Обсуждение: reproducible bug in I don't know what component

bug=# select * from example_objects where name = 'Модемы';
 object_id |  name
-----------+--------
         2 | Мебель
         2 | Модемы
(записей: 2)
bug=# select version();
                                                             version
         

---------------------------------------------------------------------------------------------------------------------------------
 PostgreSQL 7.4.2 on i386-redhat-linux-gnu, compiled by GCC i386-redhat-linux-gcc (GCC) 3.3.3 20040216 (Red Hat Linux
3.3.3-2.1)
(1 запись)

Do the following in an installation initdb'd in ru_RU.KOI8-R (It doesn't
happen if you initdb'd with UTF-8). You need to run psql in a locale
that is capable of russian letters, namely an UTF-8 locale, or a KOI8-R
locale. Then:

CREATE DATABASE bug WITH ENCODING='unicode';
\c bug
\i dump.sql
-- here you have to set client_encoding if you chose ru_RU.KOI8-R as the
locale for psql
-- set client_encoding to koi8r;
select * from example_objects where name = 'Модемы';

dump.sql is attached, the select statement is included in UTF-8.

Let me know if anything is missing.

--
Markus Bertheau <twanger@bluetwanger.de>

Am Freitag, 23. Juli 2004 11:49 schrieb Markus Bertheau:
> Do the following in an installation initdb'd in ru_RU.KOI8-R (It doesn't
> happen if you initdb'd with UTF-8). You need to run psql in a locale
> that is capable of russian letters, namely an UTF-8 locale, or a KOI8-R
> locale. Then:
>
> CREATE DATABASE bug WITH ENCODING='unicode';

That's your problem.  Your locale doesn't match your encoding.  You need to
use a compatible combination.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

=D0=92 =D0=9F=D1=82=D0=BD, 23.07.2004, =D0=B2 14:02, Peter Eisentraut =D0=
=BF=D0=B8=D1=88=D0=B5=D1=82:
> Am Freitag, 23. Juli 2004 11:49 schrieb Markus Bertheau:
> > Do the following in an installation initdb'd in ru_RU.KOI8-R (It doesn't
> > happen if you initdb'd with UTF-8). You need to run psql in a locale
> > that is capable of russian letters, namely an UTF-8 locale, or a KOI8-R
> > locale. Then:
> >
> > CREATE DATABASE bug WITH ENCODING=3D'unicode';
>=20
> That's your problem.  Your locale doesn't match your encoding.  You need =
to=20
> use a compatible combination.

What is happening in the server that this is required?

--=20
Markus Bertheau <twanger@bluetwanger.de>

Markus Bertheau <twanger@bluetwanger.de> writes:
> Do the following in an installation initdb'd in ru_RU.KOI8-R (It doesn't
> happen if you initdb'd with UTF-8).

If this is a bug, it's a bug in the ru_RU.KOI8-R locale definition.
You can prove that the locale considers the strings equal without
Postgres at all:

[tgl@rh1 tgl]$ cat ru_data
root
root
ŜÅçÅÝÅçÅ£îŒ
ŜÅÅÇÅçÅ¥î‹
[tgl@rh1 tgl]$ sort -u ru_data
root
ŜÅçÅÝÅçÅ£îŒ
ŜÅÅÇÅçÅ¥î‹
[tgl@rh1 tgl]$ LC_ALL=ru_RU.KOI8-R  sort -u ru_data
root
ŜÅçÅÝÅçÅ£îŒ
[tgl@rh1 tgl]$

(The above is on an RHL 8.0 platform.)

            regards, tom lane

Am Freitag, 23. Juli 2004 15:30 schrieb Markus Bertheau:
> > That's your problem.  Your locale doesn't match your encoding.  You need
> > to use a compatible combination.
>
> What is happening in the server that this is required?

When you ask locale-aware functions to compare strings, convert to lower-case,
or what the case may be, these functions expect the strings to have a certain
encoding (after all they just receive a stream of bytes, so they cannot check
the encoding themselves).  So if the function thinks it's comparing two
KOI8-R strings and you are actually passing UTF-8 strings, the results are
going to be close to comparing garbage.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/