MySQL 5.7.12 was released more than 2 weeks ago. New features introduced there in a form of "rapid plugins" are widely discussed, but I am more interested in bugs reported by MySQL Community users that are fixed there. Unfortunately I do not see MySQL Community Release Notes by Morgan (like this ) for quite a some time, so I have to continue describing key bug fixes and name people who reported and verified bugs in my "Fun with Bugs" series.
As usual, let’s start with InnoDB bugs fixed:
- Bug #80070 – "allocated_size and file_size differ if create general tablespace outside datadir". It was reported by my former colleague from Percona Shahriyar Rzayev and verified by Bogdan Kecman . Nice to see more people from Oracle involved in processing community bug reports!
- Bug #79185 – "Innodb freeze running REPLACE statements". This bug (that affected many users, also on versions 5.5.x and 5.6.x, and was a kind of a regression) was reported by Will Bryant and verified (probably) and fixed by Shaohua Wang . The fix is also included into versions 5.5.49 and 5.6.30.
- Bug #73816 – ”MySQL instance stalling “doing SYNC index”". It was reported by Denis Jedig and a lot of additional evidence was provided by my former colleague Aurimas Mikalauskas . This bug was fixed (and probably verified) by Shaohua Wang .
- Bug #79200 – "InnoDB: "data directory" option of create table fails with pwrite() OS error 22", is a widely noted regression (I’ve seen customer issue with a potentially related MariaDB problem this week). This bug was reported by Frank Ullrich and verified by Bogdan Kecman . It is also fixed in MySQL 5.6.30.
- Bug #79725 – "Check algorithm=innodb on crc32 checksum mismatch before crc32(big endian)". This bug was created to track the patch contributed by Daniel Black at GitHub. It was verified by Umesh .
Next, let’s review replication bugs fixed in 5.7.12:
- Bug #79504 – "STOP SLAVE IO THREAD prints wrong LOST CONNECTION message in error log file". It was reported by Venkatesh Duggirala .
- Bug #78722 – "Relay log info currently_executing_gtid is not properly initialized or protected". This bug was reported by Pedro Gomes . It contains a nice simple test case and fix suggested.
- Bug #78445 is private. So, I can only quote the release notes:
RESET SLAVE ALLcould delete a channel even when
wait_until_sql_thread_after_gtidwere still waiting for binlog to be applied. This could cause a MySQL server exit when the functions tried to access the channel that was deleted. Now, a channel reference counter was added that is increased if the channel should not be deleted when the functions are running.
RESET SLAVE ALLwill wait for no reference, and then it will delete the channel."
I am not sure this crash is a "security" bug of any kind, but what do I know…
- Bug #78352 – "Slow startup of 5.7.x slave with relay_log_recovery = ON and many relay logs". I reported it based on regression comparing to 5.6.x reported by a customer of Percona, and verified by Umesh . Nice to see it fixed, as it was really annoying for almost anyone who upgraded production replication setup to 5.7.
- Bug #78133 – "Slave_worker::write_info() incorrect DBUG_ENTER (contribution)". This bug was created to track the patch contributed by Stewart Smith at GitHub. It was verified by Umesh .
- Bug #77740 – "silent failure to start if mysql.gtids_executed gets HA_ERR_LOCK_WAIT_TIMEOUT ". It was reported and verified by Shane Bester .
- Bug #77237 – "Multi-threaded slave log spamming on failure". This bug was reported by Davi Arnaut and verified by Umesh . Fix is also included in MySQL 5.6.30.
- Bug #78963 – "super_read_only aborts STOP SLAVE if relay_log_info_repository=TABLE, dbg crash". It was reported by my former colleague in Percona Laurynas Biveinis and verified by Umesh . Check also related Bug #79328 – "super_read_only broken as a server option".
- Bug #77684 – "DROP TABLE IF EXISTS may brake replication if slave has replication filters". This bug was reported by my former colleague in Percona Fernando Laudares Camargos for MySQL 5.6.x and verified by Umesh . MySQL 5.6.30 also got this fixed.
We all remember that Performance Schema is perfect and the next greatest thing after sliced bread, but sometimes bugs are noted even there. Check Bug #79934 – "i_perfschema.table_leak random result failure" reported and verified by Magnus Blåudd . Another example is Bug #79784 – "update setup_instruments do not affect the global mutex/rwlock/cond" reported by Zhang Yingqiang and verified by Umesh . The later, IMHO, is related to or a super set of my good old report, Bug #68097 – "Manual does not explain that some P_S instruments must be enabled at startup" that remains open as a feature request (after some changes in the manual) for more than 3 years already. I truly hope 5.7.12 fixed this for a general case – it’s truly important to be able to enable instruments dynamically if we expect Performance Schema tobe used as a main tool for troubleshooting.
I’d also want to highlight a couple of fixes related to optimizer:
- Bug #77209 – "Update may use index merge without any reason (increasing chances for deadlock)". It was reported and verified by my former colleagues from Oracle, Andrii Nikitin . MySQL 5.6.30 also includes the fix.
- Bug #72858 – "EXPLAIN .. SELECT .. FOR UPDATE takes locks". This bug was reported by my former colleague in Percona (and, I hope, my colleague again soon) Justin Swanhart , who has a birthday today. Happy Birthday to you, Justin! The bug was verified by Umesh and is also fixed in MySQL 5.6.30. Justin had reported another bug fixed in 5.7.12, Bug #69375 – "LOAD DATA INFILE claims to be holding ‘System Lock’ in processlist".
Several more bugs reported by community were also fixed, but they were in the areas (or for platforms) I am not particularly interested in.
To summarize, MySQL 5.7.12 contains important bug fixes in replication and InnoDB and it makes sense to consider upgrade even if you do not care about any "rapid plugins", X protocol, encryption of data at rest, MySQL Keyring and other "cool" new shiny features.