Horton

Re: Yet another fast GiST build

Heikki Linnakangas <hlinnaka@gmail.com>

View thread View on postgr.es
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>, Darafei Komяpa 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>
Date: 2021-01-13T18:04:10Z
Lists: pgsql-hackers 

On 13 January 2021 13:53:39 EET, Heikki Linnakangas <hlinnaka@iki.fi> wrote:
>Buildfarm animal thorntail is still not happy:
>
>> --- /home/nm/farm/sparc64_deb10_gcc_64_ubsan/HEAD/pgsql.build/../pgsql/contrib/pageinspect/expected/gist.out	2021-01-13 13:38:09.721752365 +0300
>> +++ /home/nm/farm/sparc64_deb10_gcc_64_ubsan/HEAD/pgsql.build/contrib/pageinspect/results/gist.out	2021-01-13 14:12:21.540046507 +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 
>> ------+-----+------------+-------
>> - 0/1 | 0/0 | 4294967295 | {}
>> +    lsn     | nsn | rightlink  | flags 
>> +------------+-----+------------+-------
>> + 0/1B8357F8 | 0/0 | 4294967295 | {}
>>  (1 row)
>
>Looks like the LSN on the page is not set to GistBuildLSN as expected. 
>Weird.
>
>Thorntail is a sparc64 system, so little-endian, but the other 
>little-endian buildfarm members are not reporting this error. Any idea 
>what might be going on?

Status update on this: I am building Postgres in a qemu sparc64 emulated virtual machine, hoping to be able to reproduce this. It's very slow, so it will take hours still to complete.

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. Or there is something special on that animal that causes the just-created index pages to be dirtied. It does seem to happen consistently on thorntail, but not on other animals.

- Heikki

Commits

  1. Add sortsupport for gist_btree opclasses, for faster index builds.

  2. pageinspect: Fix relcache leak in gist_page_items().

  3. Fix test failure with wal_level=minimal.

  4. Enhance nbtree index tuple deletion.

  5. Fix portability issues in the new gist pageinspect test.

  6. Add functions to 'pageinspect' to inspect GiST indexes.

  7. Fix missing validation for the new GiST sortsupport functions.

  8. Fix compilation warning in xlog.c

  9. Set right-links during sorted GiST index build.

  10. Fix checksum calculation in the new sorting GiST build.

  11. Add support for building GiST index by sorting.