When ORDER BY ... LIMIT check whether it should switch from index IDX1
to index IDX2, it should not ignore the fact that IDX2 may have a
potential range or ref(const) access.
Istead, it should calculate their costs: there is now a saved range
optimizer cost and code to re-calculate what best_access_path()
calculated for ref(const).

/* in current cost model these two can be very different
numbers unfortunately */

Let range optimizer remove parts of OR-clauses for which range analysis
produced SEL_TREE(IMPOSSIBLE).
There is no need to remove parts of AND-clauses: either they are inside
of OR (and the whole AND-clause will be removed), or the AND-clause is
at the top level, in which case the whole WHERE (or ON) is always FALSE
and this is a degenerate case which receives special treatment.

The removal process takes care not to produce 1-way ORs (in that case
we substitute the OR for its remaining member).

Reverted workaround for gcc bug, which was fixed 3 years ago:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34772

IS [NOT] NULL predicate is sargable within an outer join. Range
analysis only uses predicates from ON expressions, which have
regular semantics (without null-complemented rows, etc).
There is no reason not use IS [NOT] NULL predicates.

Discover P_S tables automatically.

Most of this patch is code clean-up:
- removed tests and code responsible for P_S tables correctness verification
- always return error from ha_perfschema::create()
- install/upgrade scripts won't create P_S tables anymore

ALGORITHM=INPLACE for InnoDB

Added thd_progress calls to row_merge_sort to give
some hint how merge sort progresses.

- After PREPARE is done, save the value of thd->select_number
  (this is what will be assigned to next select_lex object)
- Restore the value at each execution of the prepared statement.

Removed raw debug output when no error on OS operation has not happened.

mysqldump:
- Added --log-queries to allow one to disable logging for the dump

sql/log_event.cc:
- Removed setting of enable_slow_log as it's not required anymore.

sql/sql_parse.cc:
- Set enable_slow_log to value of thd->variables.sql_log_slow as this will speed up tests if slow log is disabled.
- opt_log_slow_admin_statements can now only disable slow log, not enable it.

sql/sql_explain.cc:
- Minor cleanup

Other things:
- Added sql_log_slow to system variables.
- Changed opt_slow_log to global_system_variables.sql_log_slow in all files
- Updated tests to reflect changes

Removed -DSECURITY_HARDENED=OFF for debug build scripts
ignore configure

Analysis: Provided incorrect parameter to output buffer size
and incorrectly determined actual payload size after compression.

implementation.

    Merge https://github.com/kakao/mariadb-10.0 that contains Facebook's
    implementation for defragmentation

    facebook/mysql-5.6@a2d3a74
    facebook/mysql-5.6@def96c8
    facebook/mysql-5.6@9c67c5d
    facebook/mysql-5.6@921a81b
    facebook/mysql-5.6@aa519bd
    facebook/mysql-5.6@fea7d13
    facebook/mysql-5.6@09b29d3
    facebook/mysql-5.6@9284abb
    facebook/mysql-5.6@dbd623d
    facebook/mysql-5.6@aed55dc
    facebook/mysql-5.6@aad5c82

    This version does not add new SQL-syntax and new handler API function.
    Instead optimize table is mapped to defragment table if
    innodb_defragment=ON, by default the feature is off.

    Contains changes authored by Sunguck Lee (Kakao).

buf0buf.cc line 2642.

Analysis: innodb_compression_algorithm is a global variable and
can change while we are building page compressed page. This could
lead page corruption.

Fix: Cache innodb_compression_algorithm on local variable before
doing any compression or page formating to avoid concurrent
change. Improved page verification on debug builds.

Analysis: InnoDB writes also files that do not contain FIL-header.
This could lead incorrect analysis on os_fil_read_func function
when it tries to see is page page compressed based on FIL_PAGE_TYPE
field on FIL-header. With bad luck uncompressed page that does
not contain FIL-headed, the byte on FIL_PAGE_TYPE position could
indicate that page is page comrpessed.

Fix: Upper layer must indicate is file space page compressed
or not. If this is not yet known, we need to read the FIL-header
and find it out. Files that we know that are not page compressed
we can always just provide FALSE.

Added extra check for default_role column.

Also fixed wording on one message.

to avoid test failures.

# of processed files is printed every 15 sec.

compression format (Fusion-IO).

Addeed LZMA and BZIP2 compression methods.

in file buf0mtflu.cc line 570.

Analysis: Real timing bug, we should take the mutex before we
try to send those shutdown messages, that would make sure
that threads doing a unfinished flush (they have acquired
this mutex) have time to do their work before we add shutdown
messages to work queue. Currently, we just add those shutdown
messages to work queue and code assumes that at flush, there
is constant number of items to be processed and thus
leading to assertion.