[asterisk-dev] cdr_odbc.c is broken in trunk

[Nick Gorham]

Tilghman Lesher wrote:

>On Friday 11 January 2008 05:34:44 Nick Gorham wrote:
>  
>
>>Kevin P. Fleming wrote:
>>    
>>
>>>Tilghman Lesher wrote:
>>>      
>>>
>>>>Uh, no, the stack doesn't work that way.  The entire contents of
>>>>timestr[] remain valid and unchanged until the stack frame is popped. 
>>>>Which will not happen until odbc_log() exits.  The only reason this
>>>>location would be overwritten is if there's a stack overflow error (not
>>>>impossible, but we've been fairly diligent in finding those issues).
>>>>        
>>>>
>>>This is incorrect. The variable being referred to here ('timestr' in
>>>prepare_cb) is allocated on the stack when prepare_cb() is entered and
>>>then that stack frame is released when prepare_cb() exits. Since
>>>ast_odbc_prepare_and_execute() calls prepare_cb() and then later calls
>>>SQLExecute(), the memory for this bound parameter will now be used for
>>>something else (probably the stack frame for SQLExecute() itself).
>>>      
>>>
>>Given the hassle about me posting the patch, and given the licence is
>>still pending, I guess that no one looked at my patch yet?
>>    
>>
>
>Your license was still pending at the time of commit, so no, I have not looked
>at it.
>
>  
>
>>The reason I ask, is getting the current trunk I see the code creates
>>the query using
>>
>>            "VALUES ('%s',?,?,?,?,?,?,?,?,?,?,?,?,?)", table, timestr);
>>
>>My version used the ODBC timestamp escape syntax, which I think you will
>>find is a better choice than assuming the ODBC driver in use will
>>understand the format from ast_strftime, using the ODBC escape means
>>that the driver will understand {ts 'YYYY-MM-DD HH:MM:SS' }if it doesn't
>>then at least the spec says it should. I used
>>
>>            "VALUES ({ts '%s'},?,?,?,?,?,?,?,?,?,?,?,?,?)", table,
>>timestr);
>>
>>Not a big deal, but I don't think it hurts to use the standard when
>>there is one avalable.
>>    
>>
>
>Do you have an example of a database driver that does not work with the
>current syntax?
>
>  
>
Off the top of my head the first one that comes to mind is Oracle

Oracle:

SQL> create table ntest ( i integer, d timestamp );
SQLRowCount returns -1

SQL> insert into ntest values( 1, '2001-02-03 04:05:06' );
[S1000][unixODBC][Easysoft][Oracle]ORA-01843: not a valid month at offset 29
[ISQL]ERROR: Could not SQLExecute

SQL> insert into ntest values( 1, '03-FEB-01 04.05.06 AM' )
SQLRowCount returns 1

SQL> insert into ntest values( 1, {ts '2001-02-03 04:05:06' } );
SQLRowCount returns 1

-- 
Nick