Merge 'trunk' into 'sqla_1.50_compat'

Merge 'trunk' into 'sqla_1.50_compat'

r5410@vger:  mendel | 2009-01-18 20:40:15 +0100
 r5392@vger (orig r5301):  norbi | 2009-01-12 11:55:54 +0100
  r5391@vger:  mendel | 2009-01-12 11:55:48 +0100
   * Reworded $rs->current_source_alias POD (thanks to castaway).

 r5398@vger (orig r5307):  caelum | 2009-01-14 02:25:28 +0100
 auto_savepoint support for Oracle and a note on txn_do for with "AutoCommit => 0"

 r5404@vger (orig r5313):  ribasushi | 2009-01-16 11:29:16 +0100
 Proper MSSQL last_insert_id() scoping patch by abraxxa
 No tests as it is very hard to simulate the problematic behavior - it would only manifest if:
   1) An INSERT statement is issued from connection A
   2) An INSERT statement is issued from connection B
   3) last_insert_id is called from connection A

 Without this patch the last id from the second INSERT will be returned in step 3
 Tested with both DBICTEST_MSSQL_ODBC_DSN and DBICTEST_MSSQL_DSN against a MSSQL 2005
 r5405@vger (orig r5314):  plu | 2009-01-16 19:25:57 +0100
 Fixed typo in link
 r5406@vger (orig r5315):  ribasushi | 2009-01-17 19:39:50 +0100
 Stop 99dbic_sqlt_parser.t from breaking every time a new table is added
 r5407@vger (orig r5316):  ribasushi | 2009-01-17 19:47:53 +0100
 Fix borked t/bindtype_columns.t - TODO: still only tests PG, need to test Oracle as well