add a test case

vers_insert_history_row(): do not insert rows with zero or negative lifetime.

mysql_update(): properly handle error from vers_insert_history_row()

Vers SQL: TRT fix getting TRX_ID by COMMIT_TS

Fixed wrong assumption that records are ordered by COMMIT_TS.
This is anyway a quick hack until tempesta-tech#314 is done.

See also FIXME and TODO in TR_table::query(MYSQL_TIME, bool).

Test: SEES case for trx_id.test [closes #456]

Lots of changes:
* calculate the current history partition in ::external_lock(),
  not in ::write_row() or ::update_row()
* remove dynamically collected per-partition row_end stats
* no full table scan in open_table_from_share to calculate these
  stats, no manual MDL/thr_locks in open_table_from_share
* no shared stats in TABLE_SHARE = no mutexes or condition waits when
  calculating current history partition
* always compare timestamps, don't convert them to MYSQL_TIME
  (avoid DST ambiguity, and it's faster too)
* correct interval handling, 1 month = 1 month, not 30 * 24 * 3600 seconds
* save/restore first partition start time, and count intervals from there
* only allow to drop first partitions if INTERVAL
* when adding new history partitions, split the data in the last history
  parition, if it was overflowed
* show partition boundaries in INFORMATION_SCHEMA.PARTITIONS

partition_info had a bunch of function pointers to avoid if()'s
when invoking part_type specific functionality (like get_part_id, etc).
But check_range_constants() and check_list_constants() were still
invoked conditionally, with if()'s.

Create partition_info::check_constants function pointer, get rid
of if()'s

Also remove alloc argument of check_range_constants(), added
in 26a3ff0a. Broken system versioning will be fixed in
following commits.

use include/have_xxx.inc when some feature needs to be present
(because --xxx in the opt file will fail if the xxx is not compiled in)

set variables in the test, not on the command line, to avoid unnecessary
server restarts (they're must slower than SET).

* TIMESTAMP precedence fixed.

Unit-based history point (vers_history_point_t; Vers_history_point).

Update partition stats on ha_partition::write_row()

SQL: DROP PERIOD FOR SYSTEM_TIME syntax

and remove ER_VERS_SYS_FIELD_EXISTS

originally by: Eugene Kosov

also, don't rotate versioning partitions for DELETE HISTORY

originally by: Aleksey Midenkov

implement log-term TODO item, convert redundant if()-s into asserts.

don't allow to discover WITH SYSTEM VERSIONING clause

originally by: Aleksey Midenkov

table might be NULL in ha_partition::engine_name()
(test case pushed into 5.5)

There is only one lock_sys. Allocate it statically in order to avoid
dereferencing a pointer whenever accessing it. Also, align some
members to their own cache line in order to avoid false sharing.

lock_sys_t::create(): The deferred constructor.

lock_sys_t::close(): The early destructor.

Merge pull request #622

is not supported

Allowed to use recursive references in derived tables.
As a result usage of recursive references in operands of
INTERSECT / EXCEPT is now supported.

fix the compilation error.
no support for plugins yet.

ReadView::is_open() was supposed to be called by view owner thread
exclusively. Use ReadView::get_state() instead.

At innodb_force_recovery=5 or bigger, trx_lists_init_at_db_start()
no longer initialises the purge_sys. Adjust an assertion accordingly.

rocksdb and spider

There is only one purge_sys. Allocate it statically in order to avoid
dereferencing a pointer whenever accessing it. Also, align some
members to their own cache line in order to avoid false sharing.

purge_sys_t::create(): The deferred constructor.

purge_sys_t::close(): The early destructor.

undo::Truncate::create(): The deferred constructor.
Because purge_sys.undo_trunc is constructed before the start-up
parameters are parsed, the normal constructor would copy a
wrong value of srv_purge_rseg_truncate_frequency.

TrxUndoRsegsIterator: Do not forward-declare an inline constructor,
because the static construction of purge_sys.rseg_iter would not have
access to it.

trx_purge(): Remove the parameter limit or batch_size, which is
always passed as srv_purge_batch_size.

trx_purge_attach_undo_recs(): Remove the parameters purge_sys, batch_size.
Refer to srv_purge_batch_size.

trx_purge_wait_for_workers_to_complete(): Remove the parameter purge_sys.

Add innodb debug system variable, innodb_buffer_pool_load_pages_abort, to test
the behaviour of innodb_buffer_pool_load_incomplete.
(innodb_buufer_pool_dump_abort_loads.test)

This status variable indicates that an innodb buffer pool load never
completed and dumping at shutdown would result in an incomplete dump file.

This status variable is set to 1 once a buffer pool loads. Upon a successful
load this status variable returns to 0.

With this status variable set, the system variable
innodb_buffer_pool_dump_at_shutdown==1 will have no effect as dumping after
an incomplete load will generate a less complete dump file than the current
one.

If a user aborts a buffer pool load by changing the system variable
innodb_buffer_pool_load_abort=1 will cause the the status variable
innodb_buffer_pool_load_incomplete to remain set to 1.

A shutdown that occurs while innodb is loading the buffer pool will
not save the buffer pool on shutdown.

A user may indirectly set innodb_buffer_pool_load_incomplete
to 0 by:
* Forcing a load, by setting innodb_buffer_pool_load_now=ON, or
* Forcing a dump, by setting innodb_buffer_pool_dump_now=ON

This will enable the next dump on shutdown to complete.
Signed-off-by: Daniel Black <daniel.black@au.ibm.com>

Before MDEV-12288 in MariaDB 10.3.1, InnoDB used to partition
the persistent transaction undo log into insert_undo and update_undo.

MDEV-12288 repurposes the update_undo as the single undo log.
In order to support an upgrade from earlier MariaDB versions,
the insert_undo is recovered in data structures, called old_insert.

An assertion failure occurred in TrxUndoRsegsIterator::set_next()
when an incomplete transaction was recovered with both insert_undo
and update_undo log. This could be easily demonstrated by starting
./mysql-test-run --manual-gdb innodb.read_only_recovery
in MariaDB 10.2, and after the first kill, start up the MariaDB 10.3
server with the same parameters.

The problem is that MariaDB 10.3 would roll back the recovered
transaction, and finally "commit" it twice (with all changes to
data rolled back), both insert_undo and update_undo with the same
commit end identifier (trx->no).

Our fix is to introduce a "commit number" that comprises two components:
(trx->no << 1 | !old_insert). In this way, the assertion in the purge
subsystem can be relaxed so that only the trx->no component must match.