Bug #32287863 BACKPORT THE PATCH FOR BUG #27874068 ADD SUPPORT FOR -D…

…CMAKE_BUILD_TYPE=RELEASE

Follow-up patch:
re-enable -O3 by default for RelWithDebInfo (and Release) builds on Linux.

Change-Id: Ica5190aa1d6f6539ad629d93c10e4a0fd0bcdd2d
(cherry picked from commit 6cde07847a05e12640806f005d9eb7d120440b89)

Bug #32287863 BACKPORT THE PATCH FOR BUG #27874068 ADD SUPPORT FOR -D…

…CMAKE_BUILD_TYPE=RELEASE

Follow-up patch:
re-enable -O3 by default for RelWithDebInfo (and Release) builds on Linux.

Change-Id: Ica5190aa1d6f6539ad629d93c10e4a0fd0bcdd2d

Bug #32628376 BACKPORT FIX OF BUG 32335263 TO 5.6+

PROBLEM
-------
If the query contains password in plain text,
we have the query re-written immediately after
parsing and password string is replaced. However,
there is a unsafe window before rewrite is done
and in such case we should not display the plain
text password.

Affected queries are
SHOW PROCESSLIST;
SELECT ... FROM information_schema.processlist;

Solution:
---------
Fix this is to avoid displaying the
query string till re-write is over.

#rb 26092
#rb 26070

Reviewed-by: Debarun Bannerjee <[email protected]>
Reviewed-by: Reviewed by: Dyre Tjeldvold <[email protected]>

(cherry picked from commit c7d0c5b6ee1cfc2a5328e9adf058857b43e0a091)

Bug #32628376 BACKPORT FIX OF BUG 32335263 TO 5.6+

PROBLEM
-------
If the query contains password in plain text,
we have the query re-written immediately after
parsing and password string is replaced. However,
there is a unsafe window before rewrite is done
and in such case we should not display the plain
text password.

Affected queries are
SHOW PROCESSLIST;
SELECT ... FROM information_schema.processlist;

Solution:
---------
Fix this is to avoid displaying the
query string till re-write is over.

#rb 26092
#rb 26070

Reviewed-by: Debarun Bannerjee <[email protected]>
Reviewed-by: Reviewed by: Dyre Tjeldvold <[email protected]>

(cherry picked from commit c7d0c5b6ee1cfc2a5328e9adf058857b43e0a091)

Update License Book for Server 5.6.51 release

Approved-by: Bjorn Munch <[email protected]>

BUG#32019842: WRITE_SET VECTOR HAS NO UPPER MEMORY LIMIT

When the option is active, each server transaction will collect write
sets for each row, allowing to identify what changes the transaction
made, important for replication parallelization in general and Group
Replication conflict detection.

This patch improves the handling of write set collection for each
transaction in 2 ways.

First is that write sets are discarded if their size grow above
binlog_transaction_dependency_history_size. As a consequence the
transaction will no longer be marked as concurrent with previous ones
even if they touch different data.

The second one is that components like Group Replication can set
memory limits for the collection of write sets. Components like GR are
dependent on write sets being present, so they will never be
discarded, but breaking this memory limit will cause transactions to
abort.

ReviewBoard: 25520
ReviewBoard: 25592
(cherry picked from commit 933cc1c3b1dcf175de5aef0b6e2a826435481af6)