An old workaround for `std::allocator_traits<>::value_type` lints from
IWYU on `std::vector<>` usage breaks IWYU's handling of `<memory>`.
Convert the workaround to use the same approach we already use for a
workaround of `std::__decay_and_strip<>::::__type` lints. Then update
the `<memory>` inclusions to follow the now-correct IWYU lints.
Fix the logic added by commit d91b5a72cd (Ninja: Add support for CUDA
nvcc response files, 2019-05-30, v3.15.0-rc1~8^2) to always use the CUDA
compiler response file flag for response files during device linking.
Instead of passing multiple strings to the `WriteRule` and `AddRule` methods
of `cmGlobalNinjaGenerator`, pass only a `cmNinjaRule` instance reference,
that is set up beforehand.
Adapt calls to `WriteRule` and `AddRule` in multiple places.
The response file was being written but not used due to the wrong
variable being configured for Swift. Swift does compile+link in a
single phase and does not use the `<OBJECTS>` placeholder. Use the
`<SWIFT_SOURCES>` placeholder instead for the response file substitution
which serves the same purpose.
81b4d10d8f CUDA: More exhaustive checks to determine when to do device linking
3205c7c950 cmNinjaLinkLineDeviceComputer now lives in the correct source file
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !3320
Since the Swift model does a single compile/link phase, we would not
get the compile definitions, includes, and options past along to the
build of the module. Compute these for the target when constructing
the main command.
Previously CMake used fairly naive logic to determine when to do
device linking which caused unnecessary device linking to occur
frequently. We now use a more exhaustive algorithm to determine
when we have a need for device linking.
Fixes: #19238
Rather than compute the value of `SWIFT_MODULE`, use the computed value
of `SWIFT_MODULE_NAME`. This is helpful strictly for Windows where you
have multiple configuration variants with differing ABIs that need must
co-exist. In particular, this re-use permits the use of
`CMAKE_DEBUG_POSTFIX` to have the import library and the DLL be named
with the trailing suffix but have the same import name for the swift
module and interface.
a9180ccf9a Tests: add a check for the Swift compiler
d745551fb6 Help: add some initial documentation for Swift support
9a182c9e5b Auxiliary: update vim syntax highlighting
e9b0063e8e Modules: add build rules for Swift Ninja support
b6412e3e38 Ninja: add placeholders to support Swift build
7d7f31161d Ninja: add support for Swift's output-file-map.json
d688c4c19d Swift: remove unnecessary unreleased Ninja infrastructure
0723582208 Swift: Detect compiler version
...
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !3297
This cleans up the new options that were added to support Swift. This
was not released, and the proper support approach that we settled upon
does not require as much specialised support.
On macOS ranlib truncates the fractional part of the static achive
file modification time. If the archive and at least one contained
object file were created within the same second this will make look
the archive older than the object file. On subsequent ninja runs
this leads to re-achiving and updating dependent targets.
As a work-around we touch the archive after ranlib.
Closes: #19222
This adds the placeholders that are needed to merge the swift partial
modules. It permits generating the rules necessary to actually merge
the partial modules emitted for a partial compilation.
Rather than taking a number of out parameters for the various names,
create a structure that is reused for both `GetLibraryNames` and
`GetExecutableNames`. Replace uses according to the new interface.
`CUDA_RESOLVE_DEVICE_SYMBOLS` can be used with shared, module, and
executable target types. This relaxation is to allow for better
interoperability with linkers that automatically do CUDA device symbol
resolution and have no way to disable it.
With the `|| true`, a linker error before running link-what-you-use
would also use the `|| true` fragment and unconditionally succeed. Just
skip the addition since `--lwyu=` ignores the return value anyways.
Fixes#18524
After changing the ``cmGeneratedFileStream`` methods to accept
``std::string const&`` instead of ``const char*`` we don't
need to call ``std::string::c_str`` anymore when passing
a ``std::string`` to a ``cmGeneratedFileStream`` method.
This patch removes all redundant ``std::string::c_str``
calls when passing a string to a ``cmGeneratedFileStream`` method.
It was generated by building CMake with clang-tidy enabled using
the following options:
-DCMAKE_CXX_CLANG_TIDY=/usr/bin/clang-tidy-4.0;-checks=-*,readability-redundant-string-cstr;-fix;-fix-errors
Run the `clang-format.bash` script to update all our C and C++ code to a
new style defined by `.clang-format`. Use `clang-format` version 6.0.
* 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.
Previously we dropped non-target items from the device link line because
nvcc rejects paths to shared library files, and only with target items
do we know the kind of library. However, this also prevents projects
from linking to system-provided libraries like `cublas_device` that
contain device code.
Fix this by passing more link items to device linking. Items that are
not file paths, such as `-lfoo`, can simply be passed unconditionally.
Items that are targets known to be shared libraries can still be
skipped. Items that are paths to library files can be passed directly
if they end in `.a`. Otherwise, pass them using `-Xnvlink` to bypass
nvcc's front-end. The nvlink tool knows to ignore shared library files.
Issue: #16317
Brad King