"Vyacheslav Kalinin" <vka@mgcp.com> writes:
> Here is the reproduce code:
After tracing through this I see that the problem is that we don't have
statistics for inheritance trees, and so you're getting a default
estimate for the selectivity of the join condition. I'm not sure why
the similar example I tried last night didn't show a bogus-looking
rowcount estimate, but in any case the lack of stats is the real issue.
This should get fixed sometime, but don't hold your breath ...
regards, tom lane