Also fixed a wrong result for a test case for mdev-7691
(the alternative one).
The test  cases for all these bug have materialized semi-joins used
inside dependent sub-queries.

The patch actually reverts the change inroduced by Monty in 2003.
It looks like this change is not valid anymore after the implementation
of semi-joins.
Adjusted output from EXPLAIN for many other test cases.

The test was unnecessarily depending on InnoDB purge, which can
sometimes fail to proceed.

Let us rewrite the test to use BEGIN;INSERT;ROLLBACK to cause the
immediate removal of the desired records.

shouldn't have been added in the first place, MariaDB Server is GPLv2

failed with SELECT SQ, TEXT field

The functon find_all_keys does call Item_subselect::walk, which calls walk() for the subquery
The issue is that when a field is represented by Item_outer_ref(Item_direct_ref(Item_copy_string( ...))).
Item_copy_string does have a pointer to an Item_field in Item_copy::item but does not implement Item::walk method, so we are not
able to set the bitmap for that field. This is the reason why the assert fails.

Fixed by adding the walk method to Item_copy class.

force SELinux policies to be built in mysql_release RPM packages

MDEV-12207 Include windows compatibility manifest into executable to make GetVersionEx work correctly

mysql_client uses some inline assembly code to switch thread stacks.
This works, however tools that perform backtrace get confused to fix
this we write a specific constant to signify bottom of stack. This
constant is needed when compiling with CLang as well.

This is essentially a backport of the 10.0
commit 203f4d41
that fixes a bug and silences a GCC 6.3.0 warning
about a left shift of a signed integer.

Missing parenthesis in a macro definition caused wrong operation
in the Query_cache::send_result_to_client() statement
   thd->query_plan_flags= (thd->query_plan_flags & ~QPLAN_QC_NO) | QPLAN_QC;

This would expand to
   thd->query_plan_flags= (thd->query_plan_flags & ~1) << 6 | 1 << 5;

which would shift the flags by 6 and clear an unrelated flag, instead
of clearing the flag (1 << 6).

Provide more useful progress reporting of crash recovery.

recv_sys_t::progress_time: The time of the last report.

recv_scan_print_counter: Remove.

log_group_read_log_seg(): After after each I/O request,
report progress if needed.

recv_apply_hashed_log_recs(): At the start of each batch,
if there are pages to be recovered, issue a message.

The C++ standard does not allow references to be NULL.
Assign the return value of THD::alloc() to a pointer,
not to a reference.

        Server audit plugin version updated.

Disabling the cond guards during the creation of Tricond Item for
constant and NULL left expression items

The function posix_fallocate() as well as the Linux system call
fallocate() can return EINTR when the operation was interrupted
by a signal. In that case, keep retrying the operation, except
if InnoDB shutdown has been initiated.

MDEV-9635: Server crashes in part_of_refkey  or assertion `!created && key_to_save < (int)s->keys' failed in TABLE::use_index(int) or with join_cache_level>2

Do not try to create index where ref is for hash join.

The function mysql_derived_merge() erroneously did not mark newly formed
AND formulas in ON conditions with the flag abort_on_null. As a result
not_null_tables() calculated incorrectly for these conditions. This
could prevent conversion of embedded outer joins into inner joins.

Changed a test case from table_elim.test to preserve the former execution
plan.

MDEV-5999 MySQL Bug#12766319 - 61865: RENAME USER DOES NOT WORK CORRECTLY - REQUIRES FLUSH PRIVILEGES

use update_hostname() to update the hostname.
test case comes from

  commit 0abdeed1d6d
  Author: gopal.shankar@oracle.com <>
  Date:   Thu Mar 29 00:20:54 2012 +0530

    Bug#12766319 - 61865: RENAME USER DOES NOT WORK CORRECTLY -
                          REQUIRES FLUSH PRIVILEGES

FUNCTIONS/PRIVILEGES DIFFERENTLY'.

The problem was that attempt to grant EXECUTE or ALTER
ROUTINE privilege on stored procedure which didn't exist
succeed instead of returning an appropriate error like
it happens in similar situation for stored functions or
tables.

The code which handles granting of privileges on individual
routine calls sp_exist_routines() function to check if routine
exists and assumes that the 3rd parameter of the latter
specifies whether it should check for existence of stored
procedure or function. In practice, this parameter had
completely different meaning and, as result, this check was
not done properly for stored procedures.

This fix addresses this problem by bringing sp_exist_routines()
signature and code in line with expectation of its caller.

it went out of support five years ago

It's too plainful to require <my_config.h> to be included first
for third-party storage engines. And anyway, some source files
might not include <my_config.h> at all, so there is no guarantee
that all parts of the binary will see identical definitions of system
structures (e.g. struct stat).

Define _FILE_OFFSET_BITS on the compiler's command line instead.

fix_for_comment() uses a static buffer. cannot have two
fix_for_comment() calls as arguments to one printf().

it was race condition prone. instead use either a pair of my_delete()
calls with already resolved paths, or a safe high-level function
my_handler_delete_with_symlink(), like MyISAM and Aria already do.

don't realpath() twice

At least it would be a compilation failure with -DDONT_ALLOW_SHOW_COMMANDS.
Otherwise it only confuses ctags to ignore ~75% of sql_parse.cc

TOCTOU bug. The path is checked to be valid, symlinks are resolved.
Then the resolved path is opened. Between the check and the open,
there's a window when one can replace some path component with a
symlink, bypassing validity checks.

Fix: after we resolved all symlinks in the path, don't allow open()
to resolve symlinks, there should be none.

Compared to the old MyISAM/Aria code:
* fastpath. Opening of not-symlinked files is just one open(),
  no fn_format() and lstat() anymore.
* opening of symlinked tables doesn't do fn_format() and lstat() either.
  it also doesn't to realpath() (which was lstat-ing every path
  component), instead if opens every path component with O_PATH.
* share->data_file_name stores realpath(path) not readlink(path). So,
  SHOW CREATE TABLE needs to do lstat/readlink() now (see ::info()),
  and certain error messages (cannot open file "XXX") show the real
  file path with all symlinks resolved.

Remove maria_test_invalid_symlink() and myisam_test_invalid_symlink(),
introduce mysys_test_invalid_symlink(). Other engines might need it too

Don't let my_register_filename() fail because strdup() failed. Better to
have NULL for a filename, then to fail the already successful open().

Filenames are only used for error reporting and there was already code
to ignore OOMs (my_fdopen()) and to cope with missing filenames
(my_filename()).

my_realpath() ignores MY_xxx flags anyway

While DEBUG_SYNC doesn't make sense without a valid THD, it might
be put in the code that's invoked both within and outside of a
THD context (e.g. in maria_open(), there is no THD during recovery).