bk commit into 5.1 tree (tomas:1.2381) BUG#18594

[tomas]

Below is the list of changes that have just been committed into a local
5.1 repository of tomas. When tomas does a push these changes will
be propagated to the main repository and, within 24 hours after the
push, to the public repository.
For information on how to access the public repository
see http://dev.mysql.com/doc/mysql/en/installing-source-tree.html

ChangeSet
  1.2381 06/04/21 11:26:21 tomas@stripped +1 -0
  bug#18594 ndb_restore log boken in 5.1
  - corrected previous patch
  - read log entry variables out into explicit variables for siimpler code (and backwards
compatability code)

  storage/ndb/tools/restore/Restore.cpp
    1.38 06/04/21 11:26:15 tomas@stripped +39 -47
    bug#18594 ndb_restore log boken in 5.1
    - corrected previous patch
    - read log entry variables out into explicit variables for siimpler code (and
backwards compatability code)

# This is a BitKeeper patch.  What follows are the unified diffs for the
# set of deltas contained in the patch.  The rest of the patch, the part
# that BitKeeper cares about, is below these diffs.
# User:	tomas
# Host:	poseidon.ndb.mysql.com
# Root:	/home/tomas/mysql-5.1-new

--- 1.37/storage/ndb/tools/restore/Restore.cpp	2006-04-21 01:39:27 +02:00
+++ 1.38/storage/ndb/tools/restore/Restore.cpp	2006-04-21 11:26:15 +02:00
@@ -961,15 +961,15 @@
 const LogEntry *
 RestoreLogIterator::getNextLogEntry(int & res) {
   // Read record length
-  typedef BackupFormat::LogFile::LogEntry LogE;
-  typedef BackupFormat::LogFile::LogEntry_no_fragid LogE_no_fragid;
-  const Uint32 offset= 3;
-  assert(offset == (offsetof(LogE, Data) >> 2) - 1);
-  LogE * logE= 0;
-  Uint32 len= ~0;
   const Uint32 stopGCP = m_metaData.getStopGCP();
-  NdbAutoPtr<char> ap1;
+  Uint32 tableId;
+  Uint32 triggerEvent;
+  Uint32 frag_id;
+  Uint32 *attr_data;
+  Uint32 attr_data_len;
   do {
+    Uint32 len;
+    Uint32 *logEntryPtr;
     if (buffer_read_ahead(&len, sizeof(Uint32), 1) != 1){
       res= -1;
       return 0;
@@ -977,7 +977,7 @@
     len= ntohl(len);
 
     Uint32 data_len = sizeof(Uint32) + len*4;
-    if (buffer_get_ptr((void **)(&logE), 1, data_len) != data_len) {
+    if (buffer_get_ptr((void **)(&logEntryPtr), 1, data_len) != data_len) {
       res= -2;
       return 0;
     }
@@ -986,7 +986,8 @@
       res= 0;
       return 0;
     }
-    if (m_metaData.getFileHeader().NdbVersion < NDBD_FRAGID_VERSION)
+
+    if (unlikely(m_metaData.getFileHeader().NdbVersion < NDBD_FRAGID_VERSION))
     {
       /*
         FragId was introduced in LogEntry in version
@@ -994,48 +995,39 @@
         We set FragId to 0 in older versions (these versions
         do not support restore of user defined partitioned
         tables.
-
-        These log entries miss one Uint32 FragId, hence missing_len=1
-
-        Reconstruct a new log entry with old.
       */
-      const Uint32 missing_len= 1;
-      assert((offsetof(LogE, Data) - offsetof(LogE_no_fragid, Data)) >> 2 ==
-             missing_len);
-      LogE_no_fragid * logE_no_fragid= (LogE_no_fragid *)logE;
-      
-      int i;
-      LogE *tmpLogE= (LogE*)NdbMem_Allocate(data_len + missing_len*4);
-      if (!tmpLogE)
-      {
-        res = -2;
-        return 0;
-      }
-      ap1.reset((char*)tmpLogE);
-      bzero(tmpLogE, data_len + missing_len*4);
-      /* correct len to reflect new logEntry version length */
-      len+= missing_len;
-      tmpLogE->Length = logE_no_fragid->Length;
-      tmpLogE->TableId = logE_no_fragid->TableId;
-      tmpLogE->TriggerEvent = logE_no_fragid->TriggerEvent;
-      for (i = 0; i < len - offset; i++)
-        tmpLogE->Data[i] = logE_no_fragid->Data[i];
-      logE= tmpLogE;
+      typedef BackupFormat::LogFile::LogEntry_no_fragid LogE_no_fragid;
+      LogE_no_fragid * logE_no_fragid= (LogE_no_fragid *)logEntryPtr;
+      tableId= ntohl(logE_no_fragid->TableId);
+      triggerEvent= ntohl(logE_no_fragid->TriggerEvent);
+      frag_id= 0;
+      attr_data= &logE_no_fragid->Data[0];
+      attr_data_len= len - ((offsetof(LogE_no_fragid, Data) >> 2) - 1);
+    }
+    else /* normal case */
+    {
+      typedef BackupFormat::LogFile::LogEntry LogE;
+      LogE * logE= (LogE *)logEntryPtr;
+      tableId= ntohl(logE->TableId);
+      triggerEvent= ntohl(logE->TriggerEvent);
+      frag_id= ntohl(logE->FragId);
+      attr_data= &logE->Data[0];
+      attr_data_len= len - ((offsetof(LogE, Data) >> 2) - 1);
     }
-    logE->TableId= ntohl(logE->TableId);
-    logE->TriggerEvent= ntohl(logE->TriggerEvent);
-    
-    const bool hasGcp= (logE->TriggerEvent & 0x10000) != 0;
-    logE->TriggerEvent &= 0xFFFF;
     
+    const bool hasGcp= (triggerEvent & 0x10000) != 0;
+    triggerEvent &= 0xFFFF;
+
     if(hasGcp){
+      // last attr_data is gci info
+      attr_data_len--;
       len--;
-      m_last_gci = ntohl(logE->Data[len-offset]);
+      m_last_gci = ntohl(*(attr_data + attr_data_len));
     }
   } while(m_last_gci > stopGCP + 1);
-  
-  m_logEntry.m_table = m_metaData.getTable(logE->TableId);
-  switch(logE->TriggerEvent){
+
+  m_logEntry.m_table = m_metaData.getTable(tableId);
+  switch(triggerEvent){
   case TriggerEvent::TE_INSERT:
     m_logEntry.m_type = LogEntry::LE_INSERT;
     break;
@@ -1053,10 +1045,10 @@
   const TableS * tab = m_logEntry.m_table;
   m_logEntry.clear();
 
-  AttributeHeader * ah = (AttributeHeader *)&logE->Data[0];
-  AttributeHeader *end = (AttributeHeader *)&logE->Data[len - offset];
+  AttributeHeader * ah = (AttributeHeader *)attr_data;
+  AttributeHeader *end = (AttributeHeader *)(attr_data + attr_data_len);
   AttributeS *  attr;
-  m_logEntry.m_frag_id = ntohl(logE->FragId);
+  m_logEntry.m_frag_id = frag_id;
   while(ah < end){
     attr= m_logEntry.add_attr();
     if(attr == NULL) {