Attached is an updated patch that incorporates all of the review I've
done so far on the core code. This omits the contrib changes, which
I've not looked at in any detail, and the docs changes since I've not
yet updated the docs to match today's code changes. User-visible
changes are that WITH NO DATA is gone, and instead there's
pg_extension_config_dump(regclass, text) as per today's discussion.
The latter is only lightly tested but seems to work as intended.
I believe the core code is now in pretty good shape; the only open issue
I have at the moment is that pg_dump will fail to suppress dumping of
USER MAPPING objects that are in an extension. Which is something I'm
not too excited about anyway. (The reason that's an issue is that I
reverted most of the changes in pg_dump.c in favor of using pg_dump's
already existing dependency mechanisms to suppress dumping unwanted
items. The USER MAPPING code tries to bypass the DumpableObject
representation, which means it doesn't get filtered.)
The documentation still needs a good bit of work, but I hope to have
this committed within a day or so.
regards, tom lane
In response to
pgsql-hackers by date
|Next:||From: Robert Haas||Date: 2011-02-08 03:37:06|
|Subject: postponing some large patches to 9.2|
|Previous:||From: Robert Haas||Date: 2011-02-08 03:05:21|
|Subject: Re: We need to log aborted autovacuums|