MySQL Server, the world's most popular open source database, and MySQL Cluster, a real-time, open source transactional database.

The casting logic here resulted in the timeout double to be cast into a ulonglong before the multiplication happened. This resulted in all timeouts being rounded down to whole seconds.

The behavior is visible with statements like the following:

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 0.1);
+------------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 0.1) |
+------------------------------------------------------------------------------+
|                                                                            1 |
+------------------------------------------------------------------------------+
1 row in set (0.00 sec)

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 0.5);
+------------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 0.5) |
+------------------------------------------------------------------------------+
|                                                                            1 |
+------------------------------------------------------------------------------+
1 row in set (0.00 sec)

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 1.0);
+------------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 1.0) |
+------------------------------------------------------------------------------+
|                                                                            1 |
+------------------------------------------------------------------------------+
1 row in set (1.00 sec)

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 1.5);
+------------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('16d22a8b-e803-11e8-b46f-003d70c354d2:9067', 1.5) |
+------------------------------------------------------------------------------+
|                                                                            1 |
+------------------------------------------------------------------------------+
1 row in set (1.00 sec)

The query execution time here shows the problem. Everything between 0.0 and 1.0 is rounted to a wait time of 0.0 seconds. When using a value in a 1.0 to 2.0 window, it rounds down to 1 seconds as visible in the query execution time.

After this change it works as expected, again visible in the query execution time:

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('f9702b80-e826-11e8-b833-f64268411190:4', 1.0);
+---------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('f9702b80-e826-11e8-b833-f64268411190:4', 1.0) |
+---------------------------------------------------------------------------+
|                                                                         1 |
+---------------------------------------------------------------------------+
1 row in set (1.01 sec)

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('f9702b80-e826-11e8-b833-f64268411190:4', 0.5);
+---------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('f9702b80-e826-11e8-b833-f64268411190:4', 0.5) |
+---------------------------------------------------------------------------+
|                                                                         1 |
+---------------------------------------------------------------------------+
1 row in set (0.50 sec)

mysql> SELECT WAIT_FOR_EXECUTED_GTID_SET('f9702b80-e826-11e8-b833-f64268411190:4', 0.1);
+---------------------------------------------------------------------------+
| WAIT_FOR_EXECUTED_GTID_SET('f9702b80-e826-11e8-b833-f64268411190:4', 0.1) |
+---------------------------------------------------------------------------+
|                                                                         1 |
+---------------------------------------------------------------------------+
1 row in set (0.10 sec)

Replaced make_pair<string, string > with pair<string, string > to be able to compile using gcc version 6.1.1 20160510 (Ubuntu 6.1.1-2ubuntu12~16.04)

This changes enables sending along the SNI extension to the server when connecting over TLS with a client. It's best practice these days to send SNI as a TLS client.

Usage of SNI would also enable uses cases like routing of MySQL connections to different backends behind a single IP.

This change only implements the client side parts. Server side there is also value in providing different certificates on different SNI names, for example for multi homed MySQL servers with different DNS entries to connect combined with verify_identity semantics.

At first though, clients should start sending the extension before it's useful to implement anything on the server side.

Additionally, the second commit here enables usage of the system roots when no CA file or path is configured. Right now a CA chain is required to be provided in verify_ca or verify_identity mode. This makes it much more cumbersome when writing a client that connects to a server where the certificate in the system roots and when it provides a proper certificate signed by such an authority.

By falling back to the system roots, deployments are greatly simplified because it removes the need to manually specify the root chain. The root chain is highly operating system and deployment dependent, so it requires a different configuration flag on each platform.

By setting up the system roots through OpenSSL, deploying a MySQL server with a root signed installed on a system is greatly simplified for client configuration.

Many clients in other languages / on other platforms already follow a similar behavior where system roots are used if no specific CA chain is configured. This applies to for example Java, Go, Node.js, Rust & .NET.

This change also makes it simpler for drivers that use libmysqlclient such as the ones often used in Ruby and Python to simplify configuration for developers there as well.

Also give an example for RHEL 7.

I've not played with rpmbuild for some time and sat there copying the "sample" rpmbuild command to build the package only to realise later that I was missing the -ba.

I think it probably makes sense to update this to reference the latest RHEL7 version. The patch is tiny and only a documentation change but might be handy and save someone a few minutes scratching their head like I did.

These commits fixed four bugs in multi-valued index:

1. multi-valued index isn't picked in view
2. multi-valued index isn't picked in prepared statement
3. multi-valued index isn't picked in member of(col) OR member of(col)
4. MySQL wrongly gets "impossible condition" thus empty result set for WHERE func(...) AND func(...) where func is member of, json_contains or json_overlaps.

For some special,we want to know all the request more detail,many like slowlog,rather than genlog,but some time It's difficult to reconnect all client

Existing comments int statements_event.h provides outdated values for sql mode system variables. This PR updates the values and adds new variables to prevent confusion.

Include the right headers in kdf_sha_crypt.cc:

• <stddef.h> for size_t
• <cstdlib> for std::strtol
• <tuple> for std::tie (this is the one that broke the build)
• <iterator> for std::distance

Include the right headers in kdf_pbkdf2.cc:

• <stdint.h> for uint8_t
• <cstdlib> for std::strotl
• <tuple> for std::tie (this is the one that broke the build)
• <iterator> for std::distance

Include the right headers in http_auth_method_basic.cc:

• <iterator> for std::back_inserter (this is the one that broke the build)
• <system_error> for std::make_error_code and std::errc

Failing output was:

F:\gitP\mysql\mysql-server\router\src\http\src\kdf_sha_crypt.cc(71): error C2039: 'tie': is not a member of 'std' [F:\gitP\mysql\mysql-server\build_amd64\router\src\http\src\http_auth_backend_lib.vcxproj]
F:\gitP\mysql\mysql-server\router\src\http\src\kdf_sha_crypt.cc(71): error C3861: 'tie': identifier not found [F:\gitP\mysql\mysql-server\build_amd64\router\src\http\src\http_auth_backend_lib.vcxproj]
F:\gitP\mysql\mysql-server\router\src\http\src\kdf_pbkdf2.cc(79): error C2039: 'tie': is not a member of 'std' [F:\gitP\mysql\mysql-server\build_amd64\router\src\http\src\http_auth_backend_lib.vcxproj]
F:\gitP\mysql\mysql-server\router\src\http\src\kdf_pbkdf2.cc(79): error C3861: 'tie': identifier not found [F:\gitP\mysql\mysql-server\build_amd64\router\src\http\src\http_auth_backend_lib.vcxproj]

Bug #79636: CACHE_LINE_SIZE should be 128 on AArch64 Bug #79637: Hard-coded cache line size Bug #79638: Reconcile CACHE_LINE_SIZE with CPU_LEVEL1_DCACHE_LINESIZE Bug #79652: Suspicious padding in srv_conc_t

• changed CPU_LEVEL1_DCACHE_LINESIZE to default to 128 bytes on POWER and AArch64 architectures in cases when no value could be detected by CMake using getconf
• changed pfs_global.h to assume CPU_LEVEL1_DCACHE_LINESIZE is always defined (and throw a compile-time error if it's not)
• changed CACHE_LINE_SIZE definition in ut0counter.h to be an alias of CPU_LEVEL1_DCACHE_LINESIZE
• changed a number of hard-coded 64-byte cache line size values in the InnoDB code
• fixed insufficient padding for srv_conc members in srv0conc.cc

This library is needed to be explicitly included for gcc11. It shouldn't affect other versions of gcc, but without it, code won't compile in with gcc11.

You can read about it here (Header dependency changes): http://www.gnu.org/software/gcc/gcc-11/porting_to.html

Outline-atomics flag modifies builtin atomics to detect CPU support for atomic instructions in runtime, thus getting benefits of using LSE atomics without hurting compatibility. This flag has no effect if compiling for aarch64 >=v8.1.

Due to the nature of exclusive accesses, which are used when LSE atomics are not available, the effect of LSE atomics is much stronger on large core counts. We've observed as far as 4x transactions per second on sysbench read_only by adding outline-atomics to 64-core aarch64 server (AWS g6.16xlarge).

Bumps certifi from 2019.11.28 to 2022.12.7.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.