Paolo Tavalazzi <ptavalazzi@charta.it> writes:
> [ query plans after updating to 7.4.2 ]
Okay, they're certainly a lot closer than before, so I think I was right
that you were getting bitten somehow by the pg_statistic alignment
problem. It seems like there may still be a bug lurking though. The
rows-out estimates are 7 versus 9 for the middle nested-loop join.
That might just be roundoff error, or there may be something else going
on. Also the estimates for the top join are 7 versus 1 rows, which
seems too large a ratio to be explainable as roundoff error.
Could I pester you to send me a pg_dump dump of this database off-list?
I'd like to trace through the planner with a debugger and see exactly
where the results are diverging.
regards, tom lane