In commit fc3a1cbdd8 (CompilerID: Compiler extensions default detection,
2021-05-29, v3.22.0-rc1~52^2~2), extensions were assumed to be `OFF`
for Clang targeting the MSVC ABI. However, the `clang` and `clang++`
tools both seem to have extensions enabled by default even when
targeting the MSVC ABI.
This can be observed with the `RunCMake.CompileFeatures` test.
It fails with the GNU-like `clang/clang++` front-end, but removing
the above special case makes it pass. The test passes either way
with the MSVC-like `clang-cl` front-end.
This brings the patch step into line with all the others which already
had their own `USES_TERMINAL_<step>` keyword. All steps (including
patch) already have their own `LOG_<step>` keyword too, so the lack of
`USES_TERMINAL_PATCH` was inconsistent.
Before the fix the gtest_discover_tests() function would strip the
user data in test parameters (everything to the right of GetParam())
of spaces. Now the parameters aren't altered in any way.
Fixes#23058
Commit 14d8a276 (CUDA: Support nvcc 11.5 new -arch=all|all-major flags,
2021-08-17) added all and all-major options to CUDA_ARCHITECTURES. These are
fairly generic and likely to see real-world use by distributors. Thus it's
desirable to support these also for Clang and older NVCC versions.
The supported architectures are dependent on the toolkit version. We determine
the toolkit version prior to compiler detection. For NVCC we get the version
from the vendor identification output, but for Clang we need to invoke NVCC
separately.
The architecture information is mostly based on the Wikipedia list with the
earliest supported version being CUDA 7.0. This could be documented and
expanded in the future to allow projects to query CUDA toolkit version and
architecture information.
For Clang we additionally constrain based on its support.
Additionally the architecture mismatch detection logic is fixed, improved and
updated for generic support:
* Commit 01428c55 (CUDA: Fail fast if CMAKE_CUDA_ARCHITECTURES doesn't work
during detection, 2020-08-29) enabled CMAKE_CUDA_COMPILER_ID_REQUIRE_SUCCESS
if CMAKE_CUDA_ARCHITECTURES is specified. This results in
CMakeDetermineCompilerID.cmake printing the compiler error and our code for
presenting the mismatch in a user-friendly way being useless. The custom
logic seems preferable so go back to not enabling it.
* Commit 14d8a276 (CUDA: Support nvcc 11.5 new -arch=all|all-major flags,
2021-08-17) tried to support CMP0054 but forgot to add x to the interpolated
result. Thus the conditions would always evaluate to false. This is fixed as
a byproduct of removing NVIDIA specific checks, improving the error message
and replacing architectures_mode with a simpler architectures_explicit.
Visual Studio support omits testing the flags during detection due to
complexities in determining the toolkit version when using it.
A long-term proper implementation would be #23161.
Implements #22860.
Update the imported target locations added by commit a8e0a6b3e4
(FindHDF5: Port changes from VTK, 2020-06-10, v3.19.0-rc1~312^2~1) to
use config-specific locations.
Issue: #21637
The imported targets added by commit a8e0a6b3e4 (FindHDF5: Port changes
from VTK, 2020-06-10, v3.19.0-rc1~312^2~1) to use config-specific
locations set the `IMPORTED_IMPLIB` target property. That property has
no meaning for an imported library with type `UNKNOWN`. Drop it to
avoid confusion.
The possible executable names list always starts with a plain "ruby"
binary, which is scanned for first. If that does exist (which satisfies
find_program), but doesn't match the requested version, the executable
will be dismissed by _RUBY_VALIDATE_INTERPRETER (which is sane), but
searching also stops.
Fix that by keeping the search going, removing items from the list as we
go, until find_program returns a path that _RUBY_VALIDATE_INTERPRETER is
happy with or the list is empty.
Commit a9073db7 (IAR: update language specification detection, 2021-09-23)
added usage of the if() IN_LIST operation and forgot to account for it not
being available in CMP0057 OLD mode.
Push and temporarily enable the policy. Also avoid the unnecessary temporary
variable for the list.
Fixes#23147.
Use of `-fvisibility=hidden` fails:
unsupported option '-fvisibility=hidden' for target 'powerpc-ibm-aix7.2.0.0'
Apply the change from commit 4feba34d02 (GNU: Do not use -fvisibility on
AIX or HP-UX, 2016-09-03, v3.7.0-rc1~173^2~2) to IBMClang also.
Issue: #23157
8c1731546c Help: Add release note for IBM Open XL C/C++ compiler support
24da80b70a Utilities: Suppress warnings in third-party code with IBMClang
6da99e671c IBMClang: Add support for IBM Open XL C/C++
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !6785
We require CUDA_ARCHITECTURES to be set for targets (see CMP0104). If not set
anything after compiler detection such as ABI detection will fail to generate.
This means we need to error if CMAKE_CUDA_ARCHITECTURES is not set to a valid
value as a result of compiler detection.
Currently we fail to issue the error if compiler detection failed and the ID is
unset. In such a case we won't define detected_architecture making the code
responsible for the error unreachable.
Simplify the detection of architectures used during compiler detection by
always detecting all of them, which enables us to simply the check in the
"default to compiler" path if CMAKE_CUDA_ARCHITECTURES is empty.
As a result we need to move the error checking and CMAKE_CUDA_ARCHITECTURES=OFF
handling fully into the default path thus simplifying the code and unifying the
code paths for NVCC and CUDA.
This also happens to fix:
1. CMAKE_CUDA_ARCHITECTURES=OFF on Clang.
2. A theoretical issue of a compiler defaulting to multiple architectures.
I've additionally added printing of the compiler output along the error to
better reveal possible underlying compiler/system configuration issues.
Fixes#23010.
If empty we otherwise treat it the same as unset in most places, but still end
up failing eventually with a confusing "Failed to find a working CUDA
architecture".
This also detects some other basic invalid ones (e.g. "al").
The code now matches what the comment describes.
This mistake seems to have been present since the initial introduction in
commit 5df21adf (CUDA: Add support for Clang compiler, 2020-05-07).
5efb6fb516 FindThreads: Honor THREADS_PREFER_PTHREAD_FLAG when pthread is found in libc
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !6906
The `-pthread` flag tells the compiler/linker to link to additional
libraries needed for thread support (e.g. libatomic on riscv64). The
flag therefore should be used if requested using
`THREADS_PREFER_PTHREAD_FLAG` also when the pthread functions are
found in libc.
Move the `static` added by commit 8cce0ad32f (FindMPI: Fix missing
static warning in the mpi test source, 2021-06-30, v3.22.0-rc1~502^2) to
the beginning of the line.
Fixes: #23141
ee1396e29e CMP0128: Add flag in OLD mode even when standard matches the default
b2c25de8e0 CMP0128: Avoid test code duplication
3a089cd256 CMP0128: Prefix test names with mode
e13dd52535 XL: Detect default extensions mode for legacy compiler
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !6890
This makes the scripts easier to work on, since the separate files
don't require the extra level of escaping that the inlined code did.
This also means the scripts can be rendered with appropriate
syntax highlighting in IDEs, etc.
This is in preparation for moving more inline content to separate
scripts as well. Having them in a separate subdirectory is consistent
with other modules and will make them easier to find and work with.
Since version 12.0 the GCC Fortran compiler has implemented "WG5/N1942",
which causes, if link-time opmization is enabled, obfuscation of hard-coded
string values in the compiler objects and its resulting ELF-binaries.
This causes the CMake-internal detection of the mangling scheme for the
naming of subroutines to fail. Thus we must ensure to have any link-time
optimization features to be disabled on the executable file we perform the
detection on.
The static libraries, however, must be build with LTO and non-LTO objects,
as that will ensure the verify step will operate on IPO objects, if building
those is requested by the system compiler flags.
Fixes: #23123
Signed-off-by: Björn Esser <besser82@fedoraproject.org>
Since macOS 12.0, the ``hdiutil udifrez`` and ``hdiutil udifderez``
commands to embed and extract resources in a disk image are deprecated.
The CPack DragNDrop Generator uses these to attach the SLA specified by
the `CPACK_RESOURCE_FILE_LICENSE` option. Since that option is shared
by multiple CPack generators, we cannot deprecate it. Instead, add an
explicit option to control the behavior. This will give projects a way
to package on future macOS versions that remove the commands.
In order to provide a long-term transition away from attaching SLAs to
disk images, update `cpack` to default this behavior to OFF. To retain
compatibility for CMake projects, teach the CPack module to default the
option to ON. Later a policy can be added to change the default.
Issue: #22978
