Thomas Lockhart <lockhart@alumni.caltech.edu> writes:
>> In the meantime, I think the only possible patch is
>> to disable the error check in btreesel and have it return a default
>> selectivity estimate instead of complaining. Drat.
> ... and let's use this solution for the v6.5.x branch, once it comes
> into being.
I've already done that, committed it, and posted the patch on
pgsql-patches. We can reverse out the patch after something's
been done to provide reasonable selectivity estimates for rtrees.
Applying intltsel, as 6.4 did, was so bogus that it's difficult
to argue that the resulting numbers were better than the 0.5
default estimate I just put into btreesel ;-) ... so I feel no
special desire to return to the status quo ante. I have a to-do
list item to look at the whole selectivity estimation business,
and I will try to figure out something reasonable for rtrees
while I'm at it. It may be a while before that gets to the top
of the to-do list (unless someone else gets to it before I do),
but I think this patch will do fine until then.
Mostly I'm embarrassed that we didn't notice the problem during
beta testing :-(. No regression test, and no users of rtrees
in the beta population either, it would seem.
regards, tom lane