get_col_list_to_be_dropped() incorrectly returned uninteresting instantly
dropped column which was missing in a new dict_index_t

get_col_list_to_be_dropped(): rename to collect_columns_from_dropped_indexes
and stop return dropped columns

Closed PR#1383

Changes for 10.4 only, please null-merge to 10.5

Use EITS statistics to avoid changing query plans

Caused by:
MDEV-18501 Partition pruning doesn't work for historical queries (cleanup)

gcc 7.4.1

Pruning fix for SYSTEM_TIME INTERVAL partitioning.

Allocating one more element in range_int_array for CURRENT partition
is required for RANGE pruning to work correctly
(get_partition_id_range_for_endpoint()).

SYSTEM_TYPE partitioning: COLUMN properties removed. Partitioning is
now pure RANGE based on UNIX_TIMESTAMP(row_end).

DECIMAL type is now allowed as RANGE partitioning, we can partition by
UNIX_TIMESTAMP() (but not for DATETIME which depends on local timezone
of course).

Cleanup removes useless allocation.

(2)fix: bug when call create tmp table

(2) remove empty "--error ER_PARSE_ERROR"
(3) change three members in class select_unit to protected.

This patch introduces the optimization that allows range optimizer to
consider index range scans that are built employing NOT NULL predicates
inferred from WHERE conditions and ON expressions.
The patch adds a new optimizer switch not_null_range_scan.

This patch corrects the fix of the patch for mdev-19421 that resolved
the problem of parsing some embedded join expressions such as
  t1 join t2 left join t3 on t2.a=t3.a on t1.a=t2.a.
Yet the patch contained a bug that prevented proper context analysis
of the queries where such expressions were used together with comma
separated table references in from clauses.

FIXME: rocksdb.drop_table hangs on shutdown, but is not disabled in 10.2

row_upd_build_difference_binary(): Correctly handle the
case where columns (or clustered index fields) have been added
since the 'entry' was originally created. In this case,
the update vector must replace any missing columns with the
default values of the instantly added columns.

Merge the changes to include/index_merge*inc from the upstream. The changes
add this command in many places:

 +if ($engine_type == RocksDB)
 +{
 +    set global rocksdb_force_flush_memtable_now=1;
 +}

also add it in one more place to make the test truly stable.

With --skip-debug-assert, DBUG_ASSERT(false) will allow execution to
continue. Hence, we will need /* fall through */ after them.

Some DBUG_ASSERT(0) were replaced by break; when the switch () statement
was followed by DBUG_ASSERT(0).

Adjust the testcase according to the review input

(Backported to 10.3, addressed review input)

Sj_materialization_picker::check_qep(): fix error in cost/fanout
calculations:
- for each join prefix, add #prefix_rows / TIME_FOR_COMPARE to the cost,
  like best_extension_by_limited_search does
- Remove the fanout produced by the subquery tables.
- Also take into account join condition selectivity

optimize_wo_join_buffering() (used by LooseScan and FirstMatch)
- also add #prefix_rows / TIME_FOR_COMPARE to the cost of each prefix.
- Also take into account join condition selectivity

MariaDB 10.4 was crashing when thread-handling was set to
pool-of-threads and wsrep was enabled.

There were two apparent reasons for the crash:
- Connection handling in threadpool_common.cc was missing calls to
  control wsrep client state.
- Thread specific storage which contains thread variables (THR_KEY_mysys)
  was not handled appropriately by wsrep patch when pool-of-threads
  was configured.

This patch addresses the above issues in the following way:
- Wsrep client state open/close was moved in thd_prepare_connection() and
  end_connection() to have common handling for one-thread-per-connection
  and pool-of-threads.
- Thread local storage handling in wsrep patch was reworked by introducing
  set of wsrep_xxx_threadvars() calls which replace calls to
  THD store_globals()/reset_globals() and deal with thread handling
  specifics internally.

Wsrep-lib was updated to version which relaxes internal concurrency
related sanity checks.

Rollback code from wsrep_rollback_process() was extracted to separate calls
for better readability.

Post rollback thread was removed as it was completely unused.

Mroonga does not build on Windows due to log_10_int being used
(indirectly) and n ot properly decorated with MYSQL_PLUGIN_IMPORT

The test occasionally fails with different table reference count
due to purge activity after INSERT operations (MDEV-12288).
Wait for purge before accessing dict_table_t::n_ref_count.

Merge a part of commit 25af2a18.
This was forgotten in the
merge commit e41eb044.

Commit 536215e3 in MariaDB Server 10.3.1
introduced the compiler flag (not cmake option) DBUG_ASSERT_AS_PRINTF
that converts DBUG_ASSERT in non-debug builds into printouts.

For debug builds, it could be useful to be able to convert DBUG_ASSERT
into a warning or error printout, to allow execution to continue.
This would allow debug builds to be used for reproducing hard failures
that occur with release builds.

my_assert: A Boolean flag (set by default), tied to the new option
debug_assert that is available on debug builds only.
When set, DBUG_ASSERT() will invoke assert(), like it did until now.
When unset, DBUG_ASSERT() will invoke fprintf(stderr, ...)
with the file name, line number and assertion expression.

This fixed a bug that caused the amazon.test to fail