Re: VACUUM/t_ctid bug (was Re: GiST concurrency commited)

> Since 7.4 we have troubles with ltree (seldom corruption of buffer cache, not 
> on-disk), might this bug be somehow related to the ltree problem?
> 7.2 was rock-stable with ltree.

Not sure. Fixed bug was (@ - contains operation):

update wow set a = a || '{101}'::int[] where a @ '{1,2,3}';
select a from wow where a @ '{1,2,3}' and not a @ '{101}';

After update query select must not find any rows, but it did. The problem was in 
GiST code and so any GiST idexes was affected.

Can you say more about your trouble?


-- 
Teodor Sigaev                                   E-mail: teodor@sigaev.ru
  WWW: http://www.sigaev.ru/