Each call to AddDefinition has overhead for variable watches and such.
Avoid extra calls when not needed.
This decreases the configure time for ParaView by 10 seconds on my
machine. Without the change about 1,000,000 set-to-empty calls were
being made. After the change it drops to about 100,000.
The SHA512_Update implementation accesses input data via 64-bit
integers. This requires alignment of the input buffer on some
architectures. Align our stack-allocated buffer for file content to
satisfy this requirement.
Describe explicitly the scope in which enable_language must be
invoked to handle all direct and indirect uses of a language.
While at it, re-word documentation of the OPTIONAL keyword to
avoid confusion over its partial implementation.
For clang, this allows passing -target <triple> to the compiler, and
for qcc, -V<arch> using toolchain files containing something like
set(triple arm-linux-gnueabihf)
set(CMAKE_C_COMPILER "/usr/bin/clang")
set(CMAKE_C_COMPILER_TARGET ${triple})
set(CMAKE_CXX_COMPILER "/usr/bin/clang++")
set(CMAKE_CXX_COMPILER_TARGET ${triple})
or
set(arch gcc_ntoarmv7le)
set(CMAKE_C_COMPILER /opt/qnx650/host/linux/x86/usr/bin/qcc)
set(CMAKE_C_COMPILER_TARGET ${arch})
set(CMAKE_CXX_COMPILER /opt/qnx650/host/linux/x86/usr/bin/QCC)
set(CMAKE_CXX_COMPILER_TARGET ${arch})
Both clang and qcc are inherently cross compiler( driver)s.
Commit 089fe1c1 (Optimize genex evaluation for includes and
defines., 2013-02-01) introduced an optimization on DAG processing
to not reprocess properties on targets which have already been seen.
This was refactored slightly in commit 8dfdf1c7 (Fix the tests for
evaluating includes and defines., 2013-02-18), but was not extended
to cover COMPILE_OPTIONS in commit 80ca9c4b (Add COMPILE_OPTIONS target
property., 2013-05-16).
This omission causes the same performance regression in running
cmake on LLVM which 089fe1c1 fixed before, but this time for the
transitive evaluation of the COMPILE_OPTIONS property.
As CMAKE_ROOT_FIND_PATH can be a list, a new CMAKE_SYSROOT is
introduced, which is never a list.
The contents of this variable is passed to supporting compilers
as --sysroot. It is also accounted for when processing implicit
link directories reported by the compiler, and when generating
RPATH information.
Extend the signature
try_compile(RESULT_VAR <bindir> <srcfile> ...)
to allow multiple sources as
try_compile(RESULT_VAR <bindir> SOURCES <srcfile>... ...)
Process the sources to generate a CMakeLists.txt that enables all needed
languages.
Teach the TryCompile test to try cases with two sources of the same
language and of mixed languages. Teach RunCMake.try_compile to cover
error cases for the signature.
If the expression $<TARGET_PROPERTY:prop> appears in the content
of a target property, the target that prop is read from is
the 'head target' of the expression. In contexts such as evaluating
the content of a target property during generation, such
as INCLUDE_DIRECTORIES, the 'head target' is the one on which the
initial request was made.
If evaluating a generator expression which is not a target property
content, the target must be explicitly specified. Such contexts
include add_custom_command and file(GENERATE). The content might
then look like
$<TARGET_PROPERTY:tgt,prop>
However, as there is no HeadTarget set, any generator expressions
evaluated as part of reading prop from tgt which do not specify
the tgt directly report an error.
Modify the logic of the TARGET_PROPERTY generator expression so
that in such contexts, the 'head target' is set to the appropriate
target which was first encountered.
In commit 2bc22bda (Xcode: Add frameworks search paths from link
dependeny closure, 2012-12-07) we made framework search paths from the
link closure conditional on target type, skipping it on STATIC and
OBJECT library targets that do not actually link. However, the
framework search paths also influence the compile lines (-F options) so
we need them for all target types. The Makefile generator already does
this, as did the Xcode generator prior to the above-mentioned commit.
Fix the wording of the documentation for CMAKE_PARENT_LIST_FILE to
clarify that it only applies to the include stack in the current
directory. Specify exactly what commands can include files instead of
saying "somehow invoke".
It is common to specify a CMAKE_TOOLCHAIN_FILE and get a warning
for using it despite it not being used.
The WarnUnusedCliUnused test relies on the warning being emitted
each time cmake is run on an existing build. That behavior is changed
by this patch to warn only on the first invokation of CMake, and not
on subsequent invokations (because the variable is in the cache with
the same value). For that test, a clean target is added which clears
the cache and cause the warning to be emitted each time.
As the Ninja generator does not support the feature needed to test
this, it is not tested with that generator.
This command is similar to add_definitions, in that it affects
the compile options of all targets which follow it. The implementation
is similar to the implementation of the include_directories command,
in that it is based on populating a COMPILE_OPTIONS directory property
and using that to initialize the same property on targets.
Unlike the include_directories command however, the add_compile_options
command does not affect previously defined targets. That is, in
the following code, foo will not be compiled with -Wall, but bar
will be:
add_library(foo ...)
add_compile_options(-Wall)
add_library(bar ...)
Use makefile->IssueMessage() to print the unprocessed watch message in a
format consistent with other CMake messages and with a more complete
call stack for the access.
When a watch does not specify a command to call then variable_watch
prints out a message to stderr. Remove code after that which collects
all variable values to construct a message that is never printed.
Otherwise such code causes a READ_ACCESS watch to trigger on all
variables in the currents scope.
Reported-by: Yichao Yu <yyc1992@gmail.com>
In commit 9a5b4eba (All variable accesses should produce watch
callbacks, 2007-05-17) we added a new enumeration value but did not
update the list of strings matching them. Add the missing entry.
Without this, variable_watch prints REMOVED_ACCESS instead of
MODIFIED_ACCESS when set the value of the variable.
dc1d025 OS X: Add test for rpaths on Mac.
8576b3f OS X: Add support for @rpath in export files.
00d71bd Xcode: Add rpath support in Xcode generator.
94e7fef OS X: Add RPATH support for Mac.
RPATH support is activated on targets that have the MACOSX_RPATH
property turned on.
For install time, it is also useful to set INSTALL_RPATH to help
find dependent libraries with an @rpath in their install name.
Also adding detection of rpath conflicts when using frameworks.
This fixes bug #13797.
The kinds of changes applied in 373faae5 for frameworks are now
applied to CFBundle. The prefix and suffix for CFBundles are
now handled in cmTarget::GetFullNameInternal.
This corresponds to the g++ and clang++
option -fvisibility-inlines-hidden on linux. On Windows with MinGW,
this corresponds to -fno-keep-inline-dllexport. That option is
not supported by clang currently.
This is initialized by CMAKE_<LANG>_VISIBILITY_PRESET. The target
property is used as the operand to the -fvisibility= compile option
with GNU compilers and clang.
Currently it only adds the contents of the COMPILE_FLAGS target
property, but it can be extended to handle a new COMPILE_OPTIONS
generator expression enabled property.
Process all arguments in a single loop using a simple state machine.
While at it, fix some error message typos. Also allow LINK_LIBRARIES
with no actual libraries to disable use of the -DLINK_LIBRARIES=...
from the CMAKE_FLAGS. This was already possible in the old logic if
LINK_LIBRARIES was immediately followed by another keyword argument
instead of the end of the argument list, so allow it in general.
Update the RunCMake.try_compile test cases accordingly.
3aa9ce4 GenexEval: Fix evaluation of INCLUDE_DIRECTORIES target property.
0b39fef GenexEval: Extract a getLinkedTargetsContent from TargetPropertyNode.
53164ac cmTarget: Remove some hardcoding of transitive property names.
This property should come from the content of the property itself,
plus the INTERFACE_INCLUDE_DIRECTORIES of the link *implementation*.
In contrast, when the INTERFACE_INCLUDE_DIRECTORIES is evaluated for
a target, the INTERFACE_INCLUDE_DIRECTORIES of the link *interface*
is used.
Similar logic applies for the COMPILE_DEFINITIONS target properties.
If the propertyName is already an INTERFACE_ variant of the property,
ie, the expression is similar to
$<TARGET_PROPERTY:foo,INTERFACE_INCLUDE_DIRECTORIES>
then the INTERFACE_INCLUDE_DIRECTORIES of the link *interface* of foo
is used.
However, if the propertyName is not an INTERFACE_ variant, and the
interfacePropertyName is, ie, the expression is similar to:
$<TARGET_PROPERTY:foo,INCLUDE_DIRECTORIES>
then the INTERFACE_INCLUDE_DIRECTORIES of the link *implementation*
of foo is used.
This will be used to process transitive components of properties
which depend on linked targets. Currently only the link interface
of the target can be used as the source of the linked targets, but
in the next commit it will be possible to use the link implementation
as the source of link targets.
This commit does not change the semantics of the code.
Commit e65ef08b (try_compile: Fix quoting of libraries in generated
CMakeLists.txt, 2013-05-24) added quoting to entries specified in
the LINK_LIBRARIES. However, if the input entries contain whitespace
padding, that quoted whitespace causes an error in the generated
CMakeLists.txt at target_link_libraries.
Strictly, it is an error to have space separated entries in the
CMAKE_REQUIRED_LIBRARIES, as it was never properly handled by
CMakeExpandImportedTargets even prior to commit 236133e7 (Handle
targets in the LINK_LIBRARIES of try_compile., 2013-02-09). However,
it is causing a regression in KDE code which tests the next branch.
In Qt 5.1, Qt5::Core has a INTERFACE_QT_MAJOR_VERSION property
of '5', and since CMake 2.8.11, Qt4::QtCore has an
INTERFACE_QT_MAJOR_VERSION of '4'. This was introduced in
commit 4aa10cd6 (FindQt4: Set the INTERFACE_QT_MAJOR_VERSION for
Qt4::QtCore, 2013-03-16), to produce an error if Qt 4 and Qt 5
are erroneously used by the same target. This can also be used
however to determine the Qt major version, and therefore the
particular moc executable to use during automoc steps. This means
that targets in a single buildsystem can use a selection of Qt 4
and Qt 5, and still take advantage of the CMAKE_AUTOMOC feature
without conflicting.
Since commit 236133e7 (Handle targets in the LINK_LIBRARIES of
try_compile, 2013-02-09) libraries passed to the new LINK_LIBRARIES
option of try_compile are not quoted inside the generated CMakeLists.txt
file. Quote the library names so they re-parse correctly when loaded by
CMake to configure and generate the test project.
Reported-by: Bogdan Cristea <cristeab@gmail.com>
They can't be used when evaluating link libraries, but they can be
used for include directories and compile definitions. Later they can
be used for compile options.
The idea is to write to a temp file which contains generator
expressions, and at generate time, evaluate the generator expressions,
and write the result to a file.
Because executables on Windows are limited in the length of command line
it is possible to use, it is common to write command line arguments to a
file instead and specify the file as a source of arguments.
This new FILE(GENERATE) subcommand allows the use of generator
expressions to create such files so that they can be used with
add_custom_command for example.
Make handling of directory separators consistent between
non-bundle and bundle code.
Remove xcode specific flag from cmTarget when getting install_name.
Add (more) consistent convenience functions in cmTarget to get
directories inside of bundles and frameworks to add files to.
This refactor also fixes bug #12263 where frameworks
had the wrong install name when SKIP_BUILD_RPATH.
Also make install_name for frameworks consistent between Makefile
and Xcode generator.
Since commit c519bb2b (XCode: Also qoute [] as needed to set
build-configurations, 2011-04-05) we escape "[]" conditions for
XCODE_ATTRIBUTE_ settings. However, we need to handle the "variant"
condition with a special case to map it only into the settings for the
matching configuration.
a7ba452 Add the JOIN generator expression.
96ec314 Make it possible for any genex to have arbitrary content at the end.
bd638ee Rename the method determining if a genex accepts arbitrary content.
dc742fe Extract the ProcessArbitraryContent method.
Checking if a byte is 0 before checking it for something else doesn't do
anything useful. Also one can be sure that it can't be anything wanted
if the value was set to 0 immediately before.
Refactor lookup of CMAKE_(EXE|MODULE|SHARED)_LINKER_FLAGS(|_<CONFIG>)
variables to ensure all combinations are handled. Use the helper method
AddConfigVariableFlags to simplify the implementation.
.../Source/cmakewizard.cxx: In member function ‘virtual void cmakewizard::AskUser(const char*, cmCacheManager::CacheIterator&)’:
.../Source/cmakewizard.cxx:31:35: warning: conversion to ‘int’ from ‘long unsigned int’ may alter its value [-Wconversion]
if(!fgets(buffer, sizeof(buffer)-1, stdin))
^
.../Source/cmakewizard.cxx: In member function ‘virtual bool cmakewizard::AskAdvanced()’:
.../Source/cmakewizard.cxx:70:35: warning: conversion to ‘int’ from ‘long unsigned int’ may alter its value [-Wconversion]
if(!fgets(buffer, sizeof(buffer)-1, stdin))
The compiler is (partly) right here: sizeof() returns a size_t, which often is
bigger as as int as it is unsigned long or something similar. Add an explicit
cast to get rid of that warning here, the buffer has a size of 4KiB, so the
value will fit into an int on all sane platforms.
When processing link line entries we check for matches with known naming
patterns for static and shared libraries. Teach this logic to recognize
numerical suffixes after shared library names such as "libfoo.so.1".
Generate the default AssemblerListingLocation through the flag map so
that it can be overridden by a user /Fa flag. Also teach the VS 7-9
generators to map /Fa to AssemblerListingLocation.
While at it, fix the AssemblerListingLocation default value to have a
trailing slash after the configuration name. This ensures it will be
treated as a directory and not a file name.
This allows for example, the buildsystem to use names like 'boost_any'
instead of the overly generic 'any', and still be able to generate
IMPORTED targets called 'boost::any'.
Commit 429e3699 (Process COMPILE_DEFINITIONS as generator expressions
in QtAutomoc., 2013-01-25) introduced a check for the existence of the
COMPILE_DEFINITIONS property on the target before processing it.
At the time, compile definitions from linked targets were added to the
COMPILE_DEFINITIONS property by target_link_libraries. The way such
transitive compile definitions worked was changed in subsequent
commit a1c4905f (Use the link information as a source of compile
definitions and includes., 2013-02-12), so that the COMPILE_DEFINITIONS
property may be empty, despite the fact that there are definitions
which should be used.
10bc50e Tests: ignore Guard Malloc messages in MemChecker tests
159c3e9 Tests: add a test with custom options passed to valgrind
61ddb93 CTest: fix comment documenting cmBoundsCheckerParser class
cbdfcb0 Tests: add test for non-existent Valgrind suppression file
3b5b758 CTest: drop suppression for gcc 2.9.6 errors from default Valgrind flags
7752253 Tests: verify that memory checker output files are always present
abf1df4 Tests: remove code duplication in CTestTestMemCheck tests
f499422 CTest: remove unreachable code and CTestTestMemcheckUnknown test
dde6306 CTest: use an output file for Valgrind (#14110)
bcc0f3f Tests: create output files for all memory checkers
f92d7e7 cmake-gui: use shortcut F only for "Find in Output"
6e57724 cmake-gui: Add function for going to next error message in Output window
7be2d8a cmake-gui: Change shortcut of the search field from Ctrl-F to Alt-E
4ce65e8 cmake-gui: Add search functions to the context menu of the Output widget
df3663b cmake-gui: Add search functions for Output window (#9733)
ac9a5f4 ctest_build: Pass projectDir to GenerateBuildCommand
a6c0299 CTest: Simplify ctest_* command source/build dir lookup
1ca9318 VS: Add test for building MSBuild project in subdir
3cd4000 VS: Use .sln parser to build targets in subdirs with msbuild (#13623)
df035e4 VS: Create parser for Visual Studio .sln files
de8be9e Add projectDir parameter to GenerateBuildCommand
This generator expression joins a list with a separator. The separator
may contain arbitrary content, such as commas, which is ordinarily a
delimiter in the generator expression syntax.
Add support to maintain designer functionality for Visual Studio C++
Windows Forms projects. Also add a test project showing how to use
the CMakeLists.txt file and, when successfully configured, will allow
use of the designer for the included form.
At this point we do not care what the current directory is, as nothing that
requires a specific directory is done. Just make sure that at the end we are
back in the same directory as at the start.
Initialize the CMAKE_CONFIGURATION_TYPES cache entry early during
EnableLanguage like the Xcode generator does. Avoid depending on
the MSVC compiler information module to do it. Otherwise code like
project(MyProj NONE)
sets CMAKE_CONFIGURATION_TYPES late (in GenerateConfigurations), and
to only "Debug" and "Release" instead of the standard set of 4.
Reported-by: Paul Smith <paul@mad-scientist.net>
PASS_REGULAR_EXPRESSION and FAIL_REGULAR_EXPRESSION both take a list of expressions.
Stop searching if the first of those has matched the output as the result will not
change anymore.
Because of clock scew between processors or just because of someone changing
the system time the end timestamp may be before the start time. Reporting a
negative time doesn't any sense, just report zero there as it already happens
for really fast tests.
The --workaround-gcc296-bugs has been part of the default Valgrind flags since
Valgrind support was added in commit 5b232ded15
(ENH: Add initial memory check support which works for Valgrind, 2003-12-15).
The Valgrind manpage says that this option should be avoided if not really
needed as it may cause real errors to get ignored. If someone uses a compiler
that really needs the flag this flag should be set by the user explicitely.
Most users will never set any flags and probably never notice that they use a
flag they shouldn't.
The memory checker command can't be quoted at this point, because previously it
has been tested that the given file exists, which will fail if the name is
quoted. The CTestTestMemcheckUnknown test aimed to test this case, has always
failed to do so and serves no useful purpose therefore.
The phrase "flag used" is somewhat imprecise, so extend it to stress
intentions.
Also correct:
- "Flag" => "Flags"
- "[CMAKE_BUILD_TYPE]" => "<CONFIG>"
dea1df4 Memoize usage requirement include directories in a config-specific map
26dba6a Fix include dir propagation from conditionally linked targets
b8259c3 Centralize maintenance of usage requirement include directories
Commit 42ebb188 (Memoize includes and defines from interface libraries.,
2013-02-22) introduced caching of the includes. Fix the memoization to
be configuration-specific so that we do not accumulate entries across
multiple evaluations in a multi-config generator.
Generator expressions, including configuration-specific expressions may
be used as link libraries of targets. The old-style keywords of
target_link_libraries are handled in terms of new generator expressions.
However, the generator expressions expect target names to be valid
against a regular expression, whereas target_link_libraries does not
require validation. In generator expression constructed without any
action from the user we need to ensure that only valid expressions are
generated. Ensure that strings which are not valid target names are not
used in generator expressions which validate the argument.
Code like
target_link_libraries(B debug A)
generates usage requirement references such as "$<$<CONFIG:DEBUG>:A>".
When cmTarget::GetIncludeDirectories uses such references it generates
expressions like:
$<TARGET_PROPERTY:$<$<CONFIG:DEBUG>:A>,INTERFACE_INCLUDE_DIRECTORIES>
When the conditions are false such references evaluate as an empty
string and the expression fails with an error such as:
$<TARGET_PROPERTY:tgt,prop> expression requires a non-empty target name.
Fix this by teaching cmTarget::GetIncludeDirectories to wrap the above
expression inside a conditional:
$<$<BOOL:$<$<CONFIG:DEBUG>:A>>:...>
so that $<TARGET_PROPERTY:...> will not be evaluated with an empty
target.
Maintain a target's internal list of usage requirement include
directories whenever the LINK_LIBRARIES property is set by either
target_link_libraries or set_property.
Re-word the documentation to make clear that CMake integrates usage
requirements during generation and not synchronously during
configuration or execution of target_link_libraries.
In commit 236133e7 (Handle targets in the LINK_LIBRARIES of try_compile,
2013-02-09) an error return case was added without closing the file in
progress. Add the missing fclose() call.
Spotted by sevenhill.
This allows the ctest_build command's TARGET option to name a target
in a subdirectory and still build properly with msbuild.
Add test case covering use of ctest_build() with such a TARGET.
In cmCTestHandlerCommand::InitialPass call InitializeHandler after setting
the CTest "SourceDirectory" and "BuildDirectory" configuration values
instead of before. This makes the values available to InitializeHandler
methods in subclasses, so also drop the duplicate configuration of those
values from cmCTestConfigureCommand::InitializeHandler.
Use cmVisualStudioSlnParser in GenerateBuildCommand() to provide correct
command line for MSBuild even when target project is stored in a
subdirectory.
Create class cmVisualStudioSlnParser as a generic parser for Visual
Studio .sln files. Implement minimum functionality but keep class
extensible. Add tests for the class.
Extend the cmGlobalGenerator::GenerateBuildCommand virtual method
signature with a "projectDir" parameter specifying the top of the
project build tree for which the build command will be generated.
Populate it from call sites in cmGlobalGenerator::Build where a
fully-generated build tree should be available.
Use the suggestion from http://support.microsoft.com/kb/2262855 to escape
semicolons in preprocessor definitions for VS >= 10. Update the
COMPILE_DEFINITIONS documentation disclaimer list of known limitations
accordingly. Update our "Preprocess" test to cover the case.
Suggested-by: Jean-Christophe Fillion-Robin <jchris.fillionr@kitware.com>
Transform the path using ConvertToNinjaPath just as we do for all other
paths. This fixes the OutOfSource test for objects in the ../OutOfBinary
directory by computing the proper full path for the /Fd option.
The MS tools create two types of PDB files as explained here:
http://msdn.microsoft.com/en-us/library/yd4f8bd1%28v=vs.71%29.aspxhttp://msdn.microsoft.com/en-us/library/yd4f8bd1%28v=vs.80%29.aspxhttp://msdn.microsoft.com/en-us/library/yd4f8bd1%28v=vs.90%29.aspxhttp://msdn.microsoft.com/en-us/library/yd4f8bd1%28v=vs.100%29.aspx
One is created by the compiler (/Fd) and the other by the linker (/pdb).
The two options should not specify the same file. Split them up.
In the VS IDE generators, simply drop ProgramDataBaseFileName to
take the VS default "/Fd$(IntDir)vc$(PlatformToolsetVersion).pdb".
In the Makefile generators, set "/Fd" on the compile line to be
the directory containing object files (with a trailing slash the
compiler will add the "vc$(PlatformToolsetVersion).pdb" filename
automatically). Drop the /Fd option from the exe link command
line and add "/pdb" instead (already done for dll linking).
Update these rules for both MSVC and Intel tools.
Drop support for PDB_OUTPUT_DIRECTORY and PDB_NAME in STATIC
libraries because the generated .pdb files are only from /Fd
and not real linker-generated .pdb files. Update documentation to
clarify that the PDB_* properties are only for linker .pdb files.
This regresses the PDBDirectoryAndName test for STATIC libraries.
Since it is not clear at this time what should be done for STATIC
library .pdb files, comment out the relevant portion of the test
and leave a TODO comment.
In commit 0c727b90 (install(EXPORT): Force absolute paths for usr-move,
2013-03-08) and commit d4774140 (configure_package_config_file: force
absolute paths for usr-move, 2013-01-24) we supported Linux
distributions implementing the "/usr move" by assuming that installation
to (/usr)?/lib(64)? represents a non-relocatable system package.
When cross-compiling one may prepare a package for installation into a
system location on a target machine but install the package files on the
*host* machine inside another path for use with CMAKE_FIND_ROOT_PATH.
In this case the package development files must still be relocatable.
Handle "/usr move" with a new approach that works with relocatable
files. Teach configure_package_config_file and install(EXPORT) to
generate special logic in a package configuration file or targets file
for installation under (/usr)?/lib(64)?. Teach the file to recognize
when it is loaded through a symlink that refers to the same realpath as
its original install destination. In such a case, use the original
install prefix. Otherwise, compute the prefix relative to the current
file location to make it relocatable.
In VS IDE generators add a pre-build event to perform automoc instead of
using a separate custom target. This reduces the number of targets in the
.sln that need to be loaded by the IDE.
This also works around a VS 11 bug as discussed in issue 13900.
Suggested-by: Hauke Heibel <hauke.heibel@gmail.com>
When called with a non-existent LHS target name the user may be trying
to add file-level dependencies. Clarify the error message to explain
the difference between target-level and file-level dependencies. Point
the reader at the commands and options needed for the latter.
We can do this check only if the TargetName is non-empty, which means
that we're evaluating INTERFACE_INCLUDE_DIRECTORIES from a linked
dependency which was set using target_link_libraries.
It is possible to have relative paths in INCLUDE_DIRECTORIES already
in CMake 2.8.10.2, so that part will require a policy to fix.
It is considered an error if the INTERFACE_INCLUDE_DIRECTORIES contains
a directory which does not exist, which indicates a programmer error
by the upstream, or a packaging error.
One of the RunCMake.CompatibleInterface tests also needs to be updated
due to this change. Non-existant includes were used in the test, but
are not needed.
Check that source and binary directories are not part of the
INTERFACE_INCLUDE_DIRECTORIES for installed IMPORTED targets.
This is limited to directories which do not contain generator
expressions to evaluate. Such paths can only be checked at time
of use of the imported target, which will be done in a follow up
patch.
The VS version we generate in the .sln header is used by VS when opening
the file through Windows Explorer and possibly elsewhere. Fix our
generators to use version strings known to VS to avoid a drop-down box.
For VS 10, since commit 4f96af44 (Fix VS 10 .sln files for Windows
Explorer, 2009-10-22) we use "Visual Studio 2010" instead of just
"Visual Studio 10". This is correct except that for the Express edition
we need "Visual C++ Express 2010".
For VS 11, since commit f0d66ab4 (VS11: Fix comment generated at the top
of *.sln files, 2011-10-20) we use "Visual Studio 11" in the .sln header
but the preferred value is "Visual Studio 2012" (just as the first
commit mentioned above fixed for VS 10). Also for the Express edition
we need "Visual Studio Express 2012 for Windows Desktop".
The API for retrieving per-config COMPILE_DEFINITIONS has long
existed because of the COMPILE_DEFINITIONS_<CONFIG> style
properties. Ensure that the provided configuration being generated
is also used to evaluate the generator expressions
in cmTarget::GetCompileDefinitions.
Both the generic COMPILE_DEFINITIONS and the config-specific
variant need to be evaluated with the requested configuration. This
has the side-effect that the COMPILE_DEFINITIONS does not need to
be additionally evaluated with no configuration, so the callers can
be cleaned up a bit too.
In some languages the compiler may need to know the path of the final
target file for which an object is being compiled. Honor the <TARGET>
placeholder for compilation rules to support such cases.
Note that this cannot work with OBJECT library targets because the final
target path is not known during compilation (there can even be more than
one final target).
Suggested-by: Vittorio Giovara <vittorio.giovara@gmail.com>
The commit d2536579 (Automoc: fix regression #13667, broken build in
phonon, 2012-11-19) changed Automoc to try to re-add the Qt header dir
if it was stripped out as an implicit include from the moc command
line. When invoking a compiler, those directories are stripped out
because they are built-in, but for moc, there are no built-in directories.
The follow-up commit acc22400 (Automoc: get include dirs without
stripping implicit include dirs off, 2012-12-07) went further by not
removing the implicit include dirs, if they were specified specifically
by the user.
This had the remaining problem that the implicit include dirs appeared
in a different order of precedence for moc compared to the compiler.
Resolve that by stripping out the include dirs, where specified for
the moc command line to, and then appending them at the end. Note that
the order of the appended implicit include directories is the order
they are specified in the CMAKE_CXX_IMPLICIT_INCLUDE_DIRECTORIES, not
the order specified by the user.
Rename the variable added by commit 9ce1b9ef (Add
CMAKE_BUILD_INTERFACE_INCLUDES build-variable, 2012-11-25) to
CMAKE_INCLUDE_CURRENT_DIR_IN_INTERFACE to be more consistent with the
existing CMAKE_INCLUDE_CURRENT_DIR variable.
Suggested-by: Alex Neundorf <neundorf@kde.org>
If there is a Qt 5.0.3 release, it may or may not contain the patch
that fixes this issue.
http://thread.gmane.org/gmane.comp.lib.qt.releasing/882
Just use the workaround until 5.1.0 which certainly will contain the
fix. Don't use the workaround before Qt 5.0.0.
We can't find both preprocessing expressions at once, because then
the BUILD_INTERFACE will always be favored if both are present, even
if INSTALL_INTERFACE appears first.
This was affecting the behavior of install(EXPORT) because the
INTERFACE_INCLUDE_DIRECTORIES contained entries like
/foo/include;$<INSTALL_INTERFACE:/bar/include>
As the INSTALL_INTERFACE always evaluates to '0', it always needs
to be preprocessed properly.
27e6f6e TI compiler: add automatic detection of prefix and suffixes
b301586 rename TI_DSP toolchain to TI, since it works also for the ARM compiler
12b25df Determine C/CXX/Fortran compiler: fix indentation
7baa55a Determine C/CXX/Fortran compiler: minor restructuring
d90f49b CTest: Fail early without PROJECT_BINARY_DIR (#14005)
2e1c2bd build_command: Fail early without CMAKE_MAKE_PROGRAM (#14005)
4e5cb37 Refactor RunCMake.build_command test to allow more cases
ab079ee Avoid crash when checking property compatibility without link info
92a2ab7 Avoid crash when checking property link dependencies without link info
The commit in qtbase 9dfba89c (Add implementations of QAIM::sibling in
public APIs., 2012-09-26) added a buggy implementation of sibling(), and
the commit f136701b (Use the base implementation of
QAbstractItemModel::sibling in QSIM., 2013-02-21) resolves it.
Workaround the bug for Qt releases that have it.
If CMAKE_MAKE_PROGRAM is not set fail with an error message instead of
crashing. Suggest calling project() or enable_language() first to
ensure that CMAKE_MAKE_PROGRAM is set.
Commit 79568f95 (automoc: Add source file to target early to set the
linker language, 2013-02-20) changed automoc initialization to a two
step process. In the first step, the generated source file was added
to the target, which allows the link language to be determined.
However, this bypassed the check for the availability of Qt itself.
At build-time the automoc file could not be generated because the moc
tool was not available to create it.
The solution is to only add the automoc file to the target if Qt is
found.
Teach the compatibility check added by commit 042ecf04 (Add API to
calculate link-interface-dependent bool properties or error, 2013-01-06)
to return early if no link information is available. This avoids
crashing in a case that should fail with an error message.
Teach the isLinkDependentProperty helper added by commit e9879910 (Make
INTERFACE determined properties readable in generator expressions,
2013-01-19) to return early if no link information is available.
Extend the range of valid target names with the + sign. This character
can commonly be used for target names, such as those containing 'c++'.
Add a test but skip it for Borland and Watcom tools which do not support
the character.
Suggested-By: Benjamin Kloster
When the command is invoked without an explicit build directory path we
compute it from the source directory path. When either the source or
build tree is the root of a Windows drive letter it will have a trailing
slash. Handle slashes consistently when substituting the current output
directory for the current source directory.
While at it, use cmSystemTools::IsSubDirectory instead of FindLastString
to verify that the source directory is a subdirectory.
Inspired-by: Graham Menhennitt <graham@menhennitt.com.au>
* Clarify accepted values of <LANG> suffix in CMAKE_COMPILER_IS_GNU<LANG>.
* Clarify fact that target property GENERATOR_FILE_NAME usually cannot be
used at configure time.
* Clarify that enable_language() can only be used in global scope of
top-level project using language.
* State that enable_language(... OPTIONAL) currently doesn't work.
* Document regular expression operator precedence.
If the absolute install(EXPORT) destination for the CMAKE_INSTALL_PREFIX
used during configuration is under (/usr)?/lib(64)? then assume the
current build is for a system package installation instead of a
relocatable distribution. Generate an absolute path for _IMPORT_PREFIX
in the target exports file instead of generating code to compute the
value relative to the file location. This is necessary for
distributions implementing a move to /usr such as:
https://wiki.archlinux.org/index.php/DeveloperWiki:usrlib
"All files in the /lib directory have been moved to /usr/lib and now
/lib is a symlink to usr/lib."
The relative path computation is not reliable because the targets file
could be installed through cross-prefix a symlink and loaded without it
or vice versa.
A similar change was made for package configuration file generation by
commit d4774140 (configure_package_config_file: force absolute paths for
usr-move, 2013-01-24).
Teach the WriteGroup method return true if a group or any of its
children have source files. Have children write their output to a
temporay cmOStringStream. Add it to the real output only if not empty.
Only check a release version for the minimum required version. If NSIS
returns a CVS build version string (as when built by Macports), skip the
version check altogether and assume a sufficiently new version. Also
correctly handle the case where the version check fails and the
CPACK_TOPLEVEL_DIRECTORY option is not set.
Co-Author: Graham Menhennitt <graham@menhennitt.com.au>
Commit a1c4905f (Use the link information as a source of compile
definitions and includes., 2013-02-12) introduced the use of link
information as the source of target properties via the TARGET_PROPERTY
generator expression. This generator expression has a strict
interpretation of a valid target name and emits a fatal error for
invalid names.
Ensure that only targets with names valid for use with TARGET_PROPERTY
or targets which are determined by generator expressions are processed
by it. This means that at worst, invalid target names do not participate
in the transitive evaluation of properties, but the validation
generator expression can be extended where needed to resolve that.
cbf0756 Revert "Add the TARGET_DEFINED generator expression"
21a342c Remove use of TARGET_DEFINED from the target_link_libraries test.
47b8d32 Remove use of TARGET_DEFINED from the ExportImport test.
2e39d21 Remove use of TARGET_DEFINED from target_include_directories test.
Use the /STACK: flag to pass the value through flag parsing so that the
generator converts it to the StackReserveSize project file option. The
option was accidentally left out by commit 7491f529 (first pass at VS
10, 2009-06-25).
Suggested-by: goatboy160@yahoo.com
As of commit 1da75022 (Don't include generator expressions in
old-style link handling., 2012-12-23), such entries are not
included in the LinkLibraries member. Generator expressions in
LinkLibraries are not processed anyway, so port to the new way
of getting link information.
In the case of input like
foo$<1:bar>
the preGenex should be 'foo'. In that case, the search for a ';'
will not find one, and there is no preceding input to process as a
non-genex list.
Previously, the result of 'splitting' such a string would instead
be a vector containing the same string two times.
8dfdf1c Fix the tests for evaluating includes and defines.
98a6725 Fix constness of accessors.
7e70744 Expand includes and defines transitively in 'external' genexes.
d1a2729 Fix DAG checker finding cycling dependencies.
e72eaad Workaround broken code where a target has itself in its link iface.
ec2c67b Strip stray semicolons when evaluating generator expressions.
236133e Handle targets in the LINK_LIBRARIES of try_compile.
1c0597c Add a new Export generator for IMPORTED targets.
f2ab17d Keep track of all targets seen while evaluating a genex.
This reverts commit 2bee6f5ba5.
This expression is not used, and has a semantic which is not completely
optimal (namely considering utility targets to be targets, though
usually we are interested in linkable targets).
Remove it so that we have more freedom to define better expressions in
the future.
Conflicts:
Source/cmGeneratorExpressionEvaluator.cxx
Tests/CMakeCommands/target_compile_definitions/CMakeLists.txt
Tests/CMakeCommands/target_compile_definitions/consumer.cpp
This is similar in spirit to commit e48d8420 (Cache context-independent
includes on evaluation., 2013-02-03), but it is needed since commit
a1c4905f (Use the link information as a source of compile definitions
and includes., 2013-02-12), which changed how includes and defines
are determined. As they are now determined through the link interface,
we need to cache the result of evaluating them through that.
In the case of the includes, the result was already being cached
and then immediately disposed. Store the result as a member variable
instead to make use of the caching.
We should also check whether the INTERFACE_ variant of a property
is being read, and in the case of the compile definitions, we should
test the _<CONFIG> suffixed variants. That is already available
through the use of the methods.
This way, we use the ALREADY_SEEN optimization when evaluating
the includes of a target in 'external' generator expressions, ie, those
used in a add_custom_command invokation, as opposed to evaluating the
INCLUDE_DIRECTORIES of a target itself via GetIncludeDirectories.
This means that we can use expressions of the form
$<TARGET_PROPERTY:foo,INTERFACE_INCLUDE_DIRECTORIES>
to get a list of the interface include directories of foo, including
those coming from dependencies.
We can't have a test of a target which has a single include directory in
its INCLUDE_DIRECTORIES because the shell on the MSYS platforms transforms
a single include directory to include a prefix, which is not what the test
expects. We test a target with two directories instead as a means to
test a target with no link dependencies.
Before this patch, the following is reported falsely as a self-reference:
target_link_libraries(empty2 LINK_PUBLIC empty3)
target_link_libraries(empty3 LINK_PUBLIC empty2)
add_custom_target(...
-DINCLUDES=$<TARGET_PROPERTY:empty2,INTERFACE_INCLUDE_DIRECTORIES>
)
The reason is that the existing code assumed that all reading of
include directories would be done through cmTarget::GetIncludeDirectories()
and would therefore be initialized with a DagChecker. That is not the case
if reading the property with an 'external' generator expression.
There is a test for this since commit 8e756d2b (Tolerate cycles in
shared library link interfaces (#12647), 2012-01-12), so make sure
it continues to pass, even as we require no self-references in new
INTERFACE_ property generator expressions.
In commit bed6c388 (VS,Xcode: Remove unused CMAKE_GENERATOR_* variables,
2013-02-19) we removed this variable among others from the VS generator
claiming they were all not used. In fact the Fortran compiler detection
module CMakeDetermineFortranCompiler.cmake uses CMAKE_GENERATOR_FC even
for the Visual Studio IDE generators because it does not use the full
IDE compiler id detection like the C and C++ language files do.
Imported targets are re-exported so that they can be used by the
try_compile generated code with target_link_libraries.
This makes the use of the cmake_expand_imported_targets macro
obsolete. The macro is not able to expand the generator expressions
which may appear in the IMPORTED_LINK_INTERFACE_LIBRARIES content.
Instead it just sees them as 'not a target'.
Previously, GetIncludeDirectories was called before calling
target->AddSourceFile(mocCppSource). Since commit a1c4905f (Use the
link information as a source of compile definitions and
includes., 2013-02-12), the include directories are determined by
the link information.
Valid link information requires that the linker language can be
determined, which depends on the source files languages and the
dependent targets languages. In the case of the no_link_languages
target in the unit test, there are no dependencies and the additional
source file no_link_languages_automoc.cpp is added to the target
at generate-time. That file can be used to determine the linker
language, but it must be added to the target before calling
GetIncludeDirectories.
Instead of generating a whole bunch of repeated if-statements
now a foreach()-loop is generated in the targets-file.
Also now a comment is inserted in the generated file if no
targets from other export sets are used, so if somebody looks
at the file he can see whether the information about missing
imported targets has been generated or not.
Alex
e83e6a1 Test Unix Makefiles generator support for changing compilers
c307e1c Tests/RunCMake: Allow tests to control build tree behavior
2963c98 Merge branch 'empty-compiler-crash' into compiler-change-cleanup
1df09e5 Delete entire CMakeFiles directory when deleting CMakeCache.txt (#13756)
Generate a unique PBXTargetDependency and PBXContainerItemProxy for
every edge in the dependency graph as required by the Xcode format.
Xcode only accidentally loads project files that re-use the same
PBXTargetDependency for every edge leading to a single target. If one
is removed by hand in the IDE then Xcode may crash due to dangling
references from the others.
Since cmGlobalXCodeGenerator::AddDependTarget is called exactly once for
every edge in the target dependency graph we do not need to keep track
of which edges have been visited to avoid generating duplicates.
The modern cmComputeTargetDepends guarantees there will by no cycles.
Drop the Xcode generator's check which only handled length 1 and 2
cycles anyway.
f616ff2 Update generator to use new cmGeneratorTarget api.
674bd6e Merge branch 'master' into sublime-text-2-generator
4760ead Remove ability to generate sublime clang files.
da0898e Correct missing parameter to CMP0018Flags call.
badb40d Merge branch 'master' into sublime-text-2-generator
b3ae61f Merge branch 'master' into sublime-text-2-generator
089d9cc SublimeText 2 Gen: Set the sublimeclang_options_script property.
44c2eee Merge branch 'master' into sublime-text-2-generator
304b885 Sublime Text 2 Gen: Per-source Compile flags are now saved in a separate file.
5b2aa3d SublimeText2 Gen: Fixed the issue where include directory flags used -D
90bcb77 SublimeText2 Gen: Improved use of define, include flags from CMAKE_C(XX)_FLAGS
cc84072 Merge branch 'master' into sublime-text-2-generator
8670cbe Define flags in CMAKE_C(XX)_FLAGS are now included in SublimeClang settings.
d022d4e Merge branch 'master' into sublime-text-2-generator
9cd3e70 Fixed Sublime Text project generation for in-source builds
44f35f7 Added a CMAKE_SUBLIMECLANG_DISABLED variable that disables SublimeClang.
...
SublimeClang is a optional plugin to SublimeText and I felt it
shouldn't be part of the generator for the following reasons:
1. Reduces the amount of sublime and sublimeClang specific
code we have to maintain inside CMake.
2. In testing the SublimeClang commands generated did not work
for the VTK project.
For people that do want this feature I recommend that they
looking into https://gist.github.com/robertmaynard/4724705 for a
way to use CMAKE_EXPORT_COMPILE_COMMANDS to generate JSON files
that can be used by SublimeClang.
Since the topic merged to master by commit 34a02846 (Merge topic
'ide-compiler-id', 2012-08-24), these variables are not used by
CMakeDetermine*Compiler.cmake for VS and Xcode generators. Drop the
code that sets them.
Since commit e015df7d (...delete CMakeFiles directory when cache is
deleted, 2006-02-20) we deleted the files in the CMakeFiles directory
when deleting CMakeCache.txt in order to reset the build tree to a fresh
state. This allowed commit fd33bf93 (fix for bug 6102, allow users to
change the compiler, 2007-12-13) to delete CMakeCache.txt when the user
changes the compiler and CMakeFiles/CMake<lang>Compiler.cmake and other
platform information files would go with it to allow a fresh start.
Then commit 7195aca5 (Make platform information files specific to the
CMake version, 2012-08-24) moved the platform information files to a
subdirectory e.g. CMakeFiles/<version>/CMake<lang>Compiler.cmake where
<version> is the current CMake version. This causes the compiler change
logic to fail to remove all old compiler information. Then on the next
configuration CMake<lang>Compiler.cmake would set CMAKE_<lang>_COMPILER
back to the old value and re-trigger the compiler change logic. This
causes an infinite loop of cache deletion and compiler reset.
Fix this simply by teaching cmCacheManager::DeleteCache to remove the
entire CMakeFiles directory recursively whenever it removes an existing
CMakeCache.txt. This fully resets the build tree to configure with a
fresh compiler.
Since Xcode doesn't sort the files on its own, let's sort them in our
generator to make navigation easier. Visual Studio, QtCreator, and
kdevelop all display files sorted.
fde949d Don't add target-specific interface includes and defines to Qt 4 targets.
79ae968 Revert "Add a way to exclude INTERFACE properties from exported targets."
71bf96e Revert "find_package: Reword <package>_NO_INTERFACES documentation"
3df36b5 Revert "Add the $<LINKED:...> generator expression."
e1f9080 Don't populate INTERFACE includes and defines properties in tll.
567c8d1 Revert "Don't allow utility or global targets in the LINKED expression."
a1c4905 Use the link information as a source of compile definitions and includes.
5c9f5e3 Don't use LINKED where not needed.
5b88504 Rename the IncludeDirectoriesEntry to be more generic.
b030323 Fix determination of when we're evaluating compile definitions.
The GNU compiler front-ends on AIX invoke the linker with flags of the
form "-L/path/to/gnu/runtime/lib" to tell ld where to find the language
runtime libraries. They depend on the default libpath behavior
documented in "man ld" to add the -L paths also to the runtime libpath
so the dynamic loader can find the language runtime libraries. This
differs from platforms whose linkers have distinct -rpath flags that
non-system compilers can use to tell the dynamic loader where to find
their language runtime libraries.
Since commit 96fd5909 (Implement linking with paths to library files,
2008-01-22) CMake always passes "-Wl,-blibpath:" followed by any
project-defined RPATH plus "/usr/lib:/lib" in order to explicitly set
the runtime libpath and avoid getting all the project -L paths in the
runtime libpath. The explicit libpath prevents the GNU compiler runtime
library -L paths from being placed in the libpath and then the dynamic
loader fails to find the language runtime libraries.
CMake already detects the implicit link directories for each language
since commit 07ea19ad (Implicit link info for C, CXX, and Fortran,
2009-07-23). Add the implicit link directories to the explicit runtime
libpath for GNU compilers on AIX to fix this use case.
This reverts commit 2c3654c3de.
The removal of some tests added in commit 77cecb77 (Add includes and compile
definitions with target_link_libraries., 2012-11-05) are also squashed
into this commit.
This is a partial revert of commit 77cecb77 (Add includes and compile
definitions with target_link_libraries., 2012-11-05).
As the interface includes and defines are now determined by the link
closure, there is no need to populate the corresponding properties
explicitly.
After evaluating the INTERFACE_INCLUDE_DIRECTORIES, of a target in a
generator expression, also read the INTERFACE_INCLUDE_DIRECTORIES of
its link interface dependencies.
That means that code such as this will result in the 'user' target
using /bar/include and /foo/include:
add_library(foo ...)
target_include_directories(foo INTERFACE /foo/include)
add_library(bar ...)
target_include_directories(bar INTERFACE /bar/include)
target_link_libraries(bar LINK_PUBLIC foo)
add_executable(user ...)
target_include_directories(user PRIVATE
$<TARGET_PROPERTY:bar,INTERFACE_INCLUDE_DIRECTORIES>)
Also process the interface include directories from direct link
dependencies for in-build targets.
The situation is similar for the INTERFACE_COMPILE_DEFINITIONS. The
include directories related code is currently more complex because
we also need to store a backtrace at configure-time for the purpose
of debugging includes. The compile definitions related code will use
the same pattern in the future.
This is not a change in behavior, as existing code has the same effect,
but that existing code will be removed in follow-up commits.
Since commit 34c882a9 (Allow VS 7 project Rebuild and Solution Rebuild to
work, 2007-11-10) we use a "CMakeFiles/generated.stamp" and some
associated files in the build tree to avoid re-running CMake when the
inputs have not changed but VS has cleaned the outputs it knows about.
When we do not really need to re-run we restore the generated.stamp file.
The non-re-run case can happen in multiple targets in parallel in VS >= 10
so we must restore the file atomically to avoid races. Write the stamp
file to a random temporary name and then atomically rename it to the real
stamp file.
ba48e63 Generate config-specific interface link libraries propeties.
deb51a7 Remove unused forward declarations.
9712362 Don't allow utility or global targets in the LINKED expression.
faa927e Make sure INTERFACE properties work with OBJECT libraries.
510fdcb Whitelist target types in target_{include_directories,compile_definitions}
4de7178 Ensure that the build interface includes have been added.
df74bc3 Only append build interface include dirs to particular targets.
d4e5c67 Don't keep track of content determined by target property values.
1fb545a Move a special case for PIC from the genex to the cmTarget code.
57175d5 Only use early evaluation termination for transitive properties.
4cf161a Fix determination of evaluating link libraries.
3a298c0 Fix generation of COMPILE_DEFINITIONS in DependInfo.cmake.
655e98b Ensure type specific compatible interface properties do not intersect.
46e2896 The COMPATIBLE_INTERFACE does not affect the target it is set on.
5f926a5 Test printing origin of include dirs from tll().
7c0ec75 De-duplicate validation of genex target names.
...
Since commit fd33bf93 (fix for bug 6102, allow users to change the
compiler, 2007-12-13) we keep an internal ;-list of language compiler
variable and value pairs. Preserve empty values on expansion to ensure
that the key/value pairing remains consistent.
This is needed in the case that Automoc is used, as that calls
GetIncludeDirectories, which may cache the resulting include dirs
too early in the generate step.
Also, because the automoc step is so early, we can't cache the
include directories at that point. At that point the build interface
of all dependencies are not populated yet, so we'd be caching the
includes before appending the build interface. Only start caching
when we're definitely generating the buildsystem. At that point, the
includes should be stable.
We still need to invoke AppendBuildInterfaceIncludes
in the GlobalGenerator because the build interface includes affect
mostly the dependencies of targets (such as the automoc targets),
rather than the targets themselves, so the build interface needs
to be appended for all targets before generation is done.
The ctest_configure command already reads the CTEST_CMAKE_GENERATOR
variable to get the value for the cmake -G option. Read new variable
CTEST_CMAKE_GENERATOR_TOOLSET for -T.
The "ctest --build-and-test" mode already has "--build-generator" to
specify the -G option to CMake. Add a "--build-generator-toolset" option
to specify the -T value.
Implement generator toolset selection (cmake -T) for VS >= 10 by setting
the PlatformToolset. Extend the RunCMake.GeneratorToolset test case to
verify CMAKE_GENERATOR_TOOLSET when the generator supports -T.
Since commit 485a940e (VS: Simplify MSVC version reporting, 2012-08-23)
all MSVC version information is detected during the compiler id step
from the actual compiler invoked by generated build systems rather than
hard-coded in VS generators. Therefore we can set the PlatformToolset
in VS >= 10 project files and support toolsets from other VS versions.
Reject the option by default. It will be implemented on a per-generator
basis. Pass the setting into try_compile project generation. Add cache
entry CMAKE_GENERATOR_TOOLSET and associated variable documentation to
hold the value persistently.
Add a RunCMake.GeneratorToolset test to cover basic "-T" option cases.
Verify that CMAKE_GENERATOR_TOOLSET is empty without -T, that -T is
rejected when the generator doesn't support it, and that two -T options
are always rejected.
This tracking was added during the development of commit 042ecf04
(Add API to calculate link-interface-dependent bool properties
or error., 2013-01-06), but was never used.
It was not necessary to use the content because what is really
useful in that logic is to determine if a property has been implied
to be null by appearing in a LINK_LIBRARIES genex.
I think the motivating usecase for developing the feature of
keeping track of the targets relevant to a property was that I
thought it would make it possible to allow requiring granular
compatibility of interface properties only for targets which
depended on the interface property. Eg:
add_library(foo ...)
add_library(bar ...)
add_executable(user ...)
# Read the INTERFACE_POSITION_INDEPENDENT_CODE from bar, but not
# from foo:
target_link_libraries(user foo $<$<TARGET_PROPERTY:POSTITION_INDEPENDENT_CODE>:bar>)
This obviously doesn't make sense. We require that INTERFACE
properties are consistent across all linked targets instead.
We need to make sure expressions which evaluate TARGET_PROPERTY:TYPE
multiple times for example get the correct result each time, and
not an empty string instead.
As INTERFACE_COMPILE_DEFINITIONS are now possible, we can have
situations like this:
add_library(foo ...)
add_library(bar ...)
target_link_libraries(foo bar)
target_compile_definitions(bar INTERFACE SOME_DEF)
The INTERFACE_COMPILE_DEFINITIONS of bar determine how foo should be
compiled, and if they change, foo should be rebuilt.
Additionally, as of commit d1446ca7 (Append the COMPILE_DEFINITIONS
from the Makefile to all targets., 2012-09-17), we don't need to
read definitions from the makefile if we read them from the target,
so also de-duplicate the cached info.
The DependInfo for INTERFACE_INCLUDE_DIRECTORIES is already handled
correctly.
Remove the old "nocwd" behavior that attempts to provide the "--help"
output when no working directory is available. The long help output
scrolls the real error message out of typical terminal sizes. Also
execute as little code as possible when no working directory exists.
Since commit 08cb4fa4 (Process generator expressions in the
INCLUDE_DIRECTORIES property., 2012-09-18), it is possible to use
generator expressions with the include_directories command.
As that command can also have a SYSTEM argument, ensure that the
result of using that argument with generator expressions gives a
sane result.
e48d842 Cache context-independent includes on evaluation.
089fe1c Optimize genex evaluation for includes and defines.
179f495 find_package: Reword <package>_NO_INTERFACES documentation
e7b579b Test workaround of bad interface include directories from depends.
77cecb7 Add includes and compile definitions with target_link_libraries.
0b92602 Add the $<LINKED:...> generator expression.
0fa7f69 Add API to check if we're reading a includes or defines property.
2c3654c Add a way to exclude INTERFACE properties from exported targets.
d4297d5 Export targets to a targets file, not a Config file.
df4d2b2 Make it an error for INSTALL_PREFIX to be evaluated.
7ceeba9 Advance more when preprocessing exported strings.
30268b4 Handle reading empty properties defined by the link interface.
The commit 18a3195a ('Keep track of INCLUDE_DIRECTORIES as a vector
of structs.', 2012-11-19) moved the handling of includes from
cmGeneratorTarget to cmTarget, but in the process introduced this
bug.
link.exe has problems with very very long lines in rsp files too.
Use $in_newline instead of $in variable for rspcontent which
separates the arguments with a newline instead of a simple space
and was specially made for this purpose.
Since commit d46d8df0 (Re-implemented cmGeneratedFileStream to look like a
real stream and replace the destination file atomically, 2004-11-03) our
RenameFile implementation tries to deal with MoveFile not replacing
read-only files. In order to avoid the Get/SetFileAttributes pair we used
stat to test for existence of the destination file. This has a race in
which the destination could be created between the test for existence and
the MoveFile call.
Remove the stat call and use GetFileAttributes to detect whether the file
exists. This shortens the race but does not eliminate it. Use a loop to
try multiple times in case we lose the race. While at it, drop Win9x
support and always use MoveFileEx.
Generator expressions whose output depends on the configuration
now record that fact. The GetIncludeDirectories method can use
that result to cache the include directories for later calls.
GetIncludeDirectories is called multiple times for a target
for each configuration, so this should restore performance for
multi-config generators.
While porting boost to use these features, the generation step took
too long (several minutes before I stopped it). The reason was that
the boost libraries form a large interdependent mesh. The libraries
list their dependencies in their INTERFACE such as:
$<LINKED:boost::core>;$<LINKED:boost::config>;$<LINKED:boost::mpl>
As boost::core already depends on the boost::config libraries, that
expression has no impact on the end-content, as it is removed after
the generation step. There is no DAG issue though, so the generator
expression evaluation would fully evaluate them. In the case of the
config library, it also depends on the core library, so all depends
are followed through that again, despite the fact that they've just
been evaluated. After this patch, the evaluation skips libraries if
they have already been seen via depends or directly in the content.
This patch keeps track of targets whose INTERFACE has been consumed
already. The INCLUDE_DIRECTORIES and COMPILE_DEFINITIONS properties
are whitelisted because repeated content will be stripped out later
during generation. For other properties now and in the future, that
may not be the case.
This establishes that linking is used to propagate usage-requirements
between targets in CMake code. The use of the target_link_libraries
command as the API for this is chosen because introducing a new command
would introduce confusion due to multiple commands which differ only in
a subtle way.
This is both a short form of using a TARGET_DEFINED expression
together with a TARGET_PROPERTY definition, and a way to strip
non-target content from interface properties when exporting.
Projects set interface requirements upstream, and existing
downstreams use of target_link_libraries will consume those interfaces.
This can create a backward compatibility concern as the result may
be changing the order of include directories of downstreams, or another
side-effect of using the INTERFACE properties.
Provide a way for them to emulate the behavior of a version-based
policy in the config file.
Both commit 8a37ebec (Add the target_include_directories command,
2013-01-01) and commit fc61a7a7 (Add the target_compile_definitions
command, 2013-01-08) added command implementations deriving from the new
cmTargetPropCommandBase class. Fix cmTypeMacro declarations of the
inheritance relationship.
7bf490e Make subclasses responsible for joining content.
f6b16d4 Don't allow targets args in the new target commands.
b3a7e19 Make the Property name protected so that subclasses can use it.
Kitware Robot
Brad King
Bill Hoffman
Christian Maaser
Stephen Kelly
Alex Neundorf
Clinton Stimpson
Yichao Yu
John Farrier
Cédric OCHS
Rolf Eike Beer
Matthew Bentham
Robert Maynard
Andreas Mohr
Jack O'Connor
Victor Zverovich
Marc Bartholomaeus
Ian Monroe
Petr Kmoch
Paul Kunysch
Jean-Christophe Fillion-Robin
Peter Kümmel
Anton Helwart
Gerald Hofmann
Alexander Chehovsky
Mark Salisbury
Patrick Gansterer