[FEA] Bump minimum required CMake version to 3.22.3

[robertmaynard]

Is your feature request related to a problem? Please describe.
We should continue leveraging latest CMake improvements, and therefore should review
the improvements that 3.21 and 3.22 will bring to rapids-cmake.

Describe the solution you'd like
I would like to see 3.22 become the required CMake version for rapids-cmake 22.X versions

Additional context
CMake 3.21 would bring in the following improvements:

• set behavior in regards to local/cache variables is not consistent!!! This is big quality of life improvement
• WriteBasicConfigVersion supports CalVer strings, so workaround in rapids-export can be dropped
• Less false positive error reports from FindThreads
• nicer ways to specify toolchain and install directory from the command line
• improvements to install() to copy runtime dependency will need to be reviews and we can see if they can be used when thirdparty dependencies are missing install rules
• Cleanup of foreach() and variable leaking. Most likely will need to update rapids-cmake in places due to this

CMake 3.22 would bring in the following improvements:

• find_library: Infer library prefix and suffix when in script mode. This is important to rapids-cmake as we hook up to projects before the project call which is analogous to script mode.
• compile_features ignore features for non-enabled languages. This allows projects to express a cxx_std_17 requirement and not fail when used by a project that doesn't enable CUDA.
• CUDA language level isn't inferred by the C++ language level
• GNUInstallDirs now aware of conda lib directory requirements

[kkraus14]

Potential issue with moving to CMake 3.22:

• Apache Arrow currently pins the AWS SDK to to 1.8.x due to a bug: Aws::Crt::Io::ClientBootstrap destructor may launch thread at process exit, and crash aws/aws-sdk-cpp#1809
• AWS SDK 1.8.x and earlier 1.9.x versions have a bug with CMake 3.22 that prevents it from being used: Configuration file cannot be used with CMake 3.22.0 aws/aws-sdk-cpp#1820
• libcudf depends on Apache Arrow

[robertmaynard]

@kkraus14 I believe that the 3.22.0 regression that hit AWS SDK was found during the RC cycle. And the fixe is in 3.22.0 ( aws/aws-sdk-cpp#1810 ).

Are you able to confirm this by building the AWS SDK with 3.22.0?

[kkraus14]

@kkraus14 I believe that the 3.22.0 regression that hit AWS SDK was found during the RC cycle. And the fixe is in 3.22.0 ( aws/aws-sdk-cpp#1810 ).

Are you able to confirm this by building the AWS SDK with 3.22.0?

Yea I can confirm that 3.22 breaks building AWS SDK 1.8.x. I believe this is fixed in more recent 1.9.x releases, but in the issue linked above it also confirms that using 3.22 breaks building AWS SDK 1.9.152, so I assume any version prior to that will be broken.

[robertmaynard]

Weird I thought that the following fix for 3.22.0 would fix the problem and only 3.22.0-rc1 and rc2 would have the regression.

https://gitlab.kitware.com/cmake/cmake/-/merge_requests/6442

[kkraus14]

Weird I thought that the following fix for 3.22.0 would fix the problem and only 3.22.0-rc1 and rc2 would have the regression.

https://gitlab.kitware.com/cmake/cmake/-/merge_requests/6442

I've only tested CMake 3.22.2 since that's what's in conda-forge but can confirm that it reproduces the bug above.

[robertmaynard]

Weird I thought that the following fix for 3.22.0 would fix the problem and only 3.22.0-rc1 and rc2 would have the regression.
https://gitlab.kitware.com/cmake/cmake/-/merge_requests/6442

I've only tested CMake 3.22.2 since that's what's in conda-forge but can confirm that it reproduces the bug above.

I was able to reproduce and you are correct this was never properly fixed. I have re-opened the CMake issue and we should be able to get a fix for 3.22.2+

[kkraus14]

Thanks @robertmaynard!

[vyasr]

rmm-python would also benefit from this bump since when we switch to scikit-build we are currently required to include CUDA as one of the project languages due to the way that the C++ rmm component adds libcudacxx's c++17 features to its interface.

[robertmaynard]

Thanks @robertmaynard!

This regression has been fixed in 3.22.3 patch release and the upcoming 3.23.0.

CMake does consider this to be invalid while syntax, and 3.24 will have a policy that makes this an explicit syntax error ( https://gitlab.kitware.com/cmake/cmake/-/merge_requests/7045 )