Details
-
Bug
-
Status: Closed (View Workflow)
-
Blocker
-
Resolution: Fixed
-
N/A
Description
InnoDB changes between mysql-8.0.41 and mysql-8.0.42
Not applicable to MariaDB Server
Bug#37403052 Forgeting to call btr_pcur_t::close() leads to memory leak
This fixes a memory leak that does not exist in MariaDB, and generally seems to pessimize the code by inserting redundant calls to free a possibly allocated buffer. As it has been explained in MDEV-28465, such calls should actually be unnecessary except when there has been a prior call to btr_pcur_store_position(). Based on the commit message, the memory leak must occur due to apply_update_multi_value(), which is calling row_search_index_entry(), which in turn must be storing the persistent cursor position and allocating the buffer. This leak should have been introduced in MySQL 8.0.18 with the implementation of multi-valued indexes (WL#8955, WL#8763).
Bug#37426282 innodb.ibuf_sys_tablespace_extend test failing sporadically
This is a change to a debug instrumented test that was occasionally failing. This test had been added as part of an earlier fix Bug#36511673 MLOG_FILE_EXTEND log record is not redo logged for System tablespace. That bug had been fixed almost 6 years earlier in MDEV-11556 without changing any file format.
BUG#37354028 Update copyright headers to 2025
We only update Oracle copyright years when including actual code changes.
Bug#37505746 Bugfix of Bug#37318367 in 8.0.41 contains an obvious logical issue
According to MDEV-35896, this is related to a parallel index scanner, which does not exist in MariaDB Server.
Bug #34454572 Clone_persist_gtid causes memory leak
There is no Clone plugin API (MDEV-21105) in MariaDB Server yet.
Bug#37391519 Page xxx still fixed or dirty
This fixes a shutdown bug due to incorrect error handling in parallel index creation. MDEV-16281 has not been implemented yet in MariaDB Server.
Bug#37168132 Unexpected creation of an large index, causing an inaccessible table
MariaDB Server would already report the error ER_INDEX_COLUMN_TOO_LONG) in all cases that the added test case is covering. (Note: Either set default_storage_engine=innodb or explicit ENGINE=InnoDB in CREATE TABLE statements will be needed in MariaDB.)
Bug #37022679 Too Much Disk Read on Startup, ... (fixup)
This is attempting to fix yet another regression, which was caused by reverting WL#7412 in MySQL 8.0. MariaDB will even avoid unnecessarily opening .ibd files on startup; see MDEV-32027.
Bug#37517125 Simplify the APIs to save/restore the level-one progress
This is related to the parallel index scanner, which does not exist in MariaDB.
Bug#35115629 InnoDB - Add more logging during srv_shutdown_log
This is attempting to diagnose an anomaly with the buffer pool during a shutdown. We are not currently aware of any shutdown hangs in MariaDB Server. In MDEV-36226, some diagnostics was added to report buffer pool hangs. It is possible that their hang is something related to MDEV-26733, which was likely fixed by MDEV-24142 in MariaDB Server 10.6.
Bug#37510755 Performance regression in join for tables with many columns (reverted)
This is specific to the MySQL implementation of instant ADD COLUMN (MDEV-11369) and DROP COLUMN (MDEV-15562). In our implementation, the PRIMARY KEY of a clustered index record is always accessed in the same way, and accessing instantly added, dropped or reordered columns involves reading an optional record header.
Bug#37617773 Cross DDL operations on MySQL partitioned tables, parallel COUNT(*) does not report errors
This is fixing a regression in parallel index scan, which is not implemented in MariaDB Server.
Bug#37061960 : InnoDB redo recovery failure/corruption due to INSTANT ddl.+ innodb col limits
This is fixing a crash recovery bug related to the MySQL implementation of instant ADD COLUMN or DROP COLUMN. In MariaDB, no special redo log record format was introduced in MDEV-11369 or MDEV-15562. Our ib_logfile0 record format in MDEV-12353 and MDEV-21724 is agnostic to this.
Bug#37292404 Update instant dropped extern blob
Apparently, the instant DROP COLUMN in MySQL is violating ACID and allowing an UPDATE statement to run concurrently. There is no test case included, even though this is an external contribution. In MariaDB Server, any ALTER TABLE operation will be covered by exclusive meta-data lock (MDL) for at least part of the execution. Specifically, ALTER TABLE...ALGORITHM=INSTANT is covered entirely by a single exclusive MDL. Starting with MDEV-23805 and MDEV-27962, this applies to any ALTER TABLE of empty tables unless ALGORITHM=COPY is being used.
Bug#37212019 behavior related to innodb_spin_wait_delay changed in 8.0.30
This is revising an earlier change to spin loops in MySQL 8.0, something that was removed from MariaDB in MDEV-14482 already. A random exponential backoff is a reasonable choice for distributed systems, but less so so on SMP or NUMA, because maintaining the global state of a random number generator for it would form a performance bottleneck.
BUG#32288105: MYSQL CRASHES IMMEDIATELY AFTER SELECTING FROM INFORMATION_SCHEMA (fixup)
This is related to the MySQL 8.0 Global Data Dictionary and lower_case_table_names=2.
Applicable to MariaDB Server
Bug #37359538 Extend check table to verify MBR in spatial index to original geometry
MDEV-36612 was filed for this.
Bug #36993445 virtual index unstable after rollback when index_id is greater than max uint32 (fixup)
MDEV-36613 was filed for this.
Attachments
Issue Links
- is blocked by
-
MDEV-35896 Merge applicable changes from InnoDB 8.0.41
-
- Closed
-
-
MDEV-36613 Incorrect undo logging for indexes on virtual columns whose index ID does not fit in 32 bits
-
- Closed
-
- relates to
-
MDEV-36612 Implement CHECK TABLE…EXTENDED for SPATIAL INDEX
-
- Open
-