Re: Install extensions using update scripts (was Re: Remove superuser() checks from pgstattuple)

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Andres Freund <andres(at)anarazel(dot)de>
Cc: Stephen Frost <sfrost(at)snowman(dot)net>, Peter Eisentraut <peter(dot)eisentraut(at)2ndquadrant(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Install extensions using update scripts (was Re: Remove superuser() checks from pgstattuple)
Date: 2016-09-07 13:46:32
Message-ID: 1213.1473255992@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Andres Freund <andres(at)anarazel(dot)de> writes:
> On 2016-09-05 22:24:09 -0400, Tom Lane wrote:
>> Ordinarily I'd be willing to stick this on the queue for the next
>> commitfest, but it seems like we ought to try to get it pushed now
>> so that Stephen can make use of the feature for his superuser changes.
>> Thoughts?

> Seems sensible to me. I can have a look at it one of the next few days
> if you want.

Thanks for offering. Attached is an updated patch that addresses a
couple of issues I noticed:

1. When I did the previous patch, I was thinking that any parameters
specified in an auxiliary .control file for the base version would
apply to any non-base versions based on it; so I had the
pg_available_extension_versions view just duplicate those parameters.
But actually, most of those parameters are just applied on-the-fly
while processing the update script, so it *does* work for them to be
different for different versions. The exceptions are schema (which
you can't modify during an update) and comment (while we could replace
the comment during an update, it doesn't seem like a good idea because
we might overwrite a user-written comment if we did). (I think this
behavior is undocumented BTW :-(.) So this fixes the view to only
inherit those two parameters from the base version.

2. I noticed that CASCADE was not implemented for required extensions
processed by ApplyExtensionUpdates, which seems like a bad thing if
CREATE processing is going to rely more heavily on that. This only
matters if different versions have different requires lists, but that
is supposed to be supported, and all the catalog-hacking for it is there.
So I made this work. It took a fair amount of refactoring in order to
avoid code duplication, but it wasn't really a very big change.

At this point it's awfully tempting to make ALTER EXTENSION UPDATE grow
a CASCADE option to allow automatic installation of new requirements
of the new version, but I didn't do that here.

Still no SGML doc updates.

regards, tom lane

Attachment Content-Type Size
let-create-extension-follow-update-chains-2.patch text/x-diff 30.5 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Stas Kelvich 2016-09-07 13:48:53 Bug in two-phase transaction recovery
Previous Message Peter Eisentraut 2016-09-07 12:53:22 Re: ICU integration