must be updated correctly. And Recovery should not assume that
log 1 is available but just use first available log.


mysql-test/r/maria-purge.result:
  result update
mysql-test/t/maria-purge.test:
  like in other Maria tests (4294967296 is too big, causes warning)
storage/maria/ma_loghandler.c:
  When log handler starts and discovers that logs are missing though
  control file says they should be there (they have been manually
  deleted), it decides to create a log
  with number last_logno+1. It needs to update control file with
  this new number, and also reset last_checkpoint_lsn to LSN_IMPOSSIBLE
  (last checkpoint went away when log was deleted).
  Fix for when translog_init() fails before allocating id_to_share
  (translog_destroy() tried to free NULL).
storage/maria/ma_recovery.c:
  When logs have been manually deleted, translog_first_theoretical_lsn()
  always fails (it looks for log number 1). Recovery now uses
  translog_first_lsn_in_log() which just picks the first available log.