Problem was that listen_sockets where not properly protected.
Fixed by adding LOCK_start_thread around closing socket loop.

Other things:
- Removed not necessary test in close_socket(). With the current code
  'sock' can never be INVALID_SOCKET.

Main reason for this was there was a crash in shutdown of the server
in binlog_encryption.encryption_combo-mix and some other tests because
something in listen_sockets where not initialized. Changing to
Dynamic_array caused things to work.
Other reason for removing std::vector was that it is harder to debug,
and not integrated with DBUG, safemalloc, valgrind or memory
calculation and cause code explosions (extra code generated for each
std::vector type used).

Other things:
- Added size() function to Dynamic_array()

In the case of a crash directly after a creation of an Aria table,
Aria recovery would think that the table was from another system and
require a repair of the table and inform that the table is 'zerofilled".
This would cause no harm, but was confusing to see when testing atomic
alter table.

Fixed by logging the create transaction id to the log.

Other things:
- Added "show table status from sys" to maria_empy_logs. This ensures one
  does not get any zerofill warnings when sys/sys_config is used by other
  tests.
- aria_chk --describe now prints a warning if the table was moved from
  another system.
- Logging of truncate (maria_delete_all_rows) is changed to use the
  current trid for the create table.
  This is to ensure that we do not run into the same problem with truncate.
- Changed back sys_config table to Aria as this patch should fix the
  "zerofill" problem in buildbot.
- Added scripts/mysql_sys_schema.sql to .gitignore

The new default values WITH_URING:BOOL=OFF, WITH_PMEM:BOOL=OFF imply
that the dependencies are optional.
An explicit request WITH_URING=ON or WITH_PMEM=ON will cause the
build to fail if the requested dependencies are not available.

Last, to prevent a feature to be built in even though the built-time
dependencies are available, the following can be used:

cmake -DCMAKE_DISABLE_FIND_PACKAGE_URING=1
cmake -DCMAKE_DISABLE_FIND_PACKAGE_PMEM=1

This cleanup was suggested by Vladislav Vaintroub.

In commit 412533b4 (MDEV-18582),
one of the counters that was ported from XtraDB is useless.
Innodb_buffered_aio_submitted would be 0 or 1, depending on whether
is_linux_native_aio_supported() was executed to the point where
it would be incremented.

Let us remove this counter, because it has no practical value.
Even if its value were 1, io_setup() can still fail and we may
end up with innodb_use_native_aio=0.

In this platform, it s better not to rely on optional proc filesystem presence.
 Using native API to retrieve binary absolute path instead.

Fix for out-of-tree builds.

mysql_sys_schema.sql is a CMAKE_CURRENT_BINARY_DIR file so
use the buildpkgdatadir.

Fix by adding sys schema to cat_sql()

Print the build_equal_items() step for ON expression processing

Print the conditions for WHERE, HAVING, and ON.

This partially reverts 66106130

std version has an advantage of a more convenient units implementation from
std::chrono. Now it's no need to multipy/divide to bring anything to
micro seconds.

AIX detects tell in the configure however
it really isn't there. Use the my_seek aka lseek
implementation.

Don't remove (reasons given in Jira), instead add test coverage.
Improve other printout in best_access_path.

If a query with implicit grouping contains in MIN/MAX set function in the
select list over a column that is a part of an index then the query
might be subject to MIN/MAX optimization. With this optimization the
server performs a look-up into an index, fetches a value of the column C
used in the MIN/MAX function and substitute the MIN/MAX expression for this
value. This allows to eliminate the table containing C from further join
processing. In order the optimization to be applied the WHERE condition
must be a conjunction of simple equality/inequality predicates or/and
BETWEEN predicates.
The bug fixed in the patch resulted in fetching a wrong value from the
index used for MIN/MAX optimization. It may happened when a BETWEEN
predicate containing the MIN/MAX value followed a strict inequality.

Approved by dmitry.shulga@mariadb.com

Our intention is to make users aware that the ROW_FORMAT=COMPRESSED
format is on its way to deprecation.  It is an unnecessary annoyance
to users if we issue a warning when the user is converting a
ROW_FORMAT=COMPRESSED table to a supported format
(such as ROW_FORMAT=DYNAMIC).

ha_innobase::is_read_only(): Add a parameter to specify that an
ALTER TABLE to a supported format has been requested.

ha_innobase::check_if_supported_inplace_alter(): Set the parameter
when ALTER_OPTIONS to something else than ROW_FORMAT=COMPRESSED
(and without KEY_BLOCK_SIZE) is being requested.

Thanks to Elena Stepanova for suggesting this.

Until https://github.com/rr-debugger/rr/issues/2613
has been addressed, ./mtr --rr will fail to start up the
server if it has been configured with liburing.
To make the ./mtr --rr option work out of the box, we will
disable native asynchronous I/O.

Note: libaio never worked under rr, but it failed more gracefully:
the emulated io_setup() call would always return an error.

so that sys.schema_auto_increment_columns returns  correct
auto_increment_value

Fix confusing documentation. There is no installation necessary for sys schema.
it is there after mysql_install_db, or mysql_upgrade.

Tests with 4096-byte sector size confirm that it is
safe to use O_DIRECT with page_compressed tables.
That had been disabled on Linux, in an attempt to fix MDEV-21584
which had been filed for the O_DIRECT problems earlier.

The fil_node_t::block_size was being set mostly correctly until
commit 10dd290b (MDEV-17380)
introduced a regression in MariaDB Server 10.4.4.

fil_node_open_file(): Only avoid setting O_DIRECT on
ROW_FORMAT=COMPRESSED tables that use KEY_BLOCK_SIZE=1 or 2
(1024 or 2048 bytes).

fil_ibd_create(): Avoid setting O_DIRECT on ROW_FORMAT=COMPRESSED tables
that use KEY_BLOCK_SIZE=1 or 2 (1024 or 2048 bytes).

fil_node_t::find_metadata(): Require fstat() to be always invoked
outside Microsoft Windows, so that fil_node_t::block_size can be set.

fil_node_t::read_page0(): Rely on find_metadata() to assign block_size.

Thanks to Vladislav Vaintroub for testing this on Microsoft Windows
using an old-fashioned rotational hard disk with 4KiB sector size.

Reviewed by: Vladislav Vaintroub

This is a port of commit 00f620b2
and commit 6505662c from 10.2.