There is identical logic for all grant functions. If a grant grants to
non existant users they are created if the SQL mode is not
NO_AUTO_CREATE_USER *and* the user granting has rights to create users.

This check is done via test_if_create_new_users. Refactor all code such
that this check is only done in one (or three places for-now till more
refactoring happens).

There are a number of get_current_user calls which can cause a malloc.
Reorganize code within mysql_grant_xxx such that it assumes LEX_USERS
are already resolved.

This eliminates a number of mallocs and also simplifies code in
mysql_grant_xxx functions.

grant_stage0 now resolves lex_users, places it in a list copy,
copies the clone's auth structures. Sql_cmd_grant_*::execute now
uses the list for all subsequent mysql_grant_xxx functions calls.

This change splits mysql_grant based on the type of grants issued.

* mysql_grant_global -> GRANT <priv> ON *.*
* mysql_grant_db     -> GRANT <priv> ON <database>.*
* mysql_grant_proxy  -> GRANT PROXY ...

This does introduce a bit of code duplication, however it allows further
refactoring and also isolates the grant modifying logic away from other
details such as table opening, locking, binlogging.

Another side effect is that the mysql.db table is only locked for a
database level grant. Before it was locked even if there were only
global level grants issued.

The class functions as a bundle of how privileges must be altered for
the corresponding ACL_USER (and other ACL_xxx) classes.

This stops mysql_x_grant function's prototype from ever expanding with
additional flags. This is in preparation for DENY clause.

Mark the function static and clean up some leftover references.
The replacement function is acl_get_all3.

command_array wasn't particularly meaningful as a name.

Also use LEX_CSTRING instead of 2 separate arrays.

The functions are now behind Sql_cmd_grant::execute so they no longer
need to be exposed as part of sql_acl.h

* This cleanup reduces the size of Sql_cmd_grant_object because it eliminates
  an uneeded copy. Grant_privileges is stored as a reference to already
  thd->mem_root allocated Lex_grant_privileges.
* Extend member functions of Grant_privileges to provide access to
  protected members.

* This makes the table getters and setters idempotent.
  Slight performance decrease: When running on a Tabular user table, we
  set fields (in memory) that have previously been set.

The code was "hiding' a get - set - get sequence of calls for
User_table_json. This is now more apparent. The first "get" is required
to obtain the "adjusted" privileges based on which MariaDB version wrote
the privileges in global_priv.

The second get is required for User_table_tabular, to map column
privileges to actual newly introduced privileges
For example SUPER column marked as YES maps to
GLOBAL_SUPER_ADDED_SINCE_USER_TABLE_ACLS, which is a set of more
granular privileges.

Privilege checking functions should be grouped in sql_acl.

Creating an object copy would imply that the destructor would delete the
underlying same HASH object for both instances. That would lead to a
double free and thus a crash.

During insert, the value isn't mutated so we can accept a const ptr.

Delete the copy constructor for Dynamic_array and use proper move
semantics for acl_dbs.

The String isn't mutated so it can be marked as const.

* conflicting types for grn_object_columns
Make sure functions have the same signature in c and header files.

log_t::write_checkpoint(), log_t::resize_start(): Invoke buf_flush_ahead()
with buf_pool.get_oldest_modification(0)+1 so that another checkpoint will
be invoked, to complete the log resizing.

Tested by: Oleg Smirnov (on Microsoft Windows, where this hung most often)

ha_innobase::delete_table(): Also on DROP SEQUENCE, do try to drop any
persistent statistics. They should really not be created for
SEQUENCE objects (which internally are 1-row no-rollback tables),
but that is how happened to always work.

i_s_innodb_buffer_page_get_info(): Correct a condition.
After crash recovery, there may be some buffer pool pages in FREED state,
containing garbage (invalid data page contents). Let us ignore such pages
in the INFORMATION_SCHEMA output.

The test innodb.innodb_defragment_fill_factor will be removed, because
the queries that it is invoking on information_schema.innodb_buffer_page
would start to fail. The defragmentation feature was removed in
commit 7ca89af6 in MariaDB Server 11.1.

Tested by: Matthias Leich

btr_search_hash_table_validate(), btr_search_validate(): Add the
parameter THD for checking if the statement has been killed.
Any non-QUICK CHECK TABLE will validate the entire adaptive hash index
for all InnoDB tables, which may be extremely slow when running
multiple concurrent CHECK TABLE.

During the upgrade procedure on Windows mysqld.exe is started with
the named pipe connection protocol. mysqladmin.exe then pings the
server to check if is up and running. Command line looks like:
   mysqladmin.exe --protocol=pipe --socket=mysql_upgrade_service_xxx ping
But the "socket" parameter resets the "protocol" which was previously
initialized with the "pipe" value, setting it to "socket".
As a result, connection cannot be established and the upgrade
procedure fails.
"socket" in Windows is used to pass the name of the pipe so resetting
the protocol is not valid in this case.

This commit fixes resetting of the "protocol" parameter with "socket"
parameter in the case when protocol has been previously initialized
to "pipe" value

During the upgrade procedure on Windows mysqld.exe is started with
the named pipe connection protocol. mysqladmin.exe then pings the
server to check if is up and running. Command line looks like:
   mysqladmin.exe --protocol=pipe --socket=mysql_upgrade_service_xxx ping
But the "socket" parameter resets the "protocol" which was previously
initialized with the "pipe" value, setting it to "socket".
As a result, connection cannot be established and the upgrade
procedure fails.
"socket" in Windows is used to pass the name of the pipe so resetting
the protocol is not valid in this case.

This commit fixes resetting of the "protocol" parameter with "socket"
parameter in the case when protocol has been previously initialized
to "pipe" value

Set the histogram_type to be DOUBLE_PREC_HB as
it originally was.

The fix is in replacing the waiting for the whole purge finishing
with the the waiting for only delete-marked records purging finishing.

Reviewed by: Marko Mäkelä

- InnoDB bulk insert operation aborts the server for redundant
table. InnoDB miscalculates the record size in temporary file
for the redundant table. CHAR in redundant row format table
always fixed length, but in temporary file, it is variable-length
for variable-length character sets.

The issue is caused by MDEV-30400 fix.

There are two cursors in btr_estimate_n_rows_in_range() - p1 and p2, but
both share the same mtr. Each cursor contains mtr savepoint for the
previously fetched block to release it then the current block is
fetched.

Before MDEV-30400 the block was released with
mtr_t::release_block_at_savepoint(), it just unfixed a block and
released its page patch. In MDEV-30400 it was replaced with
mtr_t::rollback_to_savepoint(), which does the same as the former
mtr_t::release_block_at_savepoint(ulint begin, ulint end) but also
erases the corresponding slots from mtr memo, what invalidates any
stored mtr's memo savepoints, greater or equal to "begin".

The idea of the fix is to get rid of savepoints at all in
btr_estimate_n_rows_in_range() and
btr_estimate_n_rows_in_range_on_level(). As
mtr_t::rollback_to_savepoint() erases elements from mtr_t::m_memo, we
know what element of mtr_t::m_memo can be deleted on the certain case,
so there is no need to store savepoints.

See also the following slides for details:
https://docs.google.com/presentation/d/1RFYBo7EUhM22ab3GOYctv3j_3yC0vHtBY9auObZec8U

Reviewed by: Marko Mäkelä