lp:~tsarev/percona-server/5.5.12-786645-slow_extended

Branch merges

Propose for merging

No branches dependent on this one.

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Bug #786645: 5.5.12 valgrind warnings

Undecided

Fix Released

Link a bug report

Related blueprints

117. By Oleg Tsarev on 2011-06-08

The reason of bug #786645: patch slow_extended.patch adds option "log_slow_verbosity" with value "profiling_use_getrusage".
This option enable/disable call "getrusage", but doesn't check status when uses variable changed by "getrusage".
As result valgrind show warning "potential use of uninitialized value".
Now I check status before access to changed variables.
This is actual for tests:
 * main.profiling
 * main.show_profile

116. By Oleg Tsarev on 2011-06-07

Sometime variable "write_to_slow_log" (used in slow_extended.patch for log_slow_rate_limit) used without initialization in "prepare_new_connection", because "prepare_new_connection" call before initialization of "write_to_slow_log".
I now first call initialization and then "prepare_new_connection".
This is actual for tests:
 * main.init_connect

115. By Oleg Tsarev on 2011-06-07

Reason of bug #786645: when we run test under valgrind test work long time, and slow query log has extra query "call mtr.check_testcase();".
I disable slow query log on startup, and enable only when I need.
This is actual for tests:
 * percona_slow_extended-log_slow_filter
 * percona_slow_extended-long_query_time
 * percona_slow_extended-microseconds_in_slow_extended
 * percona_slow_extended-min_examined_row_limit
 * percona_slow_extended-use_global_long_query_time

114. By Laurynas Biveinis on 2011-06-06

Fix LP bug 791849: valgrind reported memory leaks in XtraDB
initialization.

The patch innodb_adaptive_hash_index_partitions.patch introduces
btr_search_latch_part and btr_search_sys->hash_index variables, which
are allocated in btr_search_sys_create, however corresponding
deallocation in btr_search_sys_free was incomplete. The fix is
straightforward.

113. By Alexey Kopytov on 2011-06-05

Bug #770296.

Fixed the test so that:

1. long_query_time is set 0 to capture all queries in the slow query
log to make the test more deterministic.
2. We can now get rid of using SLEEP() due to 1.
3. "# Time:" lines are skipped from the slow query log, because they are
added to the query log if the current time is different from the last
logged query time, which of course makes the test non-deterministic.

112. By Ignacio Nin on 2011-05-24

Adapt test to our directory names (debian, redhat)

This test expected documentation to be placed in /usr/share/doc/MySQL*.
Change the path in order to follow debian and RedHat's conventions.

111. By Ignacio Nin on 2011-05-18

Merge release branch, update version

Create release-5.5.12-20.3 with 5.5.12 changes and release branch from
5.5.11. Update versions to 5.5.12 and 20.3.

110. By Oleg Tsarev on 2011-05-11

port all non-Yasufumi patches

109. By kinoyasu <kinoyasu@gauntlet4> on 2011-05-10

port Yasufumi patches to 5.5.12

108. By Alexey Kopytov on 2011-04-20

LP bug #764395 / MySQL bug #60788: InnoDB crashes with an assertion
                                   failure when receiving a signal on
                                   pwrite()

The problem was that InnoDB I/O code was not fully conforming to the
standard on POSIX systems. fsync(), pread() and pwrite() calls may be
interrupted by a signal. In such a case, InnoDB would just fail with an
assertion failure rather than just restarting the interrupted call.

Fixed InnoDB I/O code so that the interrupted system calls are restarted
if they are interrupted by a signal.