Re: posgresql 8.0 RC1 missing schemas

Hi All!

Sorry, if my problem is not a bug, I've searched for a reported issue, but
I haven't found one :(

We have compiled 8.0 RC1 on SLES 7.0 S/390 2.2.16 kernel, gcc version:
2.95.2

We run initdb command without error. It created a DB directory structure,
but didn't create some schemas in template1.
I connect to template1 and

template1=# select * from pg_tables;
schemaname | tablename | tableowner | tablespace |
hasind
--------------------+-------------------------+------------+------------+-------
information_schema | sql_languages | postgres | | f
information_schema | sql_packages | postgres | | f
information_schema | sql_sizing | postgres | | f
information_schema | sql_sizing_profiles | postgres | | f
information_schema | sql_features | postgres | | f
information_schema | sql_implementation_info | postgres | | f
pg_catalog | pg_shadow | postgres | pg_global | t
pg_catalog | pg_statistic | postgres | | t
pg_catalog | pg_inherits | postgres | | t
pg_catalog | pg_index | postgres | | t
pg_catalog | pg_operator | postgres | | t
pg_catalog | pg_opclass | postgres | | t
pg_catalog | pg_am | postgres | | t
pg_catalog | pg_amop | postgres | | t
pg_catalog | pg_amproc | postgres | | t
pg_catalog | pg_language | postgres | | t
pg_catalog | pg_largeobject | postgres | | t
pg_catalog | pg_aggregate | postgres | | t
pg_catalog | pg_trigger | postgres | | t
pg_catalog | pg_listener | postgres | | f
pg_catalog | pg_cast | postgres | | t
pg_catalog | pg_namespace | postgres | | t
pg_catalog | pg_conversion | postgres | | t
pg_catalog | pg_depend | postgres | | t
pg_catalog | pg_attrdef | postgres | | t
pg_catalog | pg_constraint | postgres | | t
pg_catalog | pg_database | postgres | pg_global | t
pg_catalog | pg_description | postgres | | t
pg_catalog | pg_group | postgres | pg_global | t
pg_catalog | pg_proc | postgres | | t
pg_catalog | pg_rewrite | postgres | | t
pg_catalog | pg_type | postgres | | t
pg_catalog | pg_attribute | postgres | | t
pg_catalog | pg_class | postgres | | t
pg_catalog | pg_tablespace | postgres | pg_global | t

but

template1=# select * from pg_catalog.pg_tablespace;
ERROR: relation "pg_catalog.pg_tablespace" does not exist.

What can I do?
Is it a known or general problem?

Zsolt Pfiszter
DS System Administrator

ISC-Hungária Informatika Ltd.
33-35, Jász utca
Budapest H-1135
Hungary

Phone:(36 1) 412 4864 Mobile: (36) 20 8234864
Fax: (36 1) 412 4800
e-mail(IBM): <mailto:zsolt(dot)pfiszter(at)hu(dot)ibm(dot)com>
web: <http://www.isc-hungaria.hu/>

Zsolt Pfiszter <zsolt(dot)pfiszter(at)hu(dot)ibm(dot)com> writes:
> We have compiled 8.0 RC1 on SLES 7.0 S/390 2.2.16 kernel, gcc version:
> 2.95.2
> We run initdb command without error. It created a DB directory structure,
> but didn't create some schemas in template1.

Hmm, somebody reported something similar a couple days ago, but the
report was sufficiently confused that I kinda thought it was pilot error
of some sort. No, we don't know what's going on ...

> template1=# select * from pg_catalog.pg_tablespace;
> ERROR: relation "pg_catalog.pg_tablespace" does not exist.

Does it work if you just say "select * from pg_tablespace"? What do you
get from "select * from pg_namespace"?

regards, tom lane

On Fri, Dec 17, 2004 at 12:06:01PM -0500, Tom Lane wrote:
> Zsolt Pfiszter <zsolt(dot)pfiszter(at)hu(dot)ibm(dot)com> writes:
> > We have compiled 8.0 RC1 on SLES 7.0 S/390 2.2.16 kernel, gcc version:
> > 2.95.2
> > We run initdb command without error. It created a DB directory structure,
> > but didn't create some schemas in template1.
>
> Hmm, somebody reported something similar a couple days ago, but the
> report was sufficiently confused that I kinda thought it was pilot error
> of some sort. No, we don't know what's going on ...

Both people reporting the problem appear to be in Hungary. Coincidence?

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

Michael Fuhr <mike(at)fuhr(dot)org> writes:
> On Fri, Dec 17, 2004 at 12:06:01PM -0500, Tom Lane wrote:
>> Zsolt Pfiszter <zsolt(dot)pfiszter(at)hu(dot)ibm(dot)com> writes:
>>> We have compiled 8.0 RC1 on SLES 7.0 S/390 2.2.16 kernel, gcc version:
>>> 2.95.2
>>> We run initdb command without error. It created a DB directory structure,
>>> but didn't create some schemas in template1.
>>
>> Hmm, somebody reported something similar a couple days ago, but the
>> report was sufficiently confused that I kinda thought it was pilot error
>> of some sort. No, we don't know what's going on ...

> Both people reporting the problem appear to be in Hungary. Coincidence?

My, that's an interesting point. Zsolt, what locale are you running
initdb under? Does the problem go away if you do it in C locale?

regards, tom lane

Dear Tom, and Michael!

Thanks in advance your help!

I tried to initdb again, but problem isn't ok.

our enviroment settings:

postgres# locale
LANG=hungarian
LC_CTYPE="hungarian"
LC_NUMERIC="hungarian"
LC_TIME="hungarian"
LC_COLLATE=POSIX
LC_MONETARY="hungarian"
LC_MESSAGES="hungarian"
LC_ALL=

template1=# select * from pg_namespace;
nspname | nspowner | nspacl
--------------------+----------+-------------------------------------
pg_catalog | 1 | {postgres=UC/postgres,=U/postgres}
public | 1 | {postgres=UC/postgres,=UC/postgres}
information_schema | 1 | {postgres=UC/postgres,=U/postgres}

I think pg_toad is missing ?

If I generate schemas on my PC and copy to our S390, it will OK? ...
Workaround ...

Zsolt Pfiszter

Tom Lane
<tgl(at)sss(dot)pgh(dot)pa(dot)u
s> To
Michael Fuhr <mike(at)fuhr(dot)org>
2004.12.18 06:00 cc
Zsolt Pfiszter/Hungary/IBM(at)IBMHU,
pgsql-bugs(at)postgresql(dot)org
Subject
Re: [BUGS] posgresql 8.0 RC1
missing schemas

Michael Fuhr <mike(at)fuhr(dot)org> writes:
> On Fri, Dec 17, 2004 at 12:06:01PM -0500, Tom Lane wrote:
>> Zsolt Pfiszter <zsolt(dot)pfiszter(at)hu(dot)ibm(dot)com> writes:
>>> We have compiled 8.0 RC1 on SLES 7.0 S/390 2.2.16 kernel, gcc version:
>>> 2.95.2
>>> We run initdb command without error. It created a DB directory
structure,
>>> but didn't create some schemas in template1.
>>
>> Hmm, somebody reported something similar a couple days ago, but the
>> report was sufficiently confused that I kinda thought it was pilot error
>> of some sort. No, we don't know what's going on ...

> Both people reporting the problem appear to be in Hungary. Coincidence?

My, that's an interesting point. Zsolt, what locale are you running
initdb under? Does the problem go away if you do it in C locale?

regards, tom lane

On Sat, Dec 18, 2004 at 10:34:31AM +0100, Zsolt Pfiszter wrote:
>
> I tried to initdb again, but problem isn't ok.
>
> our enviroment settings:
>
> postgres# locale
> LANG=hungarian
> LC_CTYPE="hungarian"
> LC_NUMERIC="hungarian"
> LC_TIME="hungarian"
> LC_COLLATE=POSIX
> LC_MONETARY="hungarian"
> LC_MESSAGES="hungarian"
> LC_ALL=

What happens if you initdb in the C locale? We're wondering if the
Hungarian stuff is causing a problem.

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

Michael Fuhr <mike(at)fuhr(dot)org> writes:
> On Sat, Dec 18, 2004 at 10:34:31AM +0100, Zsolt Pfiszter wrote:
>>> I tried to initdb again, but problem isn't ok.

> What happens if you initdb in the C locale? We're wondering if the
> Hungarian stuff is causing a problem.

Also, would you send me (off-list) the contents of your postgres.bki
file from the installation share/ directory? It's hard to see how
pg_catalog and public would get in but not pg_toast, when all three
are supposed to be loaded from that file ...

regards, tom lane

Zsolt Pfiszter <zsolt(dot)pfiszter(at)hu(dot)ibm(dot)com> writes:
> our enviroment settings:

> postgres# locale
> LANG=hungarian
> LC_CTYPE="hungarian"
> LC_NUMERIC="hungarian"
> LC_TIME="hungarian"
> LC_COLLATE=POSIX
> LC_MONETARY="hungarian"
> LC_MESSAGES="hungarian"
> LC_ALL=

> template1=# select * from pg_namespace;
> nspname | nspowner | nspacl
> --------------------+----------+-------------------------------------
> pg_catalog | 1 | {postgres=UC/postgres,=U/postgres}
> public | 1 | {postgres=UC/postgres,=UC/postgres}
> information_schema | 1 | {postgres=UC/postgres,=U/postgres}

> I think pg_toast is missing ?

It sure is, and I'd expect pg_temp_1 to exist as well after initdb.
Weird.

I tried to reproduce this on a machine I have access to, namely an S390
at Red Hat (running RHEL 3AS Update 3). Everything works just fine
so far as I can tell with those locale settings. I even cheated and
said "initdb -E unicode" which isn't compatible with this locale ---
initdb gave me a stern warning, and I'm pretty sure sorting of non-ASCII
data wouldn't work well, but it got through initdb and those schemas
were there.

I'm guessing that the problem is specific to the SLES 7 environment.
That release is a bit old isn't it? We might be looking at some weird
compiler bug, or perhaps a problem with the hungarian locale definition.

regards, tom lane