Обсуждение: "Invalid multibyte character for locale" still there

Hi,

I want to report a bug on PG 8.0 (linux : Fedora Core 3)
A LIKE '%langage C%' in the WHERE clause of my SELECT statement gives
"invalid multibyte character for locale" error.
I thought the bug
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=113231 was fixed
and the patch was incorporated in the mainstream.

Details
-------
In initdb.i18n I have
LANG="en_US"
(I tried also LANG="en_US.UTF-8")

rpm -q postgresql gives postgresql-8.0.0-1PGDG (it's the official
postgresql SRPM that rebuilt with default options).

On /etc/sysconfig/i18n I have tried LANG="en_US" and LANG="en_US.UTF-8"
as well.

Real Query
----------
SELECT t0.id_profile, t0.first_name, t0.last_name, t0.email,
t0.experience, t0.study_level, t0.study_type, t0.compet_type1,
t0.compet_type2, t0.compet_type3, t0.empl_type, t0.avail_type,
t0.prof_salary, t0.comments, t1.srcfileid
FROM cv_profiles t0 LEFT JOIN cv_profiles_ocr t1 ON (t0.id_profile =
t1.id_profile)
WHERE (( upper(t1.ocrcontent) LIKE upper('%langage C%') ) ) ORDER BY
t0.last_name, t0.first_name;

Questions
---------
1. What can I do to outcome this issue (known workarounds, etc.)?
2. Will it be adressed on PG 8.0.1 ?

Thanks,
Marian

Marian POPESCU <softexpert@libertysurf.fr> writes:
> I want to report a bug on PG 8.0 (linux : Fedora Core 3)
> A LIKE '%langage C%' in the WHERE clause of my SELECT statement gives
> "invalid multibyte character for locale" error.
> I thought the bug
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=113231 was fixed
> and the patch was incorporated in the mainstream.

That bug was certainly fixed long ago, which means that whatever you
have is some other bug.  Unfortunately, you haven't given nearly enough
information to let anyone else replicate the error.  Please send a *self
contained* test case --- that includes table declarations and test data.
See the bug reporting guidelines in the documentation.

            regards, tom lane

Marian POPESCU <softexpert@libertysurf.fr> writes:
> I put up a testcase at
> http://perso.netpratique.fr/softexpert/pgsql/unicodetest.data.gz

After experimenting, the only case in which I could get that failure was
to deliberately force a database encoding incompatible with the locale,
namely initdb with LC_ALL=en_US (which is implicitly Latin1 on FC3) and
then "createdb -E unicode".  So I think this is pilot error.  You
need to be careful to use a database encoding consistent with the locale
setting.  (In 8.0, initdb should usually get this right if you don't
override it.)

            regards, tom lane