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
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 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
Curl 8.15 removed support for Secure Transport and deprecated the
corresponding `CURLSSLBACKEND_SECURETRANSPORT` constant. We still need
the constant in a workaround for a runtime-checked range of older curl
versions. Suppress the deprecation warning at compile time.
Fixes: #27086
In commit f2596dfa0e (macOS: Work around bug in system curl 8.{3,4,5}
LibreSSL backend, 2024-07-16, v3.30.1~2^2) we tried to prefer
`secure-transport` on problematic versions of curl. However, the
`curl_global_sslset` setting must be applied before every
`curl_global_init` call, not just the first one. Otherwise a
second (or subsequent) download won't apply the work-around.
Some CoreFoundation headers included by the umbrella header, which we do
not need, use syntax that GCC does not support. Avoid including them.
GCC-Issue: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115880
In commit 95323c90a1 (file(MAKE_DIRECTORY): Add optional RESULT keyword
to capture failure., 2024-06-16, v3.31.0-rc1~414^2) we computed an
updated range of arguments, that name directories to be created, before
the `RESULT` keyword. However, we forgot to use it in the loop.
Fixes: #26768
Since upstream curl commit `2ec00372a1` (curl.h: change some enums to
defines with L suffix, 2025-02-25), the `CURL_NETRC_*` constants are
integer literals instead of `enum CURL_NETRC_OPTION`. It turns out
that `curl_easy_setopt` has always expected a `long` anyway, and
that `CURL_NETRC_OPTION` is not documented for public use.
Fixes: #26754
e9c494005c cmake-gui: Fix regression that prints an incidental error on fresh build trees
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !10429
Since commit a872844908 (cmake-gui: Handle relative paths in the build
directory text input, 2024-04-24, v3.30.0-rc1~5^2), `cmake-gui` gave an
incidental error message even when a valid build directory was being
passed. Fix this by checking for a non-empty path before using it.
Fixes: #26744
Fix logic from commit b9ee79b8a1 (GenEx: Add support for custom
transitive compile properties, 2024-05-09, v3.30.0-rc1~82^2~1) to more
precisely know when we are computing the link dependency graph.
Issue: #20416
Issue: #26709
In commit e8010b67c7 (cmGeneratorExpressionDAGChecker: Make local
generator available in constructor, 2024-04-25, v3.30.0-rc1~172^2~6)
we accidentally failed to preserve `EvaluateInterfaceProperty`'s
consistency with `TargetPropertyNode::Evaluate`. Fix its local
generator selection to match the evaluation context.
Extend commit f513781bc5 (WINDOWS_EXPORT_ALL_SYMBOLS: Export vftable
symbol, 2023-02-13, v3.27.0-rc1~495^2) to classify the symbol as data
on ARM64EC.
Fixes: #26604
Since commit 159ba027b9 (Install: Add parallel installation option,
2024-06-19, v3.31.0-rc1~333^2) the file was generated relative to the
current working directory. This is not always the top of the build
tree, so specify the path explicitly.
In commit 9c0491a3e4 (cmDyndepCollation: write out scanned source
information too, 2024-03-25) via !9708, the `sf_map` gained a new job of
also being used to track non-file set sources which could import
modules. This was implemented by removing processed `FILE_SET TYPE
CXX_MODULES` sources from the map and working with the sources
remaining. When a `FILE_SET TYPE CXX_MODULES` source appeared multiple
times, this would then erroneously complain that it "was not scheduled
for compilation". Use a set of source paths to track sources that have
already been processed. If duplicates are found, trigger an author
warning and skip the duplicate file.
Fixes: #26549
Revert commit 4a11772618 (GenEx: Limit TARGET_PROPERTY transitive
closure optimization to subgraphs, 2024-05-31, v3.31.0-rc1~114^2).
The change caused substantial performance regressions in some
existing use cases. Revert it pending further investigation.
Issue: #25728Fixes: #26457
854eba0c53 target_sources: Improve error message for CXX_MODULES on INTERFACE libraries
cd179e7560 Tests/RunCMake/CXXModules: Rename FileSetModules cases to be more specific
Acked-by: Kitware Robot <kwrobot@kitware.com>
Tested-by: buildbot <buildbot@kitware.com>
Merge-request: !10090
We support non-compiled `SOURCES` on `INTERFACE` libraries, and also
support `CXX_MODULES` on *imported* `INTERFACE` libraries (via synthetic
targets that compile module interface units). However, we do not
support `CXX_MODULES` on non-imported `INTERFACE` libraries because
there is no place to hold module interface unit's object files for their
module initializers. Previously this was not explicitly rejected, and
so was diagnosed only by "CMake Internal Error" messages due to
assumption violations in the implementation.
Fixes: #26524
Co-authored-by: Ben Boeckel <ben.boeckel@kitware.com>
The `install(RUNTIME_DEPENDENCY_SET)` form processes its arguments in a
slightly different way to other forms of the command. It handles the
`POST_...`, `PRE_...` and `DIRECTORIES` keywords separately. These
keywords are not visible to the first layer of keyword processing, so
they don't terminate multi-value keyword value parsing.
Before commit 6a1fac1450 (install: Normalize DESTINATION paths,
2024-09-18, v3.31.0-rc1~73^2), the `DESTINATION` keyword was handled as
a simple string value. That commit changed it to use a function so it
could process the value before storing it, but the function returned the
wrong value. This meant it was treated as a multi-value keyword instead
of a single-value keyword.
Fixes: #26512
