"Gaetano Mendola" <mendola@bigfoot.com> writes:
> I had one row duplicated with the same login and the same id_user,
> was failing was the update of that row complaining about the duplicated
> key.
Oh. Your report was quite unclear; I thought you were saying that
REINDEX had somehow built two copies of the same index.
Is the row actually duplicated, or has it just managed to acquire two
pointers to itself in the indexes? You could check by seeing whether
the two apparent instances have the same or different ctid values
(ctid is a system column not shown unless you ask for it, like oid).
If they are different ctids, it would be useful to see whether they have
the same or different oid,xmin,cmin,xmax,cmax (more system columns).
regards, tom lane