Merge ngrishakin@bk-internal.mysql.com:/home/bk/mysql-5.1

into  mysql.com:/home/ndbdev/ngrishakin/mysql-5.1

mysql-test/t/disabled.def

@@ -18,10 +18,7 @@
 #im_life_cycle             : Bug#20368 2006-06-10 alik    im_life_cycle test fails
 ndb_autodiscover         : BUG#18952 2006-02-16 jmiller Needs to be fixed w.r.t binlog
 ndb_autodiscover2        : BUG#18952 2006-02-16 jmiller Needs to be fixed w.r.t binlog
-ndb_autodiscover3        : BUD#20843 2006-07-04 tomas ndb_autodiscover3 fails randomly
-#ndb_binlog_discover      : BUG#19395 2006-04-28 tomas/knielsen mysqld does not always detect cluster shutdown
-#ndb_cache2               : BUG#18597 2006-03-28 brian simultaneous drop table and ndb statistics update triggers node failure
-#ndb_cache_multi2         : BUG#18597 2006-04-10 kent  simultaneous drop table and ndb statistics update triggers node failure
+ndb_autodiscover3        : BUG#20872 2006-07-05 tomas ndb_autodiscover3 fails randomly
 ndb_load                 : BUG#17233 2006-05-04 tomas failed load data from infile causes mysqld dbug_assert, binlog not flushed
 partition_03ndb          : BUG#16385 2006-03-24 mikael Partitions: crash when updating a range partitioned NDB table
 ps_7ndb                  : BUG#18950 2006-02-16 jmiller create table like does not obtain LOCK_open

scripts/make_binary_distribution.sh

@@ -181,11 +181,21 @@ if [ $BASE_SYSTEM = "netware" ] ; then
 fi
 
 copyfileto $BASE/lib \
-  libmysql/.libs/libmysqlclient.a libmysql/.libs/libmysqlclient.so* \
-  libmysql/libmysqlclient.* libmysql_r/.libs/libmysqlclient_r.a \
-  libmysql_r/.libs/libmysqlclient_r.so* libmysql_r/libmysqlclient_r.* \
+  libmysql/.libs/libmysqlclient.a \
+  libmysql/.libs/libmysqlclient.so* \
+  libmysql/.libs/libmysqlclient.sl* \
+  libmysql/.libs/libmysqlclient*.dylib \
+  libmysql/libmysqlclient.* \
+  libmysql_r/.libs/libmysqlclient_r.a \
+  libmysql_r/.libs/libmysqlclient_r.so* \
+  libmysql_r/.libs/libmysqlclient_r.sl* \
+  libmysql_r/.libs/libmysqlclient_r*.dylib \
+  libmysql_r/libmysqlclient_r.* \
+  libmysqld/.libs/libmysqld.a \
+  libmysqld/.libs/libmysqld.so* \
+  libmysqld/.libs/libmysqld.sl* \
+  libmysqld/.libs/libmysqld*.dylib \
   mysys/libmysys.a strings/libmystrings.a dbug/libdbug.a \
-  libmysqld/.libs/libmysqld.a libmysqld/.libs/libmysqld.so* \
   libmysqld/libmysqld.a netware/libmysql.imp \
   zlib/.libs/libz.a
 

sql/ha_ndbcluster_binlog.cc

@@ -3442,8 +3442,8 @@ restart:
     // wait for the first event
     thd->proc_info= "Waiting for first event from ndbcluster";
     DBUG_PRINT("info", ("Waiting for the first event"));
-    int schema_res= 0;
-    Uint64 schema_gci= 0;
+    int schema_res= 0, res= 0;
+    Uint64 schema_gci= 0, gci= 0;
     while (schema_res == 0 && !abort_loop)
     {
       schema_res= s_ndb->pollEvents(100, &schema_gci);
@@ -3452,7 +3452,14 @@ restart:
     DBUG_PRINT("info", ("schema_res: %d  schema_gci: %d", schema_res, schema_gci));
     if (schema_res > 0)
     {
-      i_ndb->pollEvents(0);
+      while (res >= 0 && gci < schema_gci && !abort_loop)
+      {
+        res= i_ndb->pollEvents(100, &gci);
+      }
+      if (gci > schema_gci)
+      {
+        schema_gci= gci;
+      }
       i_ndb->flushIncompleteEvents(schema_gci);
       s_ndb->flushIncompleteEvents(schema_gci);
       if (schema_gci < ndb_latest_handled_binlog_epoch)
@@ -3466,6 +3473,17 @@ restart:
         ndb_latest_applied_binlog_epoch= 0;
         ndb_latest_received_binlog_epoch= 0;
       }
+      else if (ndb_latest_applied_binlog_epoch > 0)
+      {
+        sql_print_warning("NDB Binlog: cluster has reconnected. "
+                          "Changes to the database that occured while "
+                          "disconnected will not be in the binlog");
+      }
+      if (ndb_extra_logging)
+      {
+        sql_print_information("NDB Binlog: starting log at epoch %u",
+                              (unsigned)schema_gci);
+      }
     }
   }
   {

storage/ndb/include/mgmapi/ndbd_exit_codes.h

@@ -71,6 +71,7 @@ typedef ndbd_exit_classification_enum ndbd_exit_classification;
 #define NDBD_EXIT_INDEX_NOTINRANGE            2304
 #define NDBD_EXIT_ARBIT_SHUTDOWN              2305
 #define NDBD_EXIT_POINTER_NOTINRANGE          2306
+#define NDBD_EXIT_PARTITIONED_SHUTDOWN        2307
 #define NDBD_EXIT_SR_OTHERNODEFAILED          2308
 #define NDBD_EXIT_NODE_NOT_DEAD               2309
 #define NDBD_EXIT_SR_REDOLOG                  2310

storage/ndb/src/kernel/blocks/qmgr/QmgrMain.cpp

@@ -438,6 +438,7 @@ void Qmgr::execCONNECT_REP(Signal* signal)
 void
 Qmgr::execREAD_NODESCONF(Signal* signal)
 {
+  jamEntry();
   check_readnodes_reply(signal, 
 			refToNode(signal->getSendersBlockRef()),
 			GSN_READ_NODESCONF);
@@ -446,6 +447,7 @@ Qmgr::execREAD_NODESCONF(Signal* signal)
 void
 Qmgr::execREAD_NODESREF(Signal* signal)
 {
+  jamEntry();
   check_readnodes_reply(signal, 
 			refToNode(signal->getSendersBlockRef()),
 			GSN_READ_NODESREF);
@@ -907,9 +909,9 @@ retry:
 
   char buf[255];
   BaseString::snprintf(buf, sizeof(buf),
-		       "Partitioned cluster! check StartPartialTimeout, "
-		       " node %d thinks %d is president, "
-		       " I think president is: %d",
+		       "check StartPartialTimeout, "
+		       "node %d thinks %d is president, "
+		       "I think president is: %d",
 		       nodeId, president, cpresident);
 
   ndbout_c(buf);
@@ -941,7 +943,7 @@ retry:
   CRASH_INSERTION(932);
   
   progError(__LINE__, 
-	    NDBD_EXIT_ARBIT_SHUTDOWN,
+	    NDBD_EXIT_PARTITIONED_SHUTDOWN,
 	    buf);
   
   ndbrequire(false);
@@ -2798,7 +2800,7 @@ void Qmgr::failReportLab(Signal* signal, Uint16 aFailedNode,
       break;
     case FailRep::ZPARTITIONED_CLUSTER:
     {
-      code = NDBD_EXIT_ARBIT_SHUTDOWN;
+      code = NDBD_EXIT_PARTITIONED_SHUTDOWN;
       char buf1[100], buf2[100];
       c_clusterNodes.getText(buf1);
       if (signal->getLength()== FailRep::SignalLength + FailRep::ExtraLength &&
@@ -2809,16 +2811,14 @@ void Qmgr::failReportLab(Signal* signal, Uint16 aFailedNode,
 	part.assign(NdbNodeBitmask::Size, rep->partition);
 	part.getText(buf2);
 	BaseString::snprintf(extra, sizeof(extra),
-			     "Partitioned cluster!"
-			     " Our cluster: %s other cluster: %s",
+			     "Our cluster: %s other cluster: %s",
 			     buf1, buf2);
       }
       else
       {
 	jam();
 	BaseString::snprintf(extra, sizeof(extra),
-			     "Partitioned cluster!"
-			     " Our cluster: %s ", buf1);
+			     "Our cluster: %s", buf1);
       }
       msg = extra;
       break;

storage/ndb/src/kernel/error/ndbd_exit_codes.c

@@ -54,6 +54,8 @@ static const ErrStruct errArray[] =
    {NDBD_EXIT_ARBIT_SHUTDOWN, XAE, "Node lost connection to other nodes and "
     "can not form a unpartitioned cluster, please investigate if there are "
     "error(s) on other node(s)"},
+   {NDBD_EXIT_PARTITIONED_SHUTDOWN, XAE, "Partitioned cluster detected. "
+    "Please check if cluster is already running"},
    {NDBD_EXIT_POINTER_NOTINRANGE, XIE, "Pointer too large"},
    {NDBD_EXIT_SR_OTHERNODEFAILED, XRE, "Another node failed during system "
     "restart, please investigate error(s) on other node(s)"},

storage/ndb/src/ndbapi/DictCache.cpp

@@ -312,6 +312,8 @@ GlobalDictCache::invalidate_all()
     if (vers->size())
     {
       TableVersion * ver = & vers->back();
+      if (ver->m_status != RETREIVING)
+      {
         ver->m_impl->m_status = NdbDictionary::Object::Invalid;
         ver->m_status = DROPPED;
         if (ver->m_refCount == 0)
@@ -320,6 +322,7 @@ GlobalDictCache::invalidate_all()
           vers->erase(vers->size() - 1);
         }
       }
+    }
     curr = m_tableHash.getNext(curr);
   }
   DBUG_VOID_RETURN;

storage/ndb/src/ndbapi/Ndb.cpp

@@ -1025,14 +1025,19 @@ int Ndb::initAutoIncrement()
   setDatabaseName("sys");
   setDatabaseSchemaName("def");
 
-  m_sys_tab_0 = getDictionary()->getTableGlobal("SYSTAB_0");
+  m_sys_tab_0 = theDictionary->getTableGlobal("SYSTAB_0");
 
   // Restore current name space
   setDatabaseName(currentDb.c_str());
   setDatabaseSchemaName(currentSchema.c_str());
 
+  if (m_sys_tab_0 == NULL) {
+    assert(theDictionary->m_error.code != 0);
+    theError.code = theDictionary->m_error.code;
+    return -1;
+  }
 
-  return (m_sys_tab_0 == NULL);
+  return 0;
 }
 
 int
@@ -1043,19 +1048,19 @@ Ndb::opTupleIdOnNdb(const NdbTableImpl* table,
   Uint32 aTableId = table->m_id;
   DBUG_PRINT("enter", ("table=%u value=%llu op=%u", aTableId, opValue, op));
 
-  NdbTransaction*     tConnection;
-  NdbOperation*      tOperation= 0; // Compiler warning if not initialized
+  NdbTransaction*    tConnection = NULL;
+  NdbOperation*      tOperation = NULL;
   Uint64             tValue;
   NdbRecAttr*        tRecAttrResult;
 
-  CHECK_STATUS_MACRO_ZERO;
+  CHECK_STATUS_MACRO;
 
-  if (initAutoIncrement())
-    goto error_return;
+  if (initAutoIncrement() == -1)
+    goto error_handler;
 
   tConnection = this->startTransaction();
   if (tConnection == NULL)
-    goto error_return;
+    goto error_handler;
 
   tOperation = tConnection->getNdbOperation(m_sys_tab_0);
   if (tOperation == NULL)
@@ -1065,7 +1070,7 @@ Ndb::opTupleIdOnNdb(const NdbTableImpl* table,
     {
     case 0:
       tOperation->interpretedUpdateTuple();
-      tOperation->equal("SYSKEY_0", aTableId );
+      tOperation->equal("SYSKEY_0", aTableId);
       tOperation->incValue("NEXTID", opValue);
       tRecAttrResult = tOperation->getValue("NEXTID");
 
@@ -1130,14 +1135,21 @@ Ndb::opTupleIdOnNdb(const NdbTableImpl* table,
 
   DBUG_RETURN(0);
 
-  error_handler:
-    theError.code = tConnection->theError.code;
-    this->closeTransaction(tConnection);
-  error_return:
+error_handler:
   DBUG_PRINT("error", ("ndb=%d con=%d op=%d",
              theError.code,
-             tConnection ? tConnection->theError.code : -1,
-             tOperation ? tOperation->theError.code : -1));
+             tConnection != NULL ? tConnection->theError.code : -1,
+             tOperation != NULL ? tOperation->theError.code : -1));
+
+  if (theError.code == 0 && tConnection != NULL)
+    theError.code = tConnection->theError.code;
+  if (theError.code == 0 && tOperation != NULL)
+    theError.code = tOperation->theError.code;
+  DBUG_ASSERT(theError.code != 0);
+
+  if (tConnection != NULL)
+    this->closeTransaction(tConnection);
+
   DBUG_RETURN(-1);
 }
 

storage/ndb/src/ndbapi/NdbEventOperationImpl.cpp

@@ -942,6 +942,7 @@ NdbEventBuffer::NdbEventBuffer(Ndb *ndb) :
 {
 #ifdef VM_TRACE
   m_latest_command= "NdbEventBuffer::NdbEventBuffer";
+  m_flush_gci = 0;
 #endif
 
   if ((p_cond = NdbCondition_Create()) ==  NULL) {
@@ -1109,6 +1110,8 @@ NdbEventBuffer::flushIncompleteEvents(Uint64 gci)
   /**
    *  Find min complete gci
    */
+  // called by user thread, so we need to lock the data
+  lock();
   Uint32 i;
   Uint32 sz= m_active_gci.size();
   Gci_container* array = (Gci_container*)m_active_gci.getBase();
@@ -1127,6 +1130,10 @@ NdbEventBuffer::flushIncompleteEvents(Uint64 gci)
       bzero(tmp, sizeof(Gci_container));
     }
   }
+#ifdef VM_TRACE
+  m_flush_gci = gci;
+#endif
+  unlock();
   return 0;
 }
 
@@ -1301,7 +1308,11 @@ operator<<(NdbOut& out, const Gci_container_pod& gci)
 
 static
 Gci_container*
-find_bucket_chained(Vector<Gci_container_pod> * active, Uint64 gci)
+find_bucket_chained(Vector<Gci_container_pod> * active, Uint64 gci
+#ifdef VM_TRACE
+                    ,Uint64 flush_gci
+#endif
+                    )
 {
   Uint32 pos = (gci & ACTIVE_GCI_MASK);
   Gci_container *bucket= ((Gci_container*)active->getBase()) + pos;
@@ -1322,6 +1333,13 @@ find_bucket_chained(Vector<Gci_container_pod> * active, Uint64 gci)
 	bzero(bucket, sizeof(Gci_container));
 	bucket->m_gci = gci;
 	bucket->m_gcp_complete_rep_count = ~(Uint32)0;
+#ifdef VM_TRACE
+        if (gci < flush_gci)
+        {
+          ndbout_c("received old gci %llu < flush gci %llu", gci, flush_gci);
+          assert(false);
+        }
+#endif
 	return bucket;
       }
       move_pos += ACTIVE_GCI_DIRECTORY_SIZE;
@@ -1336,7 +1354,16 @@ find_bucket_chained(Vector<Gci_container_pod> * active, Uint64 gci)
       bucket += ACTIVE_GCI_DIRECTORY_SIZE;
       
       if(bucket->m_gci == gci)
+      {
+#ifdef VM_TRACE
+        if (gci < flush_gci)
+        {
+          ndbout_c("received old gci %llu < flush gci %llu", gci, flush_gci);
+          assert(false);
+        }
+#endif
 	return bucket;
+      }
       
     } while(pos < size);
     
@@ -1346,14 +1373,22 @@ find_bucket_chained(Vector<Gci_container_pod> * active, Uint64 gci)
 
 inline
 Gci_container*
-find_bucket(Vector<Gci_container_pod> * active, Uint64 gci)
+find_bucket(Vector<Gci_container_pod> * active, Uint64 gci
+#ifdef VM_TRACE
+            ,Uint64 flush_gci
+#endif
+            )
 {
   Uint32 pos = (gci & ACTIVE_GCI_MASK);
   Gci_container *bucket= ((Gci_container*)active->getBase()) + pos;
   if(likely(gci == bucket->m_gci))
     return bucket;
 
-  return find_bucket_chained(active,gci);
+  return find_bucket_chained(active,gci
+#ifdef VM_TRACE
+                             , flush_gci
+#endif
+                             );
 }
 
 static
@@ -1386,7 +1421,11 @@ NdbEventBuffer::execSUB_GCP_COMPLETE_REP(const SubGcpCompleteRep * const rep)
   const Uint64 gci= rep->gci;
   const Uint32 cnt= rep->gcp_complete_rep_count;
 
-  Gci_container *bucket = find_bucket(&m_active_gci, gci);
+  Gci_container *bucket = find_bucket(&m_active_gci, gci
+#ifdef VM_TRACE
+                                      , m_flush_gci
+#endif
+                                      );
 
   if (unlikely(bucket == 0))
   {
@@ -1752,7 +1791,11 @@ NdbEventBuffer::insertDataL(NdbEventOperationImpl *op,
 
   if ( likely((Uint32)op->mi_type & (1 << (Uint32)sdata->operation)) )
   {
-    Gci_container* bucket= find_bucket(&m_active_gci, gci);
+    Gci_container* bucket= find_bucket(&m_active_gci, gci
+#ifdef VM_TRACE
+                                       , m_flush_gci
+#endif
+                                       );
       
     DBUG_PRINT_EVENT("info", ("data insertion in eventId %d", op->m_eventId));
     DBUG_PRINT_EVENT("info", ("gci=%d tab=%d op=%d node=%d",

storage/ndb/src/ndbapi/NdbEventOperationImpl.hpp

@@ -463,6 +463,7 @@ public:
 
 #ifdef VM_TRACE
   const char *m_latest_command;
+  Uint64 m_flush_gci;
 #endif
 
   Ndb *m_ndb;