Some of our "cmake" code blocks do not use fully valid CMake syntax
because they have placeholders for human reference. Sphinx has
never been able to properly lex and highlight these, but now warns.
Fix each block's syntax or change to a non-cmake block as appropriate.
Attempting to use AndroidTestUtilities to simply install some local
files on device can result in the following error:
Neither ExternalData_URL_TEMPLATES nor ExternalData_OBJECT_STORES is set
We no longer require these ExternalData-specific variables to be set if
no such remote data files are requested.
Fixes: #16529
adb used to suffer from a bug where `adb shell` would return 0 irrespective
the exit status of the underlying process that it executed. This is now
fixed for newer versions of Android (N+).
As a result, attempting to `adb shell ls` a nonexistent file can now result
in a non-zero exit status. Updating check_device_file_exists to ignore the
exit status of `adb shell` fixes this bug for newer Android devices.
This compatibility variable should be set no matter which method was
used to find HDF5. Even if `HDF5_INCLUDE_DIRS` was set by copying the
`HDF5_INCLUDE_DIR` value from `hdf5-config.cmake` it will not hurt to
copy the same value back.
Issue: #16566
Since commit v3.6.1~6^2~2 (FindHDF5: Fix h5cc arg parsing to work with
homebrew on Mac, 2016-07-11) we treated `HDF5_USE_STATIC_LIBRARIES` OFF
as a requirement for shared libraries. It is just supposed to be a
preference. Even if `HDF5_USE_STATIC_LIBRARIES` is not set we should
still fall back to finding static libraries if shared libraries are not
available.
Issue: #16566
Refactoring in commit v3.6.0-rc1~72^2 (HDF5: Rework component searching
to correctly find HL for all bindings, 2016-05-12) turned off a large
amount of its logic when HDF5_ROOT is set. This caused use of the hdf5
compiler wrapper to extract all needed libraries (e.g. z, dl as
dependencies of hdf5 static libraries) to be skipped when using
HDF5_ROOT.
Fix the search logic to honor HDF5_ROOT in all code paths. Restrict the
search for hdf5-specific components to this root, but allow external
libraries to be found anywhere.
Fixes: #16566
Re-implement our internal `_HDF5_parse_compile_line` helper to process
command line arguments all at once and in order. Otherwise the
libraries named by absolute path and those named by `-l` arguments are
not kept in order.
The new implementation will not handle separate arguments like
`-I /path/to/include/dir` but I have not seen the HDF5 compiler
wrappers produce this form. If necessary the parsing loop can
be extended with a state variable to keep track of such pairs.
1679fecb CompileFeatures Test: make sure the target "CompileFeatures" is always defined
98e6d1e5 Tests/Module/WCDH: make it work with only C features defined
c8703e9d WCDH: optionally omit error code for unknown compilers or compiler versions
0de9c398 WCDH: add macro to write simple replacement defines
When using a Visual Studio generator with an Intel toolset, such as
-T "Intel C++ Compiler XE 14.0"
the generated FortranCInterface mangling detection project may fail to
build due to `devenv` not working with the `/project ALL_BUILD` option.
This seems to be a bug in `devenv` or the Intel VS integration. Work
around the problem by building with `/project FortranCInterface`
instead. We only need to build this executable and its dependencies
within the detection test project anyway.
Fixes: #16519
The CSharp compiler for VS 2017 has moved from `MSBuildTools` to
`RoslynTargets`. Account for both locations giving priority to the
`MSBuildTools` location.
This allows one to generate a header that will basically always work. In case
an unknown compiler or compiler version is encountered it simply falls back to
the unsupported case.
This fixes a regression introduced by commit v3.3.0-rc1~5^2~2
(FindBoost: Search for debug and release libraries separately,
2015-01-26). The `_Boost_CHANGE_LIBDIR` variable was split into
`_Boost_CHANGE_LIBDIR_{DEBUG,RELEASE}` but one usage site was not
updated.
560574b0 FeatureSummary: Update release notes
3cfde818 FeatureSummary: Refactor to use global properties for package types
f0165eb6 FeatureSummary: Allow lists of dependencies in ADD_FEATURE_INFO
9da7bf08 FeatureSummary: Add QUIET_ON_EMPTY option to feature_summary
4cf4aceb FeatureSummary: Add unit tests
614a97a5 FeatureSummary: Do not force OPTIONAL type in SET_PACKAGE_PROPERTIES
65a0bfd8 FeatureSummary: Add deprecation warnings to deprecated commands
4da3cae9 FeatureSummary: Clean printed output
These new global properties were added:
* FeatureSummary_PKG_TYPES: Package types accepted by FeatureSummary
(default REQUIRED RECOMMENDED OPTIONAL RUNTIME).
* FeatureSummary_REQUIRED_PKG_TYPES: Package types that will cause
FeatureSummary to abort when called with
FATAL_ON_MISSING_REQUIRED_PACKAGES and a package in these categories
is missing (default REQUIRED).
* FeatureSummary_DEFAULT_PKG_TYPE: Default package type assigned when
not explicitly assigned by the user (default OPTIONAL).
This allows to add and remove new package types that can be printed
selectively using the "WHAT" argument.
The OPTIONAL type is used anyway as default when the type is not set
Since RUNTIME < OPTIONAL, forcing it here forces the user to set
"RUNTIME" at the first SET_PACKAGE_PROPERTIES, otherwise it is
impossible to set it later.
Without this it is impossible, for example, to add package information
(URL and DESCRIPTION) in the FindPackage file, but allow the package
to be RUNTIME.
At least the scratchbox compiler for the N900, which basically is a gcc 4.2,
sets only __GNUG__ in C++ mode, but not __GNUC__. It does indeed set
__GNUC_MINOR__ and __GNUC_PATCHLEVEL__. Extend the compiler detection code for
GNU compilers in C++ mode to look at __GNUG__ in case __GNUC__ is absent.
