Hi Federico,

Could you please help me understand how your example shows that there is no SQLSTATE for 1758? From what I see, it's there all right:

MariaDB [test]> \W
Show warnings enabled.
MariaDB [test]> GET DIAGNOSTICS CONDITION 0 @a = MYSQL_ERRNO ;
Query OK, 0 rows affected, 1 warning (0.00 sec)

Error (Code 1758): Invalid condition number
MariaDB [test]> GET DIAGNOSTICS CONDITION 1 @a = MYSQL_ERRNO, @b = RETURNED_SQLSTATE ;
Query OK, 0 rows affected (0.00 sec)

MariaDB [test]> SELECT @a, @b;
+------+-------+
| @a   | @b    |
+------+-------+
| 1758 | 35000 |
+------+-------+
1 row in set (0.00 sec)

Reporting it as a warning rather than an error is apparently intentional, at least it's documented (http://dev.mysql.com/doc/refman/5.6/en/get-diagnostics.html):
"A warning occurs if the condition number is not in the range from 1 to the number of condition areas that have information. In this case, the warning is added to the diagnostics area without clearing it."

The fact that it's not handled by a handler is another story. I'd think it's not because of the SQLSTATE, it looks like the warning somehow gets lost (suppressed?) if it happens inside a procedure:

MariaDB [test]> drop procedure if exists p;
Query OK, 0 rows affected (0.01 sec)

MariaDB [test]> delimiter ||
MariaDB [test]> create procedure p()
    -> begin
    -> GET DIAGNOSTICS CONDITION 0 @a = MYSQL_ERRNO ;
    -> end ||
Query OK, 0 rows affected (0.01 sec)

MariaDB [test]> delimiter ;
MariaDB [test]> 
MariaDB [test]> call p();
Query OK, 0 rows affected (0.01 sec)

MariaDB [test]> 

(no warning caused by the procedure call).

I'm not quite sure if it's a bug or not (haven't found it in the documentation yet), if you think it is, we might try to create a bug report at Oracle and see what they say.

My hypotesys (no SQLSTATE) seems to be wrong, but what I find strange is that:

• The client doesn't display an SQLSTATE (only error code)
• While it's reported as a warning, SHOW WARNINGS tells it's an error (see Level column). (As far as I understand, only '01...' SQLSTATEs indicate a warning?)
• About the HANDLER, you seem to be totally right, but I don't know of other errors which are suppressed inside stored procedures

These facts surprised me, so I wanted to report that (in my opinion) there is something strange. If it isn't a bug, or you think it's totally unimportant, please ignore this report.

Hi Federico,

Sorry for the delay.
Not displaying an SQLSTATE is usual for warnings:

MariaDB [test]> drop table if exists nonexisting;
Query OK, 0 rows affected, 1 warning (0.01 sec)

Note (Code 1051): Unknown table 'test.nonexisting'

The warning of Level error does look weird, I asked for Serg's consult on this. If he reckons it to be a bug, we'll file it at MySQL and see what they say, and will proceed from there.

Filed as http://bugs.mysql.com/bug.php?id=70358

My conclusion that it doesn't have an SQLSTATE was wrong. The problem was that I couldn't handle that error in any way.

Here's the story of how I discovered this bug. I wrote a procedure to copy the diagnostics area into a table. There is no easy way to do this, so I loop through the conditions and used MYSQL_ERRNO, etc, to compose a prepared statement INSERT. In the initial version, I thought I could exit from the loop when this warning occurs (DECLARE HANDLER ... LEAVE `this_loop`), but this never happened.

The workaround was easy. So... I guess it's up to you decide if this is a bug.

Curiously, I've noted that this warning is written in the SQL_ERROR_LOG. But other warnings I tried are not logged.