>From the bugs I linked earlier, just creating a test table with a couple of values that total > 2 billion will show the issue with sum().
Aaron On Mon, 2007-02-19 at 21:31 -0500, Matthew T. O'Connor wrote: > I don't know, and I'm not able to reproduce this anymore since my > mailbox is no longer big enough. If anyone can create a reproduceable > test case, I'll bring it to the PostgreSQL mailing list (I hang out > there alot). > > Matt > > > Aaron Stone wrote: > > Sure is. Big sigh of relief on the table definition. Now what might be > > causing the sum() to be only 4 signed bytes if all of the columns > > involved are 8 bytes!? > > > > Aaron > > > > On Sun, 2007-02-18 at 21:44 -0500, Matthew T. O'Connor wrote: > >> I'm not sure what the problem is, but I don't think that's it. > >> > >> INT8 (or bigint) is an 8 byte integer in PG sql hence 64 bit number, so > >> I don't think that's the problem. > >> > >> > >> Aaron Stone wrote: > >>> PostgreSQL assigns the type of SUM to be the type of the column it is > >>> aggregating over. Physmessage.messagesize is INT8, so there we go: we're > >>> only getting a 32 bit signed value as the sum. > >>> > >>> I'll add an appropriate type cast to the sql fragments in db_get_sql for > >>> dbpgsql.c, a dummy in dbmysql.c, and look up how SQLite handles this > >>> data type issue. > >>> > >>> There are actually five places in db.c where a sum is taken over > >>> physmessage.messagesize, but I think the only negative effect is to > >>> disable quota checks on mailboxes above 2GB. > >>> > >>> Aaron > >>> > >>> On Thu, 2007-02-15 at 23:54 -0800, Aaron Stone wrote: > >>>> http://www.linuxforums.org/forum/linux-programming-scripting/54312-postgresql-sum-overflow-error.html > >>>> > >>>> You're on PostgreSQL, right? I wonder if someone has narrowed this > >>>> report down to either a bug or a subtlty of the column type? > >>>> > >>>> Aaron > >>>> > >>>> On Thu, 2007-02-15 at 13:21 -0500, Matthew T. O'Connor wrote: > >>>>> Unfortunately in the interim I have deleted a lot of messages from my > >>>>> account and now after I run dbmail-util the curmail_size value is again > >>>>> positive. I assume this is because I now have less than 2G in the > >>>>> matthew account. So I do think there is an issue here somewhere, sorry > >>>>> I'm not able to help more at the moment. > >>>>> > >>>>> Matt > >>>>> > >>>>> > >>>>> Paul J Stevens wrote: > >>>>>> Running 'dbmail-util -py' at trace_level=5 should reveal what's going > >>>>>> on > >>>>>> here. > >>>>>> > >>>>>> Aaron Stone wrote: > >>>>>>> On Mon, 2007-02-12 at 00:29 -0500, Matthew T. O'Connor wrote: > >>>>>>>> I was looking at dbmail_users today and noticed this: > >>>>>>>> > >>>>>>>> user_idnr | userid | curmail_size > >>>>>>>> -----------+--------------------------------+-------------- > >>>>>>>> 4 | matthew | -1886517308 > >>>>>>> That's interesting (to say the least :-P). Internally we've got a > >>>>>>> u64_t; > >>>>>>> I wonder if that's just coming into DBMail as a high 64 bit number or > >>>>>>> if > >>>>>>> something is really screwed up. > >>>>>>> > >>>>>>> Aaron > >>>>>>> > >>>>>>> _______________________________________________ > >>>>>>> Dbmail-dev mailing list > >>>>>>> [email protected] > >>>>>>> http://twister.fastxs.net/mailman/listinfo/dbmail-dev > >>>>>>> > >>>>> > >>>>> _______________________________________________ > >>>>> Dbmail-dev mailing list > >>>>> [email protected] > >>>>> http://twister.fastxs.net/mailman/listinfo/dbmail-dev > >>>> _______________________________________________ > >>>> Dbmail-dev mailing list > >>>> [email protected] > >>>> http://twister.fastxs.net/mailman/listinfo/dbmail-dev > >>> _______________________________________________ > >>> Dbmail-dev mailing list > >>> [email protected] > >>> http://twister.fastxs.net/mailman/listinfo/dbmail-dev > >> _______________________________________________ > >> Dbmail-dev mailing list > >> [email protected] > >> http://twister.fastxs.net/mailman/listinfo/dbmail-dev > > > > _______________________________________________ > > Dbmail-dev mailing list > > [email protected] > > http://twister.fastxs.net/mailman/listinfo/dbmail-dev > _______________________________________________ > Dbmail-dev mailing list > [email protected] > http://twister.fastxs.net/mailman/listinfo/dbmail-dev _______________________________________________ Dbmail-dev mailing list [email protected] http://twister.fastxs.net/mailman/listinfo/dbmail-dev
