[VTK/vtk-m] Update VTK to 9.0 and add vtk-m#11148
[VTK/vtk-m] Update VTK to 9.0 and add vtk-m#11148strega-nil merged 32 commits intomicrosoft:masterfrom
Conversation
|
Hi @Neumann-A |
strega-nil
left a comment
strega-nil
left a comment
There was a problem hiding this comment.
There's some weird things I saw; I'm not sure if these are an issue, but I'd like to at least understand the reasoning.
strega-nil
added
waiting for response
and removed
info:needs-maintainer-attention
Co-authored-by: nicole mazzuca <mazzucan@outlook.com>
strega-nil
left a comment
strega-nil
left a comment
There was a problem hiding this comment.
Alright, looks good to me! Thanks @Neumann-A :)
|
Could some of these patches be sent to our issue tracker? Some look weird or unnecessary (or have more proper fixes if there are things wrong) to me. Thanks. |
lets just say more general fixes. Most findModule.patches are with the background knowledge that we are within VCPKG and have certain things with 100% certainty available. I also already opened up Issue on VTKs Gitlab which I thought relevant: If you have questions about any of the applied patches feel free to ask. |
| + add_library(LibHaru::LibHaru UNKNOWN IMPORTED) | ||
| + set_target_properties(LibHaru::LibHaru PROPERTIES | ||
| + INTERFACE_INCLUDE_DIRECTORIES "${LibHaru_INCLUDE_DIR}") | ||
| + if(LZMA_LIBRARY_RELEASE) |
There was a problem hiding this comment.
Yes copy paste. Need to correct that in the future
There was a problem hiding this comment.
@mathstuf: And now I know why I patched it:
-- Found LibHaru: optimized;D:/para/installed/x64-windows/lib/libhpdf.lib;debug;D:/para/installed/x64-windows/debug/lib/libhpdfd.lib (found suitable version "2.4.0-dev", minimum required is "2.4.0")
CMake Error at CMake/vtkDetectLibraryType.cmake:31 (message):
Unparsed arguments for vtk_detect_library_type:
D:/para/installed/x64-windows/lib/libhpdf.lib;debug;D:/para/installed/x64-windows/debug/lib/libhpdfd.lib
Call Stack (most recent call first):
CMake/FindLibHaru.cmake:48 (vtk_detect_library_type)
D:/para/scripts/buildsystems/vcpkg.cmake:329 (_find_package)
CMake/vtkModule.cmake:4134 (find_package)
CMake/vtkModule.cmake:4690 (vtk_module_find_package)
CMake/vtkModule.cmake:4564 (vtk_module_third_party_external)
ThirdParty/libharu/CMakeLists.txt:1 (vtk_module_third_party)
There was a problem hiding this comment.
Oh, that's bad interaction with select_library_configurations. Hrm. I'll make a fix on VTK's side. Thanks.
There was a problem hiding this comment.
| + IMPORTED_IMPLIB_DEBUG "${LZMA_LIBRARY_DEBUG}") | ||
| + endif() | ||
| + | ||
| + # Guard against possible stupidity of VTK reading only LOCATION without configuration |
There was a problem hiding this comment.
VTK definitely shouldn't be doing this.
There was a problem hiding this comment.
that is why "possible" is there. But it is not only VTK it also includes all possible consumers of VTK. I did a quick search and could not find anything regarding this within VTK but I have seen real strange things in CMakeLists.txt of ports. Like getting the IMPORTED_LOCATION from a target. Extracting the filename. Removing any PREFIX and SUFFIX from it and then using the result in target_link_libraries..... When I wrote the patch I probably just encountered that problem in another port and thought about guarding against that behavior proactively.
There was a problem hiding this comment.
Ah, yes, those shenanigans should be nuked from orbit on sight :) . However, the wording makes it seem like VTK's at fault here which is why I noticed it. Might I suggest stupidity of various projects? Issues when those are found are nice (feel free to Cc me if you need help in discussions with upstreams).
| @@ -0,0 +1,269 @@ | |||
| # Distributed under the OSI-approved BSD 3-Clause License. See accompanying | |||
There was a problem hiding this comment.
We can backport CMake's FindPostgreSQL into VTK; it looks like that's what this is.
There was a problem hiding this comment.
Yes, just a copy of FindPostgreSQL from CMake. I thought you wold be updating that automatically with the VTK release. Apparently not.
There was a problem hiding this comment.
I only pull down modules from CMake I know we need to backport for specific use cases.
| set(_vtk_python_genex_include_directories | ||
| "$<TARGET_PROPERTY:${_vtk_python_target_name},INCLUDE_DIRECTORIES>") | ||
| + set(_vtk_python_genex_interface_include_directories | ||
| + "$<TARGET_PROPERTY:${_vtk_python_target_name},INTERFACE_INCLUDE_DIRECTORIES>") |
There was a problem hiding this comment.
This is actually required to run ParaView
https://gitlab.kitware.com/paraview/paraview/-/issues/19515
this patch is actually for 8.9 the code as since then slightly changed in this location.
Updated code will be available in
#9960 as soon as I am able to run ParaView.exe locally without it crashing due to missing wrapping.
|
|
||
| -list(INSERT 0 CMAKE_MODULE_PATH | ||
| - "${VTK_SOURCE_DIR}/ThirdParty/vtkm/vtkvtkm/vtk-m/CMake") | ||
| +find_package(VTKm CONFIG REQUIRED) |
There was a problem hiding this comment.
This should be vtk_module_find_package so that VTK finds VTKm for consumers if necessary.
| new file mode 100644 | ||
| index 000000000..c0fe00c0e | ||
| --- /dev/null | ||
| +++ b/CMake/FindVTKm.cmake |
There was a problem hiding this comment.
This seems redundant given the CONFIG passed to the find_package(VTKm)?
There was a problem hiding this comment.
One of the problems I had was to understand when VTK requires a FindModule.cmake for an external dependency and when not. I think I got an error without it telling me that there is no module for it and I just made one minimal on up. Probably because I did not use vtk_module_find_package ?
There was a problem hiding this comment.
We just let find_package do its precedence. CMake 3.17's --debug-find should help figure out why it isn't being found (or at least where CMake is looking). I guess VTKm's config file was not found out-of-the-box without this? All vtk_module_find_package does is let the module system know it was looked for, does sanity checks for a few things, and puts the find_package into the vtk-config code path.
| STRING(REGEX REPLACE "/W[0-4]" "/W4" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}") | ||
| else() | ||
| - set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /W4") | ||
| + set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /W4 /bigobj") |
There was a problem hiding this comment.
I think VTKm is working on a fix for this. @robertmaynard
| - find_package(VTKm | ||
| - PATHS "${CMAKE_CURRENT_LIST_DIR}/vtkm" | ||
| - NO_DEFAULT_PATH) | ||
| + find_package(VTKm CONFIG REQUIRED) |
There was a problem hiding this comment.
This whole block can probably go away if vtk_module_find_package is used.
|
@Neumann-A Thanks. I've done a pass over the code here. |
6 participants
closes #11135
most changes are extracted from #9960
should close the following:
closes #9211
closes #9200
closes #9213
closes #9023
might close #9322
vtk-m did not test any features.
vtk build with
vtk[atlmfc,core,paraview,python,qt,vtkm]:x64-windows