The upstream freetype cmake package uses generator expressions and
imported target to express its dependencies. That's fine on its own.
However, our use of it since commit d83d925045 (FindFreetype: use
`freetype-config.cmake` if available, 2023-09-06, v3.28.0-rc1~130^2)
causes the `FREETYPE_LIBRARIES` result variable to reference those
imported targets and create a package-level dependency for clients that
did not exist previously. Revert that change for now, along with its
follow-up fixes. Further investigation will be needed to solve the
motivating use case another way.
Fixes: #25635
In commit 4d90f65b24 (CUDA: Update set of architectures supported by
CUDA 12, 2022-12-19, v3.26.0-rc1~120^2) we incorrectly guarded dropping
removed architectures by compiler id, but it is actually the CUDA 12
toolkit itself that removed support.
Xcode has been reported to not work at all with this when
ExternalProject projects share download files.
Revert commit 872daff159 (ExternalProject: declare byproducts for the
download step, 2023-09-21, v3.28.0-rc1~27^2) pending further
investigation.
Fixes: #25525
Since commit f73a5bfaa9 (UsePkgConfig: Replace exec_program() with
execute_process(), 2023-07-24, v3.28.0-rc1~321^2~6), trailing whitespace
was introduced into libraries linked, which violates CMP0004.
04959e0285 Tests: Verify CUDA::cupti existence on a subset of machines
09a3c8b46b CUDAToolkit: Correctly search all include paths from compiler
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !9072
Update the list of known versions.
Run the command
cmake -DBOOST_DIR=/path/to/boost_1_84_0 \
-P Utilities/Scripts/BoostScanDeps.cmake
to extract dependencies from the 1.84.0 source tree.
They are the same as 1.83's dependencies, so just update
the version check for warning about newer versions.
Fixes: #25499
Fix logic from commit fff5c1507e (FindMatlab: refactor: use registry
query instead of execute_process, 2023-09-14, v3.28.0-rc1~82^2~7)
to work when multiple versions are found in the registry.
Issue: #25497
Since commit 6a3059e66f (FindTIFF: bridge `tiff-config` into
FindTIFF-compatible interface, 2023-09-14, v3.28.0-rc1~87^2) we use the
`if(IN_LIST)` test that required CMP0057. Expand the scope over which
we enable that policy explicitly.
Issue: #25485
Follow up commit d892dedf22 (FindFreetype: always find the config module
quietly, 2023-12-13) with a fix to the FPHSA call that reports success.
Fixes: #25485
Fix logic from commit dc9d9589e4 (FindMatlab:WIN32: return full Matlab
version when found via registry, 2023-09-14, v3.28.0-rc1~82^2~2) to
avoid assuming that a registry entry always exists and is non-empty.
Fixes: #25497
Support multi-config-providing and `IMPLIB`-using deployments with the
`tiff-config` trampoline code. Follow the pattern used in `FindFreetype`
by commit ae9890cd36 (FindFreeType: consider `IMPLIB`-using platforms,
2023-10-26, v3.28.0-rc4~10^2~3).
See: #25485
When the config module is not present, a spurious "tiff not found" is
output before the pre-existing logic is used. Instead, silence the
module and use FPHSA to report as-if `TIFF` did the search.
Fixes: #25485
When the config module is not present, a spurious "freetype not found"
is output before the pre-existing logic is used. Instead, silence the
module and use FPHSA to report as-if `Freetype` did the search.
See: #25485
Fixes#25484
PR !8835 failed to update the CUPTI header searches to use the
new internal FindCUDAToolkit search variables. This caused the
CUDA::cupti target to always not exist.
Since commit 0744c02e24 (FindCUDAToolkit: targets pointing to stubs now
use IMPORTED_IMPLIB, 2023-07-24, v3.28.0-rc1~309^2) we recognize CUDA
stub libraries and represent them in a special way. However, the logic
only works on the first configuration of a build tree when the libraries
are first found. Once the results are cached, we incorrectly revert to
the non-stub representation.
Fix this by recognizing stub libraries based on their path instead.
In commit 26bf32cdc6 (LLVMFlang: Add support for targeting MSVC ABI on
Windows, 2023-09-28, v3.28.0-rc1~10^2) we incorrectly recorded `-g` as
supporting the `ProgramDatabase` format, but it is actually `Embedded`,
matching Clang.
In order to support easy integration with C and C++ projects that use
the `.pdb` debug formats, pretend LLVMFlang supports them and just don't
actually emit any debug information.
Issue: #24840
RPM supports either whitespace with quoting or globbing without quoting.
Prior to RPM 4.19 it accepted globbing in quotes, but it only globbed
correctly without whitespace, where quoting was not necessary anyway.
Starting in RPM 4.19, glob characters in quotes are considered literal.
Fixes: #25421
Inspired-by: Ben Boeckel <ben.boeckel@kitware.com>
See: d44114f007
When `clang-scan-deps` fails to scan (e.g., bad source syntax, junk
flags, etc.), the redirection unconditionally updates the file. If this
fails, the `.ddi` file timestamp is updated. If the state is then
reverted (e.g., the command line returns to the state of the last
successful build), the updated file is not useful, but `ninja` does not
rerun because:
- the command hash matches the last successful run
- the output file is newer than its inputs
However, since the `.ddi` file has been updated with bogus contents from
a failed scan, collation fails as the `rules` array is empty (or
incomplete from a batch scan).
If `clang-scan-deps` were properly aware of its output file, it could
use this to not write the file if any inner scan fails. Requested in
https://github.com/llvm/llvm-project/issues/72875.
See: https://github.com/llvm/llvm-project/issues/72875Fixes: #25429
llvm 17 now defaults to openmp 5.1; see LLVM commit `0c6f2f629cc0`.
It makes FindOpenMP set OPENMP_FOUND and OpenMP_C_SPEC_DATE, but not
OpenMP_C_VERSION, because it doesn't know the version mapping.
Add the spec date to the version mapping for 5.1 and 5.2 to make
things work again.
Refactoring in commit deff0e638d (CUDA: Factor out helper to parse NVCC
implicit compiler and linker flags, 2023-09-15, v3.28.0-rc1~63^2~6)
incorrectly dropped detection of the default CUDA architecture and
the CUDA toolkit include directories from the code path used for
Visual Studio generators. However, CMakeNVCCParseImplicitInfo is
now general enough that we no longer need a VS-specific code path.
Fixes: #25378
