742ff97f80 Refactor language standard computation
0892c798f7 cmMakefile: Change CompileFeatureKnown to take target name instead of target
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !4803
Clang isn't very good at finding the installed CUDA toolkit.
The upstream recommendation is that we should pass the toolkit explicitly.
Additionally:
* Avoids Clang having to search for the toolkit on every invocation.
* Allows the user to use a toolkit from a non-standard location by simply
setting CUDAToolkit_ROOT. The same way as with FindCUDAToolkit.
Clang wants the directory containing the device library and version.txt as the
toolkit path.
We thus pass the newly introduced CUDAToolkit_LIBRARY_ROOT as the toolkit path.
We save CUDAToolkit_ROOT_DIR and CUDAToolkit_LIBRARY_ROOT on Clang to have them
available in try_compile() and avoid unnecessary re-searching or a possibly
different installation being found in FindCUDAToolkit.
This however means that the selected toolkit can't be changed after the initial
language enablement.
We now determine CUDA compiler ID before doing actual detection, as we don't
want to spend time finding the CUDA toolkit for NVIDIA.
Implements #20754.
b1243201e7 CUDA: Don't RunCMake generate separable compilation tests on Clang
1b4c690543 CUDA: Throw error if CUDA_SEPARABLE_COMPILATION is ON when using Clang
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !4827
9f4eb352fe Unity Builds: Support explicit specification of sources to groups
b00585adcc Unity: Refactor implementation to make it easier to extend
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !4716
Since commit a55df20499 (Multi-Ninja: Add precompile headers support,
2020-01-10, v3.17.0-rc1~136^2) the re-used PCH file object is added
once for each configuration to flags used for all configurations.
Put it in the flags for only the corresponding configuration instead.
Fixes: #20680
Since commit a9b41195d2 (Handle multi-arch sysroots on Apple platforms,
2019-07-26, v3.17.0-rc1~287^2), CMake supports "fat" builds with
different sysroots. Those are passed to the compiler with the
`-Xarch_<xyz>` parameter.
Unfortunately this breaks the Compiler Cache (ccache) because
it does not support those compiler flags:
https://github.com/ccache/ccache/blob/v3.7.9/src/ccache.c#L2700-L2705
Restore the caching ability for certain "fat" build configurations (e.g.
`arm64` and `armv7`) where the sysroot is the same for all selected
architectures and thus a plain `-isysroot` parameter could be used.
properties LINK_OPTIONS and INTERFACE_LINK_OPTIONS are propagated
to the device link step.
To control which options are selected for normal link and device link steps,
the $<DEVICE_LINK> and $<HOST_LINK> generator expressions can be used.
Fixes: #18265
Simplifies CUDA target architecture handling.
Required for Clang support as Clang doesn't automatically select a supported architecture.
We detect a supported architecture during compiler identification and set CMAKE_CUDA_ARCHITECTURES to it.
Introduces CMP0104 for backwards compatibility with manually setting code generation flags with NVCC.
Implements #17963.
Adds a flag to indicate that pipe output from a custom command should be
interpreted as UTF-8 encoded. This change does not introduce a public
way to set the flag, but generators that create internally-generated
commands know if they are calling cmake, which uses UTF-8 pipes.
MSBuild added support for interpreting output of PreBuildEvent,
PreLinkEvent, PostBuildEvent, and CustomBuildStep as UTF-8. This change
will appear in Visual Studio 16.6 Preview 3. It is opt-in, and you need
to add the StdOutEncoding tag. MSBuild treats these as property bags so
if we emit the tag for earlier versions of Visual Studio it would be
safely ignored. This change emits the StdOutEncoding tag and sets it to
UTF-8 whenever the custom command UTF-8 pipe flag is set. This fixes
globalization issues when the output from cmake contained characters
that required MSBuild to interpret as UTF-8 before displaying them.
Arguably, many of these are bugs in `clang-tidy`. An if/else tree with
other conditionals between cloned blocks may be relying on the
intermediate logic to fall out of the case and inverting this logic may
be non-trivial.
See: https://bugs.llvm.org/show_bug.cgi?id=44165
The `CMAKE_OSX_ARCHITECTURES` value is not used directly by generators.
It is used to initialize a per-target `OSX_ARCHITECTURES` property, but
that property can also be set explicitly by project code to a subset of
the full list of architectures. In order to handle this case, construct
a mapping from each `CMAKE_OSX_ARCHITECTURES` entry to the corresponding
`CMAKE_APPLE_ARCH_SYSROOTS` entry by name. Use the mapping to find the
sysroot for each entry in `OSX_ARCHITECTURES` for a given target.
If `CMAKE_APPLE_ARCH_SYSROOTS` does not have the same length as
`CMAKE_OSX_ARCHITECTURES`, error out early rather than risking a crash
or assertion failure.
Fixes: #20534
Use `<arch>-SDK-NOTFOUND` instead of an empty string as a placeholder in
`CMAKE_APPLE_ARCH_SYSROOTS` for architectures whose SDK is not found.
This ensures the length of `CMAKE_APPLE_ARCH_SYSROOTS` matches the
length of `CMAKE_OSX_ARCHITECTURES`. It also makes the missing SDKs
more visible in the value.
Issue: #20534
On machines where the gnu bin utils are prefixed, or suffixed
the file(GET_RUNTIME_DEPENDENCIES ) command would fail without
explicitly setting the location of objdump.
Now we pre-populate the variables used to find objdump based
on the gnu bin utils, so that these use cases are better supported
