Darcy Buskermolen wrote:
> On Wednesday 21 June 2006 15:27, Darcy Buskermolen wrote:
>   
>> Also I'm working on a solution for the recent datestyle bug
>> (http://gborg.postgresql.org/project/slony1/bugs/bugupdate.php?1566).
>>     
>
>
> Here is a work in progress for datestyle bug 
> http://www.dbitech.ca/slony/patches/datestyle-insert.patch
>
> At the moment it only deals with I (insert) events, and assumes that you have 
> added set datestyle to 'ISO'; to the 
> sql_on_connection option in your slon runtime config file
>
> Any feedback on this would be welcomed. Also, based on the patch, how do 
> people feel about this making it into 1.2, or should it be held for a later 
> release?
>   

I think I'd sorta like for there to be 3 functions with brief names
added so that not so much literal "bumpf" would get added to sl_log_1...

FYI, I have added a test for this condition:

cbbrowne at dba2:Slony-I/slony1-HEAD-pristine/tests/testdatestyles> ls
-alt                                                      Friday 15:37:59
total 52
drwxr-xr-x  3 cbbrowne users 4096 Jun 23 15:37 .
drwxr-xr-x  2 cbbrowne users 4096 Jun 23 15:37 CVS
-rw-r--r--  1 cbbrowne users  172 Jun 23 15:37 README
-rw-r--r--  1 cbbrowne users 2942 Jun 23 15:30 generate_dml.sh
-rw-r--r--  1 cbbrowne users   63 Jun 23 15:21 schema.diff
-rw-r--r--  1 cbbrowne users  258 Jun 23 15:20 init_schema.sql
-rw-r--r--  1 cbbrowne users  229 Jun 23 15:20 init_data.sql
-rw-r--r--  1 cbbrowne users  108 Jun 23 15:19 init_add_tables.ik
-rw-r--r--  1 cbbrowne users   52 Jun 23 14:49 init_cluster.ik
-rw-r--r--  1 cbbrowne users   57 Jun 23 14:49 init_create_set.ik
-rw-r--r--  1 cbbrowne users   66 Jun 23 14:49 init_subscribe_set.ik
-rw-r--r--  1 cbbrowne users   95 Jun 23 14:49 settings.ik
drwxr-xr-x 13 cbbrowne users 4096 Jun 23 14:49 ..
cbbrowne at dba2:Slony-I/slony1-HEAD-pristine/tests/testdatestyles> cat
README                                                   Friday 15:40:30
$Id: README,v 1.1 2006/06/23 15:37:57 cbbrowne Exp $

This test involves submitting date and timestamps in various DATESTYLE
forms, and verifying that all load in properly.


The test comprehensively goes through all the datestyles, inserting
values based on the following criteria:

1.  +/- infinity (apparently not possible for DATE?)
2.  values based on final second of last year
3.  values based on final second of last month
4.  values based on 60 days after the end of last year (which is when a
leap day would take place), for this year, and the following 4 years,
which otta hit a leap year somewhere
5.  values based on now()
6.  1000 days starting the first day of this year

It generates this set of values for each datestyle.

Replication immediately breaks on CVS HEAD *without* Darcy's patch, and
sl_log_1 contains some rather funky looking data :-).