SYMPTOM:
- The engine allocates higher than configured amount of memory
while creating secondary index on a varchar column. The memory
allocated increases with higher values of innodb_ddl_buffer_size
variable.

BACKGROUND:
- While creating the index, parallel threads scan the rows in
table and store the index records in the key_sort_buffer->m_heap.
- The fields of the index records are stored in the
Key_sort_buffer->m_heap. The address of each index records is
tracked in the metadata tuple Key_sort_buffer->m_dtuples` that uses
standard memory allocation.

ROOT CAUSE:
- We approximate the number of tuples which can fit in the size using
Buffer_size_avaliable_for_thread
     = innodb_ddl_buffer_size/(Number of parallel scan threads)
 Number of tuples which can fit
     = Buffer_size_avaliable_for_thread/std::min(1,minimum_record_size)
- The minimum size of non-fixed size column types is determined as zero
in the method dtype_get_min_size_low(), varchar is one of such column
types. Since both the primary and secondary index keys are defined on
varchar columns, the minimum record size obtained for the secondary
index is 0.
- This causes a high number of tuples to be estimated from the formula
above.
- Since we reserve the memory in the tuple container to hold the
estimated number of tuples, this caused the memory allocation spike
in the container as we estimated more number of rows than required.
For example, ~1 GB for 128 MB innodb_ddl_buffer_size.

FIX:
- Instead of reserving the memory upfront for the metadata tuple
vector, we could let the container manage its capacity. We shall
avoid over-estimation of tuples for the situation like the one
reported in this in bug.
Instead, we now account approximate capacity of tuples container
at the time of inserting the record metadata in the tuple. This
may reduce the records packed in the key_sort_buffer, that may
reduce the sorted chunks in the temp merge file. Thus it may
increase the IO overhead a little.
- Increase the IO thresholds in the test
innodb.alter_table_temp_files_io.test by approximately 15% to account
for the minor increase in IO observed.

Change-Id: I97d19bed15b9accdb9269dfef28ce662c4bd6a6f

Issue
======
Bug#37105430 was fixed in trunk, this patch revealed some failures
in trunk. Some of these issues are still hidden in 8.0 and 8.4

Fix
======
Backport the fix for Bug#37105430 into 8.0 and 8.4
Change-Id: I6b91a7bd0c614761c7fad2e43d10b258327f0033

Approved by : Pavithra Pandith <pavithra.pandith@oracle.com>

The command_services suite uses the sql server sessions as an underlying
mechanism.
This was never supposed to work recursively: i.e. a command using
sessions executed in a session should fail.

Added a check to the command service to ensure it's not usable in such
a context.
Fixed some error handling in the test UDF.

Change-Id: Idb8e4793a3363a1aed6c6b8789445dec8ac462fd

Symptom:
The `page_track.read_only` MTR test case has been failing on Windows on
`mysqltest: At line 63: Command "force-rmdir" failed with error 1. my_errno=41.`

Root cause:
The test seems to be removing a directory with files being used by the server, which leads to error deleting these on Windows.
The test seems to need to delete it before the server is started again, and to make things easier it is done before doing restart. This does not work on Windows.

Fix:
We are first stopping the server, and only then delete the files, before starting the server again.
Additionally the `restart_innodb_read_only.inc` was rewritten using existing scripts to reduce code duplication.

Change-Id: If249119b77e3dffc1ffa10faab285d8bd2c057a9

…[8.0 version]

We hit an empty pointer in Query_block::nest_derived used by
Query_block::transform_scalar_subqueries_to_join_with_derived.

[ 8.0 version: CUBE not supported.
  8.4 version: We implicitly try to
  do a prepare for secondary engine in the original repro due to GROUP
  BY CUBE(x), since CUBE is not supported by the primary engine. In the
  simplified repro, we explicitly enable the transform to see the
  error.
  9.3 version: repro gives error message due to outer reference in
  window's ORDER BY clause. If that is removed, the error doesn't occur. ]

At one stage we try to transform a query with a subquery shown as the
following by the optimizer trace:

/* select#2 */
select `t1`.`i`
from <constant table> join
     `cte2`
     on (/* select#5 */ select 3 from `cte2`) <> 0

i.e. we have a join condition, but only one table (cte2) in the from
list. This is the result of an earlier transformation.  However,
Query_block::nest_derived is not prepare for this scenario,
presupposing the there be at least two tables when a join condition is
involved: accessing Table_ref::embedding only works if we have an
embedding, i.e.  an inner table, but here we have only a single table.

Solution: test for this situation, which allows the transform to
succeed.

Note: in 8.4 the subquery can't be offloaded to secondary engine
anyway, since there are other subqueries that are not eliminated, but
will work for primary engine for the simplified repro (no CUBE) with
transform enabled.

Change-Id: If83e7a8e72372a1d3da30b9155ede594352a2845

…uery transform [ back-port ]

[ Back-ported to enable backport fix of Bug#36314993/Bug#36464947 ]

Another case of transforming subqueries to a JOIN with derived tables,
with the containing query being grouped. We create an extra derived
table in which to do the grouping. This process moves initial select
list items from the containing query into the extra derived table,
replacing all original select list items (with the exception of
subqueries which get their own derived tables) with columns from the
extra derived table.

This logic didn't handle the ANY_VALUE function correctly:
As for other functions, it was left in the outer where it doesn't help
ward off the group legality checks, hence the unexpected error seen.

The patch adds support for ANY_VALUE(expression) in queries undergoing
the mentioned transform. We inject an ANY_VALUE wrapper around all
columns moved to the select list of the extra derived table from the
original query to columns in the grouped table iff all references to
the column are made within the arguments of an ANY_VALUE call.

Example:

SELECT ANY_VALUE(i) AS i1,
       (SELECT i FROM t) AS subquery,
       SUM(i) AS summ
FROM t;

->

SELECT ANY_VALUE(`derived_1_3`.`Name_exp_1`) AS `i1`,
       `derived_1_4`.`i` AS `subquery`,
       `derived_1_3`.`summ` AS `summ`
FROM ( SELECT SUM(`test`.`t`.`i`) AS `summ`,
              ANY_VALUE(`test`.`t`.`i`) AS `Name_exp_1` # add ANY_VALUE
       FROM `test`.`t`) `derived_1_3`
     LEFT JOIN
     ( SELECT `test`.`t`.`i` AS `i` FROM `test`.`t`) `derived_1_4`
     ON(TRUE) WHERE TRUE

Change-Id: I41a08074981ea1334b7b5c0cb955488711bd7d6d

…calar_subquery [ back-port ]

Bug#36464947 Server crash with subquery_to_derived=ON

[ This is a back-port of fix for Bug#36314993 which fixes Bug#36464947
  seen on mysql-8.0.  Bug#36314993 was not seen on mysql-8.0 itself,
  due to syntax that was only added later.  The explanation below is
  taken Bug#36314993 but not accurate for the mysql-8.0 repro.
  Underlying cause probably is though. ]

I refer to the simplified repro with ROLLUP, not CUBE in the below.
There are two issues.

1.In the rollup query we have

  SELECT ( SELECT f2
           FROM (SELECT 1
                 GROUP BY ROLLUP(1)) AS dt1
           GROUP BY f1) AS a,
         MIN(f1)
  FROM t1;

In preparation for transforming the scalar subquery, we move the
aggregate MIN(f1) into a derived table, thusly:

  SELECT (SELECT derived_1_4.Name_exp_2
          FROM (SELECT 1
                GROUP BY '' WITH ROLLUP) AS dt1
          GROUP BY derived_1_4.Name_exp_1) AS a,
         derived_1_4.`MIN(f1)`
  FROM (SELECT MIN(t1.f1) AS `MIN(f1)`,
               t1.f1 AS Name_exp_1,
               t1.f2 AS Name_exp_2
        FROM t1) AS derived_1_4

As can be seen the original occurences of f1 and f2 in the scalar
subquery have been replaced with another (outer) reference:
derived_1_4.Name_exp_1 for f1 and derived_1_4.Name_exp_2 for f2. f1 is
induced as a hidden field in the select list along with visible f2,
i.e. it is scalar. We lost the hidden flag for f1 alias
derived_1_4.Name_exp_1, which lead to the first assert: it was no
longer a scalar subquery as expected. We now made sure to inherit the
hidden flag in Item_field::replace_item_field. To facilitate this we
had to make a copy of f1 in Query_block::transform_grouped_to_derived
lest we clobber its hidden status when calling add_item_to_list a few
lines further down.

2. When doing the replacement of t1's fields after moving it into the
derived table, we correctly detected that (hidden) f1 in the select
list and f1 in the group by list are outer references, cf. the logic
in Item_field::replace_item_field. However, we created two different
items for those two occurences, which caused another assert when
creating the tmp table; we now make sure we use only one field, cf. the
new member field in struct Item_field_replacement: m_outer_field.

Now, in this case, the whole exercise of moving the grouping into a
derived table turns out to be useless, as we can't transform the
scalar subquery anyway (correlated but no equality clause,
rollup). But we don't know that this early in the transform, so it is
done and should work.

The other repros no longer crash, and are included in the tests.

Change-Id: Iac3e35bdbf2f0385e1e4e33083f4e2d19060036d

Change-Id: Idb8e4793a3363a1aed6c6b8789445dec8ac462fd

Description:
------------

The UBSan reports "load of value 16, which is not a valid
value for type 'bool'" error when running
|Rpl_commit_order_queue_test.Simulate_mts| test.

Analysis:
---------

The undefined behaviour identified by UBSan is caused by using
test()/test_and_set() methods on default constructed
std::atomic_flag object as in this case it has undeterminate
initial value (until C++20).

Fix:
----

Although it is possible to fix this issue and continue using
std::atomic_flag in the problematic unit test, there is no
strong reason to prefer it over std::atomic<bool> in that
place. Switching to std::atomic<bool> simplifies the test code
and improves readability.

As a result, all instances of std::atomic_flag have been
replaced with std::atomic<bool>, ensuring that code no longer
accesses uninitialized variables.

Additionally, the missing code formatting has been applied in a
couple of places, within the same test file.

Change-Id: I4ed03bb6633d61afb7a842f9ea37e7578c42e71a

function in 8.0.

POST-PUSH FIX:
Fixed test case failure in `main.partition_exchange`.

Change-Id: Ia31857196037619237d07f7484e452c6b2341fc7

Approved-by: Balasubramanian Kandasamy <balasubramanian.kandasamy@oracle.com>

Approved-by: Balasubramanian Kandasamy <balasubramanian.kandasamy@oracle.com>

              INFORMATION_SCHEMA

Post push fix for test failures:
 1. Re-recording tests partition_upgrade_5727_mac_lctn_2
    and partition_upgrade_5727_win_lctn_2.
 2. I_S.INNODB_COLUMNS , I_S.INNODB_TABLES output varies with
    type of build, hence made the test more generic.

Change-Id: I54697bfd602c19212fe78b0af099c79223d66810

…x digits [8.0 patch]

Problem:

In semi-sync replication when the binary log suffix exceeds six digits(.999999)
and becomes binlog-name.1000000 the replication moves from semi-sync to async.

Analysis:

The binlog file naming convention in MySQL utilizes the prefix 'binlog-name.' followed
by a suffix number. When the suffix number is within six digits, it is zero-padded to
maintain the length (e.g.,binlog-name.000001). However, once the suffix exceeds six digits,
the number is not trimmed or padded, resulting in names such as binlog-name.1000000 (1 followed by six zeros).
During replication, the ActiveTranx::compare function, which is implemented using strcmp,
compares the binlog positions. Since strcmp only returns the relationship of the first
non-matching character, it may incorrectly determine that mysql-bin.1000000 comes before mysql-bin.999999.

This incorrect comparison can lead to source no longer sending "confirmation-needed" data
packets to the replica or waiting for confirmations from the replica. As the source stops
waiting for replica's acknowledgments, semi-synchronous replication effectively degrades to
asynchronous replication.

Fix:

In the ActiveTranx::compare function, when comparing the  binlog sequence
numbers, the length of the binlog filenames is compared first; if the lengths
are identical, then comparison is made using strcmp.

Change-Id: Ic6283781706fbe06ad95fe90e6e2988efdbb88c4

If the system variable MAX_JOIN_SIZE is set, the optimizer choses
another code path than normal in Query_exression::optimize - it calls
EstimateRowAccesses and compares it with the maximum allowed in
MAX_JOIN_SIZE and gives an error if too large.

Now, in this query, EstimateRowAccesses arrives in
EstimateRowAccessesInItem and the item in question is a
Item_singlerow_subselect. Its query_expression doesn't have a value
set for its root_access_path, so we next call EstimateRowAccesses with
a path argument taken from
Item_singlerow_subselect::root_access_path(), which is always
nullptr. This eventually leads to the issue, when trying to access the
empty path.

Solution: skip calling EstimateRowAccesses is the path is nullptr.

For the hypergraph optimizer, we don't get into this problem because
Query_expression::m_root_access_path is always set, cf. this code:

  if (thd->lex->using_hypergraph_optimizer()) {
    // The hypergraph optimizer also wants all subselect items to be optimized,
    // so that it has cost information to attach to filter nodes.
    for (Query_expression *unit ...)  {
          unit->optimize(thd, ...)  <-- leads to creation of access path for QE
        return true;
    }
    :
  }

so in EstimateRowAccessesInItem:
      if (qe->root_access_path() != nullptr) {
        path = qe->root_access_path();   <-------
      } else {
        path = qe->item->root_access_path();
      }

we get a non-null path from QE's m_root_access_path. This is not
always so for the old optimizer.

Change-Id: I7966d347900014925021fd74f869dee52df8f7ab

…unfixable

When SQL functions are improved, they may throw SQL errors in situations
where they previously did not. If this happens in a table's constraints,
default expressions, partitioning, or virtual columns, the table can not
be opened. This will prevent both analysing the problem (e.g. with SHOW
CREATE TABLE) and addressing it (e.g. with ALTER TABLE ... DROP ...).

* Post-push fix: Fixing interplay with some other (lctn) test cases
                 that also rely on hand-crafted databases.

Change-Id: I69b800a5b0e007894e82cf3065e582d4438d261a

…it fails on pb2 daily trunk

Problem:

The test rpl_semi_sync_binlog_suffix_exceed_six_digit.test which was added
as part of BUG#37024069 is failing when run with gtid_mode=ON

Analysis:

The statement CHANGE REPLICATION SOURCE SOURCE_LOG_FILE gives error when
running with gtid_mode=ON as SOURCE_LOG_FILE cannot be set when
SOURCE_AUTO_POSITION is active.

Fix:

We are not adding any extra coverage by running this test with gtid_mode=ON
so move the test from rpl suite to rpl_nogtid.

Change-Id: I86da78a1d5406b0e497abe8fbb0b2f596f31c85f

In the deb packaging we run a few install commands that change ownership
of files to the root user. This means the build has to run as root,
or with fakeroot.

This isn't recommended or required (downstream packaging installs these
files without changing ownership), and in Ubuntu 25.04 the default config
does not allow builds to use root, so it fails.

Removed the ownership change from these calls to let the build run rootless.

Change-Id: I318693eb11ffc193c2d8b43c746685786f8366e7

In Fedora 42+ /usr/sbin is merely a symlink to /usr/bin, try to detect
this and change paths as needed.

Change-Id: Ideba537092e21a6c809967302abc56d30ee25c87

Follow up fix for compilation problem:
  storage/ndb/src/kernel/vm/Configuration.cpp:656:2: error: extra ';' outside
    of a function is incompatible with C++98 [-Werror,-Wc++98-compat-extra-semi]

Change-Id: I8bdbf9d3f68e1805f76bf68f455d85fd570ce45a

Add a few missing includes in router, abseil and X plugin

Change-Id: Ia974396b2fc02a6fa981025c1c16f46e12493668

…any columns"

This reverts commit 8504c085229ffd9d80df826798220053d4180f17. This
commit is the fix for Bug#37510755 and is being reverted.

Change-Id: I4e874f491f856afd900bd6da6479e5e15fd9a03d

On Fedora 38:
make build_id_test
Linking CXX executable ../runtime_output_directory/build_id_test
Verifying build-id
egrep: warning: egrep is obsolescent; using grep -E

Fix: use cmake regexp instead.

Change-Id: I05b9167279b6cfcdf770d87bb5ba98792a158691
(cherry picked from commit a5451a65b642c2c940c19c1d35cd32cb6dc1d14f)

Set C locale when invoking /usr/bin/eu-readelf

Change-Id: Ib628a6c5276a1903caff75d403616e7ff4dd204a
(cherry picked from commit d9ad0253b147981b276d523a1395a6e7ea92d969)

generated_message_tctable_lite.cc:85:73: error: cannot tail-call:
address of caller arguments taken
   85 |   PROTOBUF_MUSTTAIL return GenericFallbackImpl<MessageLite, std::string>(
      |                            ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
   86 |       PROTOBUF_TC_PARAM_PASS);

There is a bug reported for this:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119376

This patch simply disables clang::musttail for non-clang compilers.
The problematic code is apparently fixed upstream, with a patch called:

"Remove spill hack from SingularBigVarint."
protocolbuffers/protobuf@bc56c34

Change-Id: Ida1a31f9027665ad4bd375b30843442bd9923b79

                          autocommit=OFF

Query rewrite plugin reload rules logic runs in an independent thread
created when a request is submitted. This thread runs the reload logic
and exits. When the server is in global autocommit OFF mode, the new
thread and its session will also run with autocommit OFF mode. At the
end of the reload execution, the code does a statement commit and exist.
Statement commits does not preserve the changes in autocommit OFF
mode. Due to this, after the thread exits, any table data changes done
will be lost.

Fix: Perform a transaction commit to ensure all changes are preserved
before exiting the session/thread.

Change-Id: Ib6f96873f90d99b2c53bb0157e91b160760b933e

… init file

Problem
=======
An init file having a single line with multiple SQL statements, like the one below:

SET character_set_client = 'cp850';
SET sql_log_bin = OFF; ALTER USER 'root'@'localhost' IDENTIFIED BY 'pass';

throws some errors during init, viz:

On Unix systems, we get to see : "1064 : You have an error in your SQL syntax; check
the manual that corresponds to your MySQL server version for the right syntax to use..."

On Windows, the behavior is different, it throws an assertion that the query size is
greater than the max query size.

Analysis
========
The problem occurs when one single line contains multiple SQL statements.

Fix
===
When dealing with a line having multiple SQL statements, the statements are
fetched one-by-one from the current line. For each statement that has been fetched
from the line, append it with a null character to fix the issue.

Change-Id: I06f56502dda88a1f41b8d40176201c517c32b985

is killed via SQL statement

Description:
------------
When the applier thread is terminated using SQL KILL statement, there
is no specific error logging indicating that the termination was
due to this command. Instead, a generic error message is logged.
   'Fatal error during execution on the Applier process...'

Analysis:
------------
The current error message does not sufficiently differentiate between
an applier application error and an intentional termination via KILL,
leading to potential confusion.
To improve clarity, the error log should explicitly indicate when
the applier thread is killed using SQL KILL statement.

Fix:
----
Enhance logging to include a specific message when the applier thread
is terminated via SQL KILL statement.

Change-Id: Ia059be50a44f122ea6792faa8633ec3ff420218f

Problem:
--------
When importing a table with a secondary index after an instant
column drop there was a failure due to using incorrect fields
mapping both in index rec_cache and in array of fields.

Fix:
----
1. Array of fields used by secondary index is rebuilt during table
import with column drop to make sure it reflects actual list of
fields after an instant drop.
2. rec_cache is set to empty after instant column drop because
it is used only for intrinsic tables.

Change-Id: I66d2c02f90f95205cc7f28b5f036656a256f15ad

…RROR 241, MSG: INVALID SCHEMA OBJECT VERSION

This is a forwardport of:

 BUG 31848270 - EXCLUDING BLOB COL FROM EVENT DEFINITION RESULTS IN ERROR 241, MSG: INVALID SCHEMA OBJECT VERSION

    3-23954393131 createEventOperation fails with 241:Invalid schema object version

    Workaround fixing skipping Blob event validation in non-MySQLD
    NdbApi applications.

Change-Id: I900e24f765747d4a032cffb736e1480926dc50f8

When router fails to resolve the destination adresses of the
cluster-nodes due to a temporary DNS outage, it closes the acceptor
sockets but sporadically fails to open the acceptor socket again.

Background
==========

If DNS comes back without 10 seconds,

1. the metadata-cache opens the acceptor socket again
2. the routing-connect waits for the PRIMARY failover for 10 seconds and
   closes the acceptor again (without anyone ever opening them again).

Resolver errors should not lead to a "wait-for-PRIMARY-failover" as the
wait should only be done if the connection to the primary failed.

In case of DNS failures the PRIMARY may be up and running without any
intention to every switch to another node.

Change
======

- handle resolver errors correctly to not lead to a wait-for-primary-failover

Change-Id: I11ff612063fa53333ad29c7ccdbaea11b9123dc1

Approved by: Erlend Dahl <erlend.dahl@oracle.com>

Change-Id: Idbd2d69c47175d12dc1a3610d846c8f20773a0ce

Approved by: Erlend Dahl <erlend.dahl@oracle.com>

Change-Id: If31440c12fee6a563e16fcdd2455cc280b9b4f5d

             m_ptr + data_size < m_bounds.second

PROBLEM:

1. After the fix for the "Bug #37233273 - DDL buffer increase has
   out of control memory usage", this assertion was seen.
2. The fix for Bug #37233273 basically accounts for meta data as
   part of innodb_ddl_buffer_size buffer  which reduces the
   number of rows which can fit in a key buffer block while
   adding an index.
3. When the key buffer is full, the records in the key buffer
   are added to the temporary file, which constitutes a chunk.
   This chunk is smaller than actual key buffer size.
4. The offsets of each chunk is recorded in a queue. The chunks
   sizes are multiples of the sector size (4KB).
5. When merging the two chunks into one sorted chunk, we read
   from this queue.
6. Another bug, Bug#36444172 "Regression in 8.0.27+ related to
   Parallel DDL IO Amplification" read the sizes of the first
   two chunks from the queue and set a limit on the input buffer
   based on the size of the first two chunks.
7. In our case, the sizes of the first two chunks were smaller
   compared to the subsequent chunk sizes.
8. Therefore, while reading the rows from subsequent chunks we
   hit this imit on the input buffer.

FIX:

1. Each ddl thread is allocated a certain amount of input buffer
   based on the number of scan threads and the DDL buffer size.
   We should use this allocation as the limit for the input
   buffer while reading the rows from a chunk.

Change-Id: Iba4a224c9973bea2b5a6a2288a23473dd2c21b06

https://perconadev.atlassian.net/browse/PS-10074

The download URL of Percona-Server-shared-56-5.6.51-rel91.0.1.el7.x86_64.rpm that percona-server-8.0 is dependent on is incorrect.
Even If Percona-Server-shared-56-5.6.51-rel91.0.1.el7.x86_64.rpm download by using current download URL, we can't expand correctly Percona-Server-shared-56-5.6.51-rel91.0.1.el7.x86_64.rpm.
Because the Percona-Server-shared-56-5.6.51-rel91.0.1.el7.x86_64.rpm is downloaded as HTML file (not RPM) currently.

https://perconadev.atlassian.net/browse/PS-10041

https://perconadev.atlassian.net/browse/PS-10041

…ng_key_grammar fix)

https://perconadev.atlassian.net/browse/PS-10041

Fixed problem with logical conflict between Oracle's fix for the Bug #37436220
"MySQL Server SEGV within JOIN::refresh_base_slice"
(commit 026aac7@mysql/mysql-server)
and Percona's implementation of the 'CLUSTERING KEY' extension which caused
'main.tokudb_clustering_key_grammar' MTR test case to fail.

Introduced new 'AT_CLUSTERING_KEY_COLUMN_ATTR' value in the 'Attr_type'
enumeration. 'PT_unique_combo_clustering_key_column_attr' class split into two:
'PT_unique_key_column_attr' and 'PT_clustering_key_column_attr'.

…ers fix)

https://perconadev.atlassian.net/browse/PS-10041

Re-recorded a bunch of MTR test cases because Oracle also introduced the
'$do_not_echo_parameters' parameter in the 'start_mysqld.inc' MTR include file
in the fix for the Bug #37697512
"InnoDB: page_track.read_only big test fails on Windows"
(commit 0f118ef@mysql/mysql-server).
However, Oracle's implementation (in conrtast to Percona's one) does not print extra
"<hidden args>" label.

For symmetry reasons similar changes made by Percon in 'kill_and_restart_mysqld.inc'
and 'restart_mysqld.inc' also modified to omit printing "<hidden args>" label.

https://perconadev.atlassian.net/browse/PS-9647

Force inlined rec_init_offsets_new() - the same as it is in
original Enchanced Mysql version.

…nary table for component_masking_functions (gen_range fix)

https://perconadev.atlassian.net/browse/PS-9148

Fixed problem with 'gen_range()' UDF not being able to properly handle
negative numbers.

This function internally used to call 'masking_functions::random_number()'
function that used to always accept 'std::size_t' arguments. In case when
negative 'long long' arguments (extracter from UDF context) were passed to
this function, they were implicitly converted to 'std::size_t' resulting
in huge numbers and causing an assertion in 'std::uniform_int_distribution'
constructor that expected 'min' <= 'max'.

Fixed by making the 'masking_functions::random_number()' a function
template.

https://perconadev.atlassian.net/browse/PS-10041

Cleaned up orphaned MTR artefacts based on the output of the "mtr_checker"
utility (https://github.com/Percona-Lab/mtr-checker).

***
*** Suite: main ***
***
[error]: inconsistent test case "m_i_db_config": [ cnf ]

***
*** Suite: group_replication ***
***
[error]: inconsistent test case "gr_gcs_psi_threads": [ result ]
[error]: inconsistent test case "gr_ssl_tls13": [ result ]
[error]: inconsistent test case "gr_ssl_mode_verify_identity_error": [ result, opt_master, opt_slave ]
[error]: inconsistent test case "gr_skip_view_change_event": [ cnf ]
[error]: inconsistent test case "gr_psi_event_names_enabled_check": [ result ]
[error]: inconsistent test case "gr_ssl_tls13_runtime_invalid_configuration": [ result ]

***
*** Suite: innodb ***
***
[error]: inconsistent test case "percona_extended_innodb_status": [ opt_master ]

***
*** Suite: ndb ***
***
[error]: inconsistent test case "ndb_grant": [ result ]

***
*** Suite: ndb_rpl ***
***
needed [error]: inconsistent test case "ndb_rpl_multi_binlog_update": [ cnf ]
needed [error]: inconsistent test case "ndb_rpl_3_cluster": [ cnf ]

***
*** Suite: rpl ***
***
kept for re-enablement [error]: found unknown file "rpl_critical_errors.result.txt"
[error]: inconsistent test case "rpl_relay_log_space_limit_deadlock": [ opt_slave ]
[error]: inconsistent test case "rpl_auto_increment": [ opt_master ]
needed [error]: inconsistent test case "rpl_row_img": [ cnf ]
[error]: inconsistent test case "rpl_invalid_replication_timestamps_multi_source": [ cnf ]
[error]: inconsistent test case "rpl_row_jsondiff_basic": [ cnf ]
needed [error]: inconsistent test case "rpl_row_jsondiff": [ cnf ]

***
*** Suite: service_sys_var_registration ***
***
needed [error]: inconsistent test case "sys_reg": [ cnf ]

***
*** Suite: x ***
***
[error]: found "macros" that is not a regular file

***
*** Suite: engines/funcs ***
***
[error]: inconsistent test case "rpl_init_replica": [ opt_slave ]

***
*** Suite: ndb_big ***
***
[error]: inconsistent test case "smoke": [ test ]
[error]: inconsistent test case "ndb_verify_redo_log_queue": [ test ]
[error]: inconsistent test case "ndb_multi_tc_takeover": [ test, cnf ]
[error]: inconsistent test case "bug37983": [ test, opt_master ]
[error]: inconsistent test case "my": [ cnf ]
[error]: inconsistent test case "rqg_spj": [ test, opt_master ]
[error]: inconsistent test case "bug13637411": [ test, cnf, opt_master ]

***
*** Suite: ndb_ddl ***
***
[error]: inconsistent test case "my": [ cnf ]

***
*** Suite: ndb_opt ***
***
[error]: inconsistent test case "my": [ cnf ]

***
*** Suite: ndbcluster ***
***
[error]: inconsistent test case "my": [ cnf ]

***
*** Suite: ndbcrunch ***
***
[error]: inconsistent test case "my": [ cnf ]
[error]: inconsistent test case "cpubind": [ cnf ]

***
*** Plugin suite: audit_log_filter ***
***
[error]: inconsistent test case "log_format_json": [ result ]
[error]: inconsistent test case "log_format_new": [ result ]
[error]: inconsistent test case "log_format_old": [ result ]

***
*** Plugin suite: keyring_vault ***
***
[error]: inconsistent test case "table_encrypt_2_keyring": [ result, opt_master ]
[error]: inconsistent test case "table_encrypt_2_directory": [ result ]

…anup)

https://perconadev.atlassian.net/browse/PS-10041

Cleaned up orphaned MTR artefacts based on the output of the "mtr_checker"
utility (https://github.com/Percona-Lab/mtr-checker).

***
*** Suite: component_keyring_kms ***
***
[error]: inconsistent test case "migration": [ opt_master ]

***
*** Suite: data_masking ***
***
[error]: inconsistent test case "error_not_loaded_module": [ opt_master ]

***
*** Suite: rocksdb ***
***
[error]: inconsistent test case "collation_exceptions_lctn_0": [ opt_master ]
[error]: inconsistent test case "collation_exceptions_lctn_1": [ opt_master ]

***
*** Suite: tokudb_rpl ***
***
[error]: inconsistent test case "rpl_tokudb": [ opt_master ]
[error]: inconsistent test case "rpl_mixed_row_tokudb": [ result, opt_master ]

***
*** Suite: component_keyring_kmip ***
***
[error]: inconsistent test case "migration": [ opt_master ]

https://perconadev.atlassian.net/browse/PS-10041

Fixed problem with invalid length parameter passed to 'inet_ntop()'
function causing warnigs under compiler's FORTIFY enforcements.

PKG-870 Packaging tasks for release - PS 8.0.43-34

…ad_url

…oad_url

PS-10074 (8.0) Fix invliad download URL for compatsrc