Enhancement: Change atomic_writes table option to enum type. Now every file can either use atomic writes, not use it or use default.

SYNTAX: ATOMIC_WRITES=['DEFAULT','ON','OFF']

Idea here is to be able to define innodb_doublewrite = 1 but with following rules:

ATOMIC_WRITES='DEFAULT' - if innodb_use_atomic_writes = 1, we do not write to doublewrite buffer the changes
                          if innodb_use_atomic_writes = 0, we write to doublewrite buffer
ATOMIC_WRITES='ON'      - do not write to doublewrite buffer
ATOMIC_WRITES='OFF'     - write to doublewrite buffer

Note that doublewrite buffer can't be used if innodb_doublewrite = 0.

written and rest of the page is trimmed. In following writes
there is no need to trim again if write_size only increases
because rest of the page is already trimmed. If actual write
size decreases we need to trim again. Need to research if this
can happen frequently enough to make any effect.

Temporal solution: In directFS using atomic writes
we must use posix_fallocate to extend the file because
pwrite past end of file fails but when compression is
used the file pages must be physically initialized with
zeroes, thus after file extend with posix_fallocate
we still write empty pages to file.

currently Fusion-io SSD drive does not support setting file
size without fysically writing pages with zeroes when fallocate
with PUCH_HOLE is used.

Added additional error message if atomic write setup does not
succeed.

for Fusion-io currently.

dictionary setting and from there it will be stored to table space.

tables using atomic write/table.

This is work in progress and some parts are at most POC quality.

Fixed.

in the code merged from MySQL-5.6:

  const CHARSET_INFO* -> CHARSET_INFO*

(CHARSET_INFO already has the "const" qualifier in MariaDB, inlike in MySQL)

Applied a patch from Philip Hazel implementing the non-standard
syntax for word boundaries in PCRE, for compatibility with the
old Henry Spencer's regex library.

modified:
  storage/connect/valblk.cpp

modified:
  storage/connect/ha_connect.cc
  storage/connect/value.cpp

- Fix the way GetPlug works (was not updating xp)

modified:
  storage/connect/ha_connect.cc

modified:
  storage/connect/odbconn.cpp

- Prepare Value and Valblk to support unsigned data types
  (not operational yet)

modified:
  storage/connect/colblk.cpp
  storage/connect/filamvct.cpp
  storage/connect/myconn.cpp
  storage/connect/plgdbutl.cpp
  storage/connect/tabdos.cpp
  storage/connect/tabodbc.cpp
  storage/connect/tabvct.cpp
  storage/connect/valblk.cpp
  storage/connect/valblk.h
  storage/connect/value.cpp
  storage/connect/value.h
  storage/connect/xindex.cpp

            MariaDB 10.0 server

modified:
  storage/connect/ha_connect.cc
  storage/connect/myconn.cpp
  storage/connect/myutil.cpp
  storage/connect/odbccat.h
  storage/connect/odbconn.cpp
  storage/connect/tabmysql.cpp

- Add tests on new MYSQL features

added:
  storage/connect/mysql-test/connect/my.cnf
  storage/connect/mysql-test/connect/r/mysql_discovery.result
  storage/connect/mysql-test/connect/r/mysql_exec.result
  storage/connect/mysql-test/connect/r/mysql_new.result
  storage/connect/mysql-test/connect/t/myconn.inc
  storage/connect/mysql-test/connect/t/myconn_cleanup.inc
  storage/connect/mysql-test/connect/t/mysql_discovery.test
  storage/connect/mysql-test/connect/t/mysql_exec.test
  storage/connect/mysql-test/connect/t/mysql_new.test

Analysis: In earlier MySQL 5.6 versions this table innodb_index_stats  used to have a foreign key referencing to innodb_table_stats. However, in newer MySQL 5.6 versions this foreign key is removed and if you upgrade, your innodb_table_stats is created by the earlier mariadb version, thus a newer version will complain that the table is incorrectly defined.

Added drop foreign key on mysql_system_tables_fix.sql to be executed on mysql_upgrade.

Fixed the following compilation errors and test failures:
- maria SE: "stage_waiting_for_a_resource" wasn't declared w/o PFS
- sql_repl.h: PSI_mutex_key is not available in non-PFS builds
- mysqld.cc: pfs_param is not available in non-PFS builds
- mysqld.cc: init_show_explain_psi_keys() is not available in non-PFS builds
- mysqld.cc: call net_before_header_psi, net_after_header_psi even if PFS
  is not available so that thread enters stage_init at proper time. Fixes
  sp-threads and a few tests in funcs_1.
- myisam_file_io.opt: added missing loose prefix

        Added the IF NOT EXISTS option to the CONSTRAINT keyword.

MDEV-5248 Serious incompatibility and data corruption of DATETIME and DATE types due to get_innobase_type_from_mysql_type refactor combined with InnoDB Online DDL

restore old innodb get_innobase_type_from_mysql_type() function,
record all mysql_type->innodb_type mapping
(as generated by mysql-5.6).
add safety code to disable online alter when internal types don't match

storage/innobase/dict/dict0stats.cc:
  revert to 5.6 state

correct bugs in mysql_system_tables_fix.sql.
Update system_mysql_db_fix* tests

Patch by Elena.

Move the table creation to the end of the file, so
mysql.innodb_stats_table has been created and the statement
does not fail.

Apply fix suggested by Igor:
- When eliminate_item_equal() generates pair-wise equalities from a 
  multi-equality,  do generate a "bridge" equality between the first 
  field inside SJM nest and the field that's first in the overall multi-equality.

  and MYSQL CONNECT tables to take care of kewords such as IGNORE.

modified:
  storage/connect/myconn.cpp
  storage/connect/odbconn.cpp
  storage/connect/tabmysql.cpp
  storage/connect/tabmysql.h
  storage/connect/tabodbc.cpp
  storage/connect/tabodbc.h

MDEV-5113 Wrong result (extra row) and valgrind warnings in Item_maxmin_subselect::any_value on 2nd execution of PS with SELECT subquery

When setting Item_func_not_all::test_sum_item or Item_func_not_all::test_sub_item,
reset the other one to NULL - they can never be set both. When a PS is reexecuted,
different executions might be optimized differently and a wrong test_su*_item
might stay set from the previous execution.

- MIN/MAX optimizer does a check whether a "field CMP const" comparison uses a constant 
  that's longer than the field it is compared to.  Make this check only for string columns,
  also compare character lengths, not byte lengths.

- Address input from the mail list: change how EXPLAIN is 
  formatted in the slow query log.

The used_tables attribute must be recalculated for the HAVING condition
if the condition is applied to the rows read from a temporary table.

allocate thd_proc_info string in thd memroot, not on the stack, so that it
won't be overwritten while another thread might be printing it