Horton

Re: Crash report for some ICU-52 (debian8) COLLATE and work_mem values

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Peter Eisentraut <peter.eisentraut@2ndquadrant.com>
Cc: Daniel Verite <daniel@manitou-mail.org>, Peter Geoghegan <pg@bowt.ie>, PostgreSQL mailing lists <pgsql-bugs@postgresql.org>
Date: 2017-08-07T22:27:21Z
Lists: pgsql-bugs, pgsql-hackers 
Peter Eisentraut <peter.eisentraut@2ndquadrant.com> writes:
> It has never been the case that there is a guarantee that a new
> operating system environment will have the same or more collations
> available as an earlier version.  Even glibc removes or renames locales.

Indeed, and one of the alleged selling points of ICU was greater stability
of collation behaviors (including naming).  I'd like to try to actually
achieve that.

> In this particular case, you can create user-defined collations to fill
> in the missing names if you want.

NO. YOU. CAN'T.  At least not with convenient upgrade methods like
pg_upgrade.  If the collation name isn't produced by the newer version's
initdb, pg_upgrade will fail.

			regards, tom lane

Commits

  1. Reject use of ucol_strcollUTF8() before ICU 53

  2. Rethink behavior of pg_import_system_collations().

  3. Faster expression evaluation and targetlist projection.