From: Christopher Powers
Date: May 22 2012 5:48pm
Subject: bzr push into mysql-trunk branch (chris.powers:3899 to 3900) Bug#13915235
List-Archive: http://lists.mysql.com/commits/144163
X-Bug: 13915235
Message-Id: <201205221752.q4MHqCvr029561@acsmt358.oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit

 3900 Christopher Powers	2012-05-22
      Bug#13915235 - THREAD_POOL.THREAD_POOL_KILL_INNODB.TEST FAILS - ASSERT IN PFS.CC ON TRUNK
      
      The thread pool processes KILL commands by circumventing the typical thread_attach/detach
      sequence, causing a discrepancy in the thread contexts that the Performance Schema uses to
      track table locking.
      
      The more robust solution is to remove the assertion in the Performance Schema rather than
      modify the thread pool code,

    modified:
      storage/perfschema/pfs.cc
 3899 Christopher Powers	2012-05-22 [merge]
      local merge from mysql-trunk

    added:
      sql/abstract_query_plan.cc
      sql/abstract_query_plan.h
    modified:
      mysql-test/collections/default.experimental
      sql/CMakeLists.txt
      sql/ha_ndbcluster.cc
      sql/ha_ndbcluster.h
      sql/handler.cc
      sql/handler.h
      sql/opt_explain.cc
      sql/opt_explain_format.h
      sql/opt_explain_json.cc
      sql/opt_explain_traditional.cc
      sql/sql_executor.cc
      sql/sql_optimizer.cc
      sql/sql_select.cc
      storage/innobase/include/sync0rw.h
      storage/innobase/include/sync0rw.ic
=== modified file 'storage/perfschema/pfs.cc'
--- a/storage/perfschema/pfs.cc	2012-05-15 09:39:47 +0000
+++ b/storage/perfschema/pfs.cc	2012-05-22 17:44:21 +0000
@@ -2612,9 +2612,6 @@ start_table_lock_wait_v1(PSI_table_locke
 
   PFS_thread *pfs_thread= pfs_table->m_thread_owner;
 
-  DBUG_ASSERT(pfs_thread ==
-              my_pthread_getspecific_ptr(PFS_thread*, THR_PFS));
-
   PFS_TL_LOCK_TYPE lock_type;
 
   switch (op)
@@ -3016,14 +3013,6 @@ start_socket_wait_v1(PSI_socket_locker_s
     if (unlikely(pfs_thread == NULL))
       return NULL;
 
-#ifdef LATER
-    /*
-      Needs refinement, because of KILL.
-    */
-    DBUG_ASSERT(pfs_thread ==
-                my_pthread_getspecific_ptr(PFS_thread*, THR_PFS));
-#endif
-
     if (!pfs_thread->m_enabled)
       return NULL;
 

No bundle (reason: useless for push emails).