Re: Transaction, Rollback and Database Corruption question,

Joe Kislo <kislo@athenium.com> writes:
>>>> FATAL 1:  my bits moved right off the end of the world!
>>>> Recreate index pg_attribute_relid_attnam_index.
>>
>> We've heard sporadic reports of this, but I've never been able to
>> reproduce it myself.  If you can submit a reproducible test script,
>> it'd be a big help.

>     Oh.. it's repeatable :)  I will upgrade to 7.0.3 tomorrow at work, and
> see if it still happens.. I had been assuming it had something to do
> with the same problem described at the top (so it may be fixed by
> 7.0.3).

Probably it's not fixed.  My suspicion is that there's some corner-case
bug lurking in the btree index routines.  With a reproducible test case,
we could find and squash it... but I've been over that code with a
flinty eye, and I haven't found it yet.

            regards, tom lane