e.g.
- dont -> don't
- occurence -> occurrence
- succesfully -> successfully
- easyly -> easily
Also remove trailing space in selected files.
These changes span:
- server core
- Connect and Innobase storage engine code
- OQgraph, Sphinx and TokuDB storage engines
Related to MDEV-21769.
Instrument new synchronization primitive with thd_wait_begin/end
to inform threadpool about waits.
This considerably improve performance on write benchmarks
(e.g sysbench update_index) with generic threadpool, of course the cost is
possibility of many newly created threads.
Lifted long standing limitation to the XA of rolling it back at the
transaction's
connection close even if the XA is prepared.
Prepared XA-transaction is made to sustain connection close or server
restart.
The patch consists of
- binary logging extension to write prepared XA part of
transaction signified with
its XID in a new XA_prepare_log_event. The concusion part -
with Commit or Rollback decision - is logged separately as
Query_log_event.
That is in the binlog the XA consists of two separate group of
events.
That makes the whole XA possibly interweaving in binlog with
other XA:s or regular transaction but with no harm to
replication and data consistency.
Gtid_log_event receives two more flags to identify which of the
two XA phases of the transaction it represents. With either flag
set also XID info is added to the event.
When binlog is ON on the server XID::formatID is
constrained to 4 bytes.
- engines are made aware of the server policy to keep up user
prepared XA:s so they (Innodb, rocksdb) don't roll them back
anymore at their disconnect methods.
- slave applier is refined to cope with two phase logged XA:s
including parallel modes of execution.
This patch does not address crash-safe logging of the new events which
is being addressed by MDEV-21469.
CORNER CASES: read-only, pure myisam, binlog-*, @@skip_log_bin, etc
Are addressed along the following policies.
1. The read-only at reconnect marks XID to fail for future
completion with ER_XA_RBROLLBACK.
2. binlog-* filtered XA when it changes engine data is regarded as
loggable even when nothing got cached for binlog. An empty
XA-prepare group is recorded. Consequent Commit-or-Rollback
succeeds in the Engine(s) as well as recorded into binlog.
3. The same applies to the non-transactional engine XA.
4. @@skip_log_bin=OFF does not record anything at XA-prepare
(obviously), but the completion event is recorded into binlog to
admit inconsistency with slave.
The following actions are taken by the patch.
At XA-prepare:
when empty binlog cache - don't do anything to binlog if RO,
otherwise write empty XA_prepare (assert(binlog-filter case)).
At Disconnect:
when Prepared && RO (=> no binlogging was done)
set Xid_cache_element::error := ER_XA_RBROLLBACK
*keep* XID in the cache, and rollback the transaction.
At XA-"complete":
Discover the error, if any don't binlog the "complete",
return the error to the user.
Kudos
-----
Alexey Botchkov took to drive this work initially.
Sergei Golubchik, Sergei Petrunja, Marko Mäkelä provided a number of
good recommendations.
Sergei Voitovich made a magnificent review and improvements to the code.
They all deserve a bunch of thanks for making this work done!
mariabackup tries to allocate a buffer of page_size*page_size/4 size.
for 64k page it means 1Gb, which doesn't work very well on 32-bit builders.
Skip the 64k page test on 32bit.
1. Refactored innobase_close_connection(). Transaction must've already
been rolled back by this time. We should expect only transactions in the
PREPARED state when MDEV-742 is done.
2. Added missing put_pins() to trx_disconnect_prepared(). Missing
put_pins() wasn't a problem because trx_disconnect_prepared() is a dead
code. But it will get revived in the main MDEV-742 patch.
3. Fixed missing reset of trx->mysql_log_file_name when RW transaction
didn't emit any log records (zero-modification RW). The problem was
detected by ASAN when disconnected XA transaction was trying to make
use of inherited mysql_log_file_name pointing into binlog data of
detached THD.
This missing reset also had user-visible side effect, when
trx_sys_print_mysql_binlog_offset() would report binlog position
not of the most recently committed transaction.
One of possible scenarios that is expected to misbehave is as following:
thr1> CREATE TABLE t1(a INT) ENGINE=InnoDB;
thr1> INSERT INTO t1 VALUES(1);
thr1> BEGIN;
thr1> UPDATE t1 SET a=1
thr1> COMMIT; -- zero-modification, misses to reset mysql_log_file_name
thr2> BEGIN;
thr2> INSERT INTO t1 VALUES(2);
thr2> COMMIT;
thr1> BEGIN;
thr1> do-some-real-changes;
thr1> ROLLBACK; -- will store binlog pos from previous COMMIT in thr1?
In this case it means if binlog is replayed from position reported by
trx_sys_print_mysql_binlog_offset(), t1 will end up with two records
containing '2'.
Part of
MDEV-742 - XA PREPAREd transaction survive disconnect/server restart
It now lives from THD constructor to THD destructor. Reset before THD is
released to a cache. Change user doesn't reset debug_sync_control anymore.
Needed to be able to make use of DEBUG_SYNC() at later stages like
ha_close_connection().
- This issue was caused by 5e62b6a5e0.
fts_optimize_callback() should free fts_optimize_wq and make it as NULL
when it receives FTS_MSG_STOP message. So that subsequent
fts_optimize_callback() doesn't fail with segmentation fault.
Starting with commit 1a6f708ec5
the function buf_pool_get_dirty_pages_count() is only used
in a debug check. It was dead code for non-debug builds.
buf_flush_dirty_pages(): Perform the debug check inline,
and replace the assertion
ut_ad(first || buf_pool_get_dirty_pages_count(id) == 0);
with another one that is executed while holding the mutexes:
ut_ad(id != bpage->id.space());
Create_tmp_table::add_fields(): Initialize uneven_delta= 0
to suppress the warning from GCC 9.2.1 and 10.0.1,
and consistently indent the code with spaces.
The -Wconversion in GCC seems to be stricter than in clang.
GCC at least since version 4.4.7 issues truncation warnings for
assignments to bitfields, while clang 10 appears to only issue
warnings when the sizes in bytes rounded to the nearest integer
powers of 2 are different.
Before GCC 10.0.0, -Wconversion required more casts and would not
allow some operations, such as x<<=1 or x+=1 on a data type that
is narrower than int.
GCC 5 (but not GCC 4, GCC 6, or any later version) is complaining
about x|=y even when x and y are compatible types that are narrower
than int. Hence, we must rewrite some x|=y as
x=static_cast<byte>(x|y) or similar, or we must disable -Wconversion.
In GCC 6 and later, the warning for assigning wider to bitfields
that are narrower than 8, 16, or 32 bits can be suppressed by
applying a bitwise & with the exact bitmask of the bitfield.
For older GCC, we must disable -Wconversion for GCC 4 or 5 in such
cases.
The bitwise negation operator appears to promote short integers
to a wider type, and hence we must add explicit truncation casts
around them. Microsoft Visual C does not allow a static_cast to
truncate a constant, such as static_cast<byte>(1) truncating int.
Hence, we will use the constructor-style cast byte(~1) for such cases.
This has been tested at least with GCC 4.8.5, 5.4.0, 7.4.0, 9.2.1, 10.0.0,
clang 9.0.1, 10.0.0, and MSVC 14.22.27905 (Microsoft Visual Studio 2019)
on 64-bit and 32-bit targets (IA-32, AMD64, POWER 8, POWER 9, ARMv8).
Some .c and .cc files are compiled as part of Mariabackup.
Enabling -Wconversion for InnoDB would also enable it for
Mariabackup. The .h files are being included during
InnoDB or Mariabackup compilation.
Notably, GCC 5 (but not GCC 4 or 6 or later versions)
would report -Wconversion for x|=y when the type is
unsigned char. So, we will either write x=(uchar)(x|y)
or disable the -Wconversion warning for GCC 5.
bitmap_set_bit(), bitmap_flip_bit(), bitmap_clear_bit(), bitmap_is_set():
Always implement as inline functions.
The extension of the record comparison functions for SPATIAL INDEX in
mysql/mysql-server@b66ad511b6
was suboptimal for multiple reasons:
Some functions used unnecessary temporary variables of the int type,
instead of the more appropriate size_t, causing type mismatch.
Many functions unnecessarily required rec_get_offsets() to be
computed, or a parameter for length, although the size of the
minimum bounding rectangle (MBR) is hard-coded as
SPDIMS * 2 * sizeof(double), or 32 bytes.
In InnoDB SPATIAL INDEX records, there always is a 32-byte key
followed by either a 4-byte child page number or the PRIMARY KEY value.
The length parameters were not properly validated.
The function cmp_geometry_field() was making an incorrect attempt
at checking that the lengths are at least sizeof(double) (8 bytes),
even though the function is accessing up to 32 bytes in both MBR.
Functions that are called from only one compilation unit are defined
in another compilation unit, making the code harder to follow and
potentially slower to execute.
cmp_dtuple_rec_with_gis(): FIXME: Correct the debug assertion
and possibly the function TABLE_SHARE::init_from_binary_frm_image()
or related code, which causes an unexpected length of
DATA_MBR_LEN + 2 bytes to be passed to this function.
Currently, if a user wants to require TLS for every connection made
over the network, then every user account on the system needs to be
created with "REQUIRE SSL" or one of the other TLS options.
Implementing a require_secure_transport system varuable (which,
in particular, can be set using the --require_secure_transport=ON
command line option) in the MariaDB Server would make it a lot
easier to require TLS (or other secure transport) system-wide.
This patch implements this new system variable, adds the ability
to set it with SQL statements, from the command line and from the
configuration file, and also contains improvements for mtr that allow
the user to establish non-secure TCP/IP connections (for example,
to verify the operation of the new option).