Merge Percona microsec_process patch.
Includes lots of after-merge fixes, Windows fixes, and other misc. small fixes.

mysql-test/suite/rpl/t/rpl_binlog_corruption.test:
  Disabled the test for Windows (see bug #47639)
mysql-test/suite/rpl/t/rpl_killed_ddl.test:
  Disabled the test for Windows (see bug #47638)
vio/viosocket.c:
  Added an implementation of vio_poll_read for Windows.
  Winsock does not support the poll function.
  So the existing generic implementation of vio_poll_read
  could not be used for Windows.

Ability to stop mysql-test-run.pl temporary (useful for them who run buildbot slave on thier work computers)

This adds a TIME_MS column to SHOW FULL PROCESSLIST and
INFORMATION_SCHEMA.PROCESSLIST that works like the TIME column, but
in units of milliseconds with microsecond precision.

Note that this also changes behaviour of the existing TIME column. In
the MySQL server, the TIME column changes when a thread sets @TIMESTAMP.
This is contrary to documentation and also potentially confusing, so
could in any case be considered a bug. With this patch, to ensure
consistency between TIME and TIME_MS, setting @TIMESTAMP has no effect
on either value.

Add a test case for the TIME and TIME_MS columns.
Update existing test cases for changed behaviour.

Author: Percona
Signed-off-by: Kristian Nielsen <knielsen@knielsen-hq.org>


mysql-test/r/create.result:
  Result file update.
mysql-test/r/information_schema.result:
  Add test case.
  Result file update (changed behaviour).
mysql-test/t/information_schema.test:
  Add test case.
  Adjust test case for changed behaviour.
mysql-test/t/not_embedded_server.test:
  Update (commented-out) test case for newly added column.
sql/sql_show.cc:
  Merge Percona microsec_process patch into MariaDB.

watched on Windows.

on Windows.

to avoid a timeout failure of it on Windows. 

to fix a failure in rpl_get_master_version_and_clock.test

    mysqlslap: setting --engine does not get replicated
    http://bugs.mysql.com/bug.php?id=46967

and
    mysqlslap: specifying --engine and --create does not
    work with --engine=<storage_engine>:<option>
    https://bugs.launchpad.net/maria/+bug/429773

Problems were that an --engine=<storage_engine> was translated
to a "set storage_engine = <storage_engine>", wich is _not_
replicated. A --engine=<storage_engine>:<option> was not always
properly parsed and in some cases crashed.

Fixed by eliminating "set storage_engine = ..." and adding
proper DDL generation. Initialized an unitialized buffer to
prevent crashes and fixed to use proper pointer for in
case of --engine=<storage_engine>:<option> being the last
element in list of --engines.

Also cleaned up code for better readability.

Q: Should MySQL's replication actually replicate a
   "set storage_engine = ..." command or not?
A: No it should not. It is documented here:
   http://dev.mysql.com/doc/refman/5.1/en/replication-features-variables.html
   ...
   "The storage_engine system variable is not replicated, which is a
   good thing for replication between different storage engines." ...

Before the patch, mysqlslap was behaving this way:

+-------------------------------+--------+-------------+
|                               | single | replication |
+-------------------------------+--------+-------------+
| Before patch                                         | 
+-------------------------------+--------+-------------+
| --engine[1]                                          |
+-----+-------------------------+--------+-------------+
| 1.1 | eng1                    |  OK    |   Not OK    |
| 1.2 | eng1,eng2               |  OK    |   Not OK    |
| 1.3 | eng1,eng2,eng3          |  OK    |   Not OK    |
| 1.4 | memory:option           |  OK    |   Not OK    |
| 1.5 | memory:option,eng1      |  OK    |   Not OK    |
| 1.6 | eng1,memory:option      | Not OK |   Not OK    |
| 1.7 | memory:option,eng1,eng2 | Crash  |   Not OK    |
| 1.8 | eng1,memory:option,eng2 |  OK    |   Not OK    |
| 1.9 | eng1,eng2,memory:option | Not OK |   Not OK    |
+-----+-------------------------+--------+-------------+
+-------------------------------+--------+-------------+
| --create --engine[2]                                 |
+-----+-------------------------+--------+-------------+
| 2.1 | eng1                    |  OK    |   Not OK    |
| 2.2 | eng1,eng2               |  OK    |   Not OK    |
| 2.3 | eng1,eng2,eng3          |  OK    |   Not OK    |
| 2.4 | memory:option           | Not OK |   Not OK    |
| 2.5 | memory:option,eng1      | Not OK |   Not OK    |
| 2.6 | eng1,memory:option      | Not OK |   Not OK    |
| 2.7 | memory:option,eng1,eng2 | Crash  |   Not OK    |
| 2.8 | eng1,memory:option,eng2 | Not OK |   Not OK    |
| 2.9 | eng1,eng2,memory:option | Not OK |   Not OK    |
+-----+-------------------------+--------+-------------+

After my final patch, mysqlslap now runs like this:

+-------------------------------+--------+-------------+
|                               | single | replication |
+-------------------------------+--------+-------------+
| After third patch                                    | 
+-------------------------------+--------+-------------+
| --engine[1]                                          |
+-----+-------------------------+--------+-------------+
| 1.1 | eng1                    |  OK    |  OK         |
| 1.2 | eng1,eng2               |  OK    |  OK         |
| 1.3 | eng1,eng2,eng3          |  OK    |  OK         |
| 1.4 | memory:option           |  OK    |  OK         |
| 1.5 | memory:option,eng1      |  OK    |  OK         |
| 1.6 | eng1,memory:option      |  OK    |  OK         |
| 1.7 | memory:option,eng1,eng2 |  OK    |  OK         |
| 1.8 | eng1,memory:option,eng2 |  OK    |  OK         |
| 1.9 | eng1,eng2,memory:option |  OK    |  OK         |
+-----+-------------------------+--------+-------------+
+-------------------------------+--------+-------------+
| --create --engine[2]                                 |
+-----+-------------------------+--------+-------------+
| 2.1 | eng1                    |  OK    |  OK         |
| 2.2 | eng1,eng2               |  OK    |  OK         |
| 2.3 | eng1,eng2,eng3          |  OK    |  OK         |
| 2.4 | memory:option           |  OK    |  OK         |
| 2.5 | memory:option,eng1      |  OK    |  OK         |
| 2.6 | eng1,memory:option      |  OK    |  OK         |
| 2.7 | memory:option,eng1,eng2 |  OK    |  OK         |
| 2.8 | eng1,memory:option,eng2 |  OK    |  OK         |
| 2.9 | eng1,eng2,memory:option |  OK    |  OK         |
+-----+-------------------------+--------+-------------+

- address review feedback (cleaner perl code)

- When the table is created with an attribute that is not supported by
  the storage engine, the attribute ought to be still kept.

Make it not to add '--console' when '--log-error' option is present
- Review feedback

Fixed a failure of the merge_innodb test on windows.


storage/myisammrg/ha_myisammrg.cc:
  Fixed a failure of the merge_innodb test on windows.

Bug #46212 safe_process: FATAL ERROR, Unknown option: --nocore
Also fixed mysqld.cc to avoid popup-boxes

- Make it not to add '--console' when '--log-error' option is present

include/my_global.h:
  MWL#17: Table elimination: fixes for windows
  - Add ALIGN_MAX_UNIT (assume malloc returns data aligned to this much)
mysql-test/r/table_elim.result:
  MWL#17: Table elimination: fixes for windows
  - Use only lower-case as EXPLAIN [EXTENDED] changes case of table names
    on windows
mysql-test/t/table_elim.test:
  MWL#17: Table elimination: fixes for windows
  - Use only lower-case as EXPLAIN [EXTENDED] changes case of table names
    on windows
sql/opt_table_elimination.cc:
  MWL#17: Table elimination: fixes for windows
  - Add extra alignment-padding-space for stack-allocated buffers.

 after the table has been converted to MyISAM and back. (PAGE_CHECKSUM is
 still there)

mysql-test/suite/maria/r/maria3.result:
  Update tests results after previous cset. 'TRANSACTIONAL=1' attribute goes 
  after the table has been converted to MyISAM and back. (PAGE_CHECKSUM is 
  still there)

Temporarily silence test failure due to MBug#430669 in test case introduced from
MySQL 5.1.38.

if the storage engine doesn't support transactions.

- Fix XtraDB to be built as a static library.

storage/xtradb/buf/buf0flu.c:
  Fix previous merge error (was harmless except for on win32)
storage/xtradb/plug.in:
  Make it sync with innobase/plug.in.disabled

mysql-test/t/mysqldump.test:
  Make test case work when build directory is not world readable
  (this is the case for Buildbot checkouts).
mysys/my_getopt.c:
  Restore bugfix which was lost in previous merge.
storage/xtradb/buf/buf0flu.c:
  Fix extranous line caused by bad merge.

Merge remaining bits of mysql-5.1.38 into MariaDB (9 commits were missing
from the original merge).

Fix version number (we are based on 5.1.38, not 5.1.39).
Fix mistaken translation in error message.

mysql-test/include/concurrent.inc:
  Merge

Fix some mistakes in the original merge found during review.

mysql-test/include/concurrent.inc:
  Add missing lines mistakenly omitted in merge.
mysql-test/include/mix1.inc:
  Revert wrong change done during merge.
mysql-test/include/wait_for_status_var.inc.moved:
  Remove file which was accidentally not removed during conflict resolution in merge.
mysql-test/r/innodb_mysql.result:
  Revert result file change following fixing root problem in merge.

- fix xtradb compile failures on windows (why do we have to do it again?)

conflict every time they make a change)

InnoDB does not bother to free explicitly one-time allocations at exit, so we get lots of
"still reachable" warnings in Valgrind with innodb-use-sys-malloc enabled.