MySQL Community Server 8.0.17 has been released (part 3/3)
Posted by: Balasubramanian Kandasamy
Date: July 22, 2019 03:09AM
Date: July 22, 2019 03:09AM
[This is part 3 (last) of the announcement ]
* The functions STATEMENT_DIGEST() and
STATEMENT_DIGEST_TEXT() used character_set_client while
parsing the function's argument instead of the character
set of the argument. (Bug #29526571)
* The server now tracks data dictionary upgrades to ensure
that INFORMATION_SCHEMA views are recreated when data
dictionary tables are changed. (Bug #29513265)
* Execution of STATEMENT_DIGEST() or
STATEMENT_DIGEST_TEXT() could have unintended side
effects on data in the
performance_schema.events_statements_summary_by_digest
table. (Bug #29512067)
* The implementation for CREATE TABLE did not provide the
table a clearly defined starting cursor when no seeks or
writes were performed. (Bug #29511318)
* When a table is const optimized, any predicate terms in
the join's ON condition should also be evaluated on the
tables marked as const to make sure that the row
qualifies for the ON condition.
The optimizer failed to mark a table as const when it was
handled as having an impossible ON condition, in which
case the resulting row from this table is a NULL extended
row.
This fix sets the const_table flag for the table and
marks the row as a NULL row. It also changes EXPLAIN
output for a table optimized as having an impossible ON
condition, where the resulting NULL values from this row
are now propagated into the rows column values, as with
other const rows. (Bug #29493830)
* A query containing an IN subquery could return
superfluous rows when optimizer_switch did not include
firstmatch=on. (Bug #29493026)
* RPM package detection of whether the operating system is
EL6 or EL7 failed on some systems. (Bug #29492896)
* SQL layer validation of keys stored in
storage-engine-private data fields in the data dictionary
was disabled to permit storage engines to add new keys as
required. Storage engines are now responsible for key
validation. (Bug #29491593, Bug #94667)
* Checks for NULL returned an incorrect result for some
expressions comparing rows. (Bug #29491083)
* When optimizing a table join which included ORDER BY and
LIMIT, the optimizer modified the select limit on the
first table by taking the fanout of the entire join into
consideration. The fanout was calculated using the
filter_effect for each table, which can be set to
COND_FILTER_STALE (-1) to indicate an unknown value. This
value was not actually checked for by the optimizer,
which could result in a negative fanout, leading to an
unexpected rows value of 0 in the EXPLAIN output for the
query.
Now COND_FILTER_STALE (filter_effect unknown) is
specifically checked for, and, if it is found, the fanout
is also assumed to be unknown, and the number of rows to
be selected from the first table is thus not modified by
the fanout for the entire join. (Bug #29487181, Bug
#29531472)
* An implicit assumption was made for the LEAST() and
GREATEST() functions that arguments to either of these
were all signed or all unsigned values. (Bug #29467577)
* Some arguments for the BIT_OR() function were not always
handled correctly. (Bug #29459549)
* For debug builds, mishandling of the user_attributes
column of the mysql.user system table could cause a
server exit. (Bug #29451897)
* MySQL Cluster-specific Debian debug packages had an
incorrect path to the debug symbols. (Bug #29446947)
* Compilation failure occurred when building with Protobuf
3.7.0. (Bug #29436791, Bug #94543)
* On Debian and Ubuntu, MySQL packages did not enable
mysql.service after upgrades from native MySQL packages.
(Bug #29435592)
* The server did not properly close shared-memory
connections when an error occurred, which could result in
unexpected server behavior. (Bug #29435426)
* The internal JOIN_CACHE::join_record() method attempted
to leave all tables on which it operated in the same
state with regard to the has_row and null_row flags by
saving the state upon entry and restoring the appropriate
flags when on completion.
The issue addressed here arose after these flags had been
restored, due to the fact that restore_last_record() was
also called to restore any records that had been
processed, which could lead to setting back a null_row
flag that had just been restored based on the saved
state. This is fixed by calling restore_last_record()
before the saved states are restored, rather than after,
as well as making sure that reset_null_row() is also
called if the saved state indicates that the null_row
flag was not set at the beginning. (Bug #29435133)
* Some ALTER TABLE statements using ADD COLUMN did not
perform as expected. (Bug #29428288)
References: This issue is a regression of: Bug #28333657.
* An incorrect error message was reported during upgrade
from MySQL 5.7 to MySQL 8.0 when a non-natively
partitioned table was encountered. The error message
referenced the -DWITH_PARTITION_STORAGE_ENGINE CMake
build option, which is not supported in MySQL 8.0. (Bug
#29426632, Bug #94518)
* MySQL Installer did not install OpenSSL DLL dependencies
if the Development component was not selected. (Bug
#29423421, Bug #94168)
* At startup, the server did not properly check the length
of user and host names in persisted variables. (Bug
#29420141)
* The parser could leak memory for certain
multiple-statement queries. (Bug #29419820)
* Installing MySQL on EL8 from RPM packages caused a
conflict with the installed MeCab library. (Bug
#29413115)
* On FreeBSD, stack traces dumped for fatal signals did not
work correctly. (Bug #29408039)
* With the derived_merge switch disabled in the value of
the optimizer_switch system variable, information
retrieved for base tables from INFORMATION_SCHEMA.TABLES
displayed inappropriate information in the TABLE_COMMENT
column. (Bug #29406053, Bug #94468)
* In certain cases, use of ORDER BY in a subselect did not
produce the expected result. (Bug #29402943)
* ROLLUP queries with wildcards (and probably also certain
other wildcard queries) could cause assertion errors.
(Bug #29396628)
* A condition inside an IF function having a constant
evaluating to TRUE was not always handled correctly. (Bug
#29394833)
* A code change in MySQL 8.0.13 related to full-text search
caused a segmentation fault. (Bug #29393105)
* mysqlpump did not check whether the argument to
--result-file was opened correctly and exited
unexpectedly if an error occurred. (Bug #29389828)
* In queries without joins, batch mode was not enabled
until after initialization was complete, so that OFFSET
rows were read outside of batch mode. This negated any
performance benefit to be had from batch mode. (Bug
#29373972)
* When renaming a functional index, the server did not
check for a possible duplicate index name. (Bug
#29360763)
* The system variable service did not check the input
buffer size for some operations. (Bug #29343505)
* For mysql system schema dumps, mysqlpump dumped certain
tables by row rather than as account-management
statements. (Bug #29343073)
* Creation of a functional index on a UNION subquery led to
an assertion in debug builds. (Bug #29342245)
* Non-numeric arguments to AVG() were not always handled
correctly. (Bug #29321764)
* Result sets with a very large number of columns could
cause client programs to use all available memory. Now
the client library allocates no more than
max_allowed_packet bytes for result set metadata,
returning an out-of-memory error if this limit is
exceeded. If this error occurs, it is fatal and the
client should disconnect. (Bug #29316814)
* A call to mysql_shutdown() could cause the client to exit
when the server and client were from different MySQL
series, due to return packet payload misinterpretation.
(Bug #29315393)
* Client programs could fail while reading result set
metadata if communicating with the server using the
pre-MySQL 4.1 protocol. (Bug #29304864)
* HANDLER statements did not always work correctly with
tables having generated columns. (Bug #29300049)
* Session-tracking information in the client/server
protocol could be mishandled. (Bug #29297652)
* Client programs that used the libmysqlclient C client
library could exit upon receipt of an OK packet
containing malformed session-tracking information. (Bug
#29297620)
* With the PAD_CHAR_TO_FULL_LENGTH SQL mode enabled,
password changes failed, with no warning or error
reported. (Bug #29287785)
* Initialization code of user-defined functions for the
keyring_aws plugin did not properly check whether the
plugin had been initialized. (Bug #29278153)
* Scalar subqueries were not always identified and handled
correctly when compared with aggregate or window
functions. (Bug #29276063)
* Ubuntu packages did not disable binary logging for
initialization scripts. (Bug #29263771)
* The internal collation map could become corrupted by use
of invalid collation names. (Bug #29258979)
* Single-table UPDATE and DELETE statements that used
indexed expressions could cause the server to exit when
executed as prepared statements. (Bug #29257254)
* The audit_log plugin did not log UNINSTALL PLUGIN
audit_log statements. (Bug #29248047)
* MySQL now uses open(O_TMPFILE) whenever applicable when
creating a temporary file that is immediately unlinked.
This is more efficient than previously and avoids the
small possibility of a race condition. Thanks to Daniel
Black for the contribution. (Bug #29215177, Bug #93937)
* audit_log filtering operations could leak memory. (Bug
#29201747)
* The GRANT OPTION privilege could be mishandled in some
cases. (Bug #29179334)
* Previously, LIKE comparisons could be incorrect if either
of the _ or % SQL wildcard characters was used as the
ESCAPE character. These characters can now be used. (Bug
#29175461, Bug #93811)
* The component_sys_variable service could fail to read
certain system variables, causing a server exit. (Bug
#29162657)
* Multiple-table UPDATE statements could search for
incorrect generated-column values when determining which
row to update. (Bug #29154379)
* If the server was started with Performance Schema
parameters that caused an out-of-memory condition,
startup failed. (Bug #29140212, Bug #93726)
* For consistency and compliance with standard SQL, the
ISNULL() function is now printed as IS NULL in warnings
generated by EXPLAIN statements. (Bug #29119455)
* Some supplemental Unicode characters could incorrectly be
flagged with a warning message as invalid. (Bug
#29110613, Bug #93626)
* For upgrades from MySQL 5.7 to 8.0, the root user was not
given all dynamic privileges. (Bug #29043233)
* When using subpartitioning, table serialized dictionary
information (SDI) was not stored in any tablespace file.
(Bug #29020745, Bug #93499)
* Privileges for dropping some Performance Schema tables
were checked incorrectly. (Bug #29010031)
* mysqldump failed to wrap SET NAMES utf8mb4 and SET
character_set_client = utf8mb4 statements within
version-specific comments, which could cause
compatibility problems. (Bug #29007506, Bug #93450)
* For a unique index created on one or more columns defined
as NULL, the optimizer failed to handle joins using that
index as eq_ref.
As part of this fix, late NULL filtering is now also
performed to avoid treating comparisons with NULL as
potential equality matches. (Bug #28965762, Bug
#29337233)
* For debug builds, an assertion could be raised by string
comparisons for which the expected and actual collation
differed. (Bug #28960901)
* The server did not handle correctly a subquery which,
prior to conversion to a semijoin, contained a merged
derived table or common table expression which in turn
had a WHERE clause with an outer reference against an
indexed column. (Bug #28955216)
* A query that employed a derived table which included an
ORDER BY was not always handled correctly. (Bug
#28942965)
* When clients were terminated for inactivity exceeding the
wait_timeout value, the message written to the error log
was unclear. Now ER_NET_WAIT_ERROR is written, which is
more specific about the cause of the problem. (Bug
#28940167, Bug #93240)
* Concurrent execution of FLUSH PRIVILEGES along with
CREATE USER or ALTER USER could cause the server to
stall. (Bug #28937018, Bug #93085)
* CASE statement comparisons that relied on index prefix
values could produce incorrect results. (Bug #28934315,
Bug #93215)
* When adjusting the query cost after sort elimination to
compensate for sorts that had not been considered in the
original cost estimate, and the estimates could sometimes
be too low, or even negative.
The sort cost is added to the total cost in
Optimize_table_order::consider_plan(), and subtracted
from the total cost in JOIN::optimize() if it is found
that sorting is not needed.
If consider_plan() finds that sorting is not needed, it
does not add the sort cost to the total cost, but still
records a sort cost in JOIN::sort_cost(), and this is set
to the sort cost of the candidate plan considered
previously. Later, JOIN::optimize() saw that the JOIN
object had an associated sort cost, and subtracted that
cost from the total cost, since it also sees that no sort
is needed. Since the sort cost came from a candidate plan
that was not the same as the chosen plan, the estimate
was incorrect.
The fix is to make consider_plan() set JOIN::sort_cost to
0 if no sort cost is added to the total cost, so that the
mistaken adjustment of the cost in JOIN::optimize() is
not performed. (Bug #28884359)
* Some GIS code failed to compile under Visual Studio 2017
15.5.6. (Bug #28861188)
References: This issue is a regression of: Bug #28842878.
* Stored generated column values and indexes on virtual
generated columns were not correctly updated after
columns on which these generated columns depended were
swapped using ALTER TABLE with RENAME COLUMN or CHANGE
COLUMN.
Renaming of base columns for generated columns, generated
defaults, and functional indexes is now prohibited unless
the same ALTER TABLE statement satisfies one of these
conditions:
+ The statement removes the generated column,
generated default, or functional index.
+ The statement updates the dependent expression in
question. This enables supporting existing scenarios
in which generation expressions are updated to
follow base column renaming.
Restrictions on dropping columns on which generated
columns, generated defaults, or functional indexes depend
were relaxed in a similar way. (Bug #28772251, Bug
#92727)
* Base columns were not excluded from index-only access by
a generated column. (Bug #28652733)
References: See also: Bug #29664369. This issue is a
regression of: Bug #23169112.
* SET PERSIST_ONLY did not properly handle the
version_tokens_session system variable. (Bug #28542569)
* For debug builds, improper error checking for CREATE
TABLE statements could cause an assertion to be raised.
(Bug #28490361, Bug #91976)
* MySQL 5.7 supported foreign keys on InnoDB tables with a
parent key for which part of the referenced columns were
hidden, but MySQL 8.0 did not. MySQL 8.0 now supports
this capability. (Bug #28480149, Bug #91952)
* INSERT ... ON DUPLICATE KEY UPDATE did not consider
privileges granted by active roles. (Bug #28395115)
* For unloaded components, component options specified at
startup with a --loose- prefix were not processed if the
component was later loaded at runtime. (Bug #28341329)
* Long passwords solicited interactively by client programs
could fail to be null-terminated. (Bug #28121400)
* When building MySQL, CMake begins the process of
downloading Boost by creating a zero-length tarball in
the destination directory, which is removed when the
download is complete. If the download was interrupted or
timed out, the presence of this file prevented CMake from
attempting to perform the download the time it was run.
Now the zero-length tarball, if present, is removed
before the download is started. (Bug #28089173)
* A thread pool group could be blocked when a thread
process tick time exceeded the maximum permitted value.
The tick time now uses a larger data type to permit
larger values. (Bug #28072609)
* Privileges were not checked correctly for ALTER USER ...
IDENTIFIED WITH ... BY. (Bug #27923149)
* MySQL does not support OpenSSL session tickets, but did
not set the SSL_OP_NO_TICKET flag to inform OpenSSL of
that. The flag is now set. (Bug #27655493)
* The audit_null plugin did not properly check for a null
event record. (Bug #27638290)
* UpdateXML() did not always free memory properly in
certain cases. (Bug #27312862)
* Empty values in the name column of the mysql.plugin
system table caused the server to exit during startup.
(Bug #27302459)
* The server did not properly check privileges for CACHE
INDEX statements. (Bug #26173827)
* With the thread_pool plugin enabled, the Performance
Schema status_by_thread table contained no data. (Bug
#25933891)
* The GRANT OPTION privilege was treated as related to
database operations. (Bug #25203933)
* REPAIR TABLE for ARCHIVE tables could result in a server
exit. (Bug #23304911)
* If an INSTALL PLUGIN statement contained invalid UTF-8
characters in the shared library name, it caused the
server to hang (or to raise an assertion in debug
builds). (Bug #14653594, Bug #23080148, Bug #27167197)
* Logging to the mysql.slow_log system table could fail
when values were to large for table columns. Now logging
proceeds on a best-effort basis, writing what information
can be provided. Otherwise, the row is discarded and a
message is written to the error log. (Bug #11748692, Bug
#37132)
* A query using WHERE date_column LIKE 'year_value' failed
with error 1525 Incorrect DATE value on Windows
platforms. (Bug #95780, Bug #29904751)
References: This issue is a regression of: Bug #29368521.
* A fix for a previous issue caused YEAR values to be
treated as unsigned, but this did not allow for possible
negative values arising during calculations involving
subtraction, which could lead to data truncation errors.
(Bug #95045, Bug #29668676)
References: This issue is a regression of: Bug #92209,
Bug #28562930.
* When working with derived tables with an aggregation
which had zero input rows, the results of the aggregate
functions were not properly copied into the temporary
table. This caused incorrect results in cases where the
derived table was evaluated multiple times, such as when
performing a lateral join. (Bug #94721, Bug #29514504)
* A window without a frame specification inheriting from a
window with an ORDER BY yielded an incorrect result. (Bug
#94251, Bug #29328529)
* The results returned by the functions REGEXP_REPLACE()
and REGEXP_SUBSTR() used UTF-16 rather than the character
set and collation of the expression searched for matches.
(Bug #94203, Bug #29308212)
* A UNION ALL query with SUM(constant) was processed very
slowly compared to the same query using SUM(column)
instead. (Bug #93922, Bug #29227464)
* JSON_OBJECTAGG() is non-deterministic in the presence of
duplicate keys unless the window has ordering on the key,
which is expected behavior, but a key-value pair that was
no longer in the window frame still appeared in the
result. (Bug #93822, Bug #29175262)
* LIMIT was applied before HAVING in queries with
subselects. (Bug #93214, Bug #28934388)
References: This issue is a regression of: Bug #25466100.
* A query involving GROUP BY on a TIMESTAMP column resulted
in a duplicate entry for key (ER_DUP_ENTRY) error. This
problem arose when TIMESTAMP values were inserted into a
table using a given setting for the time zone and these
values were later fetched after the time zone setting had
been changed, such that at least some of the inserted
TIMESTAMP values occurred during the hour that the time
changed from standard to daylight time (DST) in the new
time zone, during which time the same TIMESTAMP value can
exist twice. Now, when the server would otherwise return
the error DUPLICATE ENTRY FOR KEY 'group_key', if the
grouping involves a TIMESTAMP column, it instead raises
the error Grouping on temporal is non-deterministic for
time zones having DST. Please consider switching to UTC
for this query.
In addition, it is suggested to set
explicit_defaults_for_timestamp to ON as well as one or
more of MODE_NO_ZERO_IN_DATE, MODE_NO_ZERO_DATE, or
MODE_INVALID_DATES as part of the server SQL mode to help
avoid this issue. (Bug #90398, Bug #27970159)
On Behalf of Oracle/MySQL Release Engineering Team,
Balasubramanian Kandasamy
Subject
Views
Written By
Posted
MySQL Community Server 8.0.17 has been released (part 3/3)
2027
July 22, 2019 03:09AM
Sorry, you can't reply to this topic. It has been closed.
Content reproduced on this site is the property of the respective copyright holders. It is not reviewed in advance by Oracle and does not necessarily represent the opinion of Oracle or any other party.