Extend commit ff8c0e9c10 (ci: Clamp CUDA 'native' architecture to values
supported by its toolkit, 2022-03-14, v3.24.0-rc1~468^2) to cover the
NVHPC job too.
Extend commit ff8c0e9c10 (ci: Clamp CUDA 'native' architecture to values
supported by its toolkit, 2022-03-14, v3.24.0-rc1~468^2) to avoid
failures due to `nvcc -arch=native`.
Our CI setup has per-project runner setups, so they are now providing
the port configuration. The main goal is to have Rosetta caches separate
from the native caches on macOS, but removing this removes a bit of
coupling to Kitware's CI setup.
In flang, both -J and -module-dir support both joined and separate values. For
the -J option, these are of the form `-J<value>` and `-J <value>` respectively.
Once https://github.com/llvm/llvm-project/pull/168748 is committed, -module-dir
will only work with a separate value. Setting CMAKE_Fortran_MODDIR_FLAG to -J
ensures that both joined and separate styles will continue to work after that
change is made. As far as behavior is concerned, both -J and -module-dir are
aliases.
Also revert commit 9702a01dc9 (Help: string(LENGTH): Fix synopsis link
to command description, 2023-12-30, v3.28.2~23^2). It was only needed
to avoid re-using a named reference that we've now removed.
Issue: #25546
Since commit 5420639a8d (cmExecuteProcessCommand: Replace cmsysProcess
with cmUVProcessChain, 2023-06-01, v3.28.0-rc1~138^2~8) we have not
actually terminated child processes on an `execute_process` timeout.
Similarly for other migrations from cmsysProcess to cmUVProcessChain.
Teach cmUVProcessChain clients that implement timeouts to actually
terminate remaining child processes when the timeout is reached.
Fixes: #27378
f357fc27e5 CPack: Backport "correctly perform querytags on old versions of RPM"
1803eda9f7 CPack/RPM: Backport "Fix detection of RPM support for weak dependencies"
d2404872b2 CPack/RPM: Backport "Remove redundant conditions for presence of rpmbuild"
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !11400
Since commit 5420639a8d (cmExecuteProcessCommand: Replace cmsysProcess
with cmUVProcessChain, 2023-06-01, v3.28.0-rc1~138^2~8) we've
occasionally observed immediate timeouts on processes that take longer
than the timeout to start, even though we only start the timer after the
child processes start. The problem is that:
* `uv_loop_init` initializes `uv_loop_t`'s cached "now" time.
* Starting processes takes time but does not update the "now" time.
* `uv_timer_start` computes expiry relative the cached "now" time,
so short timers may be expired as soon as they are started.
* `uv_run` invokes expired timer callbacks before polling for I/O
or process completion, so we "timeout" immediately.
Fix this by updating the cached "now" time via `uv_update_time` just
before starting timers. This is needed only for timers that start
before the `uv_run` event loop. Update our `uv_timer_ptr` wrapper
to make all callers consider the choice when calling `start()`.
Since commit 965a12cb8a (ci: update macOS jobs to use Xcode 26.0,
2025-09-18, v4.1.2~9^2) and commit 9d302ecd47 (ci: update macOS jobs to
use Xcode 26.0 in CMake 3.31 branch, 2025-10-07) our macOS 10.13+
packages are built using the macOS 26 SDK, with which Qt 5.15.2 does not
render buttons in `cmake-gui` correctly. Revert to an older macOS SDK
to avoid the problem until we update our Qt version.
Fixes: #27325
With Xcode 16.4, run
env SDKROOT=/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk \
.gitlab/ci/repackage/macos.sh
and host `MacOSX15.5.sdk.tar.bz2` ourselves.
7f628ea04b FindPython: Add support for Python 3.15
5b78983813 Tests/FindBoost/TestPython: Improve python version list formatting
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !11350
Backport commit 79a83ddb08 (Apple: Enable linking during
iOS/tvOS/visionOS/watchOS toolchain inspection, 2024-11-14,
v4.0.0-rc1~471^2) to 3.31.
Since commit 11da882a12 (Apple: Introduce separate system name for iOS,
tvOS, and watchOS, 2018-01-15, v3.14.0-rc1~14^2~1) our toolchain
inspection steps, like ABI detection, tell `try_compile` to use a
`STATIC_LIBRARY` instead of an `EXECUTABLE`. This was needed at the
time to avoid codesign requirements. However, commit d3a64c4e3f (Xcode:
Explicitly turn off signing in try_compile projects, 2020-07-16,
v3.19.0-rc1~483^2) introduced a more general solution to that problem.
Restore linking during toolchain inspection so that we can detect and
identify the linker.
Suggested-by: Marc Chevrier <marc.chevrier@gmail.com>
Fixes: #26443
9d302ecd47 ci: update macOS jobs to use Xcode 26.0 in CMake 3.31 branch
f6f1c5ca1f Tests: Teach RunCMake to ignore Xcode an IDERunDestination warning
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !11292
On some Xcode versions, `xcodebuild` may warn:
... xcodebuild[...] IDERunDestination: ...
Teach RunCMake to drop such incidental lines before matching against
expected output.
Backport commit edaa6ed06a (Tests: Teach RunCMake to ignore Xcode an
IDERunDestination warning, 2025-09-19, v4.1.2~9^2~1) to CMake 3.31.
Since upstream curl commit `1a12663d06` (CURLOPT: bump `CURLPROXY_*`
enums to `long`, drop casts, 2025-07-28), the `CURLPROXY_*` constants
are integer literals instead of `enum curl_proxytype`. It turns out
that `curl_easy_setopt` has always expected a `long` anyway, and that
`curl_proxytype` is not documented for public use.
Fixes: #27178
