With the switch to msgpack, there was no schema anymore whereas it was
sometimes used for both automatic conversion (e.g. the last argument of
AskStoreTransaction must now be explicitly cast to list) and type checking.

This somewhat reintroduces a kind of schema that:
- is used by the test suite for type checking
- can be generated automatically from the test suite
  when one change the procotol

Not only for performance reasons (which is significant in the case of
replication; tools/matrix is ~3% faster) but also because of several ugly
things in the way packets were defined:
- packet field names, which are only documentary; for roots fields,
  they even just duplicate the packet names
- a lot of repetitions for packet names, and even confusion between the name
  of the packet definition and the name of the actual notify/request packet
- the need to implement field types for anything, like PByte to support new
  compression formats, since PBoolean is not enough

neo/lib/protocol.py is now much smaller.

Example output:

    stress: yes (toggle with F1)
    cluster state: RUNNING
    last oid: 0x44c0
    last tid: 0x3cdee272ef19355 (2019-02-26 15:35:11.002419)
    clients: 2308, 2311, 2302, 2173, 2226, 2215, 2306, 2255, 2314, 2356 (+48)
            8m53.988s (42.633861/s)
    pt id: 4107
        RRRDDRRR
     0: OU......
     1: ..UO....
     2: ....OU..
     3: ......UU
     4: OU......
     5: ..UO....
     6: ....OU..
     7: ......UU
     8: OU......
     9: ..UO....
    10: ....OU..
    11: ......UU
    12: OU......
    13: ..UO....
    14: ....OU..
    15: ......UU
    16: OU......
    17: ..UO....
    18: ....OU..
    19: ......UU
    20: OU......
    21: ..UO....
    22: ....OU..
    23: ......UU

getAddress (via __repr__) raised EBADF on closed connectors.

In functional tests (or anything reusing this framework),
the mapping could be incorrect at the beginning of logs.

Corrupted logs cause neolog to fail with the following error:

  AttributeError: 'Log' object has no attribute 'uuid_str'

This makes commit 3c7a3160
(storage: speed up reads by indexing 'obj' primarily by 'oid')
effective for SQLite.

The fake changes in test data are because we don't force upgrade
for this optimization.

Commit aa4d621d broke log rotation
and neolog sometimes failed to read in new format.

neolog has new options: -N for old behaviour, and -C to show the cluster name.

Since commit 50e7fe52,
some code can be simplified.

This fixes a bug that could manifest as follows:

  Traceback (most recent call last):
    File "neo/client/app.py", line 432, in load
      self._cache.store(oid, data, tid, next_tid)
    File "neo/client/cache.py", line 223, in store
      assert item.tid == tid, (item, tid)
  AssertionError: (<CacheItem oid='\x00\x00\x00\x00\x00\x00\x00\x01' tid='\x03\xcb\xc6\xca\xfd\xc7\xda\xee' next_tid='\x03\xcb\xc6\xca\xfd\xd8\t\x88' data='...' counter=1 level=1 expire=10000 prev=<...> next=<...>>, '\x03\xcb\xc6\xca\xfd\xd8\t\x88')

The big changes in the threaded test framework are required because we need to
reproduce a race condition between client threads and this conflicts with the
serialization of epoll events (deadlock).

This was found when stress-testing a big cluster. 1 client node was stuck:

  (Pdb) pp app.dispatcher.__dict__
  {'lock_acquire': <built-in method acquire of thread.lock object at 0x7f788c6e4250>,
  'lock_release': <built-in method release of thread.lock object at 0x7f788c6e4250>,
  'message_table': {140155667614608: {},
                    140155668875280: {},
                    140155671145872: {},
                    140155672381008: {},
                    140155672381136: {},
                    140155672381456: {},
                    140155673002448: {},
                    140155673449680: {},
                    140155676093648: {170: <neo.lib.locking.SimpleQueue object at 0x7f788a109c58>},
                    140155677536464: {},
                    140155679224336: {},
                    140155679876496: {},
                    140155680702992: {},
                    140155681851920: {},
                    140155681852624: {},
                    140155682773584: {},
                    140155685988880: {},
                    140155693061328: {},
                    140155693062224: {},
                    140155693074960: {},
                    140155696334736: {278: <neo.lib.locking.SimpleQueue object at 0x7f788a109c58>},
                    140155696411408: {},
                    140155696414160: {},
                    140155696576208: {},
                    140155722373904: {}},
  'queue_dict': {140155673622936: 1, 140155689147480: 2}}

140155673622936 should not be queue_dict

The idea is to write:

  with self.expectedFailure(...): \

just before the statement that is expected to fail. Contrary to the existing
decorator, we want to:
- be sure that the test fails at the expected line;
- be able to remove an expectedFailure without touching the code around.

When ConnectionPool._initNodeConnection fails a first time with:

  StorageError: protocol error: already connected

the following assertion failure happens when trying to reconnect before the
previous connection is actually closed (currently, only the node sending an
error message closes the connection, as commented in EventHandler):

  Traceback (most recent call last):
    File "neo/client/Storage.py", line 82, in load
      return self.app.load(oid)[:2]
    File "neo/client/app.py", line 367, in load
      data, tid, next_tid, _ = self._loadFromStorage(oid, tid, before_tid)
    File "neo/client/app.py", line 399, in _loadFromStorage
      askStorage)
    File "neo/client/app.py", line 293, in _askStorageForRead
      conn = cp.getConnForNode(node)
    File "neo/client/pool.py", line 98, in getConnForNode
      conn = self._initNodeConnection(node)
    File "neo/client/pool.py", line 48, in _initNodeConnection
      dispatcher=app.dispatcher)
    File "neo/lib/connection.py", line 704, in __init__
      super(MTClientConnection, self).__init__(*args, **kwargs)
    File "neo/lib/connection.py", line 602, in __init__
      node.setConnection(self)
    File "neo/lib/node.py", line 122, in setConnection
      attributeTracker.whoSet(self, '_connection'))
  AssertionError

This fixes:

  Traceback (most recent call last):
    File "neo/client/Storage.py", line 108, in tpc_vote
      return self.app.tpc_vote(transaction)
    File "neo/client/app.py", line 546, in tpc_vote
      self.waitStoreResponses(txn_context)
    File "neo/client/app.py", line 539, in waitStoreResponses
      _waitAnyTransactionMessage(txn_context)
    File "neo/client/app.py", line 160, in _waitAnyTransactionMessage
      self._handleConflicts(txn_context)
    File "neo/client/app.py", line 514, in _handleConflicts
      self._store(txn_context, oid, serial, data)
    File "neo/client/app.py", line 452, in _store
      self._waitAnyTransactionMessage(txn_context, False)
    File "neo/client/app.py", line 155, in _waitAnyTransactionMessage
      self._waitAnyMessage(queue, block=block)
    File "neo/client/app.py", line 142, in _waitAnyMessage
      _handlePacket(conn, packet, kw)
    File "neo/lib/threaded_app.py", line 133, in _handlePacket
      handler.dispatch(conn, packet, kw)
    File "neo/lib/handler.py", line 72, in dispatch
      method(conn, *args, **kw)
    File "neo/client/handlers/storage.py", line 143, in answerRebaseObject
      assert cached == data
  AssertionError

During rebase, writes could stay lockless although the partition was
replicated. Another transaction could then take locks prematurely, leading to
the following crash:

  Traceback (most recent call last):
    File "neo/lib/handler.py", line 72, in dispatch
      method(conn, *args, **kw)
    File "neo/storage/handlers/master.py", line 36, in notifyUnlockInformation
      self.app.tm.unlock(ttid)
    File "neo/storage/transactions.py", line 329, in unlock
      self.abort(ttid, even_if_locked=True)
    File "neo/storage/transactions.py", line 573, in abort
      not self._replicated.get(self.getPartition(oid))), x
  AssertionError: ('\x00\x00\x00\x00\x00\x03\x03v', '\x03\xca\xb44J\x13\x99\x88', '\x03\xca\xb44J\xe0\xdcU', {}, set(['\x00\x00\x00\x00\x00\x03\x03v']))

Flushing logs will help fixing NEO bugs (e.g. failed assertions).

This fixes:

  Traceback (innermost last):
    ...
    Module transaction._transaction, line 393, in _commitResources
      rm.tpc_vote(self)
    Module ZODB.Connection, line 797, in tpc_vote
      s = vote(transaction)
    Module neo.client.Storage, line 95, in tpc_vote
      return self.app.tpc_vote(transaction)
    Module neo.client.app, line 546, in tpc_vote
      self.waitStoreResponses(txn_context)
    Module neo.client.app, line 539, in waitStoreResponses
      _waitAnyTransactionMessage(txn_context)
    Module neo.client.app, line 160, in _waitAnyTransactionMessage
      self._handleConflicts(txn_context)
    Module neo.client.app, line 471, in _handleConflicts
      assert oid is None, (oid, serial)
  AssertionError: ('\x00\x00\x00\x00\x00\x02\n\xe3', '\x03\xca\xad\xcb!\x92\xb6\x9c')

With previous commit, there's no point anymore to distinguish storage nodes
for which we only check serials.

Not doing so was an incorrect optimization. Checking serials does take
write-locks and they must not be released when a client-storage connection
breaks between vote and lock, otherwise a concurrent transaction modifying such
serials may finish before.

When a client-storage connection breaks, the storage node discards data of all
ongoing transactions by the client. Therefore, a reconnection within the
context of the transaction is wrong, as it could lead to partially-written
transactions.

This fixes cases where such reconnection happened. The biggest issue was that
the mechanism to dispatch disconnection events only works when waiting for an
answer.

The client can still reconnect for other purposes but the new connection won't
be reused by transactions that already involved the storage node.

Without this new mechanism to detect oids that aren't write-locked,
a transaction could be committed successfully without detecting conflicts.
In the added test, the resulting value was 2, whereas it should be 5 if there
was no node failure.