Percona Server moved to https://jira.percona.com/projects/PS

Merge lp://qastaging/~laurynas-biveinis/percona-server/bug1390695 into lp://qastaging/percona-server/5.6

bug1390695
Merge into 5.6

Proposed by Laurynas Biveinis on 2014-11-14

Status:

Merged

Approved by:

Laurynas Biveinis on 2014-11-14

Approved revision:

no longer in the source branch.

Merged at revision:

694

Proposed branch:

lp://qastaging/~laurynas-biveinis/percona-server/bug1390695

Merge into:

lp://qastaging/percona-server/5.6

Diff against target:

36 lines (+3/-2)

3 files modified

storage/innobase/include/lock0lock.h (+1/-1)
storage/innobase/include/trx0rec.h (+1/-1)
storage/innobase/lock/lock0lock.cc (+1/-0)

To merge this branch:

bzr merge lp://qastaging/~laurynas-biveinis/percona-server/bug1390695

Critical

Fix Released

Link a bug report

Reviewer	Review Type	Date Requested	Status
Laurynas Biveinis (community)			Approve on 2014-11-14
Review via email: mp+241774@code.qastaging.launchpad.net

Description of the change

Fix bug 1390695 (percona 5.6.21-70.0 crash on data import) by fixing
the upstream bug http://bugs.mysql.com/bug.php?id=74842 (Incorrect
attribute((nonnull)) for btr_cur_ins_lock_and_undo callees)

The crash in the Percona Server bug is happening because a thr != NULL
check after the btr_cur_ins_lock_and_undo call is getting optimized
away, while NULL is a legal value for thr there. The cause for this is
the upstream bug: lock_rec_insert_check_and_lock is incorectly
annotated with attribute((nonnull)), applying to its thr arg. Thus,
after returning from the call, the compiler assumes that thr must be
!= NULL.

Fix by correcting the attribute((nonnull)) annotations for
lock_rec_insert_check_and_lock and trx_undo_report_row_operation not
to say that thr must be nonnull.

http://jenkins.percona.com/job/percona-server-5.6-param/755/