On 18/10/10 16:45, Martin J. Evans wrote: > On 18/10/10 14:58, Roode, Eric wrote: >> On Monday, October 18, 2010 9:24 AM, Martin J. Evans wrote: >> >>> I never meant to offend you with that statement and I agreed it was an >>> error. >>> >>> You are posting on a public list and presumably wanting feedback/help. >>> I only know what you've told me, nothing else. Perhaps you might think >>> about that before leaping to the conclusion you seem to have. >>> >>> I spent nearly one hour of my Saturday looking into the problem you >>> reported so obviously I took your report seriously. >>> >>> Perhaps some calm is required. >> >> I'm sorry, I didn't mean to fly off the handle. I do appreciate any >> insight you (or anyone else) can give. I'm just frustrated at what >> appears to be a bug in the driver (or possibly DBD::ODBC), and am >> looking for any way to work around it. >> >> My apologies, >> -- Eric >> >> > > I was using the Easysoft SQL Server Driver but I also tried with the MS SQL > Server Drivers: > > sqlsrv32.dll 14/04/2008 2000.85.1132.00 > sqlncli.dll 14/10/2005 2005.90.1399.00 > > and the first echoed your result and the second worked (i.e., execute failed). > > You have a newish driver but it is looking broken to me. Have you tried the > native client driver (see above). > > You could also try calling SQLMoreResults after the execute which appears to > succeed until it returns false but ensure RaiseError is on. See > odbc_more_results in the pod of DBD::ODBC and t/20SqlServer.t for examples. I > only say this is worth trying as normally the final result of a procedure is > not known until all of the procedure has finished and you have multiple > statements in your proc (but it would be at best a workaround - see below). I > tried it myself with the MS SQL Server ODBC Driver above and it did not help. > > I also checked again the ODBC spec and the only SQL_SUCCESS_WITH_INFO returns > are for statements which succeed but may have succeeded for a reason you > might like to know e.g., > > 01003, NULL value eliminated in set function > The prepared statement associated with StatementHandle contained a set > function (such as AVG, MAX, MIN, and so on), but not the COUNT set function, > and NULL argument values were eliminated before the function was applied. > (Function returns SQL_SUCCESS_WITH_INFO.) > > 01004, String data right truncated > etc > > State 23000 you are getting is "Integrity constraint violation" which is an > error but note the text on the end of the error you are getting: > > "[state was 23000 now 01000]" > > 01000 is a general warning. I don't understand why the "now 01000" in the > text of the error. > > So something is a amiss there. Having said that the MS native client reports > the same error but SQLExecute returns SQL_ERROR. > > As far as I am concerned (and I've written ODBC drivers and code to ODBC > Drivers for more years than I care to admit) the condition you have hit is an > error and when SQLError is called an error number, state and text is returned > BUT the call to SQLExecute is returning SQL_SUCCESS_WITH_INTO instead of > SQL_ERROR. I've read your comment from the Microsoft guy but I don't believe > it and in any case I have 1 of their drivers which behaves differently than > the one you've got. > > Martin I looked at this some more and I'm even more sure it is a bug.
The TDS protocol for your situation returns: tds_error - failed to insert null tds_info - statement has been terminated a load of done_procs the return argument of your procedure containing error in add feed I think the driver is seeing the error but then its idea is overidden by the tds_info packet. Later native client drivers (and our driver) see exactly the same but they acknowledge the tds_error as being the final state and don't override it with the tds_info. BTW, the state returned is 23000 so you might be able to look at that (in $sth->state). I thought you might also be able to look at the return of your procedure but that is not possible. You could also perhaps do something with the odbc_err_handler. Martin -- Martin J. Evans Easysoft Limited http://www.easysoft.com
