When a page is being requested from buf_page_get_gen() with
mode=BUF_GET_POSSIBLY_FREED, it is very well possible that the page
had been marked as freed in the tablespace and had never been written
back to the tablespace.

buf_page_t::read_complete(): Validate the page identifier before
invoking buf_zip_decompress(), to avoid noisy messages or debug
assertion failures on an obviously invalid page.

Reviewed by: Thirunarayanan Balathandayuthapani
Tested by: Saahil Alam

Problem:
=======
In multi-batch recovery, pages with LSN > recv_sys.scanned_lsn
are discarded from doublewrite buffer. Later when
reading a corrupted page from disk, corresponding dblwr page
is no longer available. This leads to failure of InnoDB recovery.

Solution:
=========
In buf_dblwr_t::recover(): Replace recv_sys.scanned_lsn
with max_lsn when checking if a page's LSN is within
the recovery range.

…ollback()

Problem was executable comment in REPLACE that caused
consistency check that is not supported for RSU and
!InnoDB tables.

Fix is to check is Online Schema Upgrade (OSU) method
something else than TOI and if it is then table storage
engine must be InnoDB or error is produced and execution
returns and no consistency check is done.

…anagement

We compile curl with vcpkg. vcpkg prefers dependencies of the vcpkg
package to come from vcpkg as well. We break its logic by setting
and ZLIB_LIBRARY in cmake/zlib.cmake to the static library that we
compile outselves. This vcpkg check is not critical or needed
when shared libraries are used.

As a workaround, set CMAKE_DISABLE_FIND_PACKAGE_ZLIB when compiling
hashicorp_key_management, in this case vcpkg skips the zlib check.

Test output was affected by incompletely closed preceding connections.

Wait for connections to leave I_S.PROCESSLIST before issuing
SHOW PROCESSLIST.

…back_stmt(THD*)

Problem was that user could drop streaming replication table
when streaming replication was used.

Fixed by not allowing user to drop streaming replication
table when galera is enabled. When it is not enabled
super can drop it (as any table in mysql database).
Added checks for wsrep_trx_fragment_unit and wsrep_trx_fragment_size
variables so that streaming replication can't be enabled
when streaming replication table does not exists.

Fixed also bug when user is trying to disable streaming
replication by wsrep_trx_fragment_size=0 when it was
not enabled.

…xisting user after setting max_error_count and wsrep_ignore_apply_errors to zero

There was two problems. Firsty, implementation assumed that
thd->wsrep_rgi is not nullptr, this is not true for TOI.
Secondly, Sql_condition_iterator was also assumed to be
not nullptr, this might not be also true.

Fixed by adding nullptr guard for thd->wsrep_rgi and
if Sql_condition_iterator is nullptr obtain error
information directly from Diagnostics_area if there
is some error there.

Test output was affected by incompletely closed preceding connections.

Make test agnostic to concurrent connections by querying
information_schema.processlist only for connections that it uses.

With bead24b all 3 processlist queries were affected.
Before bead24b - only the first one.

correct the length check.
remove assertions that a file read from disk contains a specific substring

…e-level CREATE but not with global CREATE

CHECK TABLE was inconsistently requiring SELECT privilege on global/db level
or any privilege on the table/column level.

Change to require any table-applicable privilege on any level.

…, causes replication discrepancy

don't trust the content of a file read from disk

… queries to fail

don't copy field default values and check constraints in
CREATE ... SELECT.

CREATE ... SELECT means a table is created from a *result set*
not from some other table.

For backward compatibility, though, let's keep copying constant
default values and the "compressed" attribute.

number-to-time conversion was too eagerly capping the value.

A string "9000090" was invalid time, because of 90 seconds.
But number-to-time was capping first, validating later,
to 9000090->time worked. Let's fix it.

also, let's make invalid time values in a string field include
the field name in the warning message, just like invalid time
values in a numeric field do.