cte_nonrecursive.test with --embedded.

This also fixed some problems for embedded CTEs.
Adjusted test results accordingly.

Rollback attempted to dereference DB_ROLL_PTR=0, which cannot possibly
be a valid undo log pointer. A safe canonical value would be
roll_ptr_t(1) << ROLL_PTR_INSERT_FLAG_POS
which is what was chosen in MDEV-12288.

This bug was reproduced in 10.3 only. Potentially, the problem could
have been introduced by MDEV-11415, which suppresses undo logging for
ALGORITHM=COPY operations. In those operations, we should actually
have written the safe value of DB_ROLL_PTR instead of writing 0.
However, the test in commit 5421e3ae
demonstrates that access to the rebuilt table by earlier-started
transactions should actually have been refused with ER_TABLE_DEF_CHANGED.

btr_cur_ins_lock_and_undo(): When undo logging is disabled, use the
safe value of DB_ROLL_PTR.

btr_cur_optimistic_insert(): Validate the DB_TRX_ID,DB_ROLL_PTR before
inserting into a clustered index leaf page.

ins_node_t::sys_buf[]: Replaces row_id_buf and trx_id_buf and some
heap usage.

row_ins_alloc_sys_fields(): Initialize ins_node_t::sys_buf[].

trx_undo_page_report_modify(): Assert that the DB_ROLL_PTR is not 0.

trx_undo_get_undo_rec_low(): Assert that the roll_ptr is valid before
trying to dereference it.

dict_index_t::is_primary(): Check if the index is the primary key.

Before killing the server, we have to FLUSH TABLES in order to
prevent the corruption of any MyISAM system tables.

PageConverter::adjust_cluster_record(): Instead of writing
the invalid value DB_ROLL_PTR=0, write a value that indicates
a fresh insert, that is, prevents the DB_ROLL_PTR from being
dereferenced in any circumstances.

It can be argued that IMPORT TABLESPACE should actually
update the dict_index_t::trx_id to prevent older transactions
from accessing the table, similar to what I did on table
rebuild in MySQL 5.6.6 in
https://github.com/mysql/mysql-server/commit/03f81a55f221095d397c375afe8a10c8038da339

Use unsigned comparison.

If the CRYPTO_set_mem_functions fails just return success
from check_openssl_compatibility. The only case where
CRYPTO_set_mem_functions fails is the allow_customize==0 (aka FIPS
mode). The check_openssl_compatibility isn't able to complete unless
this function returns success.

ref: https://github.com/openssl/openssl/blob/OpenSSL_1_1_0g/crypto/mem.c#L34

MDEV-14222 Unnecessary 'cascade' memory allocation for every updated row
when there is no FOREIGN KEY

This reverts the MySQL 5.7.2 change
https://github.com/mysql/mysql-server/commit/377774689bf6a16af74182753fe950d514c2c6dd
which introduced these problems. MariaDB 10.2.2 inherited these problems
in commit 2e814d47.

The FOREIGN KEY CASCADE and SET NULL operations implemented as
procedural recursion are consuming more than 8 kilobytes of stack
(9 stack frames) per iteration in a non-debug GNU/Linux AMD64 build.
This is why we need to limit the maximum recursion depth to 15 steps
instead of the 255 that it used to be in MySQL 5.7 and MariaDB 10.2.

A corresponding change was made in MySQL 5.7.21 in
https://github.com/mysql/mysql-server/commit/7b26dc98a624d5cdf79cd5eee455cb03e3bccb5a

row_upd_store_v_row(): Declare static

row_upd_clust_rec_by_insert(), row_upd_del_mark_clust_rec():
Pass the parameter 'foreign' only #ifdef WITH_WSREP

This corruption was introduced in MDEV-13331. It would have been caught
by the MySQL 5.7 test innodb.update-cascade which MariaDB was missing
until now.

row_ins_check_foreign_constraint(): Never replace err == DB_LOCK_WAIT
with other values than DB_LOCK_WAIT_TIMEOUT.

srv_mbr_print(): Remove. Geometry fields are already being output
in raw form, and we do not need bogus MBR output for non-spatial
indexes.

row_ins_cascade_calc_update_vec(): Remove the output parameter
fts_col_affected, and instead return whether any fulltext index
is affected by the cascade operation.

row_ins_foreign_check_on_constraint(): Narrow the scope of some
variables.

ib_dec_in_dtor: Remove.

If setting user variable was used in the specification of a recursive CTE
then Item_func_set_user_var::fix_fields() went into an infinite loop.

                 does not return error

Corrected the code of st_select_lex::find_table_def_in_with_clauses() for
a proper identification of CTE references used in embedded CTEs.

When identifying a table name the following should be taken into account:
a CTE name cannot be qualified with a database name, otherwise the table
name is considered as the name of a non-CTE table.

replace with awk (yuck!), avoiding gensub (which wasn't recognized)

all mmap-related functions take size_t lengh.
mmaped_length was casted to size_t *every single time*

conversion from 'my_off_t' to '::size_t', possible loss of data

get_rec_bits() must return uchar, it's result is often
assigned to uchar. And len is always less than 8.

this fixes the crash of main.fulltest

because dbug_print_item() can be invoked any time in gdb,
also before fix_fields.

this fixes a crash of main.win in debug builds

it was set before big if() that could decide not to merge and
go with materialization.

this fixes a crash in main.view test

in gcc `-Wno-unsupported-something` will not be an error or even a warning,
so cmake will think the flag is supported. But if there's any other
warning during compilation, for any reason, unknown option will
be a warning too. Or an error when -Werror, even if that "other warning"
would not be an error on itself.

So we need to detect whether `-Wno-unsupported-something` is *really*
supported. Luckily, `-Wunsupported-something` will always fail with an
error.

So, whenever there's a need to detect if -Wno-something is supported,
test -Wsomething instead.

Problem was that wrong error message was returned when insert
returned FK-error and there was no duplicate key to process.

row_ins
	If error from insert was DB_NO_REFERENCED_ROW and there was
	no duplicate key we should ignore ON DUPLICATE KEY UPDATE
	and return original error message.

Suppress some messages that are emitted rarely
(when the FIL_PAGE_FILE_FLUSH_LSN of the first page of ibdata1
does not match the latest redo log checkpoint).

Previously, the function could theoretically return an uninitialized
value if the system tablespace contained no data files. It should be
impossible for InnoDB to start up in such scenario.