BUG #5801: characters not encoded properly for column names

The following bug has been logged online:

Bug reference:      5801
Logged by:          Marc Cousin
Email address:      cousinmarc@gmail.com
PostgreSQL version: 9.0.2
Operating system:   Windows XP
Description:        characters not encoded properly for column names
Details:

I get a different behaviour between a Linux and a Windows server, when a
user creates an accentuated column name.

All tests below were done with a linux psql client, the console being set on
win1252 charset (so the input character is truly 'é' in win1252)

With the Linux server :
marc=# SET client_encoding TO 'win1252';
SET
marc=# CREATE TABLE test (nom varchar, prénom varchar);
CREATE TABLE


marc=# \d test


          Table "public.test"


 Column |       Type        | Modifiers


--------+-------------------+-----------


 nom    | character varying |


 prénom | character varying |

'prénom' is also displayed correctly if client_encoding and console are
UTF8, so the conversion is good.

With the Windows server :
test=# SET client_encoding TO 'win1252';
SET
test=# CREATE TABLE test (nom varchar, prénom varchar);
CREATE TABLE
test=# \d test
ERROR:  invalid byte sequence for encoding "UTF8": 0xe3a96e
test=# SELECT attname from pg_attribute where attrelid = (select oid from
pg_class where relname = 'test');
ERROR:  invalid byte sequence for encoding "UTF8": 0xe3a96e
test=# select version();
                           version
-------------------------------------------------------------
 PostgreSQL 9.0.2, compiled by Visual C++ build 1500, 32-bit
(1 row)


The main reason that this is a problem is that the table cannot be pg_dumped
anymore because of this.