On Friday 27 August 2004 20:21, Tom Lane wrote:
> Steve Ingram <ingram(at)samsixedd(dot)com> writes:
> > When I pg_dump out the database, it outputs all the tables, functions,
> > types as you'd expect, but it outputs table definitions that reference
> > postgis functions/types *before* the functions/types themselves.
> Yeah, this is pretty likely if you did any ALTER TABLE or similar
> commands anywhere along the line.
Duh! Of course!!
OK, this makes a lot more sense now and now I
don't regard it as a bug ;) I bet you are relieved :) :)
> The dump order is dependent on the
> original creation order of the objects (tables, functions, etc) and so
> it's pretty easy to break it by adding columns or indexes to
> pre-existing tables.
Thats cool; I can make sure (I think) that doesn't happen,
and frankly, if a reload breaks, the fact that I am reloading *at all* is
pretty drastic, so I'll be happy to shift the contents to accommodate
any of the changes you mentioned.
> 8.0's pg_dump attempts to deal with this by paying attention to
> dependencies, but no earlier release does anything at all credible.
> FWIW, pg_restore does have options to alter the reload order, which
> may be an easier workaround than editing a textual dump.
And of course, I could have read ahead :)
Thanks for the explanation Tom
> regards, tom lane
In response to
pgsql-bugs by date
|Next:||From: PostgreSQL Bugs List||Date: 2004-08-28 18:05:12|
|Subject: BUG #1236: still in use tablespaces can be removed|
|Previous:||From: Tom Lane||Date: 2004-08-28 16:37:20|
|Subject: Re: initdb errors, postgresql-8.0-beta |