Date: Wed, 13 Jan 2021 20:04:10 +0200
From: Heikki Linnakangas <hlinnaka@gmail.com>
To: Heikki Linnakangas <hlinnaka@iki.fi>,
 Andrey Borodin <x4mmm@yandex-team.ru>
CC: Peter Geoghegan <pg@bowt.ie>, Tom Lane <tgl@sss.pgh.pa.us>,
 Justin Pryzby <pryzby@telsasoft.com>,
 =?UTF-8?Q?Darafei_Kom=D1=8Fpa_Praliaskouski?= <me@komzpa.net>,
 Pavel Borisov <pashkin.elfe@gmail.com>,
 Thomas Munro <thomas.munro@gmail.com>, Michael Paquier <michael@paquier.xyz>,
 Alexander Korotkov <a.korotkov@postgrespro.ru>,
 PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Subject: Re: Yet another fast GiST build
In-Reply-To: <fdab636c-050d-5615-2080-862e1646f5a3@iki.fi>
References: <7d9158ca-d7cb-88f0-dbbd-1d5cec32b1ae@iki.fi>
 <896113.1600643169@sss.pgh.pa.us>
 <8269ca22-d275-0afe-633c-2342c7ba11e6@iki.fi>
 <e1d41317-7420-b7ba-0d88-19625012e31e@iki.fi>
 <47CD54A6-EE59-433A-96E2-6F4E5973CEF4@yandex-team.ru>
 <422beff9-0e01-9ce6-edb9-c2a2289e8eb9@iki.fi>
 <0A1DDF47-CAB7-4D6A-9677-B1800FFB6984@yandex-team.ru>
 <4D68C21F-9FB9-41DA-B663-FDFC8D143788@yandex-team.ru>
 <4098005f-e4b5-5904-4b33-335dd106c132@iki.fi>
 <059BD785-A6CE-4E34-B80B-38E9E0309B78@yandex-team.ru>
 <CAH2-WzkBwtk0Np0oZvpiywqjrVwZQtDHSjatJ54ofyezLBxzeQ@mail.gmail.com>
 <8ACF61CC-32ED-4B4B-A2BC-15264E02D24E@yandex-team.ru>
 <3E4F9093-A1B5-4DF8-A292-0B48692E3954@yandex-team.ru>
 <ea7d112f-755e-d194-1777-f10ca6f372b7@iki.fi>
 <9FCEC1DC-86FB-4A57-88EF-DD13663B36AF@yandex-team.ru>
 <c0a679f3-fc3f-4ab0-bb18-b848cd6e9aa6@iki.fi>
 <BD9D9676-A2BF-476B-9D74-95E5CC00ACFB@yandex-team.ru>
 <359ab371-5026-15b5-1d89-d729d590d766@iki.fi>
 <fdab636c-050d-5615-2080-862e1646f5a3@iki.fi>
Message-ID: <628ABAF9-FEB6-4EAC-A5F5-750887F26656@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk


On 13 January 2021 13:53:39 EET, Heikki Linnakangas <hlinnaka@iki=2Efi> wr=
ote:
>Buildfarm animal thorntail is still not happy:
>
>> --- /home/nm/farm/sparc64_deb10_gcc_64_ubsan/HEAD/pgsql=2Ebuild/=2E=2E/=
pgsql/contrib/pageinspect/expected/gist=2Eout	2021-01-13 13:38:09=2E7217523=
65 +0300
>> +++ /home/nm/farm/sparc64_deb10_gcc_64_ubsan/HEAD/pgsql=2Ebuild/contrib=
/pageinspect/results/gist=2Eout	2021-01-13 14:12:21=2E540046507 +0300
>> @@ -3,21 +3,21 @@
>>  CREATE INDEX test_gist_idx ON test_gist USING gist (p);
>>  -- Page 0 is the root, the rest are leaf pages
>>  SELECT * FROM gist_page_opaque_info(get_raw_page('test_gist_idx', 0));
>> - lsn | nsn | rightlink  | flags=20
>> ------+-----+------------+-------
>> - 0/1 | 0/0 | 4294967295 | {}
>> +    lsn     | nsn | rightlink  | flags=20
>> +------------+-----+------------+-------
>> + 0/1B8357F8 | 0/0 | 4294967295 | {}
>>  (1 row)
>
>Looks like the LSN on the page is not set to GistBuildLSN as expected=2E=
=20
>Weird=2E
>
>Thorntail is a sparc64 system, so little-endian, but the other=20
>little-endian buildfarm members are not reporting this error=2E Any idea=
=20
>what might be going on?

Status update on this: I am building Postgres in a qemu sparc64 emulated v=
irtual machine, hoping to be able to reproduce this=2E It's very slow, so i=
t will take hours still to complete=2E

I don't think this is a problem with the test, or with the new pageinspect=
 functions, but a genuine bug in the gist building  code=2E Or there is som=
ething special on that animal that causes the just-created index pages to b=
e dirtied=2E It does seem to happen consistently on thorntail, but not on o=
ther animals=2E

- Heikki