Teach the export command to handle export sets defined by invocations
of install(TARGETS ... EXPORT foo). This makes maintenance of targets
exported to both the build tree and install tree trivial.
Code such as
target_include_directories(foo INTERFACE
$<INSTALL_INTERFACE:include$<FOO>>
)
should be treated as a relative directory, despite the genex, after
the INSTALL_INTERFACE is stripped away.
Previously, this would generate a relative directory on export, which
would be an error on import, so no policy is needed.
Code such as
install(TARGETS ...
INCLUDES DESTINATION $<INSTALL_INTERFACE:include>
)
should behave as if the INSTALL_INTERFACE wrapper were not present.
In code such as
install(TARGETS ...
INCLUDES DESTINATION $<FOO>include
)
the generator expressions are evaluated at generate-time. Delay
determining whether each entry is a relative path until after
the generator expressions are evaluated. Such relative paths
are based relative to the CMAKE_INSTALL_PREFIX.
Introduce a policy to control the behavior.
The AliasTargets unit test already tests that using a
double-semicolon in the name is not an error. Change the ExportImport
test to use a namespace with a double-semicolon too.
Re-insert the semicolon which was removed during splitting.
Commit d777b8e7 (Genex: Allow relative paths in INSTALL_INTERFACE.,
2013-07-25) introduced the prefixItems method to allow relative paths
in the argument of the INSTALL_INTERFACE expression. That method was
buggy in that it did not re-introduce the semicolon separator in
the result.
This bug also affects paths which are already absolute in user code.
The new feature of install(TARGETS ... INCLUDES DESTINATION) introduced
in commit 650e61f8 (Add a convenient way to add the includes install
dir to the INTERFACE., 2013-01-05) introduced this crash. If the
new feature is used with a target which has no
INTERFACE_INCLUDE_DIRECTORIES, a segfault occurred.
Export the INCLUDES DESTINATION without appending to the
INTERFACE_INCLUDE_DIRECTORIES of the target itself. That way, a target
can be exported multiple times with different INCLUDES DESTINATION
without unintended cross-pollution of export sets.
The test uses generate_export_header(cmp0022OLD ...) to generate the
cmp0022_export.h header used by both cmp0022OLD and cmp0022NEW. In
the latter the _EXPORTS symbol does not match what the header expects
so the library does not export anything. The Watcom linker does not
like to create shared libraries that do not export any symbols.
Fix this by setting the DEFINE_SYMBOL property on cmp0022NEW to match
that of cmp0022OLD as the header expects.
3e30d9e TLL: Don't populate old link interface if CMP0022 is NEW.
574fec9 Export: Generate INTERFACE_LINK_LIBRARIES property on targets.
d0a76ea Introduce the INTERFACE_LINK_LIBRARIES property.
ddde61c Introduce the LINK_ONLY generator expression.
5aa9731 GenexEval: Add abstracted access to link interface for a target.
This property is generated only for targets which have recorded
policy CMP0022 as NEW, and a compatibility mode is added to
additionally export the old interfaces in that case too.
If the old interfaces are not exported, the generated export files
require CMake 2.8.12. Because the unit tests use a version which
is not yet called 2.8.12, temporarily require a lower version.
If a non-IMPORTED library is added to the INTERFACE_LINK_LIBRARIES
of a IMPORTED target, the non-IMPORTED target needs to become a
target dependency and link dependency of the consuming target.
This is already the case since commit 30962029 (Make targets depend
on the link interface of their dependees, 2012-12-26), and fixed in
the parent commit, so test that it works.
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'.
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.
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>
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.
Update the unit test introduced in commit 5daaa5c4 (Fix TARGET_PROPERTY
target extractions., 2013-01-26) to not use the expression, but still
test the appropriate code.
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.
We need to make sure we can export targets which have content such
as $<0:$<TARGET_PROPERTY:not_a_target,INTERFACE_INCLUDE_DIRECTORIES>
That means making not finding a target non-fatal here.
The Config and IMPORTED_ variants may also contain generator
expressions.
If 'the implementation is the interface', then the result of
evaluating the expressions at generate time is used to populate
the IMPORTED_LINK_INTERFACE_LIBRARIES property.
1) In the case of non-static libraries, this is fine because the
user still has the option to populate the LINK_INTERFACE_LIBRARIES
with generator expressions if that is what is wanted.
2) In the case of static libraries, this prevents a footgun,
enforcing that the interface and the implementation are really
the same.
Otherwise, the LINK_LIBRARIES could contain a generator
expression which is evaluated with a different context at build
time, and when used as an imported target. That would mean that the
result of evaluating the INTERFACE_LINK_LIBRARIES property for
a static library would not necessarily be the 'link implementation'.
For example:
add_library(libone STATIC libone.cpp)
add_library(libtwo STATIC libtwo.cpp)
add_library(libthree STATIC libthree.cpp)
target_link_libraries(libtwo
$<$<STREQUAL:$<TARGET_PROPERTY:TYPE>,STATIC_LIBRARY>:libone>)
target_link_libraries(libthree libtwo)
If the LINK_LIBRARIES content was simply copied to the
IMPORTED_LINK_INTERFACE_LIBRARIES, then libthree links to libone, but
executables linking to libthree will not link to libone.
3) As the 'implementation is the interface' concept is to be
deprecated in the future anyway, this should be fine.
Ancient versions of CMake required else(), endif(), and similar block
termination commands to have arguments matching the command starting the
block. This is no longer the preferred style.
Run the following shell code:
for c in else endif endforeach endfunction endmacro endwhile; do
echo 's/\b'"$c"'\(\s*\)(.\+)/'"$c"'\1()/'
done >convert.sed &&
git ls-files -z -- bootstrap '*.cmake' '*.cmake.in' '*CMakeLists.txt' |
egrep -z -v '^(Utilities/cm|Source/kwsys/)' |
egrep -z -v 'Tests/CMakeTests/While-Endwhile-' |
xargs -0 sed -i -f convert.sed &&
rm convert.sed
In cmComputeLinkDepends we compute the transitive closure of private
shared library dependencies. When a shared library is added to this
closure we must follow all of its dependencies whether they are private
or public. Previously we only followed the private dependencies. Fix
the implementation to follow the public dependencies too. Also extend
the ExportImport test to cover this case.
We test that LINK_INTERFACE_MULTIPLICITY propagates through export() and
install(EXPORT) into dependent projects. A simple cycle of two archives
that need to be scanned three times ensures that the importing project
uses the multiplicity correctly.