On Tue, 2002-08-13 at 12:22, Tom Lane wrote: > Rod Taylor <[EMAIL PROTECTED]> writes: > > I was playing with this a while back (when I had initially added CASCADE > > to tables). I believe that in the event of a crash the temp tables are > > not removed until their next use. This means that stale *real* items > > may litter the system but the temp table no longer exists in these rare > > occurrences. > > Huh? The view goes away at exactly the same time the temp table does. > If you suffer a backend crash then that may be postponed ... but the > view continues to work up till the instant that it's removed.
After a backend crash the temp tables exist, but are not usable by the current backend as it is different than the one which originally created the temp table (the crash causing a restart and everything). So non-temp items which depend on the no longer usable temp table will be broken until they are scrubbed, which does not happen until the next time a temp table is created. Not that it really matters, but moving a temp-table destruction event into the startup sequence would solve it. ---------------------------(end of broadcast)--------------------------- TIP 6: Have you searched our list archives? http://archives.postgresql.org
