(possibly intermittent) checksum mismatch.

is detected.

The checksum mismatch can be due to partial write, thus retry the read

Window is a reserved keyword according to SQL Standard 2016. However, we
can make the grammar slightly flexible by allowing WINDOW keyword everywhere
except table aliases. Change yacc grammar to separate between all keywords
and table_alias keywords.

- get rid of casts between int and HANDLE.
- store file descriptor in TP_connection_generic to get
rid of multiple mysql_socket_getfd(c->thd->net.vio->mysql_socket))

Also support named pipes (no reason not to support, since it is easy)

mariabackup fails to builds with PFS disabled. The reason was missing include,
which was included by PFS otherwise.

# Conflicts:
#	storage/innobase/include/os0file.h

When Mariabackup is invoked on an instance that uses a multi-file
InnoDB system tablespace, it may fail to other files of the system
tablespace than the first one.
This was revealed by the MDEV-14447 test case.
The offending code is assuming that the first page of each data file
is page 0. But, in multi-file system tablespaces that is not the case.

xb_fil_cur_open(): Instead of re-reading the first page of the file,
rely on the fil_space_t metadata that already exists in memory.

xb_get_space_flags(): Remove.

Update C/C to include fix for this bug.

for multi-file innodb_data_file_path.

Use fil_extend_space_to_desired_size() to correctly extend system
tablespace. Make sure to get tablespace size from the first tablespace
part.

MDEV-12012. Post-push attempt to catch failure in rpl_gtid_delete_domain failing on P8. The test is made more verbose.

Import and adjust the MySQL 5.7 tests
innodb.update_time innodb.update_time_wl6658 into MariaDB.

The functionality is present since MariaDB 10.2.2 merged
InnoDB from MySQL 5.7.9. It was implemented in MySQL 5.7.2.

List<>::last is wrong after memcpy(). Doing it on constructed objects is bad practice.

* rocksdb fails without timer_delete() - only use it when it exists

- introduce system check for posix_memalign (not available on Solaris 10)
- Disable dtrace probes, to fix weird link errors in mariabackup

we now have cmake/submodules.cmake that updates all submodules

another followup for 4c2c057d.

there are six possible cases:
--port can be set or not.
--address can be set, not set, or set but without a port number

The correct behavior is:
1 both --port and --address have a port number
  - use it if it's the same, otherwise an error
2 only --port has the number (--address isn't set)
  - use the value from --port
3 only --port has the number (--address is set, but has no port)
  - use the value from --port
4 --port is unset, --address has the port number
  - use the value from --address
5 --port is unset, --address has no port number
  - use the value from --address, that is, port is empty string
6 --port is unset, --address is unset
  - port is unset (an error somewhere later)

case 5 wasn't handled correctly

move the privilege related test to main.cte_grant

dict_stats_exec_sql(): Refuse the operation if shutdown has been
initiated.

The real fix would be to update the persistent statistics as part
of the data dictionary transactions. To do this, we should move the
storage of InnoDB persistent statistics to the InnoDB data files,
and maybe also remove the InnoDB data dictionary.