[prev] [thread] [next] [lurker] [Date index for 2004/06/02]
On Fri, May 21, 2004 at 02:14:16PM +0100, Andy Wardley wrote:
> I'm seeing some errors when trying to install Siesta. I've tried both
> 0.65 from CPAN and the latest snapshot from the web site.
>
> First, I was getting some warnings. Class::DBI (0.96) was warning about
> the 'sort' parameter being deprecated in favour of 'order_by'. The first
> patch attached makes the change, although it will presumably break
> compatability with pre-0.96 Class:DBI.
Sassen frassen Class::DBI.
>
> I was also seeing some warnings from line 127 of Siesta.pm because
> $Siesta::Config::LOG_LEVEL was undefined, due to the fact that I don't
> have a /usr/local/siesta/siesta.conf file yet.
>
> The second patch changes Config.pm.in to call a setup method to set
> the package variable to their default values even if the config file
> doesn't exist. I created a new setup() method but left the original
> load_from() intact, just in case it's currently being used in other
> parts of code. I also added C<use Siesta::Config> to Siesta.pm, as
> implemented in the third patch. I suspect that this may not be the
> right thing to do, but it seems to solve the immediate problem.
Hrm. Most of that will have been oboleted by the recent work I did
as patches #1505-1510 I'm afraid.
> Now, having got all that out of the way, I'm still seeing a couple of
> errors.
>
> Failed Test Stat Wstat Total Fail Failed List of Failed
> -------------------------------------------------------------------------------
> t/09defer.t 255 65280 22 6 27.27% 17-22
> t/09plugin_order.t 255 65280 11 2 18.18% 10-11
> t/20fullsend.t 2 512 1 1 100.00% 1
> Failed 3/18 test scripts, 83.33% okay. 9/222 subtests failed, 95.95% okay.
>
> The first two fail with a Class::DBI call_trigger() method failure.
>
> t/09defer...................ok 16/22Can't locate object method "call_trigger" via package "Class::DBI::Object::Has::Been::Deleted" at /usr/local/lib/perl5/site_perl/5.8.0/Class/DBI.pm line 723, <DATA> line 12.
>
> t/09plugin_order............ok 9/11Can't locate object method "call_trigger" via package "Class::DBI::Object::Has::Been::Deleted" at /usr/local/lib/perl5/site_perl/5.8.0/Class/DBI.pm line 723.
>
> The third complains thusly:
>
> t/20fullsend................Siesta::Plugin::process called directly at /home/abw/projects/siesta/Siesta-0.65-abw/blib/lib/Siesta/Plugin.pm line 44.
>
> A cursory glance at the code for either of these problems doesn't offer
> any enlightenment. I suspect the Class::DBI problem is another version
> compatibility problem, but I can't track it down.
>
> Any clues gratefully received. It looks like most of it works so I'll
> press on.
I've just took a quick stab on a clean perl with the lastest release
of everything, and suspect that this is a bug in the new Class::DBI
caching code.
From the 0.96 Changes file.
0.96
New Functionality
- Maintain live object index so subsequent requests for same row return
the same object (Perrin Harkins)
Now without hunting down that code I think it's showing up because of
an 'interesting' quirk in SQLites auto-incrementing id columns.
Rather than being a true sequence, it just finds the maximum value for
the key, and adds 1 to it. This means you see patterns like this:
INSERT INTO foo (bar) VALUES ('baz');
SELECT id FROM BAR;
mysql | SQLite
1 | 1
INSERT INTO foo (bar) VALUES ('baz');
SELECT id FROM BAR;
mysql | SQLite
1 | 1
2 | 2
DELETE FROM foo WHERE id = 2;
SELECT id FROM BAR;
mysql | SQLite
1 | 1
INSERT INTO foo (bar) VALUES ('baz');
SELECT id FROM BAR;
mysql | SQLite
1 | 1
3 | 2
Then all you need is for that object index to be really naiive and
sticky, and keep the version that gets blessed into
'Class::DBI::Has::Been::Deleted' when you called ->delete on it and
foop - big bang!
Certainly the 09defer.t would tickle this.
When I get a tuit I'll get onto either Tony or the Class::DBI::SQLite
maintainer.
--
Richard Clamp <richardc@xxxxxxxxx.xxx>
Generated at 13:56 on 01 Jul 2004 by mariachi 0.52