--- Begin Message ---
Subject: |
[PATCH] make postgresql find its plugins |
Date: |
Sat, 28 Jul 2018 17:58:15 +0200 |
Hi,
attached is a patch to make postgresql find its plugins. Actually, it
only makes it unable to follow symlinks, so we can provide a union of
postgresql and plugins to the postgresql service. I also attached an
example system configuration that shows how this works. Do you think
this is the right solution? The patch was taken from nixos and updated
for the current version of postgresql.
Note that the VM will have troubles loading the extensions with only
its 256MB memory by default. You should add more ram to it.
Inside the VM, you can test with:
psql -U postgres
> create database postgistest;
> \connect postgistest;
> create extension postgis;
> create extension postgis_topology;
no error, you now have an empty spatial database :)
An extension would be to have a procedure to build the union of
packages (postgresql and extensions) called by the postgresql service.
we would have an "extension" field that would contain a list of
packages that contain extensions for postgresql and the service would
build and use the union of postgresql and these extensions. WDYT?
0001-gnu-postgresql-Fix-finding-extensions.patch
Description: Text Data
postgis.scm
Description: Text Data
--- End Message ---
--- Begin Message ---
Subject: |
Re: make postgresql find its plugins |
Date: |
Thu, 4 Oct 2018 22:36:47 +0200 |
Le Mon, 24 Sep 2018 13:23:38 +0200,
Gábor Boskovits <address@hidden> a écrit :
> It works fine for me.
>
> Some things I noticed while checking this:
> 1. postgresql contrib extensions should not be listed, and are
> working even without the patches. (This includes for example hstore,
> dblink,... see
> https://www.postgresql.org/docs/current/static/contrib.html for the
> full list) Did not test if it is problem if one adds these to the
> config though. 2. I had postgis actually missing, due to a missing
> use module, but there was no complaint, system reconfigure went just
> fine, but creating extension postgis did not work. Can we add some
> way to check if the list supplied is sane?
> 3. contrib extensions also work fine with the patches applied.
> 4. I did not test if it works if we have more than one extension, but
> it looks good.
> If we have any other extension packaged, we could take a look.
> 5. Do you think we should add a test?
> 6. I got a warning that the first patch introduces whitespace errors
> on current master. Is that ok?
Pushed on staging since the changes to postgresql requires a lot of
rebuilds.
The whitespace error is located in parts of the patch I can't change.
I've also added a sentence to the manual to make it clear that you
should list packages and not extensions themselves, and that you don't
need to add contrib extensions in the field. Otherwise, I think I've
addressed all your comments :)
Thank you for your review!
--- End Message ---