binlog_savepoint_rollback() should not try to truncate a binlog
unless binlog_savepoint_set has actually remembered the position
to truncate to.

Note: some tests fail, just as they failed before the merge!

NOT NULL constraint must be checked *after* the BEFORE triggers.
That is for INSERT and UPDATE statements even NOT NULL fields
must be able to store a NULL temporarily at least while
BEFORE INSERT/UPDATE triggers are running.

* move common code to a new set_bad_null_error() function
* move repeated comparison out of the loop
* remove unused code
  * unused method Table_triggers_list::set_table
  * redundant condition (if (table) after table was dereferenced)
* add an assert

because it's locked for innodb part of SHOW STATUS,
which can happen in read-only mode too.

incorrect option name and doing ALTER TABLE under read-only:
the test "passed" in 10.0 only because $restart_parameters
was first implemented in 10.1

        Table content filled appropriately.
        Thare are still just two records as we don't have geodetics yet.

In wsrep BF we have already took lock_sys and trx
mutex either on wsrep_abort_transaction() or
before wsrep_kill_victim(). In replication we
could own lock_sys mutex taken in
lock_deadlock_check_and_resolve().

The bug was caused by accessing uninitialized fields within the LEX related to
ssl by mysql_show_grants() -> get_current_user() -> has_auth() function.

MDEV-9249 MariaDB un-buildable on linux64: fails @ "error: ‘ERR_remove_state’ was not declared in this scope" when linking against OpenSSL 1.0.2e

ERR_remove_state is deprecated, use ERR_remove_thread_state
if possible

- Add CA validation to wsrep_sst_xtrabackup-v2.sh.
- Also added a few {} around tpem for consistency.
- Abort if encryption is requested but socat is not ssl-enabled.

Patch contributed by : Klaas Demter

Relax the number-of-XA-engines check on recovery. Allow *more*
engines to be present than absolutely necessary, extra engines
cannot affect ACID guarantees of the recovery process.

As a bonus, 10.0->crash->10.1 upgrade won't complain about
wsrep being a new XA storge engine.

plugin_init() works like this:
1. init MyISAM
2. load plugins from mysql.plugin, if it's a MyISAM table
3. init all not initialized plugins
4. all done, if step 2 loaded mysql.plugin,
   otherwise:
5. load plugins from mysql.plugin
6. init all not initialized plugins

now, with --help --verbose, step 3 will not actually
initialize them, and if mysql.plugin is unreadable,
step 6 will try to initialize existing plugins again.
Fix: when skipping initialization because of --help,
change plugin status from PLUGIN_IS_UNINITIALIZED.

On shutdown feedback was sending a short report without creating
a THD. At that point current_thd was pointing to the already
destroyed THD from the previous full report.

backport from 10.1:
  commit bfe703a4
  Author: Sergei Golubchik <serg@mariadb.org>
  Date:   Tue Feb 3 18:19:56 2015 +0100

      don't let current_thd to point to a destroyed THD

A string definition is inconsistent and thus causes a compilation error.
Signed-off-by: Vicențiu Ciorbaru <vicentiu@mariadb.org>

Fixed compile warning related to if statement always being true. The if
statement can not be false, as the address of a member field is always
true.

Removed unused functions from tokudb_dump.cc.

This patch fixes another compilation error caused by specifying
attribute nonnull for all the parameters of the copyout function. This
is incorrect as the function actually gets called with null parameters
indirectly and thus only the output parameter should be nonnull.

This patch fixes one compilation error related to __db_lsn struct. The
struct can not be defined as empty according to the main C standard.
In C++, this is handled by forcing a size of 1. To eliminate the error
we add a dummy char field of size 1. This has no effect on the C++
compiled code, but also removes the compiler error.

10.0 has an "analyze table .. persistent for all" syntax. This adds
--persistent to mysqlcheck(mysqlanalyize) to perform this extended
analyze table option.
Signed-off-by: Vicențiu Ciorbaru <vicentiu@mariadb.org>

unit test for the IO_CACHE bug

While refreshing the IO_CACHE, do not update its parameters (read_pos,
read_end & pos_in_file) if there is nothing left to read.

Currently we support file block sizes 512K-4K.

Do not build optimised power crc32 on bigendian

Part I: Add diagnostics to page allocation if state is not correct
but do not assert if it is incorrect.