The re-implementation in commit v3.5.0-rc1~116^2~1 (CMakeParseArguments:
replace by native cmake_parse_arguments command, 2015-12-05) introduced
a regression when parsing the ARGN arguments with cmake_parse_arguments.
The original implementation used
foreach(currentArg ${ARGN})
to iterate over input arguments. This flattened ;-lists within the
arguments whether they were quoted or not. Fix our new implementation
to preserve this behavior and add a test case to cover it.
Signed-off-by: Dimitar Yordanov <dimitar.yordanov@sap.com>
Signed-off-by: Matthias Maennich <matthias.maennich@sap.com>
Refactoring in commit v3.3.0-rc1~29^2~1 (cmState: Host some state from
the cmGlobalGenerator, 2015-05-24) moved storage of some generator
traits over to cmState. However, it accidentally removed initialization
of the values from the cmGlobalGenerator constructor. This is needed
because generator subclasses update the settings in their constructors.
Since a single cmState instance is shared across multiple build trees by
cmake-gui, initializing the values in its constructor is not enough.
Fix this by restoring the needed initializations to the
cmGlobalGenerator constructor.
We represent target dependency sets as `set<cmTargetDepend>` which
orders by a `cmGeneratorTarget const*` pointer value. Therefore the
order of dependencies encountered in AppendTargetDepends is not
predictable. Sort them by content to make the result deterministic.
Generate custom command build statements in the order we encounter
source files specifying them. Do not depend on pointer values of
internally allocated structures for ordering.
This version introduced CURLE_NOT_BUILT_IN which we have used since
commit v3.4.0-rc1~211^2~4 (cmCurl: Tolerate lack of CURLOPT_CAPATH
support, 2015-08-12). For older versions, just define the name to the
then-unused error code so that we can compile.
a1ad098d Tests: Avoid OS X 10.5 limitation warning in RunCMake.install test
47460f3e install(EXPORT): Fix crash on target in another directory
e86383e1 Tests: Use newer policy settings in RunCMake.install test
In commit v3.5.0-rc1~272^2~6 (cmGlobalGenerator: Add FindGeneratorTarget
API, 2015-10-25) a lookup was implemented via linear search. Replace it
with an efficient data structure.
Suggested-by: Stephen Kelly <steveire@gmail.com>
Refactoring in commit v3.5.0-rc1~272^2~13 (cmGlobalGenerator: Remove
direct storage of targets, 2015-10-25) replaced an efficient data
structure mapping from target name to cmTarget instance with a linear
search. Lookups through cmGlobalGenerator::FindTarget are done a lot.
Restore the efficient mapping structure with a name indicating its
purpose.
Reported-by: Bartosz Kosiorek <gang65@poczta.onet.pl>
Previously we did not clearly document that `--target` is only supported
to be specified once. Even worse, specifying it multiple times would
silently ignore any previously specified targets and only build the last
target.
Update the documentation to specify this. Update the implementation to
reject multiple `--target` options to prevent user errors.
Refactoring merged by commit v3.5.0-rc1~299 (Merge topic
'use-generator-target', 2015-10-20) in and around
commit v3.5.0-rc1~299^2~13 (cmExportSet: Store a cmGeneratorTarget,
2015-10-17) changed export sets to delay looking up actual targets and
stores only their names. However, in InstallCommand::HandleExportMode
we need to lookup targets immediately to check them for
EXPORT_LINK_INTERFACE_LIBRARIES. The check was accidentally made local
to the current directory, so if an export set contains a target from
another directory the lookup fails and CMake crashes. Fix the check to
look up the target name globally, and tolerate when no target is found
just in case.
Reported-by: Kelly Thompson <kgt@lanl.gov>
Refactoring in commit v3.5.0-rc1~347^2~2 (Set the current dirs on the
snapshot before creating the cmMakefile) accidentally changed the
source and binary directories configured in `cmake -E cmake_depends`
for use during dependency scanning. This can cause the wrong directory
information to be loaded. It also breaks Fortran module dependency
scanning for modules provided by targets in subdirectories that do
not have Fortran_MODULE_DIRECTORY set.
Fix the dependency scanning directory configuration and add a test to
cover the Fortran module case in which the breakage was observed.
Reported-by: Kelly Thompson <kgt@lanl.gov>
Let us take an example of a project that has some tests in a component
that need to be installed into a dedicated test package. The user
expectation is that the result could be achieved by typing the
following:
make
make tests
make install
DESTDIR=/testpkgs make install-tests
However this results in test components in the default installation as
well as the testpkg.
Add an EXCLUDE_FROM_ALL option to the install() command to tell it that
the installation rule should not be included unless its component is
explicitly specified for installation.
While evaluating `if(MATCHES)` we get a `const char*` pointer to the
string to be matched. On code like
if(CMAKE_MATCH_COUNT MATCHES "Y")
the string to be matched may be owned by our own result variables.
We must move the value to our own buffer before clearing them.
Otherwise we risk reading freed storage.
The add_custom_command(TARGET) signature only works for targets defined
in the current directory. Clarify this in the error message when the
target exists but was defined elsewhere.
Inspired-by: Bartosz Kosiorek <gang65@poczta.onet.pl>
70788e92 Remove temporary allocations when calling cmHasLiteral{Suf,Pre}fix.
bd2384f5 Optimize cmMakefile::ExpandVariablesInStringNew.
ad9394f4 Remove temporary allocations in cmMacroHelper::InvokeInitialPass.
f9599ed4 Remove temporary allocations by extending the lifetime of the retval.
275f2a85 Remove temporary allocations when calling cmGeneratorTarget::GetName.
Since commit v3.4.0-rc1~5^2~1 (VS: Add support for selecting the Windows
10 SDK, 2015-09-30) the VS 2015 generator requires a Windows 10 SDK to
be available when CMAKE_SYSTEM_VERSION specifies Windows 10 (e.g. when
building on a Windows 10 host). Howewver, it is possible to install VS
2015 without any Windows 10 SDK. Instead of failing with an error
message about the lack of a Windows 10 SDK, simply tolerate this case
and use the default Windows 8.1 SDK. Since building for Windows Store
still requires the SDK, retain the diagnostic in that case.
Refactoring merged by commit 9afbb733ec (Merge topic
'use-generator-target', 2015-10-20) in and around commit 381e7afd
(cmExportSet: Store a cmGeneratorTarget, 2015-10-17) forgot to update
one place in this method. This leads to a crash in code such as
add_library(A a.c)
add_library(B b.c)
target_link_libraries(B A)
install(TARGETS B DESTINATION lib EXPORT ExpB)
install(EXPORT ExpB DESTINATION lib/cmake/test)
add_executable(C c.c)
install(TARGETS C DESTINATION bin EXPORT ExpC)
Fix the target name reference to avoid using an unpopulated Target
pointer.
When the first argument passed is a std::string, we need to take it
by const&, otherwise we copy the string and trigger a temporary
allocation. This patch removes a few 10k temporary allocations when
running the CMake daemon on the KDevelop build dir.
This hotspot was found with heaptrack.
We can remove the temporary allocations required for the
default-constructed t_lookup passed into the openstack by refactoring
the code slightly. Furthermore, we use a vector instead of a stack,
since the latter is based on a deque which is not required for a
heap / lifo structure.
This patch removes ~215k allocations.
This hotspot was found with heaptrack.
This code used to convert std::string's to raw C strings only to
put that back into a std::string. This patch thus removes ~70k
temporary allocations when running the CMake daemon on KDevelop.
This hotspot was found with heaptrack.
This happens quite often from within comparisons such as in
NamedGeneratorTargetFinder or FindGeneratorTargetImpl. It is the
top hotspot of both, number of allocations as well as number of
temporary allocations - the majority of calls lead to temporary
allocations.
In raw numbers, this patch removes ~1E6 temporary allocations of
1.5E6 temporary allocations in total when running the cmake daemon
on the KDevelop build dir. That is 2/3 of the total.
This hotspot was found with heaptrack.
Since this command was introduced in 2002 it has incorrectly constructed
the child process command line by concatenating arguments separated by
spaces with no quoting. Fix this by passing the command argument vector
directly to RunSingleCommand without an intermediate quoting and
re-parsing step.
Reported-by: Andrey Pokrovskiy <wonder.mice@gmail.com>
1040e690 cmSystemTools: Teach RunSingleCommand to merge child pipes when possible
ce3b713b cmSystemTools: Simplify RunSingleCommand output string construction
dc039cc0 cmSystemTools: Drop redundant condition in RunSingleCommand
ffa2a8c9 cmSystemTools: Rename OUTPUT_NORMAL to OUTPUT_FORWARD to clarify its purpose
92e9bb21 cmcmd.cxx: Remove unused code in __run_iwyu implementation
fb1526f5 cmake: Change `-E chdir` to pass through stdout/stderr directly
Audit the code to make sure there are no callers that use OUTPUT_MERGE
with separate capture strings. Then change RunSingleCommand to
implement output merging by giving the child process a single pipe for
both its stdout and stderr descriptors. This will more cleanly merge
the content on atomic write boundaries in the child instead of on
arbitrary buffering boundaries in the parent.
The OUTPUT_NORMAL value is not really "normal" and has only one caller.
Rename it to OUTPUT_FORWARD to clarify that we are explicitly forwarding
the output.
aea1b036 Fix export of STATIC library PRIVATE dependencies with CMP0022 NEW
e5cbec14 Tests: Use CMP0022 NEW behavior in some ExportImport cases
0ca122fc Tests: Isolate policy changes in ExportImport test
Some compilers may add their own RPATH entries when invoking the linker.
For example, a GCC installation may contain the following definition in
the specs file:
*link_libgcc:
%D -rpath <<some specific rpath in which libstdc++.so can be found>>
In this case binaries may contain RPATH entries that CMake did not add.
When we update the RPATH on installation we must preserve these entries
even if CMake thinks the INSTALL_RPATH value should be empty.
Fix this by always using file(RPATH_CHANGE) and teach it to behave as
file(RPATH_REMOVE) if the actual RPATH in the file is empty after
replacing the build-tree RPATH with the install-tree RPATH. This will
preserve any compiler-added RPATH value instead of removing it.
The target_link_libraries command records the PRIVATE dependencies of a
STATIC library in INTERFACE_LINK_LIBRARIES as "$<LINK_ONLY:dep>". This
hides the target name from export namespacing logic inside a generator
expression. When user-written generator expressions reference a target
name they must put it inside a "$<TARGET_NAME:dep>" expression to allow
the export logic to rename the target. In the case that the private
dependency is not already a generator expression, target_link_libraries
must use "$<LINK_ONLY:$<TARGET_NAME:dep>>" to allow the export logic to
rename the target.
Reported-by: Tamás Kenéz <tamas.kenez@gmail.com>
Since commit v3.3.0-rc1~410^2~3 (ctest_test: Add QUIET option,
2015-02-17) if tests fail when QUIET is used one sees:
The following tests FAILED:
but not the subsequent line(s) indicating which tests failed.
Restore the list of failed tests.
Expand the -W set of cmake options to include support for the -Werror
and -Wno-error format, which is used to control upgrading and
downgrading warning and error messages. Implement support for these new
formats for the dev and deprecated message types.
Add tests and updated documentation for new options.
a57caf7e VS: Fix Windows 10 SDK version selection (#15831)
ad594de8 cmSystemTools: Add VersionCompareEqual helper
c173e37f VS: Do not select a partial Windows 10 SDK folder (#15831)
b3677b35 VS: Map the link `/debug` to its IDE property
c22da7cf VS: Drop unused condition in link debug flag generation
4ca9df8b cmIDEOptions: Add support for case-insensitive flags
In commit v3.4.0-rc1~5^2~1 (VS: Add support for selecting the Windows 10
SDK, 2015-09-30) we added Windows 10 SDK selection choosing the most
recent SDK that is not newer than the target version. This is backward
because it should be up to the application code to not use APIs newer
than the target version. It is up to the build system to provide a SDK
that has at least the APIs expected to be available for the target
version. Furthermore, since the default target version is the host
version of Windows, the old approach breaks when the only SDK available
is for a newer version of Windows.
Fix this by always selecting a Windows 10 SDK if one exists. Use the
SDK for the exact version if is available. Otherwise use the latest
version of the SDK available because that will have at least the APIs
expected for the target version.
Fix the link flag table entries for this flag to be case-insensitive.
Also fix the VS 2015 value for the build property enumeration name.
This causes `linkOptions.Parse(...)` to correctly extract the `/debug`
flag and map it to the IDE property instead. Therefore we do not need
to look for the flag explicitly when initializing the property.
Starting with VS 2015 the GenerateDebugInformation build property is an
enumeration (`No`, `Debug`, `DebugFastLink`) instead of a boolean value
(`false`, `true`). For now we simply change to `No` and `Debug` fix
current behavior. Support for `/debug:fastlink` can be added later.
There were issues in the special-character-escaping and line-wrapping
code which caused DragNDrop packaging to fail mysteriously at a later
step with parsing errors in the `sla.r` file generated by the following
code.
d8bc26a0 Xcode: Parse variant and genex for CMAKE_XCODE_ATTRIBUTE (#14947)
dc0ddb9e Xcode: Store configuration name along with XcodeObject (#14947)
28f98cee Xcode: Make CMAKE_XCODE_ATTRIBUTE calculation last step (#14947)
28db2268 Xcode: Factor out XCODE_ATTRIBUTE_ variant filter (#14947)
Before this change backslashes in strings were escaped during compile
flags adds via AppendFlag(). But global flags like OTHER_CPLUSPLUSFLAGS
are not added as flags but as plain strings so they were not escaped
properly.
Now the escaping is performed within cmXCodeObject::PrintString() which
ensures that strings are always encoded.
Create a `CMAKE_XDGDATA_DIR` option and add a corresponding flag to the
`bootstrap` script. This is needed for multiarch layouts where the
prefix is `/usr/${host}` but where architecture-independent files (like
the XDG-specific ones) are installed to `/usr/share`.