Re: pg_dump - Mailing list pgsql-general

From garry saddington
Subject Re: pg_dump
Date
Msg-id 1164137470.21067.51.camel@localhost
Whole thread Raw
In response to Re: pg_dump  (Richard Huxton <dev@archonet.com>)
List pgsql-general
On Tue, 2006-11-21 at 19:00 +0000, Richard Huxton wrote:
> garry saddington wrote:
> >>   pg_dump -h /var/run/postgresql
> >> See the manuals for details.
> >
> > Thank you, have read the manuals and the -h section says host initially
> > but if you read on it tells about a leading / - that's why I missed it
> > first time round -skim reading!.
>
> It's not something I've needed myself, so I had to check. The only
> difference was, I knew the psql page of the manuals *had* to cover it.
>
> > Perhaps you may be able to me to explain this:
> > I have a Zope application which speaks to postgres via psycopg. I have
> > packaged all components together so that the application can be
> > installed like 'single click' or thereabouts and I have tested it on
> > several Linux distributions. Usually there are no problems but
> > occasionally the psycopg connection will default to a different unix
> > domain socket to the one that is configured in the .conf file. Nothing
> > changes apart from the OS. Is there a way to make sure that on
> > installation the same socket is used irrespective of OS?
>
> Normally, you'd set the socket directory at ./configure, compile time.
> This suggests to me that psycopg is using libpq to actually connect, and
>    that was compiled with different settings to the server you supply.
Yes, you are correct libpq is being used and this will explain the
different behaviour. However, I now include libpq in the package which
seems to work OK.
> Any reason not to use the distro's built-in package manager to fetch
> server and client for you?
The application is aimed at teachers and so it needs to be very simple
for them to test its features. Any barriers to this will reduce the take
up.
Kind Regards
Garry



pgsql-general by date:

Previous
From: Ray Stell
Date:
Subject: list archives
Next
From: Tom Lane
Date:
Subject: Re: Transaction id wraparound and autovacuum