Merge 'trunk' into 'grouped_prefetch'

Merge 'trunk' into 'grouped_prefetch'

r6844@Thesaurus (orig r6843):  abraxxa | 2009-06-29 11:02:17 +0200
fixed typo in test

r6848@Thesaurus (orig r6847):  ribasushi | 2009-06-29 19:09:00 +0200
Minor Ordered optimization (don't use count)
r6856@Thesaurus (orig r6855):  caelum | 2009-06-29 23:42:11 +0200
 r5451@hlagh (orig r6605):  caelum | 2009-06-10 09:23:44 -0700
 new branch to implement on_connect_call
 r5484@hlagh (orig r6633):  caelum | 2009-06-11 11:03:10 -0700
 on_connect_call implementation and set_datetime_format support for Oracle
 r5492@hlagh (orig r6641):  caelum | 2009-06-11 16:39:28 -0700
 connect_call_set_datetime_format for Oracle, I have no idea why this didn't get committed before...
 r5504@hlagh (orig r6655):  caelum | 2009-06-12 17:28:06 -0700
 finished up on_connect_call stuff
 r5507@hlagh (orig r6658):  caelum | 2009-06-13 04:03:36 -0700
 fixup _setup_connect_do, other minor cleanups
 r5508@hlagh (orig r6659):  caelum | 2009-06-13 04:35:33 -0700
 make the on_(dis)?connect_do accessors returnn the original structure
 r5509@hlagh (orig r6660):  caelum | 2009-06-13 08:31:52 -0700
 allow undef for _setup_connect_do
 r5522@hlagh (orig r6679):  caelum | 2009-06-14 09:56:40 -0700
 rename connect_do store
 r5621@hlagh (orig r6769):  caelum | 2009-06-23 07:38:33 -0700
 minor doc update
 r5628@hlagh (orig r6777):  caelum | 2009-06-23 16:36:12 -0700
 properly test nanosecond precision with oracle and datetime_setup
 r5669@hlagh (orig r6784):  caelum | 2009-06-24 10:49:25 -0700
 IC::DT does support timestamp with timezone
 r5768@hlagh (orig r6846):  caelum | 2009-06-29 08:20:32 -0700
 remove DateTime from 73oracle.t
 r5781@hlagh (orig r6849):  caelum | 2009-06-29 13:07:43 -0700
 remove the _store stuff for on_connect_do
 r5785@hlagh (orig r6853):  ribasushi | 2009-06-29 14:38:30 -0700
 Some beautification

r6871@Thesaurus (orig r6870):  ribasushi | 2009-06-30 10:09:03 +0200
Cleanup dependency handling a bit
r6875@Thesaurus (orig r6874):  ribasushi | 2009-06-30 12:39:06 +0200
Allow broken resultsource-class-derived objects to still work
r6876@Thesaurus (orig r6875):  ribasushi | 2009-06-30 12:40:46 +0200
clarify
r6878@Thesaurus (orig r6877):  ash | 2009-06-30 13:48:13 +0200
Update POD on Dynamic sub-classing

r6883@Thesaurus (orig r6882):  ribasushi | 2009-06-30 17:36:38 +0200
 r6815@Thesaurus (orig r6814):  ribasushi | 2009-06-28 10:32:42 +0200
 Branch to explore double joins on search_related
 r6816@Thesaurus (orig r6815):  ribasushi | 2009-06-28 10:34:16 +0200
 Thetest case that started it all
 r6817@Thesaurus (orig r6816):  ribasushi | 2009-06-28 10:35:11 +0200
 The proposed fix (do not add an extra join if it is already present in the topmost join)
 r6818@Thesaurus (orig r6817):  ribasushi | 2009-06-28 11:04:26 +0200
 Minor omission
 r6819@Thesaurus (orig r6818):  ribasushi | 2009-06-28 11:07:33 +0200
 Adjust a couple of tests for new behavior (thus all of this might be backwards incompatible to the point of being useless):
 The counts in t/90join_torture.t are now 5*3, not 5*3*3, as a second join is not induced by search_related
 The raw sql scan in t/prefetch/standard.t is just silly, won't even try to understand it
 Just to maintain the TreeLike folding, I add a 3rd children join which was inserted by search_related before the code changes