Make pt.log() more robust. Do not delete cell lists in protocol handling (what was this for?).

git-svn-id: https://svn.erp5.org/repos/neo/branches/prototype3@206 71dcc9de-d417-0410-9af5-da40c76e7ee4

neo/master/app.py

@@ -18,6 +18,7 @@ from neo.master.verification import VerificationEventHandler
 from neo.master.service import ServiceEventHandler
 from neo.master.secondary import SecondaryEventHandler
 from neo.pt import PartitionTable
+from neo.util import dump
 
 class Application(object):
     """The master node application."""
@@ -401,11 +402,15 @@ class Application(object):
 
                 if self.lptid != prev_lptid or not self.pt.filled():
                     # I got something newer or the target is dead.
+                    logging.debug('self.lptid = %s, prev_lptid = %s',
+                                  dump(self.lptid), dump(prev_lptid))
+                    self.pt.log()
                     continue
 
                 # Wait until the cluster gets operational or the Partition Table ID
                 # turns out to be not the latest.
                 logging.debug('waiting for the cluster to be operational')
+                self.pt.log()
                 while 1:
                     em.poll(1)
                     if self.pt.operational():

neo/master/recovery.py

@@ -316,6 +316,8 @@ class RecoveryEventHandler(MasterEventHandler):
             return
         if uuid != app.target_uuid:
             # If this is not from a target node, ignore it.
+            logging.warn('got answer partition table from %s while waiting for %s',
+                         dump(uuid), dump(app.target_uuid))
             return
 
         for offset, cell_list in row_list:

neo/protocol.py

@@ -790,7 +790,6 @@ class Packet(object):
                     index += 18
                     cell_list.append(cell)
                 row_list.append((offset, cell_list))
-                del cell_list[:]
         except:
             raise ProtocolError(self, 'invalid answer partition table')
         return ptid, row_list
@@ -810,7 +809,6 @@ class Packet(object):
                     index += 18
                     cell_list.append(cell)
                 row_list.append((offset, tuple(cell_list)))
-                del cell_list[:]
         except:
             raise ProtocolError(self, 'invalid send partition table')
         return ptid, row_list

neo/pt.py

@@ -152,11 +152,14 @@ class PartitionTable(object):
                             FEEDING_STATE: 'F' }
         for offset, row in enumerate(self.partition_list):
             desc_list = []
-            for cell in row:
-                i = node_dict[cell.getNode()]
-                cell_state = cell_state_dict[cell.getState()]
-                node_state = node_state_dict[cell.getNodeState()]
-                desc_list.append('%d %s %s' % (i, cell_state, node_state))
+            if row is None:
+                desc_list.append('None')
+            else:
+                for cell in row:
+                    i = node_dict[cell.getNode()]
+                    cell_state = cell_state_dict[cell.getState()]
+                    node_state = node_state_dict[cell.getNodeState()]
+                    desc_list.append('%d %s %s' % (i, cell_state, node_state))
             logging.debug('pt: row %d: %s', offset, ', '.join(desc_list))
 
     def operational(self):

neo/storage/replicator.py

@@ -252,6 +252,7 @@ class Replicator(object):
             pass
 
     def _askCriticalTID(self):
+        logging.debug('self.new_partition_list = %r', self.new_partition_list)
         conn = self.primary_master_connection
         msg_id = conn.getNextId()
         conn.addPacket(Packet().askLastIDs(msg_id))
@@ -268,7 +269,7 @@ class Replicator(object):
     def _askUnfinishedTIDs(self):
         conn = self.primary_master_connection
         msg_id = conn.getNextId()
-        conn.addPacket(Packet().askUnfinishedTIDs(msg_id))
+        conn.addPacket(Packet().askUnfinishedTransactions(msg_id))
         conn.expectMessage(msg_id)
         self.waiting_for_unfinished_tids = True