On Tue, 2 Aug 2005, Tom Lane wrote:
> "Marc G. Fournier" <scrappy@postgresql.org> writes:
>> On Tue, 2 Aug 2005, Tom Lane wrote:
>>> Hmm, odd. But maybe there are traces of a SERIAL linkage? What do
>>> you get from
>>>
>>> select * from pg_depend where objid = 'xa_url_id_seq'::regclass;
>
>> # select * from pg_depend where objid = 'xa_url_id_seq'::regclass;
>> classid | objid | objsubid | refclassid | refobjid | refobjsubid | deptype
>> ---------+--------+----------+------------+----------+-------------+---------
>> 1259 | 335539 | 0 | 16672 | 2200 | 0 | n
>> 1259 | 335539 | 0 | 1259 | 335541 | 1 | i
>> (2 rows)
>
> Well, that second line is *definitely* a SERIAL column linkage.
>
>> 'k, checking the docs ... deptype == i is an INTERNAL, and refobjid is
>> what is referencing it (in this case, xa_url, as I'd expect) ... but,
>> looking at \d for xa_url, I'm not seeing anything there to cause it ... no
>> serial values ... the only 'default nextval()' I can find in the schema
>> is something totally unrelated ...
>
> Is it possible they did "create table xa_url(id bigserial, ...)" and
> then later changed the default expression for the column?
'k, am checking into this ... is it a simple matter of removing that
second record above from pg_depend to "fix" the pg_dump issue, or
something more involved then that?
----
Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664