Re: Repeatedly breaking indexes - SOLVED
От | Martin Edlman |
---|---|
Тема | Re: Repeatedly breaking indexes - SOLVED |
Дата | |
Msg-id | Pine.LNX.4.44.0312020758300.7095-100000@worm.fortech.cz обсуждение исходный текст |
Ответ на | Re: Repeatedly breaking indexes - SOLVED (Tom Lane <tgl@sss.pgh.pa.us>) |
Список | pgsql-bugs |
> "ps" is not a reliable guide to the locale settings being used by > Postgres. Maybe it's RedHat related, I don't know. When I had en_US in /etc/sysconfig/i18n and cs_CZ in ~postgres/.bash_profile and when I did (as root) `su - postgres -s /bin/sh -c "echo $LANG"' I saw LANG=en_US (!!!), when I did `su - postgres' and then `echo $LANG' (as postgres) I saw LANG=cs_CZ. So I believe that ps gives correct info and I believe that postmaster was running with en_US locales. It seems that the problem is with `su' which may not run user's .bash* scripts when executing command with -c. > The postmaster will adopt LC_COLLATE and LC_CTYPE from the settings > recorded in pg_control (by initdb) regardless of its environment. That's what I was hoping for, but unfortunately I didn't work this way. I isolated the problem when trying Pg on different HW. I installed RH from install CD and fortunatelly installed it with cs_CZ as default. Then evertyhing worked - the same Pg, the same database. After that I looked at `ps axe' and saw the difference in LANG. > So I'm not convinced that you've correctly identified the problem. > However, it seems possible that part of the issue is misbehavior if > the various LC_xxx settings aren't all alike --- could you dig further > and try to isolate it? I set ALL LC_xxx variables to cs_CZ in postgres.conf but it didn't help. Only setting LANG to cs_CZ in systemwide settings. Regards, Martin -- Martin Edlman Fortech s.r.o, Litomysl Public PGP key: http://edas.visaci.cz/#keys
В списке pgsql-bugs по дате отправления: