MDEV-10087 mysqld_update()/mysql_delete() continues execution even after subquery with JOIN gets error from storage engine

The issue is that record_should_be_deleted() returns true in
mysql_delete() even if sub-select with join gets error from storage
engine when DELETE FROM ... WHERE ... IN (SELECT ...) statement is
executed.

The same is true for mysql_update() where select->skip_record() returns
true even if sub-select with join gets error from storage engine.

In the test case if sub-select is chosen as deadlock victim the whole
transaction is rolled back during sub-select execution, but
mysql_delete()/mysql_update() continues transaction execution and invokes
table->delete_row() as record_should_be_deleted() wrongly returns true
in mysql_delete() and table->update_row() as select->skip_record(thd)
wrongly returns 1 for mysql_update().

record_should_be_deleted() wrogly returns true because thd->is_error()
returns false SQL_SELECT::skip_record() invoked from
record_should_be_deleted().

It's supposed that THD error should be set in rr_handle_error() called
from rr_sequential() during sub-select JOIN::exec_inner() execution.

But rr_handle_error() does not set THD error because
READ_RECORD::print_error is not set in JOIN_TAB::read_record.

READ_RECORD::print_error should be initialized in
init_read_record()/init_read_record_idx(). But make_join_readinfo() does
not invoke init_read_record()/init_read_record_idx() for
JOIN_TAB::read_record.

The fix is to set JOIN_TAB::read_record.print_error in
make_join_readinfo(), i.e. in the same place where
JOIN_TAB::read_record.table is set.

Reviewed by Sergey Petrunya.

This is a backport of commit 8b6a308e
from MariaDB Server 10.6.11. No attempt to reproduce the hang
in earlier an earlier version of MariaDB Server than 10.6 was made.

In each caller of fseg_n_reserved_pages() except ibuf_init_at_db_start()
which is a special case for ibuf.index at database startup, we must hold
an index latch that prevents concurrent allocation or freeing of index
pages.

Any operation that allocates or free pages that belong to an index tree
must first acquire an index latch in non-shared mode, and while
holding that, acquire an index root page latch in non-shared mode.

btr_get_size(), btr_get_size_and_reserved(): Assert that a strong enough
index latch is being held.

dict_stats_update_transient_for_index(),
dict_stats_analyze_index(): Acquire a strong enough index latch.

These operations had followed the same order of acquiring latches in
every InnoDB version since the very beginning
(commit c533308a).

The hang was introduced in
commit 2e814d47 which imported
mysql/mysql-server@ac74632293bea967b352d1b472abedeeaa921b98
which failed to strengthen the locking requirements of the function
btr_get_size().

* MDEV-29397 Fix note turning into error

ODBC Rewind triggered an error with no SQL, but this is sometimes a
valid condition (as can be seen with other classes). Setting this to a 0
return stops errors firing when they shouldn't.

Also fixes ASAN hits from in MDEV-29687 tabext.cpp.

spatial_index_info: Replaces index_tuple_info_t. Always take
a memory heap as a parameter to the member functions.
Remove pointer indirection for m_dtuple_vec.

spatial_index_info::add(): Duplicate any PRIMARY KEY fields that would
point to within ext->buf because that buffer will be allocated in
a shorter-lifetime memory heap.

Every operation that is going to write redo log is supposed to
invoke log_free_check() before acquiring any latches. If there
is a risk of log buffer overrun, a log checkpoint would be
triggered by that call.

ibuf_merge_space(), ibuf_merge_in_background(),
ibuf_delete_for_discarded_space(): Invoke log_free_check()
when the current thread is not holding any page latches.

Unfortunately, in lower-level code called from ibuf_insert()
or ibuf_merge_or_delete_for_page(), some page latches may be
held and a call to log_free_check() could hang.

ibuf_set_bitmap_for_bulk_load(): Use the caller's mini-transaction.
The caller should have invoked log_free_check() while not holding
any page latches.

Something appears to be broken in the DBUG subsystem.
Let us remove frequent calls to it from the InnoDB internal SQL interpreter
that is used in the purge of transaction history.

The DBUG_PRINT in que_eval_sql() can remain for now, because those
operations are much less frequent.

InnoDB crash recovery can run out of memory before
commit 50324ce6 in
MariaDB Server 10.5.

Let us disable some frequently failing recovery tests
in earlier versions.

Not creating explicit record locks will speed up the test.

Also, disable the use of InnoDB persistent statistics in the test of
MDEV-27270 to avoid intermittent failures in 10.6 or later
(after commit 9608773f)
due to the nondeterministic scheduling of STATS_AUTO_PERSISTENT.

The file plugin_exports became unused in
commit fec844ac

- mariabackup fails to assign srv_undo_space_id_start when the
dedicated undo tablespaces are disabled

When trying to output stacktrace, and addr2line is not installed, the
child process forked by start_addr2line_fork() will fail to do exec(),
and finish with exit(1).

There is a problem with exit() though - it runs exit handlers,
and for the forked copy of crashing process, it is a bad idea.

In 10.5+ code for example, exit handlers include
tpool::task_group static destructors, and it will hang infinitely
waiting for completion of the outstanding tasks.

The fix is to use _exit() instead, which skips the execution of exit
handlers

Make sure that EXPLAIN object allocated on runtime arena.

/usr/ports/pobj/mariadb-10.9.3/mariadb-10.9.3/mysys/my_lock.c:183:7: warning: incompatible function pointer types assigning to 'sig_return' (aka 'void (*)(void)') from 'void (*)(int)' [-Wincompatible-function-pointer-types]
      ALARM_INIT;
      ^~~~~~~~~~
/usr/ports/pobj/mariadb-10.9.3/mariadb-10.9.3/include/my_alarm.h:43:16: note: expanded from macro 'ALARM_INIT'
                        alarm_signal=signal(SIGALRM,my_set_alarm_variable);
                                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/ports/pobj/mariadb-10.9.3/mariadb-10.9.3/mysys/my_lock.c:189:7: warning: incompatible function pointer types passing 'sig_return' (aka 'void (*)(void)') to parameter of type 'void (*)(int)' [-Wincompatible-function-pointer-types]
      ALARM_END;
      ^~~~~~~~~
/usr/ports/pobj/mariadb-10.9.3/mariadb-10.9.3/include/my_alarm.h:44:41: note: expanded from macro 'ALARM_END'
                                              ^~~~~~~~~~~~
/usr/include/sys/signal.h:199:27: note: passing argument to parameter here
void    (*signal(int, void (*)(int)))(int);
                             ^
2 warnings generated.

The prototype is the same for all of the *BSD's.

void
(*signal(int sigcatch, void (*func)(int sigraised)))(int);

Per the code my_set_max_open_files 3 lines earlier, we attempt
to set the nofile (number of open files), rlimit to max_open_files.

We should use this in the warning because wanted_files may not
be the number.

Fix a couple of build warnings that fire with CONNECT engine.

More checks for bad geometry data added.

return an error on invalid gis data

remove useless if()

with C/C.
The patch introduces mariadb_capi_rename.h which is included into
mysql.h. The hew header contains macro definitions for the names being
renamed. In versions 10.6+(i.e. where sql service exists) the renaming
condition in the mariadb_capi_rename.h should be added with
&& !defined(MYSQL_DYNAMIC_PLUGIN)
and look like
The patch also contains removal of mysql.h from the api check.

Disabling false_duper-6543 test for embedded.

ha_federated.so uses C API. C API functions are being renamed in the server,
but not renamed in embedded, since embedded server library should have proper
C API, as expected by programs using it.
Thus the same ha_federated.so cannot work both for server and embedded
server library.

As all federated tests are already disabled for embedded,
federated isn't supposed to work for embedded anyway, and thus the test
is being disabled.

Abort startup, if SSL setup fails.
Also, for the server always check that certificate matches private key
(even if ssl_cert is not set, OpenSSL will try to use default one)

Read the version of the view share when we read definition to prevent
simultaniouse access to a view table SHARE (and so its MEM_ROOT)
from different threads.

MDEV-16549 Server crashes in Item_field::fix_fields on query with view and subquery, Assertion `context' failed, Assertion `field' failed

Add one-table-resolve context for items created with an aim of switching
to temporary table because then it can be cloned in push-down-condition.

The problem was fixed earlier. Adding an MTR test only.

GRANT ROLE can update db-level privileges -> must invalidate acl_cache

* to "clear hostname cache" one needs to use hostname_cache->clear()
* no need to clear acl_cache for SET DEFAULT ROLE

storage/connect/libdoc.cpp:603:17: error: 'void xmlXPathInit()' is deprecated [-Werror=deprecated-declarations]

on Linux this pthread_attr_setstacksize() fails with EINVAL
"The stack size is less than PTHREAD_STACK_MIN (16384) bytes".

But on FreeBSD it succeeds and causes a crash later, as 8196 is too little.

Let's keep the stack at its default size in the timer thread.

(and for 64-bit big endian)