PXB-3403: Merge 9.1 #1635
Open
PXB-3403: Merge 9.1 #1635
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
A query in main.derived_condition_pushdown occasionally failed when running with the hypergraph optimizer. The problem was that one of the input tables to a hash join was an internally generated temporary table, and the hash join code was not prepared for this possibility. In such tables, table->pos_in_table_list is nullptr, so the hash join code shouldn't unconditionally dereference table->pos_in_table_list. Fixed by adding checks for nullptr. Change-Id: I299933c95a1b1330f821ee7f2d62f1b01b98cd47
Change-Id: Ia22eb89411136d20f385ececf8c7ed34d8d9fdb9
…noclose] Move the test which covers the DERIVED_CONDITION_PUSHDOWN and NO_DERIVED_CONDITION_PUSHDOWN table-level hints to an include file which can be reused for a hypergraph optimizer variant of the test in a later patch. Change-Id: I396cbdbc7bab63cdbf17d6a7718c7dd92b19bf55
Enable a test which covers the DERIVED_CONDITION_PUSHDOWN and NO_DERIVED_CONDITION_PUSHDOWN table-level hints with the hypergraph optimizer. Change-Id: Icb95ccf6a3e64882da368864e8320a3a753f0333
Add a cmake library ext::rpc with necessary INTERFACE properties.
Get rid of the global
INCLUDE_DIRECTORIES(BEFORE SYSTEM "${TIRPC_INCLUDE_DIR}")
and
ADD_DEFINITIONS(-DHAVE_TIRPC)
Also some cosmetic cleanup of cmake code.
Change-Id: I0b2e1885bde584bfc4c22b9f8f7590a43e8b4225
…045 are never hit
Various issues here :
1. Those error inserts are not hit as they are in SUMA execSUB_CREATE_REQ()
which is only executed on SUMA when a Subscriber (aka EventOperation) is
created.
Just creating the Event is not enough to trigger them, so the test
covers nothing and naturally they would be left behind when it finishes.
Solution :
- Add create EventOperation and check that it fails
2. The other error inserts (error2) are related to SUMA Scans, where it
asks DIH to get distribution info. The error inserts are intended
to give code coverage of some 'many fragments' continueB cases.
a) This has nothing to do with Events
b) Creating and Dropping events does not involve scanning
Solution :
- Split into separate testcase
- Add index creation which *will* trigger a SUMA scan and will
cause these error insertions to be hit.
3. Error insert 13049 causes a crash when hit.
This error insert triggers a CONTINUEB, but then carries on with
the scan.
When the CONTINUEB arrives, the scan is not in the correct state
causing a crash.
Fix by having the code return after queueing the CONTINUEB.
Summary :
Testcase was a mess, testing nothing.
Error insert being left behind was a good indicator of a problem.
New test :
test_event -n SumaScanGetNodesContinueB T1 added to daily-basic--03
Change-Id: I4c4470cc6ef2a53dd470a4e4ac90ca1085504140
Change-Id: Iaab86d1a56ee6cb1c101ba64b23c693d06e6dbe9
Change-Id: I6402e68df1d2cd2a1ee20d6116667d39c583b976
…045 are never hit
Various issues here :
1. Those error inserts are not hit as they are in SUMA execSUB_CREATE_REQ()
which is only executed on SUMA when a Subscriber (aka EventOperation) is
created.
Just creating the Event is not enough to trigger them, so the test
covers nothing and naturally they would be left behind when it finishes.
Solution :
- Add create EventOperation and check that it fails
2. The other error inserts (error2) are related to SUMA Scans, where it
asks DIH to get distribution info. The error inserts are intended
to give code coverage of some 'many fragments' continueB cases.
a) This has nothing to do with Events
b) Creating and Dropping events does not involve scanning
Solution :
- Split into separate testcase
- Add index creation which *will* trigger a SUMA scan and will
cause these error insertions to be hit.
3. Error insert 13049 causes a crash when hit.
This error insert triggers a CONTINUEB, but then carries on with
the scan.
When the CONTINUEB arrives, the scan is not in the correct state
causing a crash.
Fix by having the code return after queueing the CONTINUEB.
Summary :
Testcase was a mess, testing nothing.
Error insert being left behind was a good indicator of a problem.
New test :
test_event -n SumaScanGetNodesContinueB T1 added to daily-basic--03
Change-Id: I4c4470cc6ef2a53dd470a4e4ac90ca1085504140
Change-Id: Ia1809ab6933cf4614f612716c996028f13f58922
Change-Id: I29b679727270f635e34f6c3a33559504b6f2a7a4
Change-Id: I259b97c51f98dc252886e84b8411151c99ae674a
Context: Test inject error 8064 to delay LQH_TRANSREQ during TC takeover. To force the takeover master node is restarted in 4 different ways: normal node restart, error insert 7007 (that triggers 9999), dump code 7213 and dump code 7214. The last 2 are sent to other node from where master is restarted (via 9999 or CRASH_INSERTION). Problem is with 7213 that is used to kill the master during a COMMIT. Node sends 5048 to master (that should crash it) followed by a 9999 (probably to ensure that master will be restarted if 5048 does not hit immediately). Problem: Between send of EI 5048 and 9999 error_insert_value is cleared via CLEAR_ERROR_INSERT_VALUE macro but, (since Bug#36082229 [1/2] Check there are no active error inserts set after NDBAPI test run. f0ec71b05ea6bc5e131e7d409ba31a1bc7995545), CLEAR_ERROR_INSERT_VALUE also clears the error_insert_extra variable that, in this context, is used to store the nodeId of master node. This way 9999 is sent to self instead of master node. Solution: Remove unnecessary CLEAR_ERROR_INSERT_VALUE call between send of EI 5048 and EI 9999. Change-Id: Ia630b8e07660720886cab54ece620a0ff32ac5a5
Change-Id: Ic63531538a4cd63b5f1ea39e22c60928273d7a35
Change-Id: If24942795f493ccebbe769190958e8f4355400a7
When a user variable is used in a DML statement, the variable and its value are allocated in the THD::user_var_events_alloc. Before the DML statement is written to the binlog, the user variable and its corresponding values from THD::user_var_events_alloc are also recorded in binlog. For stored procedures(SP), memory for user variables and their associated values is allocated within the procedure's execution memory root as each DML statement is written separately to the binlog. At the end of stored procedure execution, THD::user_var_events_alloc is set to nullptr and it is reset to THD::mem_root before executing the next statement. However, when executing statements using the Statement handle interface, THD::user_var_events_alloc is not reset to THD::mem_root after statement execution or before processing the next statement. This leads to the reported issue when any statement is executed following a stored procedure using the Statement handle interface. To address this issue, THD::user_var_events_alloc is now reset to the statement execution memory root after executing a sub-statement from the statement handle interface. Change-Id: Ie943d36e6f75dfa673e8d9ca5afd9125d21533a9
Change-Id: I91a6ac30c94432080a829b921495aff9a8470fd7
Fixed incorrect object length in unit test causing read past allocated data. Updated copyright notice to new format. Fixed doc comments. Change-Id: Id10d46c67a8a875e19aeb55e5c6d0be2d5f423d4
Change-Id: I2b710dfbb42533ee5e5d57a7a90804e7a24d9e08
…runk Change-Id: I4c5cc97eea7a2522103e153e4422d63ebb19133b
…real_item() The preparation of table value constructors doesn't keep track of the number of hidden columns added for expressions in the ORDER BY clause. This could cause problems later in the resolving and make the server exit. Fixed by counting the number of hidden items in table value constructors the same way as it's done in regular query blocks. Change-Id: Icb62214afb9a4a56c3bb92c648c1460cbf4b82c0
- Bug#36804093 ERROR 4153 (HY000): Incorrect double value:
'-1.7976931348623157e308' for column
- In debug code, there is a basic validation of characters allowed in
a double column. Included character e.
- In release code, use DBL_MAX for double.
- Added a test case to cover the scenario.
Change-Id: I610299dc07a5dc846d691a453ffb3263958bbcc1
Unpack curl-8.9.1.tar.gz $rm -rf buildconf.bat config.guess config.sub configure docs ltmain.sh m4 Makefile packages plan9 projects src tests winbuild git add curl-8.9.1 Change-Id: I643cbfe146e964e965122e9cb58055aec19ffb7c
Update CURL_VERSION_DIR Changes to curl cmake files: - cmake_minimum_required - CMAKE_DEBUG_POSTFIX - Keep FILE protocol for HTTP_ONLY - find_package(OpenSSL) - optional_dependency(ZLIB) to reuse results from cmake/zlib.cmake - Do not add "/MP" for clang on Windows Change-Id: If0d63d027b2ea319767e207cd493d7dbba23cf0f
Delete old source files. Change-Id: Ic7f63b43a8c2e744b3fc802ede76f947a0d6a53a
Add a cmake library ext::rpc with necessary INTERFACE properties.
Get rid of the global
INCLUDE_DIRECTORIES(BEFORE SYSTEM "${TIRPC_INCLUDE_DIR}")
and
ADD_DEFINITIONS(-DHAVE_TIRPC)
Also some cosmetic cleanup of cmake code.
Change-Id: I0b2e1885bde584bfc4c22b9f8f7590a43e8b4225
(cherry picked from commit 8efb8b269bac6a1b617ec5ae7c89a278e0a467b9)
Unpack curl-8.9.1.tar.gz $rm -rf buildconf.bat config.guess config.sub configure docs ltmain.sh m4 Makefile packages plan9 projects src tests winbuild git add curl-8.9.1 Change-Id: I643cbfe146e964e965122e9cb58055aec19ffb7c (cherry picked from commit 9de4385bffbe66f4718d6dddbfbf3dd8fae75ffb)
Update CURL_VERSION_DIR Changes to curl cmake files: - cmake_minimum_required - CMAKE_DEBUG_POSTFIX - Keep FILE protocol for HTTP_ONLY - find_package(OpenSSL) - optional_dependency(ZLIB) to reuse results from cmake/zlib.cmake - Do not add "/MP" for clang on Windows Change-Id: If0d63d027b2ea319767e207cd493d7dbba23cf0f (cherry picked from commit c6863849c0d9a8a9b6760bdeb1904da26a69a006)
Delete old source files. Change-Id: Ic7f63b43a8c2e744b3fc802ede76f947a0d6a53a (cherry picked from commit 17111ab258762c5ac206d00ed1f93b8a11b8c4c7)
…de() An SQL statement had heap-use-after-free when run as a prepared statement. The failing statemen was a functional index in a create table statement, and objects created ended up in the execution MEM_ROOT. The solution is to switch to the prepare MEM_ROOT when doing resolving of functional indexes. Change-Id: Ic54ef5eaaa4b90e1018cb82b9c73f5f6c5e29e01
…-grant-tables=1 Problem: When the server starts with skip-grant-tables and then privileges checking is re-enabled e.g. by FLUSH PRIVILEGES it is done in 2 steps: ACL loading and grant loading. The authorization cache was set to initialized after 1. step, so the authorizations done between these steps used not fully initialized cache. Fix: 1) Moved setting initialized flag after grant loading. 2) Added assert before hash table searches. 3) ALL_ACCESS returned in get_*_grant when the flag is not set. Test: Run an SQL command when FLUSH PRIVILEGES is stopped between the described steps. Change-Id: Idafa39296f4fd33aae1725f3f544c5af4724d23b
Additional patch, removing MINIMAL_RELWITHDEBINFO from the RPM spec. Change-Id: I40a76fcd1c3228ff41a68c6d4954ce6cd63319ad
causes service to shut down There are two problems here: With a production binary, we see failures in check_column_privileges() or failing to obtain a virtual function for an Item_func_eq object. With a debug binary, execution fails earlier in an assert: Query_expression::set_prepared(), Assertion `!is_prepared()' failed. Due to the assert failing in debug mode, it is difficult to get to the actual source of the problem, but we can narrow the debug issue down to establishing a proper execution context for an independent item to be evaluated in function sp_prepare_func_item(). The assumption for this function is that the current LEX and query expression found through thd->lex can be set to "prepared" and "executing" state. However, sp_prepare_func_item() can be used for two kinds of items. Some of these are standalone and have an associated LEX object and a query expression object. Examples of these are the procedure instructions sp_instr_set and sp_instr_jump_if_not. Others are expressions that are created below a LEX that is already in an executing state. Examples are items used to assign values to procedure arguments after procedure execution, and assignment to local variables after query execution. The latter is the problem in this case. The solution to the first problem is to distinguish the two cases and pass a bool argument for the standalone items and false for the remaining ones. In sp_prepare_func_item(), we then ensure that preparation and execution state is set properly for the standalone items, whereas we ensure by inspecting the current LEX object that we are already in an executing context for the dependent items. The second problem is related to the first one. It started appearing after the fix for bug#35856247 was pushed, which changes the way base_ref_items are saved in Query_block::save_properties(). But the solution also fixes this problem, since we no longer save properties for dependent items, which will interfere with the already saved properties for the current query block. We also had to set proper execution state in Sql_cmd_get_diagnostics::execute() and Sql_cmd_load_table::execute() because otherwise these functions would fail the newly added asserts. Change-Id: Ib5d31c07994340006818e180eed74e1d6aa23280
…load to HeatWave - rollback [3/3] Rollback of commit 708cd63af81dc35cfaba2417a8c3a186cadfa8bf due to HW regression. Change-Id: Ia401e25e0bd277a4e875ca35d0570ef8e376eee2
binlog_transaction_dependency [1/2] Changed the data structure from Tree map to ankerl::unordered_dense::map of variable m_writeset_history in sql/rpl_trx_tracking.h file which enhances performance. Changed CMakeFiles to support #include <ankerl/unordered_dense.h> Change-Id: I59835525559a41d47305463d07e903b88d6a8696
Change-Id: I3663f4e7cb9201ef767a3d8b680c0de92ef57470
Post-push fix: The shared library mysqlrouter_mysqlclient was installed in the wrong folder on windows. Change-Id: I446e73dc365d6690f84ae0205f08d4d46f7d70ce
Approved-by: Erlend Dahl <[email protected]> Change-Id: I8297ae3ef968412549bf7acd2e1e2aa7701cef91
Approved-by: Erlend Dahl <[email protected]> Change-Id: I276ded1dbb0647b1521b06c8ebde5278cb6e303d
Move the service API header file from the GPL sources to a cloud-only directory. Part 1: Delete header file from ./include/mysql/components Change-Id: I311a85e5b0808c70b2dbc0d37a5aae501de478ce
Used an alternative RapidJSON sequence that doesn't cause the macosx dyld to keep the shared obbject around post- dlclose(). Also stabilized the firewall.firewall-binlog-entries test by excluding the autoinc id that is non-consequential for the test Change-Id: Iaf1df12540f6a61a3827a6d293940bf93a068dd6
group_replication_group_name anymore when storing usage data Removed the reading of the group_replication_group_name system variable by the option_tracker component. Updated the doxygen about this. Removed setting the variable to some weird value during tests. Re-recorded the affected test result. Reading the row updates now ignores the cluster ID too. Change-Id: I76eba54d52931c3ba7b7a892bac7ab400eb1eb8d
|
aibek.bukabayev [email protected] seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account. You have signed the CLA already but the status is still pending? Let us recheck it. |
aybek
force-pushed
the
merge9.1
branch
2 times, most recently
from
6161279 to
e50d308
Compare
components/keyrings/common/config/config_reader.cc upstream: Migrate the keyring_aws plugin to a component 9dde815 9dde815 ps: Implementing KMIP keyring component, added some helper methods 5e1fd79 5e1fd79 components/keyrings/common/config/config_reader.h upstream: Migrate the keyring_aws plugin to a component 9dde815 9dde815 ps: Implementing KMIP keyring component, added some helper methods 5e1fd79 5e1fd79 components/keyrings/common/json_data/json_writer.cc upstream: fixed static analysis issues in keyring code 8a7dede components/keyrings/keyring_file/CMakeLists.txt components/keyrings/keyring_file/keyring_file.cc upstream: added Option and option usage tracker component 01e4c24 man/CMakeLists.txt upstream: Remove FILE globbing, use explicit list of man pages to install. 1c91597 storage/innobase/dict/dict0dd.cc upstream: added a new line below the code where we converged dfecee6 added new include file 28eb1ff Increase INSTANT ADD/DROP MAX_ROW_VERSION e64e217 storage/innobase/log/log0recv.cc upstream: solved the bug of parallel tablespace scan in 8.0 not working as expected 742fd15 storage/innobase/include/ut0log.h pxb: Use new logging mechanism in xtrabackup 78d62a7 upstream: Upgrade to clang-format 15 8742743 storage/innobase/os/os0file.cc pxb: introduced new method during keyring_file removal in order to fix the failing test case 6ed6532 upstream: added comment /* _WIN32 */ 5e4407d upstream: The `ulint purpose` is removed from signature of `os_file_create()`. The `OS_FILE_AIO` and `OS_FILE_NORMAL` are removed. The `ulint type` of `os_file_create()` method is renamed to `purpose` 3881f88 storage/innobase/fil/fil0fil.cc upstream: Remove extra support for FusionIO 688a682 Test fixes: storage/innobase/xtrabackup/test/t/xb_export_check_cfg.sh For EXPORT, the CFG version is bumped to 8 as previously 255 was being written to indicate an invalid version e64e217 storage/innobase/xtrabackup/test/t/xb_duplicate_sdi.sh storage/innobase/xtrabackup/test/t/duplicate_sdi_datadir.tar.gz storage/innobase/xtrabackup/test/t/dictionary_sdi.sh updated the test datadir `duplicate_sdi_datadir.tar.gz` with mysql 8.4 version due to 8.1 can not be upgraded to 9.1 storage/innobase/xtrabackup/test/run.sh added slash to the end of the basedir variable because upstream updated the keyring components loading path 0fe2378 storage/innobase/xtrabackup/test/t/simulate_different_version.sh storage/innobase/xtrabackup/src/backup_mysql.cc updated simulate higher version to 9.1 storage/innobase/xtrabackup/test/suites/keyring/innodb_encryption_mix_plugins.sh storage/innobase/xtrabackup/test/suites/keyring/pxb-1793.sh storage/innobase/xtrabackup/test/suites/keyring/shared_tablespace_encryption.sh use keyring file component instead of keyring file
…e() method on upstream
satya-bodapati
requested changes
Jan 23, 2025
storage/innobase/log/log0recv.cc
Outdated
|
|
||
| #ifndef UNIV_HOTBACKUP | ||
| } else { | ||
| recv_sys->missing_ids.insert(space_id); |
There was a problem hiding this comment.
missing_ids logic is moved else where or completely removed?
There was a problem hiding this comment.
it is completely removed
There was a problem hiding this comment.
Can you please share the commit? missing ids is about handling MLOG_FILE_DELETE entries. So if a tablespace is not found but later a MLOG_FILE_DELTE entry is there, it should n't complain. But if a tablespace is missing and no MLOG_FILE_DELETE entry exists, it should error out.
| @@ -12900,6 +12900,16 @@ ER_KEYRING_COMPONENT_AWS_READ_ONLY_LIMIT | |||
| ER_KEYRING_COMPONENT_AWS_CANNOT_ENCRYPT_READ_ONLY | |||
| eng "CMK was changed and the keyring must be re-encrypted, but the keyring is read only. Set the read_only config parameter to false for re-encryption and reload the component." | |||
|
|
|||
| ER_KEYRING_COMPONENT_NO_CONFIG | |||
There was a problem hiding this comment.
Please check the PS 9.1 code base and then adjust the xtrabackup error messages after that.
Comment on lines
421
to
425
| // we will use xtrabckup 9.1 for all 9.x versions | ||
| mysql9x = pxb_version == "9.1" && | ||
| (version_number > 90000 && version_number < 100000); | ||
|
|
||
| if (!mysql9x && pxb_version != server_version) { |
There was a problem hiding this comment.
I would say this appraoch is fine until the last innovation release until the 9.7 LTS.
So please adjust it to fail on 9.7 version
| @@ -2634,7 +2634,7 @@ static bool xtrabackup_read_info(char *filename) { | |||
| xb_server_version = | |||
| xtrabackup::utils::get_version_number(mysql_server_version_str); | |||
|
|
|||
| // ut_ad(xb_server_version > 80000 && xb_server_version < 90000); | |||
| ut_ad(xb_server_version > 80000 && xb_server_version < 91000); | |||
There was a problem hiding this comment.
how will this pass if server is 9.2 or 9.3? 🤔 Please ensure it is OK until 9.7. 9.7 should fail.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.