pginfo <pginfo(at)t1(dot)unisoftbg(dot)com> writes:
> As I wrote it happens by indexing.
> And all the data are new inserted (after dumping from 7.2 from anoder box
> before).
> I do not belive that it is data corruption.
Hm. Another theory that would fit those facts is that you're getting
bitten by the bug in strcoll() that used to exist in glibc. I think you
said you are using Red Hat somethingorother --- are you up2date on glibc
fixes?
regards, tom lane