Applied lost in a merge revision 7f38a070:

MDEV-10043 - main.events_restart fails sporadically in buildbot (crashes upon
             shutdown)

There was race condition between shutdown thread and event worker threads.

Shutdown thread waits for thread_count to become 0 in close_connections(). It
may happen so that event worker thread was started but didn't increment
thread_count by this time. In this case shutdown thread may miss wait for this
working thread and continue deinitialization. Worker thread in turn may continue
execution and crash on deinitialized data.

Fixed by incrementing thread_count before thread is actually created like it is
done for connection threads.

Also let event scheduler not to inc/dec running threads counter for symmetry
with other "service" threads.

On FreeBSD liblz4 is installed in /usr/local/lib.
Groonga uses pkg_check_modules to check for liblz4 (that is, pkg-config),
and then it used to set for libgroonga.a

  link_directories({$LIBLZ4_LIBRARY_DIRS})
  target_link_libraries(... ${LIBLZ4_LIBRARIES})

Now groonga is a static library, linked into ha_mroonga.so. CMake won't
link dynamic liblz4.so into libgroonga.a, instead it'll pass it as a
dependency and will link it into ha_mroonga.so. Fine so far. But it will
not pass link_directories from the static library as a dependency,
so ha_mroonga.so won't find liblz4.so

As suggested on cmake mailing list (e.g.
here: http://public.kitware.com/pipermail/cmake/2011-November/047468.html)
we switch to use the full path to liblz4.so, instead of the -l/-L pair.

in ALTER TABLE ... DROP KEY, ADD KEY, don't forget to compare old
and new keys' block sizes. If they differ - the key definition has changed.

The problem was that waiting for pause_for_ftwrl was done before
event_group was completed. This caused rpl_pause_for_ftwrl() to wait
forever during FLUSH TABLES WITH READ LOCK.
Now we only wait for FLUSH TABLES WITH READ LOCK when we are changing
to a new event group.

Protection added to reopen_file() and new_file_impl().

Without this we could get an assert in fn_format() as name == 0,
because the file was closed and name reset, atthe same time
new_file_impl() was called.

 is starting.

This is needed as if we kill the START SLAVE thread too early during
shutdown then the IO_THREAD or SQL_THREAD will not have time to properly
initlize it's replication or THD structures and clean_up() will try to
delete master_info structures that are still in use.

Added mutex_lock/mutex_unlock of mutex that is to be destroyed in
wait_for_commit::reinit() in a similar fashion that we do in
~wait_for_commit

This solves some possible dead locks when one calls stop slave while slave
is starting.

The reason for this is that stop slave takes LOCK_active_mi over the
whole operation while some slave operations will also need LOCK_active_mi
which causes deadlocks.

Fixed by introducing object counting for Master_info and not taking
LOCK_active_mi over stop slave or even stop_all_slaves()

Another benefit of this approach is that it allows:
- Multiple threads can run SHOW SLAVE STATUS at the same time
- START/STOP/RESET/SLAVE STATUS on a slave will not block other slaves
- Simpler interface for handling get_master_info()
- Added some missing unlock of 'log_lock' in error condtions
- Moved rpl_parallel_inactivate_pool(&global_rpl_thread_pool) to end
  of stop_slave() to not have to use LOCK_active_mi inside
  terminate_slave_threads()
- Changed argument for remove_master_info() to Master_info, as we always
  have this available
- Fixed core dump when doing FLUSH TABLES WITH READ LOCK and parallel
  replication. Problem was that waiting for pause_for_ftwrl was not done
  when deleting rpt->current_owner after a force_abort.

- Removed not used variables
- Added __attribute__()
- Added static to some local functions
  (gcc 5.4 gives a warning for external functions without an external definition)

As suggested in MySQL Bug#58536, increase the limit in this
debug assertion in order to avoid false positives on heavily
loaded systems.

This removes functionality of where ./mtr --mem /tmp/dir could be a directory.

Now MTR_MEM=/tmp/dir ./mtr is needed.

The case where MTR_MEM=/tmp/dir ./mtr --mem has the equivalent effect.
Signed-off-by: Daniel Black <daniel.black@au.ibm.com>

--mem works better as a pure flag, because it can be followed by command-line arguments (test names). If the option is allowed to have a value, the test name which directly follows it will be treated as the option value instead. It is possible to implement workarounds to avoid this, but they would not be completely reliable, and there is no practical purpose of such extension of functionality to justify them.

MDEV-11935: Queries in stored procedures with and EXISTS(SELECT * FROM VIEW) crashes and closes hte conneciton.

Use correct start point even for taken out from subselect items in process of exists2in conversion.

        ALTER_CONVERT flag removed and replaced for ALTER_OPTIONS
        for the CONVERT TO CHARACTER SET command.

(with blocks are still reachable).
There was a similar suppression already, but it had an extra line
comparing to failures which we are getting, so it wasn't applied.
Added another variant of the suppression.

fil_extend_space_to_desired_size(): Use a proper type cast when
computing start_offset for the posix_fallocate() call on 32-bit systems
(where sizeof(ulint) < sizeof(os_offset_t)). This could affect 32-bit
systems when extending files that are at least 4 MiB long.

This bug existed in MariaDB 10.0 before MDEV-11520. In MariaDB 10.1
it had been fixed in MDEV-11556.

a large memory buffer on Windows

fil_extend_space_to_desired_size(), os_file_set_size(): Use calloc()
for memory allocation, and handle failures. Properly check the return
status of posix_fallocate(), and pass the correct arguments to
posix_fallocate().

On Windows, instead of extending the file by at most 1 megabyte at a time,
write a zero-filled page at the end of the file.
According to the Microsoft blog post
https://blogs.msdn.microsoft.com/oldnewthing/20110922-00/?p=9573
this will physically extend the file by writing zero bytes.
(InnoDB never uses DeviceIoControl() to set the file sparse.)

I tested that the file extension works properly with a multi-file
system tablespace, both with --innodb-use-fallocate and
--skip-innodb-use-fallocate (the default):

./mtr \
--mysqld=--innodb-use-fallocate \
--mysqld=--innodb-autoextend-increment=1 \
--mysqld=--innodb-data-file-path='ibdata1:5M;ibdata2:5M:autoextend' \
--parallel=auto --force --retry=0 --suite=innodb &

ls -lsh mysql-test/var/*/mysqld.1/data/ibdata2
(several samples while running the test)

Increase tokudb_lock_timeout for the test

The function trx_purge_stop() was calling os_event_reset(purge_sys->event)
before calling rw_lock_x_lock(&purge_sys->latch). The os_event_set()
call in srv_purge_coordinator_suspend() is protected by that X-latch.

It would seem a good idea to consistently protect both os_event_set()
and os_event_reset() calls with a common mutex or rw-lock in those
cases where os_event_set() and os_event_reset() are used
like condition variables, tied to changes of shared state.

For each os_event_t, we try to document the mutex or rw-lock that is
being used. For some events, frequent calls to os_event_set() seem to
try to avoid hangs. Some events are never waited for infinitely, only
timed waits, and os_event_set() is used for early termination of these
waits.

os_aio_simulated_put_read_threads_to_sleep(): Define as a null macro
on other systems than Windows. TODO: remove this altogether and disable
innodb_use_native_aio on Windows.

os_aio_segment_wait_events[]: Initialize only if innodb_use_native_aio=0.

MDEV-11802 aims to fix the root cause instead.

The failure happens due to a race condition between processing
a row event (INSERT) and an automatically generated event
DROP TEMPORARY TABLE. Even though DROP has a higher GTID, it can
become visible in @@gtid_slave_pos before the row event with
a lower GTID has been committed. Since the test makes the slave
to synchronize with the master using GTID, the waiting stops
as soon as GTID of the DROP TEMPORARY TABLE becomes visible,
and if changes from the previous event haven't been applied yet,
the error occurs.

According to Kristian (see the comment to MDEV-10631), the real
problem is that DROP TEMPORARY TABLE is logged in the row mode
at all. For this particular test, since DROP does not do anything,
nothing prevents it from competing with the prior transaction.

The workaround for the test is to add a meaningful event
after DROP TEMPORARY TABLE, so that the slave would wait on its
GTID instead of the one from DROP.

Additionally (unrelated to this problem) removed FLUSH TABLES,
which, as the comment stated, should have been removed after
MDEV-6403 was fixed.

The standalone warning is not a sign of a problem, just of slowness,
so it should be added to global suppressions. If a real problem
happens, there will be other errors

* Revert "Make --mem a pure flag. If there is need to specifically set the location"

This reverts commit 716621db.

* MDEV-11619: mtr: when --mem is pure flag, conflicts with $MTR_MEM

Conflicts occurs when MTR_MEM=/xxx/yy ./mtr --mem is invoked. Here
the --mem option overrides opt_mem leaving the default path to be chosen.

This change makes when MTR_MEM set, opt_mem, the flag, is also
set. Both the environment and flag can no be set without conflicting.
Signed-off-by: Daniel Black <daniel.black@au.ibm.com>

* MDEV-11619: if opt_mem is a path include it first

* MDEV-11619: MTR_MEM locations - don't follow symlinks

From Bjorn Munch it seems symlinks can confuse some
tests. Lets just avoid those.

(ref: https://github.com/mysql/mysql-server/pull/116#issuecomment-268479774)
Signed-off-by: Daniel Black <daniel.black@au.ibm.com>

Use the ib_int64_t type alias instead of the standard type int64_t,
so that the code will compile on Microsoft Visual Studio 2013.

recv_writer_thread(): Do not assign recv_writer_thread_active=true
in order to avoid a race condition with
recv_recovery_from_checkpoint_finish().

recv_init_crash_recovery(): Assign recv_writer_thread_active=true
before creating recv_writer_thread.

Remove the debug parameter innodb_force_recovery_crash that was
introduced into MySQL 5.6 by me in WL#6494 which allowed InnoDB
to resize the redo log on startup.

Let innodb.log_file_size actually start up the server, but ensure
that the InnoDB storage engine refuses to start up in each of the
scenarios.

If InnoDB is started in innodb_read_only mode such that
recovered incomplete transactions exist at startup
(but the redo logs are clean), an assertion will fail at shutdown,
because there would exist some non-prepared transactions.

logs_empty_and_mark_files_at_shutdown(): Do not wait for incomplete
transactions to finish if innodb_read_only or innodb_force_recovery>=3.
Wait for purge to finish in only one place.

trx_sys_close(): Relax the assertion that would fail first.

trx_free_prepared(): Also free recovered TRX_STATE_ACTIVE transactions
if innodb_read_only or innodb_force_recovery>=3.

srv_release_threads(): Actually wait for the threads to resume
from suspension. On CentOS 5 and possibly other platforms,
os_event_set() may be lost.

srv_resume_thread(): A counterpart of srv_suspend_thread().
Optionally wait for the event to be set, optionally with a timeout,
and then release the thread from suspension.

srv_free_slot(): Unconditionally suspend the thread. It is always
in resumed state when this function is entered.

srv_active_wake_master_thread_low(): Only call os_event_set().

srv_purge_coordinator_suspend(): Use srv_resume_thread() instead
of the complicated logic.

There is only one select that returns a result set in this test.
It needs to be sorted