On 15/10/2010 22:20, Roode, Eric wrote:
Hello all,
DBI docs says that statement handle method execute() returns undef on
error. I have found a situation where that does not hold. Was hoping
someone could help me out, maybe there's a workaround.
First, I'm running ActiveState Perl 5.10.0 on Windows 7, connecting to a
SQL Server 2000 database on a different machine. I have DBI version
1.613, and DBD::ODBC version 1.24.
In a nutshell, here is what happens. I connect to the database, using
the ODBC driver, and setting RaiseError to 1 and PrintError to 0. I
prepare a statement which calls a procedure. I bind parameters to the
statement (some input, some output, various types), I call execute().
There's a problem with one of the parameters; because of that, the
stored procedure tries to insert a null value into a column that does
not accept nulls. So SQL Server gives an error ("Cannot insert the
value NULL into column 'enabled'"), and the DBI system stores this
string in errstr. However, the execute() method returns -1, which is
what it would return on success. Also, no error is thrown (I have the
$sth->execute call wrapped in an eval, and $@ is empty afterward).
You can see the definition of the table and the stored procedure at
http://nopaste.gamedev.pl/?id=8272.
The Perl code that demonstrates the problem is at
http://nopaste.gamedev.pl/?id=8273.
And the DBD trace output is at http://nopaste.gamedev.pl/?id=8274.
I hope someone can shed some light on what's going on here. Should I
check errstr and ignore the return value? Thanks in advance.
Eric
(Tim Bunce, if you see this could you clarify what dbd_st_execute is
supposed to return as I could not find the full details in DBI::DBD. It
would appear DBD::ODBC returns -2 for error, -1 for rowcount not known
and a positive number for rows affected).
The status returned was SQL_SUCCESS_WITH_INFO:
!!dbd_error2(err_rc=1, what=st_execute/SQLExecute,
handles=(3194c98,2b5a848,30d4cd0)
that err_rc=1 which is not an error.
If SQLRowCount returns -1 then dbd_st_execute will return -1, that may
be a bug but really the code should not have got to this point anyway.
The are differences between what DBI documents for the execute method
and what DBD::ODBC returns from the dbd_st_execute and I cannot see for
instance the -2 (for errors) documented in DBI::DBD so I'm not sure if
this is right or wrong (Tim?).
As far as I can see this looks like a bug in your driver (but I'll try
and reproduce here). Surely failing to insert into a column should be an
error not success with info (an example of the latter is say an insert
which worked but truncated your data). What ODBC Driver manager and
driver versions are you using - you can find these from the
Administrative tools, data sources or in windows 7 by searching for data
sources in the control panel top right.
Martin
