Horton

Re: 039_end_of_wal: error in "xl_tot_len zero" test

Anton Voloshin <a.voloshin@postgrespro.ru>

View thread View on postgr.es
From: Anton Voloshin <a.voloshin@postgrespro.ru>
To: Tom Lane <tgl@sss.pgh.pa.us>, David Rowley <dgrowleyml@gmail.com>
Cc: Thomas Munro <thomas.munro@gmail.com>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2024-05-13T09:49:54Z
Lists: pgsql-hackers 
On 13/05/2024 00:39, Tom Lane wrote:
> Hm.  It occurs to me that there *is* a system-specific component to
> the amount of WAL emitted during initdb: the number of locales
> that "locale -a" prints translates directly to the number of
> rows inserted into pg_collation. [...]

Yes. Another system-specific circumstance affecting WAL position is the 
name length of the unix user doing initdb. I've seen 039_end_of_wal 
failing consistently under user aaaaaaaa but working fine with aaaa, 
both on the same machine at the same time.

To be more precise, on one particular machine under those particular 
circumstances (including set of locales) it would work for any username 
with length < 8 or >= 16, but would fail for length 8..15 (in bytes, not 
characters, if non-ASCII usernames were used).

-- 
Anton Voloshin
Postgres Professional, The Russian Postgres Company
https://postgrespro.ru

Commits

  1. Stabilize 039_end_of_wal test.

  2. Create syscache entries for pg_extension