Problematic mutex is dict_sys.mutex.
Idea of the patch: unlink() fd under that mutex while
it's still open. This way unlink() will be fast and
actual file removal will happen on close().
And close() will be called outside of dict_sys.mutex.
This should be safe against crash which may happen between
unlink() and close(): file will be removed by OS anyway.
The same applies to both *nix and Windows.

I created and removed a 4G file on some NVMe SSD on ext4:

write(3, "\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1"..., 1048576) = 1048576 <0.000519>
fdatasync(3)                            = 0 <3.533763>
close(3)                                = 0 <0.000011>
unlink("file")                          = 0 <0.411563>

write(3, "\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1"..., 1048576) = 1048576 <0.000520>
fdatasync(3)                            = 0 <3.544938>
unlink("file")                          = 0 <0.000029>
close(3)                                = 0 <0.407057>

Such systems can benefit of this patch.

fil_node_t::detach(): closes fil_node_t but not file handle,
  returns that file handle

fil_node_t::prepare_to_close_or_deatch(): 'closes' fil_node_t

fil_node_t:close_to_free(): new argument detach_handle

fil_system_t::detach(): now can detach file handles

fil_delete_tablespace(): now can detach file handles

row_drop_table_for_mysql(): performs actual file removal

To install Spider one can simply drop a /etc/mysql/conf.d/spider.cnf like

  [mariadb]
  plugin-load-add=ha_spider.so

This is automatically generated and installed when plugin is correctly
registered to plugin.cmake with its own component name. Many other plugins
such as Connect and RocksDB install in the same way.

This solved MDEV-19917 as the mere adding and removing of spider.cnf
automatically installs and uninstalls it.

Remove the overly complex and uncecessary install.sql from Spider,
if should not be needed in modern times anymore.

With this change there is no need for a uninstall.sql either.

- Recommend max_allowed_packet=1G which is the same as the default client
  value.
- Remove thread_concurrency removed in 10.5.
- Remove query cache, not recommended practice anymore.
- Remove binlog_*, should not recommend those too easily but rather require
  the database administrator to read up on those themselves.
- Remove chroot setting, not relevant in modern container era.
- Show explicitly innodb_buffer_pool_size example as the most likely thing
  a database administrator should change.
- Don't recommend rate limiting in slow log, logging once in a 1000
  would not be optimal for the basic case, hence bad example.
- Install the example configs in /usr/share/mysql.
- Use correct path /run/ instead of /var/run/.

Split the big my.cnf into multiple smaller files with the same filenames
and contents as official Debian/Ubuntu packaging has.

The config contents stays the same apart from following additions
which the original MariaDB upstream configs had and probably needs
to be kept:
- lc-messages=en_US and skip-external-locking in server config

Configs the original MariaDB upstream had that are seemingly
unnecessary and thus removed:
- port=3306 removed from the client config
- log_warnings=2 removed from server config

Also adopt update-alternatives system using
mysql-common/configure-symlinks. This way it is aligned with
downstream Debian/Ubuntu packaging.

Analysis:
========
List of values provided for "replicate_ignore_table" and "replicate_do_table"
are stored in HASH.  When an empty list is provided the HASH structure doesn't
get initialized. Existing code treats empty element list as an error and tries
to clean the uninitialized HASH. This results in above MSAN issue.

Fix:
===
The clean up should be initiated only when there is an error while parsing the
'replicate_do_table' or 'replicate_ignore_table' list and the HASH is in
initialized state. Otherwise for empty list it should simply return success.

This corrects build failures on ppc64{,le} with the
WITH_EMBEDDED_SERVER option enabled.

MDEV-22641 added an unusual case in which the same object
file in was included twice with a different function
defination. The original cmake/merge_archives_unix.cmake
did not tolerate such eventualities.

So we move to the highest voted answer on Stack Overflow
for the merging of static libraries.
https://stackoverflow.com/questions/3821916/how-to-merge-two-ar-static-libraries-into-one

Thin archives generated compile failures and the libtool
mechanism would of been another dependency and using .la
files that isn't part of a normal cmake output. The straight
Apple mechanism of libtool with static archives also failed
on Linux.

This leaves the MRI script mechansim which was implemented
in this change.

To install Spider one can simply drop a /etc/mysql/conf.d/spider.cnf like

  [mariadb]
  plugin-load-add=ha_spider.so

This is automatically generated and installed when plugin is correctly
registered to plugin.cmake with its own component name. Many other plugins
such as Connect and RocksDB install in the same way.

This solved MDEV-19917 as the mere adding and removing of spider.cnf
automatically installs and uninstalls it.

Remove the overly complex and uncecessary install.sql from Spider,
if should not be needed in modern times anymore.

With this change there is no need for a uninstall.sql either.

- Recommend max_allowed_packet=1G which is the same as the default client
  value.
- Remove thread_concurrency removed in 10.5.
- Remove query cache, not recommended practice anymore.
- Remove binlog_*, should not recommend those too easily but rather require
  the database administrator to read up on those themselves.
- Remove chroot setting, not relevant in modern container era.
- Show explicitly innodb_buffer_pool_size example as the most likely thing
  a database administrator should change.
- Don't recommend rate limiting in slow log, logging once in a 1000
  would not be optimal for the basic case, hence bad example.
- Install the example configs in /usr/share/mysql.
- Use correct path /run/ instead of /var/run/.

Split the big my.cnf into multiple smaller files with the same filenames
and contents as official Debian/Ubuntu packaging has.

The config contents stays the same apart from following additions
which the original MariaDB upstream configs had and probably needs
to be kept:
- lc-messages=en_US and skip-external-locking in server config

Configs the original MariaDB upstream had that are seemingly
unnecessary and thus removed:
- port=3306 removed from the client config
- log_warnings=2 removed from server config

Also adopt update-alternatives system using
mysql-common/configure-symlinks. This way it is aligned with
downstream Debian/Ubuntu packaging.

Analysis:
========
List of values provided for "replicate_ignore_table" and "replicate_do_table"
are stored in HASH.  When an empty list is provided the HASH structure doesn't
get initialized. Existing code treats empty element list as an error and tries
to clean the uninitialized HASH. This results in above MSAN issue.

Fix:
===
The clean up should be initiated only when there is an error while parsing the
'replicate_do_table' or 'replicate_ignore_table' list and the HASH is in
initialized state. Otherwise for empty list it should simply return success.

This corrects build failures on ppc64{,le} with the
WITH_EMBEDDED_SERVER option enabled.

MDEV-22641 added an unusual case in which the same object
file in was included twice with a different function
defination. The original cmake/merge_archives_unix.cmake
did not tolerate such eventualities.

So we move to the highest voted answer on Stack Overflow
for the merging of static libraries.
https://stackoverflow.com/questions/3821916/how-to-merge-two-ar-static-libraries-into-one

Thin archives generated compile failures and the libtool
mechanism would of been another dependency and using .la
files that isn't part of a normal cmake output. The straight
Apple mechanism of libtool with static archives also failed
on Linux.

This leaves the MRI script mechansim which was implemented
in this change.

Change how lookup for the "auto" PSI_memory_keys is done.
Lookup for filename hashes (integers), instead of C strings

Generate these hashes at the compile time with constexpr,
rather than at runtime.

Let us invoke the debug member functions of mtr_t directly.

mtr_t::memo_contains(): Change the parameter type to
const rw_lock_t&. This function cannot be invoked on
buf_block_t::lock.

The function mtr_t::memo_contains_flagged() is intended to be invoked
on buf_block_t* or rw_lock_t*, and it along with
mtr_t::memo_contains_page_flagged() are the way to check whether
a buffer pool page has been latched within a mini-transaction.

xdes_get_state(), fseg_get_nth_frag_page_no(),
fseg_find_free_frag_page_slot(), fseg_find_last_used_frag_page_slot(),
fseg_get_n_frag_pages(), fseg_n_reserved_pages_low(),
fseg_print_low(): Remove the unused parameter mtr, and add
a const qualifier to the pointer to the buffer pool page frame.

This should have been part of
commit 70d4e55d.

debug_sync_set_action(): Declare the dummy function inline,
to silence a warning about declared-but-unused static function.
This amends commit 3ccd6766.

MDEV-22399: Remove multiple calls to enable and disable Handler::keyread and perform it after the plan refinement phase is done

Introduce a function to enable keyreads for indexes and use this
function when all the decision of plan refinement phase are done.

svr_n_page_hash_locks: Increase from 16 to 64. Before MDEV-15058,
we used to have the buf_pool.page_hash partitioned to each instance.

rw_lock_lock_word_decr(): Sleep a little in the spinloop.

rw_lock_s_lock_low(): Correct a comment. The function does perform
spinning.

This improves scalability in read-only workloads on a 32-CPU system
when the number of concurrent connections exceeds the CPU core count.

Thanks to Axel Schwenke for running benchmarks.

The issue here is charset for Sort_param::tmp_buffer is cleared when bzero is done for Sort_param.
Make sure to set the charset explicitly in the constructor for tmp_buffer.

reduce the amount of engine-specific code in the server,
particularly as it does not serve any purpose now.

may be needed for VP engine,
to be reconsidered in MDEV-7795

Disallow BIT_AND(), BIT_OR(), BIT_XOR() for data types GEOMETRY and INET6,
as they cannot return any useful integer values.

MDEV-22325 ib_logfile0 is too small for innodb_thread_concurrency=0. The size of ib_logfile0 should be bigger than 200 kB * innodb_thread_concurrency.

Correct log message. IMO, we shouldn't be very precise in that
message as the formula behind it is not trivial.

Also performed a little cleanup.

buf_LRU_make_block_young(): Merge with buf_page_make_young().

buf_pool_check_no_pending_io(): Remove. Replaced with
buf_pool.any_io_pending() and buf_pool.io_pending(),
which do not unnecessarily acquire buf_pool.mutex.

buf_pool_t::init_flush[]: Use atomic access, so that
buf_flush_wait_LRU_batch_end() can avoid acquiring buf_pool.mutex.

buf_pool_t::try_LRU_scan: Declare as bool.

When MDEV-22769 introduced srv_shutdown_state=SRV_SHUTDOWN_INITIATED in
commit efc70da5
we forgot to adjust a few checks for SRV_SHUTDOWN_NONE.

In the initial shutdown step, we are waiting for the background
DROP TABLE queue to be processed or discarded. At that time,
some background tasks (such as buffer pool resizing or dumping
or encryption key rotation) may be terminated, but others must
remain running normally.

srv_purge_coordinator_suspend(), srv_purge_coordinator_thread(),
srv_start_wait_for_purge_to_start(): Treat SRV_SHUTDOWN_NONE
and SRV_SHUTDOWN_INITIATED equally.

dict_table_t::parse_name(): Strip any partition or subpartition
name before copying the name to the decoding buffer.

MDEV-19320 Sequence gets corrupted and produces ER_KEY_NOT_FOUND
           (Can't find record) after ALTER .. ORDER BY

MDEV-19977 Assertion `(0xFUL & mode) == LOCK_S ||
           (0xFUL & mode) == LOCK_X' failed in lock_rec_lock

create_table_info_t::create_foreign_keys(): Make the create_name buffer
long enough for both the database and table name. It is still not long
enough to hold partition or subpartition names. Because we do never
supported FOREIGN KEY constraints on partitions, we can simply skip
the call to innobase_convert_name() on CREATE TABLE.

MDEV-22719 Long unique keys are not created when individual key_part->length < max_key_length but SUM(key_parts->length) > max_key_length

Make UNIQUE HASH key in case when key_info->key_length > max_key_length

MDEV-21804 Assertion `marked_for_read()' failed upon INSERT into table with long unique blob under binlog_row_image=NOBLOB

Problem:- Calling mark_columns_per_binlog_row_image() earlier may change the
result of mark_virtual_columns_for_write() , Since it can set the bitmap on
for virtual column, and henceforth  mark_virtual_column_deps(field) will
never be called in mark_virtual_column_with_deps.

This bug is not specific for long unique, It also fails for this case
   create table t2(id int primary key, a blob, b varchar(20) as (LEFT(a,2)));