The problem was that multi_range_read_info_const() called
multi_range_key_create_key() which changed m_part_spec.start_part,
while there was an activ table scan ongoing.

Fixed by copying and restoring m_part_spec around
multi_range_key_create_calls()

Part #2: Take into accont the fact that three binary collations that
we handle are technically NOPAD collations

Disallow use of NOPAD collations in indexed columns.

first, and we do not not care whether client has received all data.

This is a TCP optimization to avoid TIME_WAIT in TCP connection teardown.

This patch would abort connection on timeout, which usually happens when
client reads a large result set, at slower pace then the server can
write.

The patch also cleans up socket timeout handling, so that Windows
is consistent with another platforms (using nonblocking socket IO
 + waiting in poll/select on single socket, rather than setsockopt).
This makes identifying timeouts easier.


Also removed the superficial shutdown() before closesocket() in a few
places where it was used, because it  was never needed , and
reportedly breaks SO_LINGER on Windows.

When the transaction isolation level is SERIALIZABLE, or when
a locking read is performed in the REPEATABLE READ isolation level,
InnoDB must lock delete-marked records in order to prevent another
transaction from inserting something.

However, at READ UNCOMMITTED or READ COMMITTED isolation level or
when the parameter innodb_locks_unsafe_for_binlog is set, the
repeatability of the reads does not matter, and there is no need
to lock any records.

row_search_mvcc(): Skip locks on delete-marked committed records upfront,
instead of invoking row_unlock_for_mysql() afterwards. The unlocking
never worked for secondary index records.

dict_stats_rename_table(): After DB_LOCK_WAIT_TIMEOUT
or DB_DUPLICATE_KEY, reset the trx->error_state before retrying.
Also, properly treat DB_DEADLOCK as a hard error.

Part #2: Adjust other parts of testsuite to take into account
that RocksDB files are in "#rocksdb" now.

Part2: make MyRocks add its directory into @@ignore_db_dirs when starting.

This is necessary because apparently not everybody are using plugin's my.cnf
So load ha_rocksdb.{so,dll} manually and then hit MDEV-12451, MDEV-14461
etc.

Rename  Use '#rocksdb' instead of '.rocksdb' as subdirectory for RocksDB
datadir.

MDEV-14614 InnoDB: Failing assertion: trx->error_state == DB_SUCCESS or lock wait timeout upon saving statistics

The assertion failure was caused by
MDEV-14511 Use fewer transactions for updating InnoDB persistent statistics

We are reusing a transaction object after commit, and sometimes,
even after a successful operation, the trx_t::error_state may be
something else than DB_SUCCESS. Reset the field when needed.

    commit 6d63a034
    MDEV-11297: Add support for LIMIT clause in GROUP_CONCAT()

  * Note: breaking change; since this commit, a plugin that has
    worked so far might get rejected due to plugin maturity
  * mariabackup is not affected (allows all plugins)
  * VERSION file defines SERVER_MATURITY, which defines the
    corresponding numeric value as SERVER_MATURITY_LEVEL in
    include/mysql_version.h
  * The default value for 'plugin_maturity' is SERVER_MATURITY_LEVEL - 1
  * Logs a warning if a plugin has maturity lower than
    SERVER_MATURITY_LEVEL
  * Tests suppress the plugin maturity warning
  * Tests use --plugin-maturity=unknown by default so as not to fail
    due to the stricter plugin maturity handling

Before MySQL 5.7 or MariaDB 10.2.2, there used to be some
magic InnoDB table names that would assign some InnoDB flags
on CREATE TABLE or DROP TABLE.

Starting with MySQL 5.7 (or MariaDB 10.2.2) InnoDB no longer contains
the "table monitor" or "tablespace monitor". The conditions on
srv_print_innodb_tablespace_monitor, srv_print_innodb_table_monitor
never hold. So, the code was dead.

Also, remove a bogus reference to dict_print(), which used to implement
the InnoDB table monitor.

fil_get_space_names(): Remove.

fts_drop_orphaned_tables(): Iterate fil_system->space_list directly.

Relax memory barrier for lock_word.

rw_lock_lock_word_decr() - used to acquire rw-lock, thus we only need to issue
ACQUIRE when we succeed locking.

rw_lock_x_lock_func_nowait() - same as above, but used to attempt to acquire
X-lock.

rw_lock_s_unlock_func() - used to release S-lock, RELEASE is what we need here.

rw_lock_x_unlock_func() - used to release X-lock. Ideally we'd need only RELEASE
here, but due to mess with waiters (they must be loaded after lock_word is
stored) we have to issue both ACQUIRE and RELEASE.

rw_lock_sx_unlock_func() - same as above, but used to release SX-lock.

rw_lock_s_lock_spin(), rw_lock_x_lock_func(), rw_lock_sx_lock_func() -
fetch-and-store to waiters has to issue only ACQUIRE memory barrier, so that
waiters are stored before lock_word is loaded.

Note that there is violation of RELEASE-ACQUIRE protocol here, because we do
on lock:

  my_atomic_fas32_explicit((int32*) &lock->waiters, 1, MY_MEMORY_ORDER_ACQUIRE);
  my_atomic_load32_explicit(&lock->lock_word, MY_MEMORY_ORDER_RELAXED);

on unlock

  my_atomic_add32_explicit(&lock->lock_word, X_LOCK_DECR, MY_MEMORY_ORDER_ACQ_REL);
  my_atomic_load32_explicit((int32*) &lock->waiters, MY_MEMORY_ORDER_RELAXED);

That is we kind of synchronize ACQUIRE on lock_word with ACQUIRE on waiters.
It was there before this patch. Simple fix may have negative performance impact.
Proper fix requires refactoring of lock_word.

Relax memory barrier for waiters: these 2 stores must be completed before
os_event_set() finishes. This is guaranteed by RELEASE barrier issued by
mutex.exit() of os_event_set().

Remove volatile modifier from waiters: it's not supposed for inter-thread
communication, use appropriate atomic operations instead.

Changed waiters to int32_t, my_atomic friendly type.

Remove volatile modifier from lock_word: it's not supposed for inter-thread
communication, use appropriate atomic operations instead.

Change lock_word from lint to int32_t: the latter is my_atomic_* friendly type.

- Changed rocksdb.2pcgroup_commit.test to print information on error
- Updated myisam_views-big.result

LOCK_thd_data was used to protect both THD data and
ensure that the THD is not deleted while it was in use

This patch moves the THD delete protection to LOCK_thd_kill,
which already protects the THD for kill.

The benefits are:
- More well defined what LOCK_thd_data protects
- LOCK_thd_data usage is now much simpler and easier to verify
- Less chance of deadlocks in SHOW PROCESS LIST as there is less
  chance of interactions between mutexes
- Remove not needed LOCK_thread_count from
  thd_get_error_context_description()
- Fewer mutex taken for thd->awake()

Other things:
- Don't take mysys->var mutex in show processlist to check if thread
  is kill marked
- thd->awake() now automatically takes the LOCK_thd_kill mutex
  (Simplifies code)
- Apc uses LOCK_thd_kill instead of LOCK_thd_data

This will allow show processlist to continue, without blocking
all new connections, if some threads gets stuck while holding
LOCK_thd_data or mysys_var->mutex

Connections that has mutex 'stuck' are marked as 'Busy' in 'Command'

Todo:
Make F_BACKOFF to do 'pause' instead of just (1)