Date: Fri, 9 Feb 2001 16:39:35 +0100
From: "Thies C. Arntzen" <thies@thieso.net>
To: Bruce Momjian <pgman@candle.pha.pa.us>
Cc: "Thies C. Arntzen" <thies@digicol.de>, rasmus@php.net,
	PostgreSQL-interfaces <pgsql-interfaces@postgresql.org>
Subject: Re: PostgreSQL and PHP persistent connections
Message-ID: <20010209163935.A23742@schnuffel.thieso.net>
References: <20010209144154.A23220@schnuffel.thieso.net>
	<200102091519.KAA02853@candle.pha.pa.us>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.5i
In-Reply-To: <200102091519.KAA02853@candle.pha.pa.us>;
	from pgman@candle.pha.pa.us on Fri, Feb 09, 2001 at 10:19:51AM
	-0500

On Fri, Feb 09, 2001 at 10:19:51AM -0500, Bruce Momjian wrote:
> > > We discussed using 'ROLLBACK' before passing a connection to a new user,
> > > but the problem was that ROLLBACK with no open transaction causes a
> > > server log error message.  We discussed adding 'ROLLBACK SILENT' to fix
> > > this, but I believe a better, more portable solution is a simple "BEGIN
> > > WORK;ROLLBACK".  This will do nothing if there is no open transaction,
> > > and will ROLLBACK any open transaction.  I propose this be sent by PHP
> > > as the first query when passing persistent connections.
> > 
> >     was there a conclusion made yet (that i missed somehow) on
> >     this subject? i'd like to implement *something* in the PHP
> >     core for this soon.
> 
> One person objected that BEGIN;ROLLBACK will throw an error to the log
> file if there was an open transaction, but I think that is fine as this
> is a rare occurance, and an argument can be made that leaving an open
> transaction on session close in a pooled connection environment should
> throw some kind of error anyway.  Also, this fix will work for all
> versions of PostgreSQL, so it can be implemented right away with no
> backend changes or version checks.

    ok, i'll start playing around with this solution soon. i'll
    commit it to the PHP 4 cvs as soon as i'm happy with it.

    thies
> 
> We still need a global RESET, but I don't think that can be done until
> 7.2.  I will add that to our project TODO list.