See http://public.kitware.com/Bug/view.php?id=10346.
The proposed patch for the issue could not be applied as is
because the SOURCE_DIR always exists for an ExternalProject_Add
call by the time we get to the place to emit the potential error.
The fix is to emit the error only if the source dir is empty.
By which, I mean devoid of files and subdirectories. If
SOURCE_DIR is used by itself, without any DOWNLOAD_COMMAND
or repository info, then it implies that the SOURCE_DIR is ready
to build as-is without need for a download step. Clearly, if it
is empty, then it is not ready to build as is. So complain if
the SOURCE_DIR is empty.
Commit 82c081ba (Fix rpath-link flag for SunPro C++ on Linux,
2009-07-13) taught CMake to pass '-rpath-link' because SunPro C++ 5.9
does not support '-Wl,'. Now SunPro C++ 5.11 does not recognize the
option without using '-Wl,'. Detect whether to use '-Wl,' based on the
output of "sunCC -flags".
Use it from ExternalProject and the ExternalProject test's
CMakeLists file rather than having duplicate find_program calls.
Add logic so that we do not try to use *.cmd variants of git
programs when using the MSYS Makefiles generator. Should fix
the last remaining dashboard issue with the new ExternalProject
git support additions.
Also, correct minor problem regarding placement of the local git
repo during test execution. On clean builds, it was being placed
incorrectly because of the ../.. relative reference. Use an absolute
path to place the local git repo in the proper directory, and only
use the relative reference when referring to it.
Add archives of these file types and add to the test
cases covered in the ExternalProject test.
Also add an "Example" directory in the Tests/ExternalProject
directory containing the canonical simplest example of
ExternalProject usage.
The ENABLE_EXPORTS property exports all symbols from executables on
UNIX-like platforms, typically for use by plugins. Honor this behavior
on Cygwin. See issue #10122.
Improve FILE(DOWNLOAD ...):
- Add percent complete progress output to the FILE DOWNLOAD
command. This progress output is off by default to
preserve existing behavior. To turn it on, pass
SHOW_PROGRESS as an argument.
- Add EXPECTED_MD5 argument. Verify that the downloaded
file has the expected md5 sum after download is complete.
- Add documentation for SHOW_PROGRESS and EXPECTED_MD5.
When the destination file exists already and has the
expected md5 sum, then do not bother re-downloading
the file. ("Short circuit" return.)
Also, add a test that checks for the status output
indicating that the short circuit behavior is actually
occurring. Use a binary file for the test so that the
md5 sum is guaranteed to be the same on all platforms
regardless of "shifting text file line ending" issues.
Improve ExternalProject:
- Add argument URL_MD5.
- Add verify step that compares md5 sum of .tar.gz file
before extracting it.
- Add md5 check to download step, too, to prevent
unnecessary downloads.
- Emit a warning message when a file is not verified.
Indicate that the file may be corrupt or that no
checksum was specified.
Fixes issue http://public.kitware.com/Bug/view.php?id=10258
Also, fix complaint that DOWNLOAD_COMMAND cannot contain arguments
consisting entirely of upper case letters. It validly does when,
for example, you construct a custom cvs command line and the module
name is all upper case, like VTK.
Map to the platform and compiler information for GNU because the
compilers are command-line compatible for common operations. Later we
can add Clang-specific features as necessary. We honor the preferred
capitalization is "Clang", not the common mis-spelling "CLang".
CUDA_VERSION_MAJOR and CUDA_VERSION_MINOR were only computed when CUDA_VERSION was first
computed. Subsequent runs of FindCUDA would not have CUDA_VERSION_MAJOR/MINOR set. We
now extract the major and minor versions from the CUDA_VERSION cache variable every run.
Was not removing definition flags (-D...) from cxx flags,
when the definition flag was last in the list returned from
wx-config.
--HG--
extra : rebase_source : 3452c52b92717f181e902abef38c1e2718ce3b27
Was not removing definition flags (-D...) from cxx flags,
when the definition flag was last in the list returned from
wx-config.
--HG--
extra : rebase_source : 3452c52b92717f181e902abef38c1e2718ce3b27
Synced from KDE svn: pkg_check_modules() and pkg_search_module() now
both support a QUIET keyword. When given, no messages will be printed (except the REQUIRED ones)
This also fixes#10469 (confusing output of FindLibXml2.cmake)
Alex
After configuring CMakeFiles/CMake<lang>Compiler.cmake in the build tree
the second time (to store ABI information), include it immediately.
This allows any logic and settings in the compiler information files to
be used without duplicating it in CMakeDetermineCompilerABI.cmake.
The change in commit "Use Fortran ABI detection results conservatively"
(2010-05-05) needs this to use the same logic to set CMAKE_SIZEOF_VOID_P
during first and later runs of CMake.
We set CMAKE_Fortran_SIZEOF_DATA_PTR in the Fortran compiler information
file after detecting the compiler ABI. However, since Fortran does not
really have pointers, the preprocessor-based detection is unreliable.
The result is needed to set CMAKE_SIZEOF_VOID_P only for Fortran-only
projects because the value can come from C or C++ compilers otherwise.
Therefore when CMAKE_SIZEOF_VOID_P is available from another language we
should defer to it.
The expectation of users of the MSVC60, MSVC70, MSVC71, MSVC80, MSVC90
and the new MSVC10 variables is that at most one of them will be set
for any given build tree. This change enforces that expectation for
build trees using Makefile generators. It also fixes the one mismatch
in that expectation to be found in the Visual Studio generator world:
previously, the VS 7.1 generator would set *both* MSVC70 and MSVC71;
now, it only sets MSVC71.
With these changes, user expectations are now met, and the recently
introduced CheckCompilerRelatedVariables test should pass everywhere.
This means that the user no longer sees this value _but_ this is backwards compatible because setting JASPER_LIBRARIES had no effect previously because we would override it using set()
As reported on the mailing list, find_path/file/library/program() basically don't work
at all if CMAKE_FIND_ROOT_PATH is set and searching in the host system directories
is disabled. This patch adds /include, /lib and /bin to the search directories, so they
will be appended to CMAKE_FIND_ROOT_PATH so this will work for the "Generic" platform (embedded
systems without OS)
Alex
(I accidentially removed ExternalProject.cmake from git by doing
mv ExternalProject.cmake ExternalProject.cmake.save
git checkout master
which I hoped would basically do a revert as it does with svn and cvs, but it
deleted the file from git)
Alex
Detect the runtime linker's search path and add to the compile time
linker's search path. This is needed because OpenBSD's static linker
does not search for shared library dependencies in the same places as
the runtime linker.
Teach compiler identification to support values such as
export CC='gcc -g -O2'
by separating the arguments on spaces. We already do this for the
values of CFLAGS, CXXFLAGS, and FFLAGS.
This allows the user not to link to the common libraries,
which are regularly required. The user must specify all
libraries that he does want to link in the find_package
line (png tiff jpeg zlib regex expat).
--HG--
extra : rebase_source : df29f96c957600629a34a1c5fafb8b3d6f274e22
Allow the user to set the CMake variable CTEST_COST_DATA_FILE, which will be used to store the cost data from test runs. If not set, defaults to the original location in the build tree Testing/Temporary dir.
Commit "Modernize GNU compiler info on Windows" (2009-12-02) reorganized
GNU flags on Windows but let -fPIC slip through for compilation of
objects in shared libraries. While this flag is valid on most GNU
compiler platforms we need to suppress it in Windows-GNU.cmake just as
we already do in CYGWIN-GNU.cmake.
This modifies the behavior of PYTHON_WRITE_MODULES_HEADER, should be backwards
compatible. Also marked a couple of the variables generated by adding Python
modules as advanced.
PathScale Fortran mangles module symbols as "MY_SUB.in.MY_MODULE" and
also requires "my_module_" when the module is imported. We cannot
provide the symbol with ".in." mangling so we should not provide
"my_module_" because it would duplicate the one in the Fortran-provided
object file.
Commit "FortranCInterface: Fix PathScale detection" (2010-01-22) already
made the same fix for the non-underscore module case.
In the CTest module we previously warned if the source directory did not
contain known version control directories. The message was:
"CTest cannot determine repository type. Please set UPDATE_TYPE
to 'cvs' or 'svn'. CTest update will not work."
This was confusing when building sources from a tarball. Furthermore,
we now support many more version control tools. This feature is now
mature enough that the warning causes confusion more than it provides
real help. We simply remove it.
The compiler documents symbols _DF_VERSION_ and _VF_VERSION_ but they do
not seem to be available to the preprocessor. Instead we add a vendor
query table entry for Compaq. Running "f90 -what" produces
Compaq Visual Fortran Optimizing Compiler Version ...
This clearly identifies the compiler.
At least one Fortran compiler does not provide a preprocessor symbol to
identify itself. Instead we try running unknown compilers with version
query flags known for each vendor and look for known output. Future
commits will add vendor-specific flags/output table entries.
PathScale Fortran mangles module symbols as "MYSUB.in.MYMODULE" and also
requires "mymodule_" when the module is imported. We cannot provide the
symbol with ".in." mangling so we should not provide "mymodule_" because
it would duplicate the one in the Fortran-provided object file.
In commit "use export all symbols on cygwin" (2003-01-21) we started
passing -Wl,--export-all-symbols when linking shared libraries. Now
cygwin exports all symbols automatically if no symbols are explicitly
exported. When symbols are explicitly exported we want to honor that
narrow interface. Therefore this flag should not be passed.
Change based on patch from issue #10122.
The variable should contain the name of a library needed to link the
symbol equivalent to dlopen. On Cygwin no special library is needed,
and certainly not "gdi32".
Change based on patch from issue #10122.
After discussing with Brad and Clinton:
-the namespace for the imported targets is now "Qt4::", tested with Makefiles, Visual Studio and XCode projects
-the imported targets are always created
-if QT_USE_IMPORTED_TARGETS is set to TRUE (it defaults to FALSE), the QT_QTFOO_LIBRARY variables are set to point to these imported
targets, otherwise the old behaviour is used.
-on OSX if Qt has been found as framework, disable QT_USE_IMPORTED_TARGETS, since cmake doesn't handle the framework directory as location of the library correctly
Alex
-set the type of the IMPORTED libraries to UNKNOWN, this way also on Windows
only the "LOCATION" property has to be set
-the if() around the SET(QT_${basename}_FOUND 1) was useless (always true)
-the mapping of the configuration types DEBUG and PROFILE did not belong here
Alex
This commit syncs FindQt4.cmake again with KDEs version.
Now for every Qt library an imported target with the name
Qt4ImportedTarget__<LIBNAME> is created.
This way we can now finally handle the release and debug versions of the Qt
libraries correctly.
Also, if a Qt-using project A installs a file with exported targets, these
targets now depend on the imported Qt targets, e.g.
Qt4ImportedTarget__QtCore. The location of QtCore is then resolved at
buildtime of project B, which uses the exported targets from project A.
Before this patch the full path to the QtCore on the original build machine
of project A was stored, so this had to match the directory layout on the
build machine for project B.
Alex
While Cygwin supports linking directly to .dll files, the behavior is
now discouraged. All Cygwin packages now provide import libraries of
the form lib*.dll.a and CMake has built the import libraries for years.
We believe it is now safe to stop explicitly searching for .dll files
because their import libraries will always be available when the
corresponding header files are available. Users can always set
find_library cache entries to point at a .dll file by hand if they
really must use one.
Change based on patch from issue #10122.
Commit "Find locally installed software first" made /usr/local the first
prefix searched to be consistent with the Filesystem Hierarchy Standard:
http://www.pathname.com/fhs/
The standard also implies that the root prefix "/" should not have any
package or development files. The "/bin" and "/lib" directories should
have only minimal contents to boot the system. No "/include" ever
exists. This commit re-orders the search path prefix list from
/usr/local
/
/usr
to
/usr/local
/usr
/
to prefer package and development files over low-level system files.
See issue #10136.
On Cygwin /usr/lib == /lib and /usr/bin == /bin. This change also makes
search results report locations as "/usr/..." instead of "/lib/...".
See issue #10122.
We add compiler information files
Compiler/PathScale-<lang>.cmake
to specify PathScale compiler information for C, C++, and Fortran
languages. We use a macro in Compiler/PathScale.cmake to consolidate
the information common to all languages.
The commit "Drop -rdynamic from Linux build rules" removed default use
of the flag on Linux. It was expected to be compatible because any
project using plugins should set ENABLE_EXPORTS on its executables to
export their symbols for use by the plugins in a cross-platform way.
However, it is possible to build without ENABLE_EXPORTS and load plugins
that do not link to any symbols from the executable explicitly. These
plugins may need to see RTTI and other executable symbols needed by the
language implementation. Executables using such plugins were broken by
the change.
If we want to remove the -rdynamic flag in the future we should do so in
a compatible way. At that time we should also remove equivalent flags
on other platforms (like -bexpall on AIX). We will either need a policy
or an explicit API to disable symbol exports on executables.
The primary purpose of the above-mentioned commit was to avoid passing
the -rdynamic flag to compilers on Linux that do not support it. In
this commit we restore the flag but only on GNU and Intel compilers
which are known to support it.
See issue #9985.
Implement Fortran 32/64-bit ABI detection on some platforms. We need to
set CMAKE_SIZEOF_VOID_P correctly in Fortran-only projects so that the
find_library() command knows whether to look for 64-bit binaries. We
also detect ELF binaries to enable RPATH replacement. See issue #10119.
Now the case that both the release- and the debug-version of a library is
handled first, because otherwise we always ran into this branch, since the
debug-only and the release-only branch also set both variables.
Alex
Clinton Stimpson
David Cole
Brad King
Zach Mullen
Yaakov Selkowitz
Kovarththanan Rajaratnam
James Bigler
Alex Neundorf
Mathieu Malaterre
Bill Hoffman
Miguel A. Figueroa-Villanueva
C. Bergström
Eric NOULARD
Chuck Atkins
Michael Wild
David Partyka
Marcus Hanwell
James Bigler
Philip Lowman