Question about Encoding a Custom Type

Howdy,

Possibly showing my ignorance here, but as I'm working on updating
citext to be locale-aware and to work on 8.3, I've run into this
peculiarity:

try=# \encoding
UTF8
try=# select setting from pg_settings where name = 'lc_collate';
setting
-------------
en_US.UTF-8
(1 row)

try=# create table try (name citext);
try=# insert into try (name) values ('aardvark'), ('AAA');
try=# select name, name = 'aaa' from try;
name | ?column?
----------+----------
aardvark | f
AAA | t
(2 rows)

try=# insert into try (name) values ('aba'), ('ABC'), ('abc');
try=# select name, name = 'aaa' from try;
name | ?column?
----------+----------
aardvark | f
AAA | t
aba | f
ABC | f
abc | f
(5 rows)

try=# insert into try (name) values ('AAAA');
try=# select name, name = 'aaa' from try;
ERROR: invalid byte sequence for encoding "UTF8": 0xf6bd
HINT: This error can also happen if the byte sequence does not match
the encoding expected by the server, which is controlled by
"client_encoding".

I've no idea what could be different about 'AAAA' vs. any other value.
And if I do either of these:

select name, name = 'aaa'::text from try;
select name, name::text = 'aaa' from try;

It just works. I'm mystified.

My casts:

CREATE CAST (citext AS text) WITHOUT FUNCTION AS IMPLICIT;
CREATE CAST (citext AS varchar) WITHOUT FUNCTION AS IMPLICIT;
CREATE CAST (citext AS bpchar) WITHOUT FUNCTION AS IMPLICIT;
CREATE CAST (text AS citext) WITHOUT FUNCTION AS ASSIGNMENT;
CREATE CAST (varchar AS citext) WITHOUT FUNCTION AS ASSIGNMENT;
CREATE CAST (bpchar AS citext) WITHOUT FUNCTION AS ASSIGNMENT;

Question about the code? It's all here (for now):

https://svn.kineticode.com/citext/trunk/

Hrm. Fiddling a bit more, I find that this fails, too:

try=# select citext_smaller( 'aardvark'::citext,
'AARDVARKasdfasdfasdfasdf'::citext );
ERROR: invalid byte sequence for encoding "UTF8": 0xc102
HINT: This error can also happen if the byte sequence does not match
the encoding expected by the server, which is controlled by
"client_encoding".

So I guess that something must be up with citext_smaller(). It's quite
simple, though:

PG_FUNCTION_INFO_V1(citext_smaller);

Datum citext_smaller (PG_FUNCTION_ARGS) {
text * left = PG_GETARG_TEXT_P(0);
text * right = PG_GETARG_TEXT_P(1);
PG_RETURN_TEXT_P( citextcmp( PG_ARGS ) < 0 ? left : right );
}

Context:

https://svn.kineticode.com/citext/trunk/citext.c

Anyone have any idea? Feedback would be *most* appreciated.

Thanks,

David