originally based on the patch by Antony T Curtis

originally based on the patch by Antony T Curtis

cleanup sql_yacc.yy: rename rules for loops with and without label to follow BEGIN...END rule naming

preserve the Item_param type of the param_marker in the parser,
don't upcast cast it to Item

* split my_var class in three - base my_var and two descencants, move
  properties into descendants, remove if(), create a virtual method instead
* factor out the common code in the select_var_ident parser rule

instead of copy-pasting its methods into a new class

prepand() -> prepend()
concat() -> append()

make Item_param::get_settable_routine_parameter() public, because it's
public in the parent Item class and all other Item descendants too

move SP invocation code out of the huge switch in mysql_execute_command

instead of having it 1K everywhere, make it 1K on 32-bit and 2K on 64-bit.
As the latter has larger pointers (and larger sizeof(SHOW_VAR),
it needs a larger buffer to store the same amount of SHOW_VARs

also disable galera-specific output in mysql_tzinfo_to_sql,
it'll be enabled later.

* fix debian patch
* update the copyright
* rename include guards to follow conventions
* restore incorectly deleted test file, add clarification in a comment
* capitalize the first letter of the status variable

Provide Protocol_discard::storeXXX functions for all kinds
of objects that could be stored. Failure to so would caused
crashes in embedded server which attempted to collect the
stored info.

This also reduces the amount of copy operations for non-embedded
case.

Update the testcase so it is still meaningful when fix for MDEV-6657
is present in the code.

Merge with main branch

Removed compiler warnings

Added MAX_STATEMENT_TIME user variable to automaticly kill queries after a given time limit has expired.

- Added timer functions based on pthread_cond_timedwait
- Added kill_handlerton() to signal storage engines about kill/timeout
- Added support for GRANT ... MAX_STATEMENT_TIME=#
- Copy max_statement_time to current user, if stored in mysql.user
- Added status variable max_statement_time_exceeded
- Added KILL_TIMEOUT
- Removed digest hash from performance schema tests as they change all the time.
- Updated test results that changed because of the new user variables or new fields in mysql.user

This functionallity is inspired by work done by Davi Arnaut at twitter.
Test case is copied from Davi's work.

Documentation can be found at
https://kb.askmonty.org/en/how-to-limittimeout-queries/

mysql-test/r/mysqld--help.result:
  Updated for new help message
mysql-test/suite/perfschema/r/all_instances.result:
  Added new mutex
mysql-test/suite/sys_vars/r/max_statement_time_basic.result:
  Added testing of max_statement_time
mysql-test/suite/sys_vars/t/max_statement_time_basic.test:
  Added testing of max_statement_time
mysql-test/t/max_statement_time.test:
  Added testing of max_statement_time
mysys/CMakeLists.txt:
  Added thr_timer
mysys/my_init.c:
mysys/mysys_priv.h:
  Added new mutex and condition variables
  Added new mutex and condition variables
mysys/thr_timer.c:
  Added timer functions based on pthread_cond_timedwait()
  This can be compiled with HAVE_TIMER_CREATE to benchmark agains timer_create()/timer_settime()
sql/lex.h:
  Added MAX_STATEMENT_TIME
sql/log_event.cc:
  Safety fix (timeout should be threated as an interrupted query)
sql/mysqld.cc:
  Added support for timers
  Added status variable max_statement_time_exceeded
sql/share/errmsg-utf8.txt:
  Added ER_QUERY_TIMEOUT
sql/signal_handler.cc:
  Added support for KILL_TIMEOUT
sql/sql_acl.cc:
  Added support for GRANT ... MAX_STATEMENT_TIME=#
  Copy max_statement_time to current user
sql/sql_class.cc:
  Added timer functionality to THD.
  Added thd_kill_timeout()
sql/sql_class.h:
  Added timer functionality to THD.
  Added KILL_TIMEOUT
  Added max_statement_time variable in similar manner as long_query_time was done.
sql/sql_connect.cc:
  Added handling of max_statement_time_exceeded
sql/sql_parse.cc:
  Added starting and stopping timers for queries.
sql/sql_show.cc:
  Added max_statement_time_exceeded for user/connects status in MariaDB 10.0
sql/sql_yacc.yy:
  Added support for GRANT ... MAX_STATEMENT_TIME=# syntax, to be enabled in 10.0
sql/structs.h:
  Added max_statement_time user resource
sql/sys_vars.cc:
  Added max_statement_time variables
mysql-test/suite/roles/create_and_drop_role_invalid_user_table.test
  Removed test as we require all fields in mysql.user table.
scripts/mysql_system_tables.sql
scripts/mysql_system_tables_data.sql
scripts/mysql_system_tables_fix.sql
  Updated mysql.user with new max_statement_time field

thd->variables' table_plugin & tmp_table_plugin should be set to
NULL for wsrep system threads.
Also made a minor change to skip checking of wsrep options if
wsrep_on is not set.

- Simplified test cases in wsrep.variables
- Fixed a condition in wsrep_check_opts.cc
- Fixed an "unbound variable" in wsrep_sst_rsync

(lock != ctx->wait_lock)

References: lp:1364840 lp:1280896 - reverted a part of fix for
lp:1280896 (updating a unique key can cause parallel applying to hang )
in revision #4105. This "BF (brute force) lock skipping" caused
regression which surfaced in randgen test for bug lp:1364840

on CentOS 5 x86 (at least) this results in a binary
that cannot load plugins. Draw the line, quite
arbitrarily, at gcc 4.6.

instead of exec'ing /usr/sbin/ifconfig|grep|sed|awk

remove if() over many COM_xxx values

there is no reason to initialize wsrep start position
and auth from inside the get_one_option() callback