1. 19 Aug, 2013 1 commit
    • Brad King's avatar
      COMP: Cleanup CMP0022 and CMP0023 warnings · ed6c07d0
      Brad King authored
      CMake 2.8.12 introduces policies CMP0022 and CMP0023 that trigger
      warnings in some of VTK's CMake code.  Fix them as follows.  The result
      is a cleaner and more explicit link specification anyway.
      CMP0022 warns when we set the LINK_INTERFACE_LIBRARIES property directly
      without also setting the newly introduced INTERFACE_LINK_LIBRARIES
      property directly.  Simply drop these settings because they are achieved
      already by use of the keyword-based target_link_libraries signature.
      CMP0023 warns when we use both the plain and keyword signatures of the
      target_link_libraries command for a single target.  The module macros
      use the keyword signature (LINK_PUBLIC/LINK_PRIVATE), so any direct
      calls to the plain target_link_libraries signature are doing "private"
      linking.  Add the LINK_PRIVATE keyword to such calls to make this
      explicit and avoid the warning.
      Change-Id: I1e5bf687733e3f19462da698c0d5e3899439bf02
  2. 16 Aug, 2013 7 commits
  3. 15 Aug, 2013 8 commits
  4. 14 Aug, 2013 6 commits
  5. 13 Aug, 2013 12 commits
    • Berk Geveci's avatar
      Performance improvement to CellToPoint. · 038024b7
      Berk Geveci authored
      Using faster API to import performance of vtkCellDataToPointData.
      vtkCellDataToPointData was using ds->GetCell(..)->GetPointIds() which is
      considerably slower than directly calling ds->GetCellPoints(...). That change
      considerably improves the speed of this filter.
      Change-Id: I99660cde38a1b359b76d8c71b673d5ae9a229507
    • David Gobbi's avatar
      Implement multi-sample antialiasing on OS X. · 23f6ba79
      David Gobbi authored
      Set the necessary attributes to create a sample buffer for the
      OpenGL context of the vtkCocoaRenderWindow.  This allows VTK to
      do anti-aliased rendering in the window.  However, multisampling
      is turned off by default on OS X because it causes rendering
      problems on some Mac models.
      Change-Id: I0ff8073e38d8a63e637cfe4e7146e2d1327d65eb
    • David Gobbi's avatar
      Merge topic 'ccs-edge-locator-free' into master · 4eb4a238
      David Gobbi authored
      9b46579d Perform explicit list deletion in CCSEdgeLocator.
    • David Thompson's avatar
      Merge topic '14188_polyhedron_fixes' into master · 2b64cfa3
      David Thompson authored
      0627519c Terminate early when possible.
      fc0ee6cf Reset point ID list when appending polyhedra.
      ee3a7c30 Fix polyhedron-ray intersection...
      31d834a4 Don't forget to copy polyhedron cell-data...
      2f4a9cad Fix problems processing polyhedron connectivity.
    • Bill Lorensen's avatar
      Merge topic 'ExpectEOFsIn3DS' into master · d132ee85
      Bill Lorensen authored
      631d795b Reads past EOF are expected
    • Kevin H. Hobbs's avatar
      Reads past EOF are expected · 631d795b
      Kevin H. Hobbs authored
      The vtk3DSImporter reports errors like :
      ERROR: In /home/kevin/kitware/VTK/IO/Import/vtk3DSImporter.cxx, line
      vtk3DSImporter (0x75cd20): Pre-mature end of file in read_word
      on almost every machine in the dashboard.
      I did a bit of poking around with gdb and std::cerr, and it looks like
      these reads past the end of the file are not at all unexpected.
      For example parse_face_array uses read_word to read the number of faces
      and then to read what I assume are the face vertex ids: a failure here
      is catastrophic.
      Then even though importer->GetFileFD() could be at the end of the file
      it calls start_chunk which calls both read_word and read_dword to see if
      there is another (sub?)chunk to parse into.
      This commit adds two new functions peek_word and peek_dword that are
      exactly like read_word and read_dword except that they do not report an
      error when they read past the end of the file.
      peek_word and peek_dword replace read_word and read_dword in start_chunk
      where the read past the end of the file is not an error.
      read_dword is now unused
      Because read_dword is now unused, it should be removed.
      Stop leaking render windows
      Hopefully X and Mesa have advanced to the point where we can delete the
      render window without error.
      If any tested configurations still have trouble let us work around them
      more precisely.
      Change-Id: I90a68d1241379886d38d150ba00816d869c6675f
    • David Thompson's avatar
      Terminate early when possible. · 0627519c
      David Thompson authored
      Change-Id: Ibcd1bb931c4f72960a3c997bf4d84653516d70f4
    • David Thompson's avatar
      Reset point ID list when appending polyhedra. · fc0ee6cf
      David Thompson authored
      Change-Id: I0d5ef2af0323e57ef08de64235a34f4672f3bb9f
    • David Thompson's avatar
      Fix polyhedron-ray intersection... · ee3a7c30
      David Thompson authored
      ... by properly resizing point and point-id arrays for
      each polygonal face before intersecting with the ray.
      Change-Id: I24b7053428e50152c8e10fcb240a56a924e772ce
    • David Thompson's avatar
      Don't forget to copy polyhedron cell-data... · 31d834a4
      David Thompson authored
      ... when appending datasets.
      Also, clear out point-list when GetFaceStream fails.
      Change-Id: Icf502a95ec8bfacc634291c6ef749d3cbdff5df9
    • Zack Galbreath's avatar
      Merge topic 'CMake-Immediate' into master · 8497b42d
      Zack Galbreath authored
      1f965eec ENH: Remove superfluous IMMEDIATE option from CMake CONFIGURE_FILE
    • Brian Helba's avatar
      ENH: Remove superfluous IMMEDIATE option from CMake CONFIGURE_FILE · 1f965eec
      Brian Helba authored
      In CMake 2.0 and higher (since 2005), the IMMEDIATE option to
      CONFIGURE_FILE is implicit:
      Change-Id: I60e48f767d33e0a364dfc9cafd9821836b2f3f43
  6. 11 Aug, 2013 1 commit
    • Berk Geveci's avatar
      Replaced use of critical section with runtime initialization. · 9b06556c
      Berk Geveci authored
      vtkPolyData class uses a class variable that it was initializing
      lazily - which required a mutex to do in a thread safe way. Since
      mutexes are expensive, I replaced this with a run-time
      initialization, which is thread safe and much less expensive.
      Change-Id: I34bcee4b4e5f105a00ebd87129a5debff0170705
  7. 09 Aug, 2013 1 commit
    • David Thompson's avatar
      Fix problems processing polyhedron connectivity. · 2f4a9cad
      David Thompson authored
      Several filters incorrectly process polyhedra because of
      misunderstandings of how their connectivity is stored or
      simple math errors in computing connectivity offsets.
      Fix them.
      Change-Id: Icce8becea56567b869277d0d265e3f9092b7eb98
  8. 08 Aug, 2013 3 commits
    • David Gobbi's avatar
      Merge topic 'arrayweights-copy-ctor' into master · 004944f0
      David Gobbi authored
      e6d1e307 Add a copy constructor to vtkArrayWeights.
    • David Gobbi's avatar
      Add a copy constructor to vtkArrayWeights. · e6d1e307
      David Gobbi authored
      The vtkArrayWeights header declares a destructor, therefore it
      must also declare a copy constructor and assignment operator
      according to the "rule of three".  In C++11 mode, clang gives
      a warning when the "rule of three" is broken.
      Change-Id: I0b5041100d5f00cdb0ddd709853ee49b151bba8e
    • David Gobbi's avatar
      Perform explicit list deletion in CCSEdgeLocator. · 9b46579d
      David Gobbi authored
      The vtkClipClosedSurface helper class vtkEdgeLocatorNode would
      delete the whole node list in its destructor.  This is a bad
      idea, because because if the implicit copy constructor is called
      on a node, and the copy destructs, then the list will partially
      destruct.  Add a FreeList() method that is called explicitly when
      the list is to be freed.
      Change-Id: I8945dc1682b6a70f66561b1f4155ed22be6fb0d6
  9. 07 Aug, 2013 1 commit