commit 854c219a7f (MDEV-17301)
broke a constraint: Fixed-length columns cannot be extended in InnoDB
without rebuilding the table.
ha_innobase::can_convert_string(): Correct the condition. We must
not allow any instantaneous change to the length of CHAR columns
measured in characters. For any format other than ROW_FORMAT=REDUNDANT,
we can allow the length in bytes to be extended if mbminlen<mbmaxlen held
before the change of the character set.
In commit 0f90728bc0 (MDEV-16809)
we introduced the configuration option innodb_log_optimize_ddl
for controlling whether native index creation or table-rebuild
in InnoDB should avoid writing full redo log.
Fungo Wang reported that this option is causing occasional failures.
The reason is that pages may be written to data files in an
inconsistent state. Applying log records to such inconsistent pages
may fail.
The solution is to always invoke PageBulk::finish() before page latches
may be released, to ensure that the page contents is in a consistent
state.
Something similar was implemented in MySQL 8.0.13:
mysql/mysql-server@d1254b9473
buf_block_t::skip_flush_check: Remove. Suppressing consistency checks
is a bad idea.
PageBulk::needs_finish(): New predicate: Determine whether
PageBulk::finish() must fix up the page.
PageBulk::init(): Clear PAGE_DIRECTION to ensure that needs_finish()
will hold. We change the field from PAGE_NO_DIRECTION to 0
and back without writing redo log. This trick avoids the need
to introduce any new data member to PageBulk.
PageBulk::insert(): Replace some high-level accessors to bypass
debug assertions related to PAGE_HEAP_TOP that we will be violating
until finish() has been executed.
PageBulk::finish(): Tolerate m_rec_no==0. We must invoke this also
on an empty page, to ensure that PAGE_HEAP_TOP is initialized.
PageBulk::commit(): Always invoke finish().
PageBulk::release(), BtrBulk::pageSplit(), BtrBulk::storeExt(),
BtrBulk::finish(): Invoke PageBulk::finish().
In fsp_path_to_space_name(), we would access a byte right before
the start of the string, tripping AddressSanitizer.
This reverts commit d87006a1c1
and commit a7634281aa.
This version is not optimized yet. It could have bugs because I didn't
check it with unit tests. Also, std::char_traits are not really supported.
So, now it's not possible to create f.ex. a case insensitive string_view.
fil_path_to_space_name(): renamed, moved to another file
and refactored to use string_view
The only InnoDB changes between Percona XtraDB Server 5.6.47-87.0
and 5.6.48-88.0 are related to InnoDB changes between MySQL 5.6.47
and MySQL 5.6.48, which we had already applied.
This issue was originally reported by Fungo Wang, along with a fix, as
MySQL Bug #98990.
His suggested fix was applied as part of
mysql/mysql-server@a003fc373d
and released in MySQL 5.7.31.
i_s_metrics_fill(): Add the missing call to Field::set_notnull(),
and simplify some code.
- Problem is that test case creates iblogfile* files. So existing
ibdata pages could point to future LSN. Fix is that taking the
backup of data before iblogfile* creation and apply it before
exiting the test case.
Problem:
=======
- Read operations are always allowed to hold a secondary index leaf
latch and then look up the corresponding clustered index record.
Flush table operation acquires secondary index latch while holding
a clustered index latch. It leads to deadlock violation.
Fix:
====
- Flush table operation should acquire secondary index before taking
clustered index to avoid deadlock violation with select operation.
Allocate space for fields inside the window function (arguments, PARTITION BY and ORDER BY clause)
in the ref pointer array. All fields inside the window function are part of the temporary
table that is required for the window function computation.
This form of corruption was only reproduced on MariaDB 10.5.4
after the MDEV-22867 fix was applied in
commit 431200090e.
While we do not know how to reproduce this corruption in
MariaDB 10.4, we are applying the code fix without a test case.
btr_cur_pessimistic_update(): Invoke btr_set_instant() if needed.
Error state is not stored in check_and_do_in_subquery_rewrites() when there is
illegal combination of optimizer switches. So all the functions eventually
return false. Thus the assetion failure.
The auth_pam_tool that is executed from pam_auth() can be still
not finished by the time we do the waitpid() there.
As we use WNOHANG option for the waitpid(), it didn't wait and
left the zombie process. So let's do the loop of waitpid() with the
limited number of sleeps.
The client can only find out if the server has disconnected when it tries to
read or send something. If the server gets disconnected before
send_client_reply_packet(), the client will try sending authentication
information but it will fail. But, if the client is fast enough to send
autentication information before disconnecting, it will notice that when
reading the ok packet. So the client can fail on read or on write.
It is unpredictable because, the process are unsynchronized and this
could happen in any order.
In mi_check_index_tuple(), when rowid filter check returns
CHECK_OUT_OF_RANGE, set info->lastpos= HA_OFFSET_ERROR, like
it is done above for the ICP check.
failed in Diagnostics_area::set_ok_status on FUNCTION replace
When there is REPLACE in the statement, sp_drop_routine_internal() returns
0 (SP_OK) on success which is then assigned to ret. So ret becomes false
and the error state is lost. The expression inside DBUG_ASSERT()
evaluates to false and thus the assertion failure.
While trying to detect datadir, take into account that one can use
Windows service name as section name in options file, for Windows service.
The historical obscurity is being used by WAMP installations.
Make sure that the sort_buffer that is allocated has atleast space for MERGEBUFF2 keys.
The issue here was that the record length is quite high and sort buffer size is very small,
due to which we end up with zero number of keys in the sort buffer. The Sort_param::max_keys_per_buffer
was zero in such a case, due to which we were flushing empty sort_buffer to the disk.
accept might return an error, including SOCKET_EAGAIN/
SOCKET_EINTR. The caller, usually handle_connections_sockets
can these however and invalid file descriptor isn't something
to call fcntl on.
Thanks to Etienne Guesnet (ATOS) for diagnosis,
sample patch description and testing.
commit 484931325e included a
workaround for a 10.5 merge issue that should now be properly
addressed in commit ab4069909d.
buf_chunk_init(): Remove an unnecessary MEM_MAKE_ADDRESSABLE().
We might invoke MEM_UNDEFINED() here, but actually the allocated
memory ought to be guaranteed to be zero-initialized.
commit 484931325e was a necessary
fix for the buffer pool resizing tests in 10.5 in
AddressSanitizer. However, that change would break the tests
innodb.innodb_buffer_pool_resize and
innodb.innodb_buffer_pool_resize_with_chunks
when run in MemorySanitizer, or presumably in Valgrind as well.
(Those tests run "forever" in Valgrind.)
buf_pool_resize(): Cancel the effect of MEM_NOACCESS() in Valgrind
and ASAN. In MSAN, MEM_NOACCESS() is a no-op, and hence we must do
nothing special here.
MEM_MAKE_ADDRESSABLE() would declare the memory contents undefined.
In this particular case, we must actually declare the contents
defined for Valgrind.
