"Matthew T. O'Connor" <matthew(at)zeut(dot)net> writes:
> I made a patch to fix this, but in testing it I noticed that the stats
> system doesn't work on shared tables as I was expecting it too (as my
> latest patch requires it too :-). It treats instances of shared tables
> in separate databases as totally unique tables.
Hmm. The bufmgr and lower levels handle shared tables by always
associating a database number of zero with them, but I'll bet that the
pg_stats stuff does not do that. I'd call that a bug, yes. Jan, any
thoughts on how complex to fix?
regards, tom lane