Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-25466

Merge new release of InnoDB 5.7.34 to 10.2

    XMLWordPrintable

    Details

      Description

      MySQL 5.7.33 included only two InnoDB changes that were not relevant to MariaDB, and no MDEV was created for applying those.

      MySQL 5.7.34 includes a few more changes:

      Not applicable to MariaDB Server

      Bug #31529221 ALTER TABLE FAILS AND REPORTS INCORRECT KEY FILE FOR TABLE T1
      This is specific to the MySQL reinvented counterpart of MariaDB 10.1 encryption or page_compressed.
      Bug #32124113 MYSQL CRASH FOR GENERATED COLUMN DELETE WITH FOREIGN KEY
      This duplicates MDEV-18366. The test case does not crash in MariaDB, with or without ASAN.
      Bug #31899685 INNODB HANDLER SHARE'S INDEX_MAPPING ACCESS INVALID MEMORY OF DICT INDEX
      The buggy code was removed in MDEV-16557 Remove INNOBASE_SHARE::idx_trans_tbl already in 10.2.18.
      BUG#32368342: UPDATE COPYRIGHT HEADERS TO 2021
      We only update Oracle copyright years when applying their code changes.
      Bug #32354435 BACKPORT TO 5.7: BUG #32216281 REMOVE DBUG_OFF AND DBUG_ASSERT (2, 3)
      Using NDEBUG and assert() could be better than reinventing our own wheels, but we are not going to do such a massive change in a minor release.
      Bug#32225367: RW_LOCK_STATS OPERATION HAS REGRESSION FROM 5.6, LOWER CACHE EFFICIENCY
      This appears to be a performance improvement. The rw_lock_t implementation in MariaDB 10.1 and 10.2 is slightly different from MySQL 5.6 and 5.7, and we would need extensive benchmarking to evaluate whether this would improve things. Recent benchmarking for MDEV-25404 shows that the latch rewrite in the upcoming MariaDB Server 10.6.0 release is improving performance over 10.5. We will reduce risk by not considering this change.

      Applicable to MariaDB Server

      Bug #32032897 DEADLOCKING WAIT GRAPH ON BUSY SERVER
      This only affects MariaDB 10.2 when using the non-default setting innodb_safe_truncate=OFF. It is a bug in the MySQL 5.7 specific reimplementation of TRUNCATE TABLE (WL#6501).

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              marko Marko Mäkelä
              Reporter:
              marko Marko Mäkelä
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved: