I wrote:
> If we remove the RECHECK printout then this will stop being true ---
> and worse, it will break silently and in a way that will only cause
> you to get occasional wrong answers from your queries.
But actually, that only says it's a bad idea to drop RECHECK from the
pg_dump output. It doesn't say that an 8.4 server has to throw error
when it sees RECHECK. The way things are set up now, ignoring it seems
fairly failsafe. The default assumption in 8.4 is that operators are
lossy, so the worst case consequence is that you waste cycles on
unnecessary rechecks. You don't get wrong answers, even if you
neglected to update your opclass support code.
Off to make a proposal to pghackers ...
regards, tom lane