lp:~gl-az/percona-server/5.5-bug1053342

Branch merges

Propose for merging

Merged into lp:percona-server/5.5 at revision 316

Laurynas Biveinis (community): Approve on 2012-10-11

Diff: 11 lines (+1/-1)

1 file modified

Percona-Server/sql/log_event.h (+1/-1)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Bug #1053342: Slave server crash after an update statement

High

Fix Released

Link a bug report

Related blueprints

314. By George Ormond Lorch III on 2012-10-08

Fixed upstream slave crash bug http://bugs.mysql.com/bug.php?id=66890.

Issue was due to the thd->mem_root heap being released prematurely during the
processing of deferred events.

In sql/log_event.cc:6060 User_var_log_event::do_apply_event(), a test is made
to determine if deferred events are being collected and processed. If true, a
call to this->set_deferred is made to set a flag indicating that 'this' is a
deferred event. This means that the thd->mem_root should not be freed on the
completion of User_var_log_event::do_apply_event().

The problem lies in the way User_var_log_event::set_deferred() is implemented.
The class User_var_log_event has an internal member 'bool deferred' that
should be set to true when User_var_log_event::set_deferred() is called. The
original code was setting 'deferred' to the state of another member
'char* val'. At the time User_var_log_event::set_deferred() is being called,
the 'val' member is NULL, therefore leaving the 'deferred' member at 0/false
instead of the true that was intended.

Changing User_var_log_event::set_deferred() in sql/log_event.h:2596 to be
'{ deferred= true; }' instead of '{ deferred= val; }' fixes the issue. This
is easily seen by running the slave with Valgrind before and after.
Valgrind reports 15434 errors of both 'Invalid read' and 'Invalid write'
before the fix and 0 errors and 0 leaks after the fix.

313. By Hrvoje Matijakovic on 2012-10-05

* bug fix for Bug #1054090

312. By Stewart Smith on 2012-10-03

empty merge for UNIV_DEBUG fix from 5.1

311. By Stewart Smith on 2012-10-01

merge: Null-merge already-present bug 978036 fix from 5.1

310. By Stewart Smith on 2012-10-01

merge: Null-merge bug 1046389 fix from 5.1.

309. By Stewart Smith on 2012-10-01

merge fix for Bug #1018264: Multiple hash index partitions causes overly large hash index

308. By Stewart Smith on 2012-10-01

Merge fixes for:
- Bug #686534: innodb_lru_dump_restore.patch: buf_LRU_file_dump() holds LRU_list_mutex while doing file I/O
- Bug #881001: Make innodb_blocking_lru_restore / innodb_blocking_buffer_pool_restore accept arguments

307. By Stewart Smith on 2012-10-01

merge: null merge 5.1 -> 5.5 for lp:1050387

306. By Stewart Smith on 2012-09-28

merge max-binlog-files

305. By Stewart Smith on 2012-09-28

merge fix for Bug #1013644: innodb-sys-stats-root-page option conflicts with innodb-sys-stats