Date: Sat, 15 Jan 2022 00:39:25 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: Maxim Orlov <orlovmg@gmail.com>
Cc: pgsql-hackers@postgresql.org, Pavel Borisov <pashkin.elfe@gmail.com>,
	Alexander Korotkov <aekorotkov@gmail.com>,
	Teodor Sigaev <teodor@sigaev.ru>,
	Nikita Glukhov <n.gluhov@postgrespro.ru>,
	Konstantin Knizhnik <knizhnik@garret.ru>,
	Yura Sokolov <y.sokolov@postgrespro.ru>
Subject: Re: Add 64-bit XIDs into PostgreSQL 15
Message-ID: <20220115063925.GS14051@telsasoft.com>
References: 
 <CACG=ezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe=pyyjVWA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CACG=ezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe=pyyjVWA@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20220115063925.GS14051%40telsasoft.com>
Precedence: bulk

I tried to pg_upgrade from a v13 instance like:
time make check -C src/bin/pg_upgrade oldsrc=`pwd`/13 oldbindir=`pwd`/13/tmp_install/usr/local/pgsql/bin

I had compiled and installed v13 into `pwd`/13.

First, test.sh failed, because of an option in initdb which doesn't exist in
the old version: -x 21000000000

I patched test.sh so the option is used only the "new" version.

The tab_core_types table has an XID column, so pg_upgrade --check complains and
refuses to run.  If I drop it, then pg_upgrade runs, but then fails like this:

|Files /home/pryzbyj/src/postgres/src/bin/pg_upgrade/tmp_check/new_xids.txt and /home/pryzbyj/src/postgres/src/bin/pg_upgrade/tmp_check/old_xids.txt differ
|See /home/pryzbyj/src/postgres/src/bin/pg_upgrade/tmp_check/xids.diff
|
|--- /home/pryzbyj/src/postgres/src/bin/pg_upgrade/tmp_check/new_xids.txt        2022-01-15 00:14:23.035294414 -0600
|+++ /home/pryzbyj/src/postgres/src/bin/pg_upgrade/tmp_check/old_xids.txt        2022-01-15 00:13:59.634945012 -0600
|@@ -1,5 +1,5 @@
|  relfrozenxid | relminmxid 
| --------------+------------
|-            3 |          3
|+        15594 |          3
| (1 row)

Also, the patch needs to be rebased over Peter's vacuum changes.

Here's the changes I used for my test:

diff --git a/src/bin/pg_upgrade/test.sh b/src/bin/pg_upgrade/test.sh
index c6361e3c085..5eae42192b6 100644
--- a/src/bin/pg_upgrade/test.sh
+++ b/src/bin/pg_upgrade/test.sh
@@ -24,7 +24,8 @@ standard_initdb() {
 	# without increasing test runtime, run these tests with a custom setting.
 	# Also, specify "-A trust" explicitly to suppress initdb's warning.
 	# --allow-group-access and --wal-segsize have been added in v11.
-	"$1" -N --wal-segsize 1 --allow-group-access -A trust -x 21000000000
+	"$@" -N --wal-segsize 1 --allow-group-access -A trust
+
 	if [ -n "$TEMP_CONFIG" -a -r "$TEMP_CONFIG" ]
 	then
 		cat "$TEMP_CONFIG" >> "$PGDATA/postgresql.conf"
@@ -237,7 +238,7 @@ fi
 
 PGDATA="$BASE_PGDATA"
 
-standard_initdb 'initdb'
+standard_initdb 'initdb' -x 21000000000
 
 pg_upgrade $PG_UPGRADE_OPTS --no-sync -d "${PGDATA}.old" -D "$PGDATA" -b "$oldbindir" -p "$PGPORT" -P "$PGPORT"
 
diff --git a/src/bin/pg_upgrade/upgrade_adapt.sql b/src/bin/pg_upgrade/upgrade_adapt.sql
index 27c4c7fd011..c5ce8bc95b2 100644
--- a/src/bin/pg_upgrade/upgrade_adapt.sql
+++ b/src/bin/pg_upgrade/upgrade_adapt.sql
@@ -89,3 +89,5 @@ DROP OPERATOR public.#%# (pg_catalog.int8, NONE);
 DROP OPERATOR public.!=- (pg_catalog.int8, NONE);
 DROP OPERATOR public.#@%# (pg_catalog.int8, NONE);
 \endif
+
+DROP TABLE IF EXISTS tab_core_types;