Merge lp:~tsarev/percona-server/innodb_deadlock_788569 into lp:percona-server/5.1

[Oleg Tsarev]

Move tests from mysql-test/ to mysql-test/innodb_deadlock_count.patch according to bug #788569
Now test explicit return "ERROR 40001: Deadlock found when trying to get lock; try restarting transaction"

[Valentine Gostev (longbow)]

good to go

review: Approve (qa)

[Alexey Kopytov (akopytov)]

Please revert this merge. It has certain issues that must be addressed.

review: Needs Fixing

[Alexey Kopytov (akopytov)]

The only reason why the current test does not show the "deadlock found" error is because you use "send" without "reap", thus preventing mysqltest from processing the error.

So here's the patch against the .test file that achieves *absolutely* the same as this MP. Everything else in the MP is just irrelevant noise.

I would also remove disable_abort_on_error. A more clean way would be to define the expected errors explicitly, rather than disabling them all, that is adding "--error 0, ER_LOCK_DEADLOCK" before calling "reap".

@@ -38,7 +38,12 @@
 connection con2;
 SEND SELECT b FROM t WHERE a=1 FOR UPDATE;

-SLEEP 0.2;
+connection con1;
+reap;
+
+connection con2;
+reap;
+
 --echo # Switch to connection con3
 connection con3;
 let $result = `SELECT VARIABLE_VALUE FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_deadlocks'`;

review: Needs Fixing

[Alexey Kopytov (akopytov)]

review: Approve

=== added directory 'mysql-test/innodb_deadlock_count.patch'
=== renamed file 'mysql-test/percona_innodb_deadlock_count.result' => 'mysql-test/innodb_deadlock_count.patch/percona_innodb_deadlock_count.result'
--- mysql-test/percona_innodb_deadlock_count.result	2010-06-07 10:32:56 +0000
+++ mysql-test/innodb_deadlock_count.patch/percona_innodb_deadlock_count.result	2011-06-21 16:40:18 +0000
@@ -22,6 +22,11 @@
 SELECT b FROM t WHERE a=2 FOR UPDATE;
 # Switch to connection con2
 SELECT b FROM t WHERE a=1 FOR UPDATE;
+# Switch to connection con1
+b
+1
+# Switch to connection con2
+ERROR 40001: Deadlock found when trying to get lock; try restarting transaction
 # Switch to connection con3
 1
 # Drop test table

=== renamed file 'mysql-test/percona_innodb_deadlock_count.test' => 'mysql-test/innodb_deadlock_count.patch/percona_innodb_deadlock_count.test'
--- mysql-test/percona_innodb_deadlock_count.test	2010-06-07 10:32:56 +0000
+++ mysql-test/innodb_deadlock_count.patch/percona_innodb_deadlock_count.test	2011-06-21 16:40:18 +0000
@@ -38,7 +38,14 @@
 connection con2;
 SEND SELECT b FROM t WHERE a=1 FOR UPDATE;
 
-SLEEP 0.2;
+--echo # Switch to connection con1
+connection con1;
+reap;
+
+--echo # Switch to connection con2
+connection con2;
+reap;
+
 --echo # Switch to connection con3
 connection con3;
 let $result = `SELECT VARIABLE_VALUE FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_deadlocks'`;

=== modified file 'series'
--- series	2011-04-20 07:13:25 +0000
+++ series	2011-06-21 16:40:18 +0000
@@ -55,4 +55,3 @@
 mysql_remove_eol_carret.patch
 innodb_expand_fast_index_creation.patch
 innodb_bug60788.patch
-