On Thu, May 22, 2008 at 5:27 PM, Dee Ayy <dee.ayy@xxxxxxxxx> wrote: > Has anyone ever seen a partial execution of an UPDATE statement? > > I'm using Perl DBI to update a table in MySQL. > > The strangest thing to me is that a varchar(35) field which holds a custom > date/time format was updated correctly, but other tinyint(1) fields were > not > set correctly, and 1 of 3 text fields which were supposed to add the new > "log" to existing logs in the 3 fields did not appear in the 1 field, but > did in the other 2. > > I found the query in the mysql query log file and some of the text had > newlines. So I thought the query may have failed at this point, BUT when > manually executing the query in phpMyAdmin, it executed fine. I thought > the > whole query would fail due to possibly not escaping some characters. BUT, > like I said before, even this logic is incorrect because some fields were > updated before my manual attempt at testing the query. > > The only similar behavior I've seen is using C/C++ with printf and > sometimes > you need to add a "flush" statement so the text really prints. > > Basically for 5 years, this script has used the following sequence > (prepare, > execute, finish). But there have been recent changes to the script. Only > one record is showing this strange behavior (but it most likely touches > code > that other records do not). > > After the successful manual execution of the statement, the 3 text fields > do > not verify my claim that 1 in 3 had previously failed. BUT the custom > date/time field was confirmed by a few people to have been updated (when > other fields were not), so I may not be going crazy. > > This is a single UPDATE statement. No transactions are involved. > > TIA > If its one record, could you just have corrupted data on that record? Based on your description, perhaps thats all it is. Try cleaning the record and seeing if that fixes the problem. -- Bastien Cat, the other other white meat
