Thread: BUG #4180: PANIC while PQExec on Client with differen locale from database

The following bug has been logged online:

Bug reference:      4180
Logged by:          bishop
Email address:      polimax@mail.ru
PostgreSQL version: 8.3.1
Operating system:   Vista Home Premium
Description:        PANIC while PQExec on Client with differen locale from
database
Details:

The code

PQsetClientEncoding(conn,"WIN1251");
res = PQexec(conn, "INSERT INTO \"table_name\" (\"name\") VALUES
('test');");

return such error message (sorry for localised part of message)

PANIC:  ERRORDATA_STACK_SIZE exceeded
сервер неожиданно закрыл соединение
Скорее всего это означает, что сервер
завершил работу со сбоем до или в
процессе выполнения запроса.

Server encoding UTF8, database encoding UTF8 client encoding WIN1251

The previouse version 8.2 work well with this code.

"bishop" <polimax@mail.ru> writes:
> PQsetClientEncoding(conn,"WIN1251");
> res = PQexec(conn, "INSERT INTO \"table_name\" (\"name\") VALUES
> ('test');");

What do you get, exactly, if you do the same command but with
client_encoding = UTF8?  I suppose this must be a problem with
failing to convert the encoding of an error message, but without
knowing what the original error message is, it's hard to investigate.

            regards, tom lane

[ please keep the mailing list cc'd on discussions ]

"bishop.gis" <bishop.gis@gmail.com> writes:
> This is result executing the same command with client_encoding = UTF8.
> ERROR:  =CE=C5=D7=C5=D2=CE=C1=D1 =
> =D0=CF=D3=CC=C5=C4=CF=D7=C1=D4=C5=CC=D8=CE=CF=D3=D4=D8 =C2=C1=CA=D4 =
> =C9=CD=D1 =CB=CF=C4=C9=D2=CF=D7=CB=C9 "UTF8": 0xcdee (on
> English -- invalid UTF-8 byte sequence detected near byte 0xcdee)
> =F0=EF=E4=F3=EB=E1=FA=EB=E1:  This error can also happen if the byte =
> sequence does not match
> the encoding expected by the server, which is controlled by
> "client_encoding".

Hmm ... I don't read Russian, but I'm quite certain that what you sent
isn't in UTF8.  Perhaps it is in KOI8-R?

This lends some weight to the theory I suggested that gettext() is
failing to return UTF8 on Windows.  That could be a red herring, since
it's possible the encoding got changed when you copied the text into
your mail message, but the theory seems to explain the observed facts
so far.

What locale settings are you using --- particularly lc_ctype and
lc_messages?

            regards, tom lane

I got almost the same error:

I dumped the database, which was in Latin9, with version 8.2.6 using
UTF8 as encoding.
Then loaded the database to version 8.3.1.
Now the database was running with UTF8 while on the client still was
iso-8859-15.
As as result postgresql..exe did crash very often.

Then I tried:
$ pg_dump -U postgres -f pxdump-2008-05-19a-Latin9.txt -E='Latin9' px

and got the following message:

pg_dump: SQL command failed
pg_dump: Error message from server: ERROR:  character 0xc2a4 of encoding
"UTF8"
has no equivalent in "LATIN9"
pg_dump: The command was: COPY public.p_behbl_log (lfdnr_log,
lfdnr_behbl, patnr
, datum, zahn_, zahn, behandlung, gebuehrenpos, ltyp, efs_bemerkung,
anz, faktor
, bemerkung, begruendung, material, punktzahl, punktsumme, punktwert,
betrag, wa
ehrung, fallnr, behandler, assistenz, createtime, sitzungskennz,
unfallkennz, no
tfallkennz, digest, digest_seq, abgerechnet, log_timestamp, log_user,
log_reason
) TO stdout;

in the Logfile I found:

2008-05-20 14:49:58 CEST ERROR:  character 0xc2a4 of encoding "UTF8" has
no equivalent in "LATIN9"
2008-05-20 14:49:58 CEST STATEMENT:  COPY public.p_behbl_log (lfdnr_log,
lfdnr_behbl, patnr, datum, zahn_, zahn, behandlung, gebuehrenpos, ltyp,
efs_bemerkung, anz, faktor, bemerkung, begruendung, material, punktzahl,
punktsumme, punktwert, betrag, waehrung, fallnr, behandler, assistenz,
createtime, sitzungskennz, unfallkennz, notfallkennz, digest,
digest_seq, abgerechnet, log_timestamp, log_user, log_reason) TO stdout;
2008-05-20 14:49:58 CEST LOG:  could not receive data from client:
Unknown winsock error 10061
2008-05-20 14:49:58 CEST LOG:  unexpected EOF on client connection

dumping just in utf8 did work.

Obviously, the converison from Latin9  (and other encodings?) to UTF8
did produce some characters which can not be encoded back to the
original encoding.

Regards,
Christoph Becker



Tom Lane schrieb:
> [ please keep the mailing list cc'd on discussions ]
>
> "bishop.gis" <bishop.gis@gmail.com> writes:
>
>> This is result executing the same command with client_encoding = UTF8.
>> ERROR:  =CE=C5=D7=C5=D2=CE=C1=D1 =
>> =D0=CF=D3=CC=C5=C4=CF=D7=C1=D4=C5=CC=D8=CE=CF=D3=D4=D8 =C2=C1=CA=D4 =
>> =C9=CD=D1 =CB=CF=C4=C9=D2=CF=D7=CB=C9 "UTF8": 0xcdee (on
>> English -- invalid UTF-8 byte sequence detected near byte 0xcdee)
>> =F0=EF=E4=F3=EB=E1=FA=EB=E1:  This error can also happen if the byte =
>> sequence does not match
>> the encoding expected by the server, which is controlled by
>> "client_encoding".
>>
>
> Hmm ... I don't read Russian, but I'm quite certain that what you sent
> isn't in UTF8.  Perhaps it is in KOI8-R?
>
> This lends some weight to the theory I suggested that gettext() is
> failing to return UTF8 on Windows.  That could be a red herring, since
> it's possible the encoding got changed when you copied the text into
> your mail message, but the theory seems to explain the observed facts
> so far.
>
> What locale settings are you using --- particularly lc_ctype and
> lc_messages?
>
>             regards, tom lane
>
>