CI failure: postgres_fdw_get_connections

From: Robert Haas <robertmhaas(at)gmail(dot)com>
To: "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: CI failure: postgres_fdw_get_connections
Date: 2026-06-11 23:12:14
Message-ID: CA+Tgmoax3cHXHsm9OidN4F-xiu16y8q2W8T5dTNFic1Zoo2cOw@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

I just tried the new GitHub Actions CI for the first time. Thanks to
everybody who worked on making that happen. However, I got a failure,
on Linux-Meson 32 bit only:

# --- /__w/postgresql/postgresql/contrib/postgres_fdw/expected/postgres_fdw.out
2026-06-11 19:31:44.347832846 +0000
# +++ /__w/postgresql/postgresql/build/testrun/postgres_fdw-running/regress/results/postgres_fdw.out
2026-06-11 19:47:06.867590217 +0000
# @@ -12983,8 +12983,7 @@
# FROM postgres_fdw_get_connections(true);
# server_name | closed | remote_backend_pid
# -------------+--------+--------------------
# - loopback | t | t
# -(1 row)
# +(0 rows)

Run is here: https://github.com/robertmhaas/postgresql/actions/runs/27372091232/job/80887169394

Apparently, this comment isn't always correct:

-- is not available. Despite the termination, remote_backend_pid should
-- still show the non-zero PID of the terminated remote backend.

The issue seems to be that for the entry to appear in the output of
postgres_fdw_get_connections, it must not yet have been removed from
ConnectionHash. However, pgfdw_inval_callback can blow away
connections freely if they haven't yet been used in the current
transaction, so invalidation processing at the beginning of any
statement subsequent to the "SELECT 1 FROM ft1 LIMIT 1;" at
postgres_fdw.sql line 4607 can close the connection if a relevant
invalidation message has been received. So I guess maybe there's
enough DDL happening elsewhere concurrently with this test that a
sinval reset is possible?

--
Robert Haas
EDB: http://www.enterprisedb.com

Browse pgsql-hackers by date

  From Date Subject
Next Message Chao Li 2026-06-11 23:20:46 Re: amcheck: fix bug of missing corruption in allequalimage validation
Previous Message Peter Smith 2026-06-11 22:57:33 Re: Redundant/mis-use of _(x) gettext macro?