could not reattach to shared=
 memory (key=3D813803, addr=3D0xffff8c3a5000): Invalid argument
2023-02-06 06:07:02.132 GMT [1503976] FATAL:  could not reattach to shared=
 memory (key=3D813803, addr=3D0xffff8c3a5000): Invalid argument

in its logs.

The reason it's okay in v15 and up is presumably this:

Author: Thomas Munro <tmunro@postgresql.org>
Branch: master Release: REL_15_BR [f3e78069d] 2022-01-11 00:04:33 +1300

    Make EXEC_BACKEND more convenient on Linux and FreeBSD.
    =

    Try to disable ASLR when building in EXEC_BACKEND mode, to avoid rando=
m
    memory mapping failures while testing.  For developer use only, no
    effect on regular builds.

Is it time to back-patch that commit?  The alternative would be
to run the animal with an ASLR-disabling environment variable.
On the whole I think testing that f3e78069d works is more
useful than working around lack of it.

			regards, tom lane