In commit 9d44a77454 (find_*: Explicitly normalize found paths as they
exist on disk, 2024-10-17, v4.0.0-rc1~597^2~1), we removed path
normalization from the internal `cmSearchPath::AddPathInternal` helper.
Most call sites were updated to normalize input paths first, but search
paths derived from `<PackageName>_ROOT` CMake variables are no longer
normalized. Instead we normalize the path to the file found after
searching.
When `find_package` "config" mode considers a candidate CMake package
configuration file, normalize its path before loading the adjacent
package version file so that the latter is loaded by absolute path.
Otherwise `cmMakefile::ReadDependentFile` interprets a relative path
with respect to the current source directory rather than the current
working directory.
Fixes: #27279
In commit e90f60f864 (find_package: Don't glob certain macOS paths,
2024-10-23, v4.0.0-rc1~579^2~1) we changed the name matching logic of
`find_package` to check if a possible match is a directory before
checking whether the name is a match. In some situations, this results
in unnecessarily calling `stat` for a very large number of files, which
can be extremely slow on some systems (especially Windows). Fix this by
making the check the last thing we do before accepting a possible match.
Fixes: #26817
Do not try to load a CPS file if we've already loaded the same file.
This only works if the current `find_package` call selects the same file
as the previous most recent call, and if the user has not meddled with
the `<name>_CONFIG` variable.
Issue: #26731
Fix logic to populate required and optional components from CMake
variables when `find_package` is called in a nested context.
This was broken in commit e2a6416622 (find_package: Refactor in support
of recursion, 2024-11-29, v4.0.0-rc1~356^2), which promoted the
component sets from locals (in cmFindPackageCommand::InitialPass) to
member variables. Previously, in a nested context, these sets were
simply not filled, and we relied on the variables indicating component
requirement to already be set. When logic was added to properly fill the
sets (which is needed for CPS), it blindly dumped all components into
the required set, without actually checking whether the context had
marked the components as required or optional.
Fixes: #26824
Modify how CMake handles required components of a CPS transitive
dependency to not pass them as COMPONENTS if a CMake-script package is
found as the resolved dependency. This is necessary as many CMake-script
package description files do not treat component requests as target
requests (which, in CPS-land, they effectively are), but do implement
logic to mark themselves 'not found' if requested components are
missing. As a result, passing in the required targets as required
components is likely to cause the dependency to be spuriously not found
if it is only available via a CMake-script package configuration file.
Fix this by introducing a new 'required targets' concept, and by passing
CPS component requirements as both required targets and optional
components. The latter serves as a hint for packages that might provide
only a subset of themselves. The former is used to post-validate a
CMake-script package, or is folded on-the-fly into required components
when considering CPS packages.
Note that this functionality is not exposed to the user at this time,
and is only used when resolving transitive dependencies for a CPS
package.
Update cmPackageInfoReader's version parsing to more fully conform to
the specification and to reject non-conforming version strings. Start
adding framework to support version schemas other than "simple". Fix how
cmFindPackageCommand extracts version parts to not fail if more than
four parts are present.
The logic to extract the version of a CPS file into the location used to
record files that were considered but rejected was happening too late,
resulting in rejected files unnecessarily reporting their version as
"unknown". Fix this by filling the variable sooner.
Implement finding components of CPS packages. Specifically, reject any
candidate packages that don't provide all required components, and
ignore appendices that don't provide requested (required or optional)
components. This applies to both top-level searches and also searching
for package dependencies.
Fix find_package to set an error message when trying to find
dependencies of a CPS package fails. This fixes the command previously
reporting "find_package unknown error" for such failures.
Run the `clang-format.bash` script to update all our C and C++ code to a
new style defined by `.clang-format`, now with "east const" enforcement.
Use `clang-format` version 18.
* If you reached this commit for a line in `git blame`, re-run the blame
operation starting at the parent of this commit to see older history
for the content.
* See the parent commit for instructions to rebase a change across this
style transition commit.
Issue: #26123
Implement finding dependencies of CPS packages. This is done by setting
up additional `cmFindPackageCommand` instances which are used to look
for a parent package's dependencies.
A root path like `/` or `c:/` needs to end in a slash. Revise our
prefix search logic to maintain a trailing slash instead of removing one
just to add it again.
Implement logic (partly adapted from the 2023 proof-of-concept) to
actually parse CPS files and generate imported targets. Implement logic
to locate and load supplemental files. Adjust prefix handling to require
that the CPS file provides sufficient information to translate the
prefix placeholder into a meaningful path. (Note that this corresponds
to a change in the specification.)
Add a helper class to read CPS files. Use this to teach find_package how
to consider and accept .cps files in its search. (Note that no version
testing is performed at this time.) Add a simple test that we can find a
package from a .cps file and correctly extract the version information.
Note that this doesn't actually import anything from CPS yet.
Teach find_package to search CPS search paths, and to look for CPS file
names. Modify the set of file names to also include the file type (CPS
or CMake-script). Modify the search function to allow specifying which
file type(s) to consider.
During full path search, each possible path is searched for only one of
the two possible file types. However, subsequent runs, or when
considering a user-specified path (<name>_DIR), CMake will look for both
file types.
Note that this only adds the new path search logic as described above;
CMake does not yet know how to read CPS files, and there is a high
likelihood that Bad Things will happen if it tries. However, this seemed
like a good place to checkpoint.
Tweak a helper function to avoid an implicit cast from unsigned to
signed. This is hardly the only -Wsign-conversion we are tripping, but
cmFindPackage is now free of such.
Separate argument handling and actual operation of the find_package
command into separate methods. This is preparatory to adding improved
support for transitive dependencies and will allow nested calls to be
set up using structured data types, rather than having to perform all
communication via a constructed argument list.
As a consequence, the sets of required and optional components, as well
as the component list string, are now members rather than local
variables. This also allows us to drop some parameters and simplify how
we set the variables indicating which components are required.
Replace map::insert(make_pair(...) with map::emplace(...). This should
be slightly more efficient, but also removes one of only two uses of
std::pair. (Upcoming changes are going to remove the other, which will
let us drop use of <utility>.)
Some error messages (Windows registry related) of the `find_xxx` and
`cmake_host_system_information` commands, reported keywords in quotes,
while most commands did not.
Change cmDirectoryListGenerator::Names to be a pointer rather than a
reference wrapper. This allows it to be a null pointer, which allows
cmAnyDirectoryListGenerator to pass a nullptr rather than needing to
construct and hold an empty list just to satisfy the reference being
non-null. Importantly, it also make it more obvious that anyone
constructing a cmDirectoryListGenerator or subcless thereof needs to
pass something that isn't going to immediately go out of scope.
Tweak `find_package` to not compare an already-specified `<name>_DIR`
against the set of ignored paths. This is a minor behavior change in
that, if a previously found package is in a location that is NEWLY
ignored (i.e. because the user modified the ignored paths since the
previous run of CMake), we won't throw out the old result. However, it
also means that a user specifying `<name>_DIR` takes precedence over the
set of ignored paths, which seems like the desired behavior.
Note that the current behavior was introduced in commit 11f97d1968
(find_package: Refactor CMAKE_[SYSTEM_]IGNORE_PATH, 2022-01-28,
v3.23.0-rc1~31^2) and appears to have been unintentional.
Add cmAnyDirectoryListGenerator, which matches any directory but also
sorts matches in the same manner as cmProjectDirectoryListGenerator.
Modify SearchFrameworkPrefix to use this, in combination with a literal
path (cmAppendPathSegmentGenerator), instead of cmFileListGeneratorGlob
(which is removed, as it is no longer used). This improves the
consistency of when sorting is available.
This uses cmDirectoryListGenerator's new ability to match anything, as
mentioned in the previous commit.
Modify cmDirectoryListGenerator to support exact matching as well as
'starts-with' matching. The latter is used to implement '<name>*' search
paths, but cmDirectoryListGenerator is also used for '<name><suffix>'
matching (via cmMacProjectDirectoryListGenerator, mainly for macOS
paths), which resulted in the latter actually matching '<name><suffix>*'
even though that is not documented and unlikely to be useful.
This also adds the ability for cmDirectoryListGenerator to match
anything (by giving an empty list of names), which isn't in use yet, but
which we will use in the future.
find_package()'s debug mode provides information about which
prefixes are searched, but not which roots are prepended to each
prefix. Display this information if debugging is enabled.
Upstream Boost 1.70 and above provide a proper `BoostConfig.cmake`
package configuration file. Packages for all major distros now
provide it in at least one LTS release. Add a policy to pretend
that the `FindBoost` module does not exist so that projects calling
`find_package(Boost)` use the upstream package directly.
Closes: #19402
41f4e1c457 CMakePackageConfigHelpers: Document PACKAGE_PREFIX_DIR for public use
c5231ba29e find_package: Save/restore PACKAGE_PREFIX_DIR
8ac7958e3a generate_apple_*_selection_file: Save/restore PACKAGE_PREFIX_DIR
bf88879f1f generate_apple_architecture_selection_file: Avoid early returns
a4ac2c92f4 Help: Add missing section heading for apple architecture selection
b7fcc44be9 Help: Fix CMakePackageConfigHelpers typos, grammar and formatting
f1cacaa830 Tests/RunCMake/CMakePackage: Define variable closer to where it is used
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !9430
Package configuration files generated by `configure_package_config_file`
set this variable in `@PACKAGE_INIT@` and then use it later. There may
be intervening calls to `find_package`, e.g., via `find_dependency`.
If the loaded package also used `configure_package_config_file`, it
may change the value of `PACKAGE_PREFIX_DIR`. Explicitly save and
restore the value to avoid this.
Fixes: #25827
`include-what-you-use` diagnostics, in practice, are specific to
the environment's compiler and standard library. Update includes
to satisfy IWYU for our CI job under Debian 12.
Robert Maynard