when looking for tables to delete from

Fix for LP BUG#908269 Wrong result with subquery in select list, EXISTS, constant MyISAM/Aria table.

Problem: When building the condition for JOIN::outer_ref_cond the optimizer forgot to take into account
that this condition could depend on constant tables as well.

mysql-test/r/information_schema_all_engines.result:
  Update result
mysql-test/t/information_schema_all_engines.test:
  Added --sorted-results as tables in information_schema are not sorted.

Split IO threads into ones that handle only read completion and ones that handle only write completion, as it was originally done, but got lost with "completion port" patch. The reason we need to have dedicated read and dedicated write threads is that read completion routine can block waiting for write io to complete, and in rare cases where all io threads are handling async reads, it can deadlock.

- Create/use do_copy_nullable_row_to_notnull() function for ref access, which is used 
  when copying from not-NULL field in table that can be NULL-complemented to not-NULL field.

geometry fields are blobs too.

- ensure that mtr supressions table is flushed before doing controlled crash and restart
- use DBUG_SUICIDE() rather than abort() in partition tests - avoids a crash message/warning
- disable  perfschema all_instances test on Windows- there are  legitimate reasons  for output to be different on  Unix (some different threads, some different locks), the differences are expected to grow in the future, e.g with threadpool.

The patch for bug 685411 erroneously removed a call of engine->set_thd()
from Item_subselect::fix_fields().

If init_command was incorrect, we couldn't let users execute
queries, but we couldn't report the issue to the client either
as it does not expect error messages before even sending a
command. Thus, we simply disconnected them without throwing
a clear error.

We now go through the proper sequence once (without executing
any user statements) so we can report back what the problem
is. Only then do we disconnect the user.

As always, root remains unaffected by this as init_command is
(still) not executed for them.

mysql-test/r/init_connect.result:
  We now report a proper error if init_command fails.
  Expect as much.
mysql-test/t/init_connect.test:
  We now report a proper error if init_command fails.
  Expect as much.
sql/sql_connect.cc:
  If init_command fails, throw an error explaining this to
  the user.

Note: to build with -fno-rtti as we currently build the server, boost version  1.45 or later is required.
(without -fno-rtti, 1.40 is enough)

from a heap temptable, which uses pointers to records (that is, byte*
pointers) as rowids.  
This meant that for rows with the same sort key value, the order 
was determined by memory layout.

Port vasil.dimov@oracle.com-20111205083046-jtgi1emlvtfnjatt from mysql-trunk

Port vasil.dimov@oracle.com-20111205082900-lx9om1joscejr25e from mysql-trunk

Port vasil.dimov@oracle.com-20111205082831-7v1qu50hvd9hjr3g from mysql-trunk

Port vasil.dimov@oracle.com-20111205082756-wtlg8isyn4yohyny from mysql-trunk

Port vasil.dimov@oracle.com-20111205082626-87j5f48dq1ouk86r from mysql-trunk

Fix oqgraph build . Plugin does not need rtti, and does not load if rtti compile settings are different from server's

InnoDB: Remove HAVE_purify, UNIV_INIT_MEM_TO_ZERO, UNIV_SET_MEM_TO_ZERO.

The compile-time setting HAVE_purify can mask potential bugs.
It is being set in PB2 Valgrind runs. We should simply get rid of it,
and replace it with UNIV_MEM_INVALID() to declare uninitialized memory
as such in Valgrind-instrumented binaries.

os_mem_alloc_large(), ut_malloc_low(): Remove the parameter set_to_zero.

ut_malloc(): Define as a macro that invokes ut_malloc_low().

buf_pool_init(): Never initialize the buffer pool frames. All pages
must be initialized before flushing them to disk.

mem_heap_alloc(): Never initialize the allocated memory block.

os_mem_alloc_nocache(), ut_test_malloc(): Unused function, remove.

rb:813 approved by Jimmy Yang