Bug fixes made since ParaView 5.9.0 are listed in this document. The full list of issues addressed by this release is available here.

Catalyst

• Fixed Catalyst build error on Windows with CMake 3.15 and before (details).
• Fixed Catalyst V2 API to avoid adding the same timestep twice and getting warnings about it (details).
• Fixed issue with Catalyst V2 scripts missing time at first call (details).
• Ensure all fields are requested in time step 0. When using V2 Catalyst Python scripts, since the script is not imported when the first RequestDataDescription is called, we need to ensure all fields/meshes are requested. The comment already mentioned that, we were just toggling all the necessary flags for the fields to be requested correctly (details).

Reader changes

• Added HDF5-based XRage reader. This reader loads HDF5 data produced by xRage (details).
• Several fixes for the PIO reader. Default selected variables have been fixed (details). Fixed a bug where sometimes a variable was determined to be a string when it was actually numerical data. Enabled comments that start with '!' and '#'. Improved some error messages (details).
• EnSight Gold reader now ignores 'maximum time steps:' lines in .case files (details).
• Fixed bug in NetCDF reader where only the first array would be displayed in the array selection (details).
• Motion FX reader now has a universal transform (details).

Ray tracing fixes

• Ray tracing is initialized upon request, not at program startup, to avoid the cost of initialization and potentially platform-specific problems encountered when ray tracing is not requested (details).
• On macos, OSPRay now detects whether it is running on Rosetta and reports that it is not available instead of crashing (details).
• Silenced a JSON parser warning issued when an OSPRay-enabled client connected to a server built without OSPRay support (details).
• Fixed a bug where once made visible, axes labels would always be visible under ray tracing (details).
• Reduce severity of VisRTX warning messages to avoid extraneous output (details).

Bug fixes

• The Annotate Global Data filter now validates the format string and warns when the format is not valid for the selected array type (details).
• The PassArrays filter now no longer passes "vtkGhostType" arrays if they are not in the array selection. Previously, the PassArray filter would pass "vtkGhostType" arrays even if they were excluded from the "*DataArrays" properties (details).
• A warning that could appear during interactive cell selection with labels turned on has been fixed (details).
• Fixed a segmentation fault that could occur when disconnecting from a remote server (details).
• cinemasci: use relative paths when loading modules (details).
• The Surface LIC representation adds backface representation support (details).
• Fixed Surface LIC representation when the data has a Normals array (details).
• Fixed memory leak in value pass (details).
• Subdivision for non-linear curves has been implemented (details).
• Array buffer and element array buffer bindings are now cached in OpenGLState.
• Disabled progress events in the "Projected tetra" volume rendering algorithm. This speeds up rendering and prevents problems caused by code executed when progress events are invoked (details).
• Re-enable point merging when requested for unstructured grids (details).
• Updated vtkSphereTree to use thread-safe method to get cell IDs (details).
• Fixed some compiler errors on IBM XLC compilers (details).
• NVIDIA IndeX plugin bugfixes (details)
  • Fixed loading of the NVIDIA IndeX plugin in pvbatch when PV_PLUGIN_PATH is set (details).
  • Fixed "z-buffer precision" error message in the NVIDIA IndeX plugin when the Axes Grid was enabled.
  • Fixed NVIDIA IndeX rendering when the color map was rescaled to a custom range.
  • Runtime compilation of NVIDIA IndeX Visual Elements is now faster.
  • Prevent runtime issues in the NVIDIA IndeX library when a stub version of libcuda.so.1 is in the library search path.