Re: Bug #18326

Jack writes: [ some reformatting ] } } Larry writes: } } } } Richard Ridley asks if anyone in the real world has seen the evil done } } by bug 18326. You bet, and its bad news. We recently upgraded a customer } } from 4.1 to 5.0, and were bit by this bug. The way it manifested itself } } in our application may not be typical, however, the way we solved the } } problem may be of interest. } } } } In our application, a query from a certain screen is done using a } } three table join. We select and store the rowid of the rows } } returned from the query, then use it to retrieve the actual record } } when the user does a "Next", "Previous", and so forth. Anyway, this } } query worked fine under 4.1 for about a year, but with 5.0, it broke } } immediately. The bug caused the rowids to be returned in the wrong } } order. So a query where the user was looking for an address on } } "Budd Road" returned a record on "Smith Lane". Talk about confused users! } } [fix/workaround deleted] } } } } We did not get bitten by this one per se, but did hit something simliar. } Join 2 tables - load the results into an array, when the user selects a } row take some action against the table based on the ROWID. Guess what? } The rowid returned from the join statement for the second table was the } same rowid for every row. We had to turn around and use three or four } fields from the row to read the table instead of the rowid. Supposedly } fixed in 5.01 - don't know yet. The ROWID bug that Jack is seeing is Bug #13029. I have OnLine 5.01.UD1 on our HP platform and it does fix that bug. However, I suspect that the bug originally mentioned above (18326) still exists, because we found something very similar even in 5.01. -- Bryan Klopfenstein bryan@carsinfo.com CARS Information Systems Corp. Cincinnati, OH Standard Disclaimers Apply