On Wed, Mar 20, 2013 at 11:17 AM, Daniele Varrazzo
<daniele.varrazzo@gmail.com> wrote:
> Hello,
>
> Matthew Woodcraft and me have worked on exposing PQresultErrorField on
> the Error object, allowing to get detailed information from a server
> error. This includes the new diagnostics fields to be released in
> PostgreSQL 9.3.
>
> Ticket: http://psycopg.lighthouseapp.com/projects/62710-psycopg/tickets/149
> Docs:
> http://initd.org/psycopg/docs/module.html#psycopg2.Error.diag
> http://initd.org/psycopg/docs/extensions.html#psycopg2.extensions.Diagnostics
>
> The feature needs the PQresult stored into Error, making it a
> StandardError subclass with a customized C structure. Unfortunately
> this breaks Python 2.4, where the exceptions were old style classes.
>
> Working around the issue seems a messy matter of #ifdefs scattered all
> around the code: I think it would be better to just pull the plug on
> Python 2.4 for psycopg release 2.5. If there was a 2.4.7 release (it
> could be: there is a couple of bug fixed and, sigh, another Zope
> problem) it would still be compatible with Python 2.4.
>
> Issues? Objections? Let me know.
Python 2.4 is still the default on RHEL5, I believe? That's always a
potential issue - but python 2.6 at least is available as a second
package, so it's something that can reasonably easily be worked
around.
And since psycopg 2.5 is certainly not going to be in a rhel5 default
package :), I doubt it's a problem that actually needs much attention.
Bottom line is that as long as we're aware of that whenever someone
tries to install it via whatever manual operation on rhel5, it's
perfectly reasonable to drop support for python 2.4.
--
Magnus Hagander
Me: http://www.hagander.net/
Work: http://www.redpill-linpro.com/