Optimizing cascade node list traversal, which could turn out as
performance bottleneck
Even this current cascade node check could be skipped, but a dedicated
mtr test is needed to confirm that

* created tests focusing in multi-master conflicts during cascading foreign key
  processing
* in row0upd.cc, calling wsrep_row_ups_check_foreign_constraints only when
  running in cluster
* in row0ins.cc fixed regression from MW-369, which caused crash with MW-402.test

MySQL 5.7 introduced some optimizations to avoid file I/O during
ALGORITHM=INPLACE operations. While both innodb-index-online and
innodb-table-online will exercise both the merge sort files and
the online log files in 10.1, in 10.2 they would only exercise the
online log files.

Modify one test case in innodb.innodb-table-online so that
skip_pk_sort will not hold. In this way, this test case will
write and read the merge sort files. The other instrumented tests
in innodb-index-online and innodb-table-online will only write
and read online_log files.

This should also fix the MariaDB 10.2.2 bug
MDEV-13826 CREATE FULLTEXT INDEX on encrypted table fails.

MDEV-12634 FIXME: Modify innodb-index-online, innodb-table-online
so that they will write and read merge sort files. InnoDB 5.7
introduced some optimizations to avoid using the files for small tables.

Many collation test results have been adjusted for MDEV-10191.

Compilation got confused  about 2 wsrep.h headers in include path
Rename backup's wsrep.h to backup_wsrep.h to fixO

Introduce innodb_encrypt_log.combinations and prove that
the encryption and decryption take place during both
online ADD INDEX (WL#5266) and online table-rebuilding ALTER (WL#6625).

Apply the same changes to both InnoDB and XtraDB.

message was written.

The messages are getting lost because they are written with fprintf()
but without fflush(), so abort() would lose buffered text.

Applied fix from 10.2, which makes in_logf() use
sql_print_information(), which in turn does proper flush after each message

MDEV-13807 mariabackup --apply-log-only does generate redo log by performing rollback and possibly other tasks

Skip rollback and other redo-log-generating tasks if
srv_apply_log_only is set.

Instead of assigning the debug variable recv_no_log_write = FALSE,
assign it to srv_apply_log_only, so that any unwanted writes are caught.

The underlying data members are declared public, so there is no
point in hiding these const accessors.

Use dict_index_is_auto_gen_clust() instead.

The problem was introduced by the patch for MDEV-7661,
which (in addition to the fix itself) included an attempt to make
CONVERT/CAST work in the same way with fields
(i.e. return NULL in strict mode if a non-convertable character found).
It appeared to be a bad idea and some users were affected by this
behavior change. Changing CONVERT/CAST not depend on sql_mode
(restoring pre-10.1.4 behavior).

use ut_get_name() for formatting database/table names.

* recorded new result file, original was copied from MariaDB
  which produces syntactically a bit different result
* semantics are same in MariaDB and ours result file

* changed thd_binlog_format to return configured binlog format in wsrep execution,
  regardless of binlogging setting (i.e. with or without binlogging)
* thd_binlog_format is used in innobase::write_row(), and would return confusing
  result there when log_bin==OFF

* ported the partition.test from MariaDB 10.3 code tree
* fix to TOI replicate ALTER TABLE EXCHANGE PARTITION statement

Import the changes to innodb.innodb-index innodb.innodb-index-debug

Note: As noted in MDEV-13613, due to the behaviour change in MDEV-11114,
DROP COLUMN will not imply DROP/ADD PRIMARY/UNIQUE KEY, like it does
in MySQL. The tests have been adjusted accordingly.

…porary file

Fixed by removing writing key version to start of every block that
was encrypted. Instead we will use single key version from log_sys
crypt info.

After this MDEV also blocks writen to row log are encrypted and blocks
read from row log aren decrypted if encryption is configured for the
table.

innodb_status_variables[], struct srv_stats_t
	Added status variables for merge block and row log block
	encryption and decryption amounts.

Removed ROW_MERGE_RESERVE_SIZE define.

row_merge_fts_doc_tokenize
	Remove ROW_MERGE_RESERVE_SIZE

row_log_t
	Add index, crypt_tail, crypt_head to be used in case of
	encryption.

row_log_online_op, row_log_table_close_func
	Before writing a block encrypt it if encryption is enabled

row_log_table_apply_ops, row_log_apply_ops
	After reading a block decrypt it if encryption is enabled

row_log_allocate
	Allocate temporary buffers crypt_head and crypt_tail
	if needed.

row_log_free
	Free temporary buffers crypt_head and crypt_tail if they
	exist.

row_merge_encrypt_buf, row_merge_decrypt_buf
	Removed.

row_merge_buf_create, row_merge_buf_write
	Remove ROW_MERGE_RESERVE_SIZE

row_merge_build_indexes
	Allocate temporary buffer used in decryption and encryption
	if needed.

log_tmp_blocks_crypt, log_tmp_block_encrypt, log_temp_block_decrypt
	New functions used in block encryption and decryption

log_tmp_is_encrypted
	New function to check is encryption enabled.

Added test case innodb-rowlog to force creating a row log and
verify that operations are done using introduced status
variables.

Fix memory leaks, and add a missing newline to a message.

The type uint16_t is apparently promoted to int in a comparison,
causing a sign mismatch when comparing to ulint. Convert to ulint
before comparison.

The type uint16_t is apparently promoted to int in a comparison,
causing a sign mismatch when comparing to ulint.
Convert both operands to uint16_t.

In online table-rebuilding ALTER TABLE (LOCK=NONE), virtual column values
are being written to the online_log. WL#8149 in MySQL 5.7 changed some
low-level functions that are also being used outside row0log.cc, causing
performance penalty to other code.

We revert those changes, and introduce separate functions for writing
the virtual column values.

The only functional change should be the one that is mentioned in
MDEV-13795: row_log_table_low_redundant() will no longer write
virtual column values along with old_pk, just like row_log_table_low().

As noted in MDEV-13795, some forms of table-rebuilding ALTER with
virtual columns is broken. At least DROP PRIMARY KEY, ADD PRIMARY KEY
is broken.

rec_get_converted_size_temp(), rec_convert_dtuple_to_temp():
Remove the parameter for passing virtual column values.

rec_get_converted_size_temp_v(), rec_convert_dtuple_to_temp_v():
New functions for appending virtual column values to the online_log.

rec_get_converted_size_comp_prefix_low(),
rec_convert_dtuple_to_rec_comp(): Remove the v_entry parameter,
and do not allow n_fields=0.

The InnoDB purge subsystem can be best stopped by opening a read view,
for example by START TRANSACTION WITH CONSISTENT SNAPSHOT.

To ensure that everything is purged, use wait_all_purged.inc,
which waits for the History list length in SHOW ENGINE INNODB STATUS
to reach 0. Setting innodb_purge_run_now never guaranteed this.

This also avoids hitting the hang at server shutdown.
That hang should be fixed, of course.

        set_charset() calls added for Item_func_json_arran
        and Item_func_json_object::val_str-s.

        Set default charset for temporary paths so UDF call don't crash.

        JSON_QUOTE return type set to be JSON.

        Tests added.

        '=0' added to meke the compiler happy.