Fix problem with AttributeHeader::init() seen with gcc 4.2.1.

Using the same object as both Uint32 and class AttributeHeader violates
strict aliasing rule.

  dont assert if 2 LCP's are being run during a node recovery

  Not very clever fix for DIH incorrect REDO handling
  - Dont report GCP_SAVE_CONF until first LCP has been complete during NR

  In TC init node status for already started nodes during node restart
  (not present in 5.1)

  Fix 3 problems in REDO validation that could(will) lead to "Error while reading REDO log"

into  perch.ndb.mysql.com:/home/jonas/src/50-work

into  perch.ndb.mysql.com:/home/jonas/src/50-work

  review comment 2, atleast 1 signal need for test prg

into  perch.ndb.mysql.com:/home/jonas/src/50-work

  review comment
  if some tcp-transporter has data, then do select with timeout 0

into  perch.ndb.mysql.com:/home/jonas/src/50-work

  Make sure that data can not e left lingering in receive buffer

  redo extra verification code so that tupkeyref is reset just before tupkeyreq

  bug in interpretedupdate (used for auto-increment) that made TUPKEYREF be sent twice
  this was however "normally" shadowed, but made LQH crash when LCP was about to start

into  perch.ndb.mysql.com:/home/jonas/src/50-work

  fix commit trigger on multi-update

  make sure not to leave partially initialized pagerage-records

  put64 for 64-bit variables

  fix test prg

    Additional fix for 2-node case

  make sure master is sendable

    Handle race condtition between MASTER_GCPCONF and execGCP_NODEFINISH

  Fix correct min-value for LockPagesInMemory

  reset rowcount on temporary error during scan of base table

  Handle random(not in order) LQHKEYREQ failures during node-restart

  Make StartREDO come up as default in cluster log

  make sure logPartPtr is correctly initialized

  forgott to add testcase for bug#26481

  fix for killed node during initial node restart

  master failure during master take over

  backport fix from 5.1

  make lockpages=1 run a bit earlier in startup

  fix cpu peak in big clusters during unpack of config

  add support for doing mlockall before mallc instead of after
  (recommit in 5.0)

  fix bug when master failure during nr
  (recommit against 5.0)

  on master node failure during qmgr-commitreq
    make sure to remove all committed failed nodes from failed/prepfailed arrays