Merge 'trunk' into 'mssql_top_fixes'

r6932@Thesaurus (orig r6931):  ribasushi | 2009-07-02 08:08:33 +0200
Another candidate for somethingawful.com (fix left join-ed count)
r6934@Thesaurus (orig r6933):  ribasushi | 2009-07-02 09:04:13 +0200
Changelog
r6935@Thesaurus (orig r6934):  ribasushi | 2009-07-02 11:23:48 +0200
cleanup
r6936@Thesaurus (orig r6935):  ijw | 2009-07-02 12:41:01 +0200
Check fetched rows == count for related resultsets
r6937@Thesaurus (orig r6936):  ijw | 2009-07-02 12:43:47 +0200
Confirm prefetch doesn't affect main row fetch, and main row fetch works with and without counting
r6938@Thesaurus (orig r6937):  ribasushi | 2009-07-02 12:52:51 +0200
More fail (fix is known but needs work)
r6939@Thesaurus (orig r6938):  ribasushi | 2009-07-02 13:07:22 +0200
And more fail
r6940@Thesaurus (orig r6939):  ribasushi | 2009-07-02 13:16:46 +0200
These tests are in prefetch/count.t
r6941@Thesaurus (orig r6940):  ribasushi | 2009-07-02 13:38:31 +0200
cleanup
r6942@Thesaurus (orig r6941):  ribasushi | 2009-07-02 13:38:49 +0200
Solve more prefetch inflation crap
r6943@Thesaurus (orig r6942):  ribasushi | 2009-07-02 13:47:41 +0200
Make the code readable
r6944@Thesaurus (orig r6943):  ribasushi | 2009-07-02 15:52:35 +0200
Everything works, just need to fix join-path chaining over search_related (to guard against obscure db quirks)
r6946@Thesaurus (orig r6945):  caelum | 2009-07-02 21:06:32 +0200
add sybase reconnect test
r6948@Thesaurus (orig r6947):  ribasushi | 2009-07-02 22:20:21 +0200
Last part of the join handling puzzle
r6951@Thesaurus (orig r6950):  ribasushi | 2009-07-03 00:14:50 +0200
 r6360@Thesaurus (orig r6359):  arcanez | 2009-05-21 20:18:52 +0200
 branch to work on prefetch/select
 r6361@Thesaurus (orig r6360):  arcanez | 2009-05-21 20:32:46 +0200
 failing test
 r6373@Thesaurus (orig r6372):  ribasushi | 2009-05-22 11:07:26 +0200
 Simplify unresolvable test by arcanez
 r6905@Thesaurus (orig r6904):  ribasushi | 2009-07-01 12:54:03 +0200
 Extend test
 r6950@Thesaurus (orig r6949):  ribasushi | 2009-07-03 00:14:09 +0200
 Apparent fix - simply delay the in_storage flagging of the main object until all prefetched objects are inflated. The rest of the changes are just cosmetics, preparing for the collapse_result rewrite

r6953@Thesaurus (orig r6952):  ribasushi | 2009-07-03 00:17:22 +0200
Changes

diff --cc lib/DBIx/Class/ResultSet.pm
index 3ebb162,58ccaa3..32f6b5a
--- 1/lib/DBIx/Class/ResultSet.pm
--- 2/lib/DBIx/Class/ResultSet.pm
+++ b/lib/DBIx/Class/ResultSet.pm
@@@ -1254,14 -1274,13 +1274,20 @@@ sub _count_subq_rs 
  
    $sub_attrs->{select} = $rsrc->storage->_subq_count_select ($rsrc, $sub_attrs);
  
+   # read the comment on top of the actual function to see what this does
+   $sub_attrs->{from} = $self->_switch_to_inner_join_if_needed (
+     $sub_attrs->{from}, $sub_attrs->{alias}
+   );
+ 
 +  # this is so that ordering can be thrown away in things like Top limit
 +  $sub_attrs->{-for_count_only} = 1;
 +
 +  my $sub_rs = $rsrc->resultset_class->new ($rsrc, $sub_attrs);
++
    $attrs->{from} = [{
 -    count_subq => $rsrc->resultset_class->new ($rsrc, $sub_attrs )->as_query
 +    -alias => 'count_subq',
 +    -source_handle => $rsrc->handle,
 +    count_subq => $sub_rs->as_query,
    }];
  
    # the subquery replaces this

diff --cc lib/DBIx/Class/Row.pm
Simple merge