Some callers may want the raw byte vector instead of the hex character
string. Convert the internal implementation to use this so that we
can later add public APIs to get it.
With the Makefile generator one can use `cd $subdir; make install` to build and
install targets associated with a given subdirectory. This is not possible to
do with the Ninja generator since there is only one `build.ninja` file at the
top of the build tree. However, we can approximate it by allowing one to run
`ninja $subdir/install` at the top of the tree to build the targets in the
corresponding subdirectory and install them.
This also makes sense for `test`, `package`, and other GLOBAL_TARGET targets.
It was already done for `all` by commit v3.6.0-rc1~240^2~2 (Ninja: Add
`$subdir/all` targets, 2016-03-11).
In cmGlobalNinjaGenerator::AppendTargetOutputs we previously
handled GLOBAL_TARGET outputs specially in order to avoid adding
directory components to the output. However, this is not necessary
because cmNinjaTargetGenerator::New already filters out copies of
these targets that are not at the top level. Instead we can simply
follow the same output computation code path as UTILITY targets.
CMake used to put all header search paths into HEADER_SEARCH_PATHS
attribute. Unfortunately this attribute does not support to declare
a search path as a system include.
As a hack one could add a -isystem /path to the cflags but then include
ordering is not deterministic. A better approach was chosen with this
patch by not filling HEADER_SEARCH_PATHS at all and to populate
the C, C++, and Fortran flags directly. The include paths used by
Xcode should be now identical to the ones used by Unix Makefiles and
Ninja generator.
The implementation of `install(EXPORT)` generates files into a staging
directory for later installation. We use the full install destination
in the path to the staging directory to avoid collisions. In order to
avoid exceeding maximum path lengths (especially on Windows) we compute
a hash of the install destination when it is too long. Fix this logic
to account for the length of the file name(s) when deciding whether to
switch to the hashed name.
Reported-by: Alan W. Irwin <irwin@beluga.phys.uvic.ca>
The `CMAKE_<LANG>_SIMULATE_ID` variables are not set to "GNU" for a
GNU-like Clang compiler on Windows. They are only set to "MSVC" for a
MSVC-like Clang. Revise our response file format selection accordingly.
Reported-by: Chaoren Lin <chaorenl@google.com>
This will allow additional information about the availability
and capabilities of extra generators to be queried without
actually creating them.
Instead of a static NewFactory() method like the main generator
factories have, use a static GetFactory() method to get a pointer to a
statically allocated extra generator factory. This simplifies memory
management.
f951d0ad Add tests for BUNDLE_EXTENSION
c63380b1 Update documentation about bundle extensions
134d5c1f Honor BUNDLE_EXTENSION also for Frameworks (#14742)
2b909c08 Honor BUNDLE_EXTENSION also for App Bundles (#16148)
Teach Visual Studio generators to include object files from object
libraries in the list of objects whose symbols are to be exported.
The Makefile and Ninja generators already did this. Update the
test to cover this case.
Reported-by: Bertrand Bellenot <Bertrand.Bellenot@cern.ch>
Revert these commits:
* v3.6.0-rc1~134^2
Tests: QtAutogen: Same source name in different directories test, 2016-04-13
* v3.6.0-rc1~134^2~1
Autogen: Generate qrc_NAME.cpp files in subdirectories, 2016-04-19
* v3.6.0-rc1~134^2~2
Autogen: Generate not included moc files in subdirectories, 2016-04-19
They regress existing builds that depend on the paths/symbols generated
previously. Another approach will be needed to solve the name collision
problem they were intended to solve. Leave the error diagnostics for
the colliding cases that were added in the same topic as the above
commits because they provide a useful early failure in relevant cases.
Fixes#16209.
Factor CMAKE_<LANG>_USE_RESPONSE_FILE_FOR_{OBJECTS,LIBRARIES} lookup out
into a common helper. Use a separate helper for each because more
specific logic may be added to each later.
Custom command dependencies are followed for each target's source files
and add their transitive closure to the corresponding target. This
means that when a custom command in one target has a dependency on a
custom command in another target, both will appear in the dependent
target's sources. For the Makefile, VS IDE, and Xcode generators this
is not a problem because each target gets its own independent build
system that is evaluated in target dependency order. By the time the
dependent target is built the custom command that belongs to one of its
dependencies will already have been brought up to date.
For the Ninja generator we need to generate a monolithic build system
covering all targets so we can have only one copy of a custom command.
This means that we need to reconcile the target-level ordering
dependencies from its appearance in multiple targets to include only the
least-dependent common set. This is done by computing the set
intersection of the dependencies of all the targets containing a custom
command. However, we previously included only the direct dependencies
so any target-level dependency not directly added to all targets into
which a custom command propagates was discarded.
Fix this by computing the transitive closure of dependencies for each
target and then intersecting those sets. That will get the common set
of dependencies. Also add a test to cover a case in which the
incorrectly dropped target ordering dependencies would fail.
Although we provide a `VS_GLOBAL_ROOTNAMESPACE` option to both set
the `RootNamespace` value and reference it, some users may try to
set `VS_GLOBAL_RootNamespace` to set `RootNamespace` as a variant
of the `VS_GLOBAL_<variable>` property. In this case we still
need to add the reference to `$(RootNamespace)`.
Since commit v3.6.0-rc1~339^2 (CPack: Fix CPACK_INSTALL_CMAKE_PROJECTS
SubDirectory (4th) option, 2016-02-18) we now honor the "Directory"
option of this variable. Prior to that commit the value was not used
so projects may have used a placeholder value like `.` instead of `/`.
Treat `.` the same as `/` to restore compatibility with such projects.
Fixes#16205.
The cmGetPropertyCommand::StoreResult expects NULL for unset
properties. Make ALIASED_TARGET align with that expectation.
Additional corrections to the unit tests are necessary because
get_property removes variables for unset properties (in contrast
to get_target_property which stores a -NOTFOUND value).
Add an explicit `<Natvis>` element in VS project files for `*.natvis`
files. These enable custom debug visualizers for project-specific
types.
Fixes#16043.
Use `sysconf(_SC_ARG_MAX)` wherever `_SC_ARG_MAX` is available instead
of hard-coding an incomplete list of operating systems that implement
it. In particular, it is available on BSD platforms that were not
previously listed.
Commit v3.6.0-rc1~174^2~1 (cmGlobalNinjaGenerator: Clarify logic for
forcing use of response files, 2016-04-06) started using negative
command line length values to represent forced use of response files but
forgot to update `calculateCommandLineLengthLimit()` accordingly. Teach
it to return `0` instead of `-1` when no limit is computed from the
system to avoid forcing response files.
Reported-by: Raphael Kubo da Costa <rakuco@FreeBSD.org>
4db08807 CMake: Report whether generators support platforms
43a68a6d cmGlobalGeneratorFactory: Use CM_OVERRIDE for all derived classes
af0e1cd4 Make CMake version dirty state available to code
6a077b5d Make CMake version suffix available to code
2aadb02f QtIFW: Don't show component selection page if it don't need
53992808 QtIFW: process USE_LGPL when CMake_INSTALL_COMPONENTS
8ae035a5 CMake: install COMPONENT cmcldeps
This code has accidentally not been compiled since commit
v3.0.0-rc1~556^2 (cmake: Split -E command implementation into separate
source file, 2013-10-03) because CMAKE_HAVE_VS_GENERATORS was local to
`cmake.cxx`.
Fixes#16195.
This code has not been compiled since commit v3.0.0-rc1~556^2 (cmake:
Split -E command implementation into separate source file, 2013-10-03).
It appears the code was never used since it was added anyway.
Files generated by `install(EXPORT)` and `export_library_dependencies()`
may be installed with packages and consumed by dependents. In order to
avoid re-building dependents only because the version of CMake changed,
drop the CMake version from the export file content.
The fix in commit commit v3.6.0-rc1~82^2 (Drop find_(library|file|path)
prefixes from PATH on non-Windows, 2016-05-09) aggressively dropped
search of the entries in PATH itself in addition to the prefixes derived
from it. This regresses find modules that (incorrectly) depended on the
behavior, including some of our own modules such as FindImageMagick.
Restore the search of entries in PATH itself. If we want to drop it
later we will need to do so with a policy. Fixes#16192.
Factor the implementation out of cmMakefileLibraryTargetGenerator
into a helper method in cmMakefileTargetGenerator so it can be
re-used elsewhere later.
Drop our `HaveExportedObjects` check before dumping exports for an
object file. It is possible for only a subset of needed symbols to have
explicit markup, and re-exporting the marked symbols does not hurt.
This leaves no callers of `HaveExportedObjects`, but leave the
method in place anyway because it may be useful in the future.
Fixes#16161.
Replace use of cmsys::auto_ptr with a CM_AUTO_PTR macro that maps to
our own implementation adopted from the KWSys auto_ptr implementation.
Later we may be able to map CM_AUTO_PTR to std::auto_ptr on compilers
that do not warn about it.
Automate the client site conversions:
git grep -l auto_ptr -- Source/ | grep -v Source/kwsys/ | xargs sed -i \
's|cmsys::auto_ptr|CM_AUTO_PTR|;s|cmsys/auto_ptr.hxx|cm_auto_ptr.hxx|'
In the `try_compile` source file signature we propagate the caller's
value of `CMAKE_<LANG>_FLAGS` into the test project. Extend this to
propagate `CMAKE_<LANG>_FLAGS_<CONFIG>` too instead of always using the
default value in the test project. This will be useful, for example, to
allow the MSVC runtime library to be changed (e.g. `-MDd` => `-MTd`).
However, some projects may currently depend on this not being done,
so we need to activate the behavior using a policy.
This change was originally made by commit v3.6.0-rc1~160^2 (try_compile:
Honor CMAKE_<LANG>_FLAGS_<CONFIG> changes, 2016-04-11) but without the
policy and so had to be reverted during the 3.6 release candidate cycle.
Fixes#16174.
Clang refuses to default initialize an instance of a class that does not
have a default constructor. Fix the check by adding default
constructors. Don't use brace initialization like it is proposed in the
error message. We want to test the override support independent from
the support for brace initialization.
Revert commit v3.6.0-rc1~160^2 (try_compile: Honor
CMAKE_<LANG>_FLAGS_<CONFIG> changes, 2016-04-11). The behavior it
introduced can break projects that depend on the lack of such behavior.
We will have to introduce a policy or other mechanism to enable the
behavior in a compatible way. Simply revert it for now.
See issue #16174.
059a6ca0 Merge branch 'unknown-aliased-target' into compiler-features
1d6909a2 use CM_NULLPTR
b4b73f56 cxx features: add check for nullptr
a7a92390 mark functions with CM_OVERRIDE
9e2d6f0c CM_OVERRIDE: mark destructor overridden in the feature test.
2ca76a66 Validate target name in ALIASED_TARGET property getter
This is important for two reasons:
1. A compiler might warn about a class that has a virtual member
function but no virtual destructor. We don't want to treat the feature
as incomplete in this case.
2. MSVC10 supports the override identifier except on destructors. In
this case, the feature really is incomplete and we want to detect it as
such.
In commit v3.6.0-rc1~174^2 (Ninja: Honor CMAKE_NINJA_FORCE_RESPONSE_FILE
for compile rules, 2016-04-06), Ninja learned to look for
`CMAKE_NINJA_FORCE_RESPONSE_FILE` in the current scope or the
environment in order to force response file usage for all compilation
rules.
However, on Windows, the RC compiler goes through cmcldeps which does a
`replace(output, output + ".dep.obj")` on the command line. However,
with a response file (which we name `output + ".rsp"`), the response
file path is replaced instead causing the compiler to (correctly)
complain that the response file `output + ".dep.obj.rsp"` does not
exist.
What needs to happen is for cmcldeps to look through the response file,
replace *its* contents and place it in the `output + ".dep.obj.rsp"`
file.
Also add a test which actually compiles an RC file into a library and
executable for all generators on Windows and additionally test
`CMAKE_NINJA_FORCE_RESPONSE_FILE` for Ninja generators.
Fixes#16167.
7647f6af Add CM_OVERRIDE to some functions
5286110d cxx features: add check for override
09aa2c94 Use <unordered_set> where available
ea5477e4 Make C++ feature checks extensible
Run clang-tidy's modernize-use-override checker. This checker must have
issues in version 3.8. It has way too little matches. And it adds
override to destructors. Revert the changes on the destructors and
change override to CM_OVERRIDE.
Turn the feature check for cxx11_unordered_map into a function such that
we can use it for other features as well. Drop the 11 suffix, as we may
want to check features from other standards.
Refactoring in commit 49f10f0d (cmGeneratorTarget: Adopt Fortran module
directory generation, 2016-06-10) accidentally made a local variable
declared `static` causing results to be re-used incorrectly.
Remove the 'optional' paramenter from the second overload of the Convert
function. This parameter is used from one single location. Inline the
codepath for which the argument is true to the callsite.
The cmDependsJavaParserHelper tries to implement a "deep copy" in the
assignment operator of the internal class CurrentClass. To do that, it
uses std::copy and std::back_inserter. The copy constructor is
implemented in terms of the assignment operator but it does not
initialize the member NestedClasses, a pointer to vector. This pointer
is dereferenced in the assignment operator. Change the pointer to a
value and rely on the compiler generated special functions.
This property allow to specify a specific Visual Studio tool for a
source file overriding the default tool behavior. For example, a
`.resw` file being processed as a `PriResource` file. This has the
advantage of being able to teach CMake to process new file types without
code modifications.
Download http://www.gnu.org/licenses/lgpl.txt and place it as
Licenses/LGPLv3.txt in our source tree. When building cmake-gui, use
option CMake_GUI_DISTRIBUTE_WITH_Qt_LGPL to enable notification in the
"About" dialog of how the distribution of Qt is licensed. Install the
license file as ${CMAKE_ROOT}/Licenses/LGPLv3.txt so that the dialog can
display a path to it.
Factor the flag generation out of cmCommonTargetGenerator::GetFlags
into a new cmLocalGenerator::GetTargetCompileFlags method.
This will allow it to be used without a target generator available.
Add a cmLocalGenerator::GetTargetFortranFlags virtual method to get
generator-specific generation of Fortran-specific flags. Implement it
in cmLocalCommonGenerator by moving the implementation from
cmCommonTargetGenerator::AddFortranFlags. This will allow it to be used
without having a target generator available.
Inspired-by: Tobias Hunger <tobias.hunger@qt.io>
Move code to create/get the fortran module directory from the
cmCommonTargetGenerator to cmGeneratorTarget.
Rename the ComputeFortranModuleDirectory method to
CreateFortranModuleDirectory as this method *creates* the directory if
it is missing.
Even in relatively small projects using `--trace` (and `--trace-expand`)
may produce a lot of output. When developing a custom module usually
one is interested in output of only a few particular modules.
Add a `--trace-source=<file>` option to enable tracing only a subset of
source files. The final output would be only from requested modules,
ignoring anything else not matched to given filename(s).
Create a LINK_WHAT_YOU_USE target property and corresponding
CMAKE_LINK_WHAT_YOU_USE variable to enable this behavior.
Extend link commands by running `ldd -u -r` to detect shared
libraries that are linked but not needed.
In commit v3.6.0-rc1~279^2~10 (VS: in Clang/C2 toolset, setup correct
compiler settings, 2016-02-18) a flag mapping was added for the clang
`-Os` flag. However, this collides with a mapping we already had for
the MSVC flag of the same name. This is a symptom of a larger problem
in that the VS generators need a per-toolset flag map (issue #16153).
For now, simply drop the new mapping and drop `-Os` from clang compiler
flags in the MinSizeRel configuration.
Reported-by: Felix Bruns <felixbruns@gmail.com>
Since 2.6.3 the UTILITY target may have source files. A defect was filed
that these files are now visible in the source tree. A fix later removed
all generated files from the source tree, regardless of the target type.
You can't even include them by using the SOURCES option. This fix adds
generated files again, except for the UTILITY target which cluttered the
source tree.
Fixes#14272.
ed5fa48d cmXMLWriter: use ifstream from KWSys
24ab29b8 Prefer istringstream and ostringstream over stringstream.
ab8b77dd Remove redundant arguments from fstream constructors
eb79fa72 Access std::ios_base with std::ios
At the moment, cmStandardIncludes.h needs to be included before any
standard includes because it disables some warnings that are caused
by the standard library of some compilers. Move this responsibility
to the cmConfigure.h file.
Also add include guards to cmConfigure.h to make sure the file can be
included multiple times.
63c0e92c cmState: Expose list of properties of values in the cache
6eee2463 cmCacheEntry: Retrieve all properties of cache entries
120899c6 cmPropertyList: Add a way to retrieve all properties
7066218e cmake: Kill cmake::CacheManager and its getter
ea5324cd cmMakefile: Port messages for compile features to cmake
df8c3130 cmGlobalGenerator: Don't use cmMakefile::IssueMessage after configure
946d1e50 cmMakefile: Avoid IssueMessage after configure is finished
096c7754 cmLocalGenerator: Store Backtrace for the directory
f62ed322 cmLocalGenerator: Add GetTargetDefines to get all defines for a target
853b1bb4 cmLocalGenerator: Constify AppendDefines and AddCompileDefinitions
The force parameter is ugly and makes the method harder to reason about
(issues the message ... but maybe it doesn't ... but then again you can
force it). It is a violation of
https://en.wikipedia.org/wiki/Interface_segregation_principle
and is the kind of thing described in a recent blog here:
http://code.joejag.com/2016/anti-if-the-missing-patterns.html
"Any time you see this you actually have two methods bundled into one.
That boolean represents an opportunity to name a concept in your code."
The makefile is only used when called by the cmMessageCommand, so inline
the use of it there. It otherwise creates an undesirable dependency on
cmMakefile for issuing messages in the cmake instance, a violation of
the Interface Segregation Principle.
https://en.wikipedia.org/wiki/Interface_segregation_principle
This also makes it more explicit that the variable definitions only
affect the message() command. If an AUTHOR_WARNING is issued for any
other reason, it is not affected. To affect that, it is necessary to
set the cache variable instead of the regular variable.
This is an unfortunate interface quirk, but one which can't be fixed
easily now.
Add a ``FIND_LIBRARY_USE_LIB32_PATHS`` global property analogous to the
``FIND_LIBRARY_USE_LIB64_PATHS`` property. This helps find commands on
multilib systems that use ``lib32`` directories and either do not have
``lib`` symlinks or point ``lib`` to ``lib64``.
When we are on OSX and we are launching cmake-gui from a symlink, the
application will fail to launch as it can't find the qt.conf file which
tells it what the name of the plugin folder is. We need to add this path
BEFORE the application is constructed as that is what triggers the
searching for the platform plugins
CMakeFindBinUtils sets CMAKE_RANLIB to `:` if it is not available in
order to get a no-op. This does not work on a Windows host build
environment that runs commands in `cmd` instead of `sh`. Teach the
Ninja and Makefile generators to simply skip the command if it is `:`.
This this was already done by the Makefile generator since commit
v2.6.0~3161 (BUG: Do not write link script lines that use the ':',
2006-06-18), but only when using a link script.
Reported-by: Michael Jäntsch <Michael.Jaentsch@gmx.de>
After finding it in `foo.Framework/Headers/dir/header.h`, we should
report the `foo.Framework/Headers` directory, not
`foo.Framework/Headers/dir`, because the former is what actually
contains the path the caller wishes to include.
63e5eb5f Help: Add notes for 'productbuild' topic
2e3c67d1 productbuild: Add new productbuild cpack generator.
50a3d340 PackageMaker: factor out common code for creating pkg files.
Policy CMP0026 deprecated the LOCATION property, and we have long
provided a $<TARGET_FILE:...> generator expression. However, if
a project tries to use $<TARGET_PROPERTY:...,LOCATION> we should
at least not crash.
The compatibility implementation of the LOCATION property uses
cmGlobalGenerator::CreateGenerationObjects to create the structures
needed to evaluate the property before generation starts. The
implementation assumed that accessing the property could only be done
during configuration (via the typical get_property command use case).
The $<TARGET_PROPERTY:...,LOCATION> genex causes the LOCATION property
to be accessed during generation. Calling CreateGenerationObjects
during generation blows away all the objects currently being used for
generation and is not safe. Add a condition to call it only when
configuration is not finished.
Use clang-tidy's readability-simplify-boolean-expr checker.
After applying the fix-its, revise all changes *very* carefully.
Be aware of false positives and invalid changes.
Use clang-tidy's performance-unnecessary-copy-initialization checker.
After applying the fix-its (which turns the copies into const&), revise
the changes and see whether the copies can be removed entirely by using
the original instead.
Use clang-tidy's performance-unnecessary-value-param checker to find
value parameter declarations of expensive to copy types that are not
modified inside the function. Ignore findings in kwsys.
After applying the fix-its, manually change `const T&` to `T const&`.
d256ba07 try_compile: Optionally forward custom platform variables to test project
fb4791b3 cmCoreTryCompile: Refactor forwarding of variables to test project
Add a `CMAKE_TRY_COMPILE_PLATFORM_VARIABLES` variable to specify a list
of custom variables to be forwarded to a `try_compile` test project.
This will be useful for platform information modules or toolchain files
to forward some platform-specific set of variables from the host project
(perhaps set in its cache) to the test project so that it can build the
same way.
De-duplicate the logic that constructs the cmake `-D` flag used to pass
variables into the test project cache. Also subsume variables that were
propagated by generating `set()` commands in the project and pass them
as cache entries instead.
5784747d Improve string find: prefer character overloads.
5cec953e Use std::replace for replacing chars in strings.
2a1a2033 cmExtraEclipseCDT4Generator: use std::replace.
34bc6e1f cmCTestScriptHandler: don't call find repeatedly.
c1340827 Add a variable to specify language-wide system include directories
44199097 cmMakefile: Optimize AddSystemIncludeDirectories for empty set
a896043b GHS: Compute include directories consistently with other generators
Create a `CMAKE_<LANG>_STANDARD_INCLUDE_DIRECTORIES` variable to specify
system include directories for for `<LANG>` compiler command lines.
This plays a role for include directories as the existing
`CMAKE_<LANG>_STANDARD_LIBRARIES` variable does for link libraries.
This patch preserves backward compatibility of
deb package names with previous CMake versions
but similarly to CPack/RPM allows to change
package name format and supports DEB-DEFAULT
setting that produces proper Debian package names.