Title: AW: [HACKERS] Strange left join problems in 8.1
I've tried this with cvs , -r REL8_1_STABLE (hope this was the right one), and works flawless now.
Thanks a lot!
Best regards
Mario Weilguni
-Ursprüngliche Nachricht-
Von: Tom Lane [mailto:[EMAIL PROTECTED]]
Gesendet: Sa
I've quite
interesting results in 8.1, I'm not sure if the queries itself are beyond SQL
specs, but the results are quite interesting:
This is the most
simple I found for the query that still has the problem, the second left join is
not really necessary and can be replaced
Correct
Mario Weilguni [EMAIL PROTECTED] writes:
It seems like the planner is pulling the last where condition into the
second left join, evaluating it in wrong order.
Can you try this on 8.1 branch tip? It sounds suspiciously similar to
an already-fixed issue (see reports from Sebastian Böck around
Can you try this on 8.1 branch tip? It sounds suspiciously similar to
an already-fixed issue (see reports from Sebastian Böck around
mid-November). You haven't provided a complete test case so I really
can't investigate for myself ...
regards, tom lane
Of course I
Mario Weilguni [EMAIL PROTECTED] writes:
Of course I can give it a try, do I need to re-initdb when I use cvs to
checkout 8.1 tip?
No, not if you're on 8.1 final (or anything post-beta4, IIRC). Just be
sure to use the same configure arguments as before (pg_config can help
remind you if you
