1. 31 May, 2016 4 commits
  2. 30 May, 2016 4 commits
  3. 27 May, 2016 2 commits
  4. 25 May, 2016 1 commit
  5. 23 May, 2016 1 commit
    • olivier R-D's avatar
      event history api tweak · 4f8f92dc
      olivier R-D authored
      take event as arg not evfilter
      return EventResult obj and not fields
      update command line to read events
      4f8f92dc
  6. 22 May, 2016 1 commit
  7. 21 May, 2016 3 commits
  8. 20 May, 2016 2 commits
    • Dominik Luntzer's avatar
      Correctly handle period and count limits in both history implementations · bace92d0
      Dominik Luntzer authored
      * Added missing handling of the count parameter to the sqlite implementation
      * Add tests for both limits and both history implementations
      * Fixed some minor issues
      bace92d0
    • Dominik Luntzer's avatar
      Fix broken history tests · ef0cba63
      Dominik Luntzer authored
      The test code is currently setting a limit of 10 history data values.
      However, the limit handling is not implemented in both history
      implementations and therefore the limit is never enforced.
      Removed the limit from the history tests to ensure that no tests are broken
      when the limit handling is actually implemented.
      ef0cba63
  9. 18 May, 2016 2 commits
  10. 17 May, 2016 3 commits
  11. 16 May, 2016 7 commits
  12. 15 May, 2016 2 commits
  13. 14 May, 2016 1 commit
  14. 13 May, 2016 2 commits
  15. 11 May, 2016 1 commit
    • Dominik Luntzer's avatar
      Perform lazy loading when restoring a cached address space · 2fcbbf85
      Dominik Luntzer authored
      When starting an opcua server, the creation of the address space is
      currently a performance bottleneck. The startup process can be accelerated
      by loading a pre-generated address space pickle.
      However, the startup process still takes ~25 seconds on a raspberry pi
      model b (compared to ~125 seconds when generating the address space from code).
      Storing the address space in a shelve, where the data for each node is
      pickeled individually, allows to further improve the startup performance
      since only the nodes that are actually accessed are loaded from
      disc (all other nodes are loaded later when they are accessed).
      Since the default address space contains thousands of nodes but just a
      small amount is actually accessed during startup, the startuptime could
      be improved to ~3.5 seconds.
      2fcbbf85
  16. 10 May, 2016 4 commits