Problem converting £ and € symbols

Question

Following conversion of a database to al32utf8 I noticed problems with converted currency symbols. This is a sample test I ran to reproduce:
1. Set the code page to 1252 as this is how the data was inserted in the win1252 database and create test data
sql> prompt set codepage
set codepage
sql> host chcp 1252
Active code page: 1252
sql>
sql> prompt create test data
create test data
sql> create table bc_test
  2  as select '£ 1' col1
  3           ,'£'   col2
  4     from    dual;
sql>
2. create procedures for checking characters:
sql> create test procs:
  2
sql> create or replace procedure check_chars (v1 in varchar2)
  2  as
  3    v2 varchar2(1 char);
  4    n  integer := 0;
  5  begin
  6    dbms_output.put_line('input: "'||v1||'"');
  7    dbms_output.put_line('pos  chr  ascii');
  8    dbms_output.put_line('---  ---  -----');
  9    if v1 is not null then
10      for i in 1..length(v1)
11      loop
12        v2 := substr(v1,i,1);
13        dbms_output.put_line(lpad(i,3)||'   '||v2||'   '||asciistr(v2));
14      end loop;
15    end if;
16  end;
17  /
sql> sho err
No errors.
sql>
sql>
sql> create or replace procedure check_col1
  2  as
  3    v1 varchar2(1000);
  4  begin
  5    select col1
  6    into   v1
  7    from   bc_test
  8    where  rownum = 1;
  9    check_chars (v1);
10  end;
11  /
sql> sho err
No errors.
sql>
sql> create or replace procedure check_col2
  2  as
  3    v1 varchar2(1000);
  4  begin
  5    select col2
  6    into   v1
  7    from   bc_test
  8    where  rownum = 1;
  9    check_chars(v1);
10  end;
11  /
sql> sho err
No errors.
3. Run procedure to check each column, then convert to al32utf8 and recheck:
sql>
sql> prompt check col1
check col1
sql> exec check_col1
input: "£ 1"
pos  chr  ascii
  1   £   \00A3
  2
  3   1   1
sql>
sql> prompt check col2
check col2
sql> exec check_col2
input: "£"
pos  chr  ascii
  1   £   \00A3
sql>
sql> prompt convert col1
convert col1
sql> update bc_test
  2  set    col1 = sys_op_csconv(col1,'AL32UTF8','WE8MSWIN1252');
sql>
sql> prompt check col1
check col1
sql> exec check_col1
input: "£ "
pos  chr  ascii
  1      \00C2
  2   £   \00A3
  3
sql>
sql> prompt convert col2
convert col2
sql> update bc_test
  2  set    col2 = sys_op_csconv(col2,'AL32UTF8','WE8MSWIN1252');
sql>
sql> prompt convert col2
convert col2
sql> exec check_col2
input: ""
pos  chr  ascii
sql>
sql> rollback;
sql>
The 1st column has had two extra characters added and the second has had the £ converted to null
4. Try with code page 65001
sql> prompt change codepage
change codepage
sql> host chcp 65001
Active code page: 65001
sql>
sql> prompt update col1 again
update col1 again
sql> update bc_test
  2  set    col1 = sys_op_csconv(col1,'AL32UTF8','WE8MSWIN1252');
sql>
sql> prompt check col1
check col1
sql> exec check_col1
input: "£ "
pos  chr  ascii
  1   �   \00C2
  2   �   \00A3
  3
sql>
sql> prompt update col2
update col2
sql> update bc_test
  2  set    col2 = sys_op_csconv(col2,'AL32UTF8','WE8MSWIN1252');
sql>
sql> prompt update col2
update col2
sql> exec check_col2
input: ""
pos  chr  ascii
Same result, and matches what the DMU did
I get similar results with Euro symbols, they are converted to multiple multibyte characters
I can fix the data by running a replace, but what is happening? The data in most cases was added by running a script in sql plus, though some could have been via application tools.
BANNER
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production
PL/SQL Release 11.2.0.3.0 - Production
CORE    11.2.0.3.0      Production
TNS for 64-bit Windows: Version 11.2.0.3.0 - Production
NLSRTL Version 11.2.0.3.0 - Production

Answer 1

The behavior from your testcase is expected.
sql> create table bc_test
  2  as select '£ 1' col1
  3           ,'£'   col2
  4     from    dual;
When run in a WE8MSWIN1252 database, this is effectively creating a table with 2 columns of 3 bytes and 1 byte in length.
sql> update bc_test
  2  set    col1 = sys_op_csconv(col1,'AL32UTF8','WE8MSWIN1252');
When you update the column by converting its value from WE8MSWIN1252 to AL32UTF8, the 8-bit currency symbol expands in size which causes the converted value to be truncated to the column limit.
sql> exec check_col1
input: "£ "
pos  chr  ascii
  1      \00C2
  2   £   \00A3
  3
Here you're seeing the truncated conversion result due to the column size limit (3 bytes). The Pound sign is 0xC2A3 in AL32UTF8.
The DMU scan will detect size expansion over limit issues and block the conversion in such cases.