Re: [GENERAL] PGXLOG variable worthwhile? - Mailing list pgsql-hackers

From Marc G. Fournier
Subject Re: [GENERAL] PGXLOG variable worthwhile?
Date
Msg-id 20020919133205.S53125-100000@hub.org
Whole thread Raw
In response to Re: [GENERAL] PGXLOG variable worthwhile?  ("Robert Treat" <xzilla@users.sourceforge.net>)
List pgsql-hackers
On Thu, 19 Sep 2002, Robert Treat wrote:

> I don't know if I agree with that. Most servers (apache for instance) have
> configuration variables on where files are going to live, not command line
> options.

Not where it involves *critical* files:

OPTIONS
       -R libexecdir
                   This option is only available  if  Apache  was
                   built  with the SHARED_CORE rule enabled which
                   forces the Apache core code to be placed  into
                   a  dynamic shared object (DSO) file. This file
                   is searched in a hardcoded path under  Server-
                   Root  per default. Use this option if you want
                   to override it.

> Well, as with most (all?) GUC variables, wouldn't you have the option of
> doing postmaster -o "pgxlog=/dev/null" and have the same functionality
> as -X ?

True, but then that negates the whole argument about not having a command
line option, no?  Which I believe was the whole argument on this ... no?

> Shouldn't this work the other way around? Use what's in the conf file
> unless I explicitly state otherwise? IIRC that's how it works with -i

God, I wish I had thought to note it at the time ... one of the things I
did when I dove into this was to check how various Unix daemons were doing
it, now I can't recall which I was looking at that mentioned the config
file overriding the command line options, but you are correct, the command
line should override the conf file ...



pgsql-hackers by date:

Previous
From: "Marc G. Fournier"
Date:
Subject: SIMILAR TO syntax (Was: Re: The TODO List (Was: Re: O...)
Next
From: Tom Lane
Date:
Subject: Re: [GENERAL] Memory Errors...