Horton

longfin and tamandua aren't too happy but I'm not sure why

Robert Haas <robertmhaas@gmail.com>

View thread View on postgr.es
From: Robert Haas <robertmhaas@gmail.com>
To: "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>, Dilip Kumar <dilipbalaut@gmail.com>
Date: 2022-09-27T18:55:18Z
Lists: pgsql-hackers 
Hi,

longfin and tamandua recently begun failing like this, quite possibly
as a result of 05d4cbf9b6ba708858984b01ca0fc56d59d4ec7c:

+++ regress check in contrib/test_decoding +++
test ddl                          ... FAILED (test process exited with
exit code 2)     3276 ms
(all other tests in this suite also fail, probably because the server
crashed here)

The server logs look like this:

2022-09-27 13:51:08.652 EDT [37090:4] LOG:  server process (PID 37105)
was terminated by signal 4: Illegal instruction: 4
2022-09-27 13:51:08.652 EDT [37090:5] DETAIL:  Failed process was
running: SELECT data FROM
pg_logical_slot_get_changes('regression_slot', NULL, NULL,
'include-xids', '0', 'skip-empty-xacts', '1');

Both animals are running with -fsanitize=alignment and it's not
difficult to believe that the commit mentioned above could have
introduced an alignment problem where we didn't have one before, but
without a stack backtrace I don't know how to track it down. I tried
running those tests locally with -fsanitize=alignment and they passed.

Any ideas on how to track this down?

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

Commits

  1. meson: Split 'main' suite into 'regress' and 'isolation'

  2. ci: Add 32bit build and test

  3. Fix InitializeRelfilenumberMap for 05d4cbf9b6ba708858984b01ca0fc56d59d4ec7c

  4. Fix alignment problems with SharedInvalSmgrMsg.

  5. In BufTagGetForkNum, cast to the correct type.

  6. Increase width of RelFileNumbers from 32 bits to 56 bits.

  7. Harden pg_filenode_relation test against concurrent DROP TABLE.