Description:
------------
I have Oracle 8.1.7 working on linux. Some Oracle params:
NLS_RDBMS_VERSION => 8.1.7.2.0
NLS_CHARACTERSET => CL8ISO8859P5

I have table named utf8_test.
SQL> desc utf8_test;
Name Type         Nullable Default Comments 
---- ------------ -------- ------- -------- 
ID   NUMBER       Y                         
NAME VARCHAR2(10) Y                         

SQL> select * from utf8_test;

        ID NAME
---------- ----------
         1 ????
         2 ????????

The result of fetching data is depended from NLS_LANG.

When I fetched data from this table a result is depended from NLS_LANG variable.

1) When I fetch data with NLS_LANG=RUSSIAN_CIS.CL8MSWIN1251 character set, I receive 2 rows with correct data.

2) When I fetch data with NLS_LANG=RUSSIAN_CIS.UTF8 character set, I receive 2 rows,  with incorrect data. Data in NAME field in 2-nd row is incorrect (FALSE instead of '????????'). But 1-st row is correct.


I think the problem maybe in the folowing:
This field is defined in database as varchar2(10);
Length of the 1-st row NAME in UTF8 character set is 4*2=8 byte. (8<10)
Length of the 2-nd row NAME in UTF8 character set is 8*2=16 byte. (16>10).


Reproduce code:
---------------
<?
// oracle client NLS_LANG=RUSSIAN_CIS.UTF8
$conn = oci_new_connect( 'scott', 'tiger', 'DB' );
$cursor = oci_parse( $conn, 'select id, name from utf8_test' ); // tested with all oci_fetch_* functions
oci_execute( $cursor, OCI_DEFAULT );
oci_fetch_all( $cursor, $result, 0, -1, OCI_FETCHSTATEMENT_BY_ROW );
print "<pre>";
var_dump($result);
?>

Expected result:
----------------
array(2) {
  [0]=>
  array(2) {
    ["ID"]=>
    string(1) "1"
    ["NAME"]=>
    string(8) "????"
  }
  [1]=>
  array(2) {
    ["ID"]=>
    string(1) "2"
    ["NAME"]=>
    string(16) "????????"
  }
}

Actual result:
--------------
array(2) {
  [0]=>
  array(2) {
    ["ID"]=>
    string(1) "1"
    ["NAME"]=>
    string(8) "????"
  }
  [1]=>
  array(2) {
    ["ID"]=>
    string(1) "2"
    ["NAME"]=>
    bool(false)
  }
}