Update VTK from slicer-v9.2.20230607-1ff325c54 to a recent version

[jcfr]

Is your feature request related to a problem? Please describe.

The version currently integrated in Slicer corresponds to the branch Slicer/VTK@slicer-v9.2.20230607-1ff325c54-2¹ and is based of Kitware/VTK@1ff325c from June 2023.

While selected fixes and improvement have been integrated through the pull requests listed below, we are missing out on additional fixes and improvements as well as increasing the maintenance burden as time goes.

The last VTK update was done through this pull request:

• ENH: Update VTK from 9.1.20220125 to 9.2.20230607 #7010

Describe the solution you'd like

We should create a new branch names like slicer-v9.3.2024MMDD-abcdefghi with only a few Slicer specific changes.

⚠️ While working on #7487, a performance regression related to vtkRender::GetZ() was identified, we should make sure to review its status when working on updating VTK:

• ENH: Update VTK backporting selected Rendering, OpenVR and OpenXR patches #7487 (comment)
• https://gitlab.kitware.com/vtk/vtk/-/merge_requests/10391#note_1459483
• [slicer-v9.2.20230607-1ff325c54-2] Backports related to Rendering & VR VTK#46 (comment)

Describe alternatives you've considered

NA

Additional context

Here are the pull requests backporting changes (from the most recent to the oldest) since the last update through #7010:

• ENH: Update VTK backporting selected Rendering, OpenVR and OpenXR patches #7487
• BUG: Update VTK to to fix vtkSMPToolsAPI static initialization order issue #7453
• ENH: Update VTK and vtkAddon for ColorizeVolume module in Sandbox extension #7305
• BUG: Update VTK to fix segfault when enabling volume rendering #7286
• BUG: Update VTK to fix RGBA volume rendering #7274
• BUG: Update VTK to remove unnecessary messages in vtkTransformPolyDataFilter #7227
• ENH: Update VTK backporting RenderingVR and RenderingOpenXR fixes #7195
• ENH: Update VTK to backport OpenXR error handling improvements #7193
• ENH: Update VTK to support distribution of OpenXRRemoting in extension #7190
• BUG: Update VTK to fix regression in Dynamic modeler Curve cut #7140
• COMP: Update VTK to workaround external project output flagged as error #7104
• BUG: Update VTK to fix regression preventing from loading older vtp files #7080

Below a comment originally posted by @lassoan in #7487 (comment)

The change in Kitware/VTK@cff3d9a60 is not very useful, as it is extremely expensive to instantiate a new hardware picker for every position. This is also something that anyone can easily do outside the renderer and actually do that picker more efficiently (e.g., cache the Z buffer content). I think the original behavior of vtkRenderer::GetZ should be restored, and its documentation should be amended with a note that the method should be called at the right time (when the Z buffer is not cleared).

The proposed GetZ implementation might be considered to be added to the vtkRenderer as some convenience function if Paulo Carvalho found it useful, but it should be reviewed, because for example creating a new hardware selector from scratch for every point position looks terribly wasteful, especially when the renderer already has a cached selector (and maybe vtkRenderer::PickProp could be enough). I think overall what happened is not Paulo's fault but that VTK developers did not review his proposed change for a really long time and then just rushed a merge without anyone really thinking about the implications of this change. Probably the change looked insignificant because nobody realized that the method was heavily used in vtkWorldPointPicker and vtkSelectVisiblePoints (and it worked correctly there because the method was used at the right time).

Footnotes

1. https://github.com/Slicer/VTK/tree/slicer-v9.2.20230607-1ff325c54-2 ↩